Lets do Multiple:
FIRST ONE:
Hydraulic Water Redistribution by Silver Fir (Abies alba Mill.) Occurring under Severe Soil Drought
by Paul Töchterle 1,2,*,†ORCID,Fengli Yang 3,†,Stephanie Rehschuh 1,Romy Rehschuh 1,Nadine K. Ruehr 1,Heinz Rennenberg 3 andMichael Dannenmann 1
1
Institute of Meteorology and Climate Research, Atmospheric Environmental Research (IMK-IFU), Karlsruhe Institute of Technology (KIT), Kreuzeckbahnstrasse 19, 82467 Garmisch-Partenkirchen, Germany
2
Institute of Geology, University of Innsbruck, Innrain 52f, 6020 Innsbruck, Austria
3
Institut für Forstbotanik und Baumphysiologie, Freiburg University, Georges-Köhler Allee 53/54, 79085 Freiburg, Germany
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this study.
Forests 2020, 11(2), 162; https://doi.org/10.3390/f11020162
Submission received: 20 December 2019 / Revised: 24 January 2020 / Accepted: 29 January 2020 / Published: 31 January 2020
(This article belongs to the Section Forest Ecophysiology and Biology)
Downloadkeyboard_arrow_down Browse Figures Versions Notes
Abstract
Hydraulic redistribution (HR) of water from wet- to dry-soil zones is suggested as an important process in the resilience of forest ecosystems to drought stress in semiarid and tropical climates. Scenarios of future climate change predict an increase of severe drought conditions in temperate climate regions. This implies the need for adaptations of locally managed forest systems, such as European beech (Fagus sylvatica L.) monocultures, for instance, through the admixing of deep-rooting silver fir (Abies alba Mill.). We designed a stable-isotope-based split-root experiment under controlled conditions to test whether silver fir seedlings could perform HR and therefore reduce drought stress in neighboring beech seedlings. Our results showed that HR by silver fir does occur, but with a delayed onset of three weeks after isotopic labelling with 2H2O (δ2H ≈ +6000‰), and at low rates. On average, 0.2% of added ²H excess could be recovered via HR. Fir roots released water under dry-soil conditions that caused some European beech seedlings to permanently wilt. On the basis of these results, we concluded that HR by silver fir does occur, but the potential for mitigating drought stress in beech is limited. Admixing silver fir into beech stands as a climate change adaptation strategy needs to be assessed in field studies with sufficient monitoring time.
Keywords: hydraulic redistribution; drought; silver fir; European beech; mixed stand
- Introduction
Hydraulic redistribution (HR) is the passive flux of water between wet- and dry-soil zones through plant roots as conduits. It is driven by soil-water potential gradients between dry- and wet-soil layers, and between roots and soil matrix [1,2]. Typically, HR occurs during the night, when transpiration has ceased [3,4,5]. Water can be redistributed in the upward (i.e., hydraulic lift [2,6,7]), downward (i.e., hydraulic descent [8,9,10,11]), and lateral directions [12,13,14,15]. Field observations showed that HR plays an important role in terrestrial ecohydrological cycles. Plants can benefit from HR through enhanced photosynthesis and transpiration [16], alleviated soil-moisture loss during the dry season [17], and a prolonged growing season [18,19]. These immediate benefits of HR consequently enhance nutrient acquisition [20], increase nutrient mobility, and facilitate root-litter decomposition [21,22].
HR has been documented in more than 100 species [23], including agricultural crops and grasses [24,25,26], and forest trees and shrubs [27,28,29]. Although HR has been observed in diverse climatic settings [16,23,30,31], it is most prevalent in arid and semiarid ecosystems, such as savannas [16,32,33], arid climates and semideserts [5,34,35,36], Mediterranean-type ecosystems [10,37,38,39], and tropical forests [40,41,42]. However, HR may become increasingly relevant in temperate ecosystems that are subject to intensified drying–wetting cycles due to extreme drought events in projected climate-change scenarios [27,43].
Several techniques can be used to identify HR under laboratory or field conditions. Reverse water flow in roots can be identified by sap-flow techniques like the heat-balance [29,44] or -ratio method [45,46]. However, quantification of reverse water flow in roots by sap-flow measurements can easily be disrupted by fluctuations in ambient temperature [46,47], or small-scale soil and geomorphic variability [10]. Furthermore, sap flow can only be measured in individual roots and upscaling to the root system is a significant source of error [48]. Contrastingly, measurements of soil-water potential near plant roots, accompanied by water stable isotope analyses, have been used to track water movement in soil [30,49,50]. In this context, stable isotopes, either at natural abundance or by using heavy isotope enriched water (i.e., H218O, 2H2O) as a tracer, are used as a novel technique to quantify water flow between dry- and wet-soil layers through plant roots and water uptake by adjacent plants [3,27,51].
In temperate forests, HR was only detected in a few tree species in the field, such as Norway spruce (Picea abies (L.) Karst.), Douglas fir (Pseudotsuga menziesii), ponderosa pine (Pinus Ponderosa), loblolly pine (Pinus taeda), and sessile oak (Quercus petraea) [51,52,53,54,55]. In an adult mixed oak/European beech forest, Zapater et al. [51] showed HR by oaks using an 18O-labelling approach but did not find any tracer material in European beech. However, both HR and the uptake of redistributed water by neighboring plants was detected in studies with seedlings of English oak, Norway spruce, and European beech under moderate drought in split-root systems in the greenhouse [27].
In Central Europe, European beech, being both an abundant natural tree species and a key species in forestry, was reported to be particularly vulnerable to drought [56,57]. As extreme drought events and intensified drying–wetting cycles are projected to become more prevalent [43,58,59], beech forests in Central Europe face consequences such as declining growth and drastic economic losses for forestry [60,61,62]. Admixing deep-rooting tree species could potentially increase the resilience of beech stands. In this context, silver fir was proposed due to its high productivity and presumably higher drought resistance [63,64]. Furthermore, recent studies indicate that water supply to European beech in mixed forest stands may be supported by the presence of silver-fir neighbors [65,66].
The aim of this study was to show if silver fir can perform HR under extreme drought conditions. For this purpose, we applied an improved split-root approach under controlled conditions, combined with the 2H2O labelling of water and in situ stable isotope analysis of soil moisture. We hypothesized that fir roots were able to allocate 2H2O from moist- to dry-soil zones by HR.
- Materials and Methods
2.1. Mesocosm Setup
This study was conducted using plant-soil mesocosms under controlled conditions in the scientific greenhouse at the KIT Campus Alpin in Garmisch-Partenkirchen, Germany. Temperature (T) and relative humidity (rH) were controlled and underwent daily cycles (T = 20.5 ± 4°C, and rH = 58 ± 12 % on average). Ambient CO2 concentration showed diurnal fluctuations between 380 and 450 ppm without long-term trends during the timespan of the experiment.
Six mesocosms, each comprising 2 nested polyvinylchloride (PVC) compartments (dimensions of inner and outer compartment: length × width × height = 90 × 38 × 40 cm³ and 20 × 20 × 10 cm³, respectively) were set up as shown in Figure 1 and Figure 2. Soil for the mesocosms was collected in autumn 2015 in the Black Forest close to Emmendingen (SW Germany). The material was taken from the Ah horizon of a Dystric Cambisol that originated from Triassic sandstone and showed a sandy loam texture (see [67] for site and soil details). The soil material was mixed with perlite at a volume ratio of 1:1. Perlite is a highly porous mineral that improves soil drainage and aeration properties while retaining moisture. These properties helped with the homogenization of soil moisture and isotope equilibration.
Forests 11 00162 g001 550
Figure 1. Experiment mesocosms. Two fir and 1 beech saplings were planted in polyvinylchloride (PVC) compartments, whereas a beech sapling was isolated within a smaller compartment in the center. A root strand of each fir was redirected into the beech compartment. Soil moisture was extracted via diffusion into dry air passing through gas-permeable tubing connected to a water isotope analyser (Picarro L 2130-i). Experiment comprised 6 such mesocosms; 3 were used as replicates and 3 as controls (see main text for more detail).
Forests 11 00162 g002 550
Figure 2. Photo of two mesocosms with silver-fir and European beech saplings in separate compartments.
Two silver-fir seedlings (3.5 years of age) were planted in the outer compartment (hereafter referred to as the fir compartment), and a single European beech seedling (2 years of age) in the inner compartment (hereafter referred to as the beech compartment). Root length of the fir and beech seedlings was, on average, 30 and 15 cm, respectively. A first-order coarse root with intact second- and third order fine roots of each silver fir seedling was redirected to the beech compartment. The mesocosms were then wrapped with plastic film to avoid evaporation and consequent contamination of the lab environment with 2H2O vapor. The fir-root strands were the only hydraulic connection between beech and fir compartments. As experiment control, 1 mesocosm was left unplanted but otherwise received identical treatment (Figure 1).
2.2. Environmental Parameters
Volumetric soil moisture and temperature were measured using DECAGON EM50 loggers (Decagon Devices, Inc., Pullman, Washington, DC, USA). A soil-moisture sensor (Type GS1 and 5TM, Decagon Devices, Inc., Pullman, WA, USA) was vertically installed in each beech compartment at approximately 10 cm depth. Volumetric water content (VWC) was recorded every 2 hours. Reported soil-moisture data were calibrated against gravimetric measurements of the same soil material used in the HR experiment. We observed erratic readings from soil-moisture sensors at very low soil-water contents. This had implications on the calculation of soil-water potential, as is explained in the following section.
2.3. Soil-Water Potential
Soil-water potential was calculated for the beech compartments with a widely used model for water retention in soils [68]:
𝜃=𝜃𝑟+(𝜃𝑠−𝜃𝑟)(1+(𝛼|𝛹|)𝑛)𝑚
(1)
where |𝛹|= 1𝛼×{((𝜃−𝜃𝑟)(𝜃𝑠−𝜃𝑟))−1𝑚−1}1𝑛
, and Ψ is the absolute value of soil-water potential (hPa/cm water column) at a specific volumetric water content θ (cm3 cm−3). θs and θr were the saturated and residual water content of the soil, respectively. Parameters α (hPa), n, and m are shape parameters. For fitting, θs and θr were determined from measurements, and α and n were estimated using the soil texture of the potting substrate (sandy loam texture) according to Hodnett et al. [69]. For calculation, we used gravimetrically calibrated volumetric soil moisture. We conducted predawn water potential (Ψpredawn) measurements on 2 October to validate model results. Ψpredawn was measured on beech branches using a Scholander pressure chamber [70] (Model 1000 Pressure Chamber Instrument, PMS Instrument Company, Albany, Oregon, USA). Plant Ψpredawn could be used as an estimate for soil Ψ on the basis of the assumption that plant Ψpredawn was in equilibrium with soil Ψ adjacent to roots. Water potential is expressed as pF values, which is the logarithm of the absolute values of Ψ (hPa). Due to potential biases inherent to VWC measurements at very dry soil conditions, pF calculation can be subject to large uncertainty leading to pF values above the physical limit of ~6.9.
2.4. Leaf–Gas Exchange Measurements
Leaf–gas exchange measurements on beech seedlings were performed from August to October, and captured the drought response of beech. We measured light-saturated photosynthesis (Asat) and stomatal conductance (gs) using a portable leaf–gas exchange system (Li-6400, LI-COR Inc., Lincoln, NE, USA) equipped with a light source (Li-6400-02B LED, LI-COR Inc., Lincoln, NE, USA). One green leaf per beech tree was measured under predetermined saturated light conditions (PAR = 1200 µmol m−2s−1), an average leaf temperature of 27.7 °C, and average relative humidity of 54%.
2.5. Hydraulic-Redistribution Experiment
Prior to the experiment, the VWC of beech compartments was maintained at roughly 15% by irrigation. Each mesocosm was then exposed to drought conditions by withholding irrigation for 25 days. On the night from 27 to 28 August, 2 L (i.e., 5.8 L m−2) of deuterated water (δ2H ≈ +6000‰) was injected into the bottom-soil zone of the fir compartment of 3 mesocosms containing trees and the unplanted control mesocosm. For each fir compartment, 6 evenly spread-out irrigation tubes (Figure 1) were used as injection ports in order to ensure the homogeneous distribution of deuterated water. The two remaining mesocosms containing trees received no label injection. Isotopic composition of soil moisture in the beech compartment was measured on 15 separate days during the 10 following weeks. Measurements generally took place during the morning hours until noon. Each beech compartment was re-watered on 29 October with 1.5 L of tap water in order to weaken the water-potential gradient between the fir and beech compartments, and, thus, the required conditions for HR.
2.6. Deuterated-Water Tracing
We used a membrane-inlet water-vapor sampling technique in soil coupled to a cavity ringdown laser spectrometer [71,72] to trace 2H2O transport from the fir to the beech compartment. For this purpose, the beech compartment of each mesocosm was equipped with gas-permeable tubing (Accurel PP V8/2HF polypropylene tubing, Membrana GmbH, Germany: 30 cm long sections buried at 5 and 10 cm depth) that was flushed with synthetic dry air (20% O2 in N2) during measurements (Figure 2). The dry air absorbed moisture upon passing the gas-permeable section. This approach resembled in situ soil air probes [72], but with double the length of gas-permeable tubes in each beech compartment. Accurel tubes are suitable for isotopic measurements, as they do not cause isotopic fractionation across a wide range of soil-moisture contents, and render possible vapor measurements in equilibrium with soil water [71].
A Picarro L 2130-i cavity ringdown spectrometer was used to analyze the stable isotopic composition (δ2H) of the gas stream. On measurement days, gas streams exiting each of the 6 mesocosms were continuously measured for at least 15 min. To minimize carryover artefacts, lines were flushed for 20 min between each measurement, and the values for the first 5 min of each measurement were ignored. Stable isotope data reported in the results section of this study refer to the average values of the remaining measurement time. Results of each day were scale-normalized to 2 in-house standards (δ2H = –235.0 ± 1.8 ‰ and 1.8 ± 0.9 ‰ VSMOW) that were calibrated against international reference materials (VSMOW2, SLAP2).
HR was quantified in this study as a percentage of excess deuterium (d) added to the fir compartment by labelling, that was recovered in the beech compartment. In other words, 2H excess recovery (∆d) corresponds to the needed amount of 2H from labelled water to reach the measured isotopic composition of a defined volume of water. We used isotopic measurements to calculate a mixing ratio between labelled water (i.e., δ2H = +6000‰) and background moisture using the isotopic composition of the control mesocosm as a baseline to account for natural variability. This mixing ratio is used to determine the amount of excess 2H in beech compartments, relative to excess 2H in the label water based on the VWC (θ) of the beech compartment and the amount of added label water (Equation 2):
𝛥𝑑= 𝛥𝑑𝑏𝑒𝑒𝑐ℎ𝛥𝑑𝑙𝑎𝑏𝑒𝑙×100= (𝜀𝐻2𝑏𝑒𝑒𝑐ℎ−𝜀𝐻2𝑐𝑜𝑛𝑡𝑟𝑜𝑙) 𝑚𝑏𝑒𝑒𝑐ℎ𝑚𝑂 𝜃(𝜀𝐻2𝑙𝑎𝑏𝑒𝑙−𝜀𝐻2𝑐𝑜𝑛𝑡𝑟𝑜𝑙) 𝑚𝑙𝑎𝑏𝑒𝑙𝑚𝑂 𝑀𝑙𝑎𝑏𝑒𝑙×100,
(2)
where ∆d is the percentage of 2H excess recovered in beech compartments, ∆dbeech and ∆dlabel are the 2H excess [g] in the respective water reservoir (i.e., beech compartment or deuterated label water), ε2Hbeech, ε2Hbeech, and ε2Hbeech are the isotopic enrichment of the respective water reservoir (given in atomic % of deuterium calculated from δ2H values), 𝑚𝑏𝑒𝑒𝑐ℎ𝑚𝑂
and 𝑚𝑙𝑎𝑏𝑒𝑙𝑚𝑂
are the molar weight ratio between hydrogen and oxygen in the respective water reservoir at the measured isotopic composition, θ is the VWC of beech compartments, and Mlabel is the absolute amount of label water added to the fir compartment [g].
- Results and Discussion
After 25 days of drought treatment, the VWC in the planted beech compartments declined to 6% ± 2%, while VWC in the unplanted control compartment was still at 16%. Soil-water potential in the beech compartments of the mesocosms ranged between pF values of ca. 2.0 and 3.5 (Figure 3c). Net beech photosynthesis had declined to an average of 8 µmol m−2 s−1 (Figure 3a), and stomatal conductance showed values of around 0.14 mol m−2 s−1 (Figure 3b). After the fir compartments were labelled with 2 L of deuterated water, the net photosynthesis and stomatal conductance of beech continued to decline to 4 µmol m−2 s−1 and 0.05 mol m−2 s−1, respectively, one week after labelling.
Forests 11 00162 g003 550
Figure 3. (a) Time series of light-saturated photosynthesis (Asat) and (b) stomatal conductance (gs) of beech seedlings as measurement averages ± standard error (n = 5). (c) Soil matrix potential in beech compartments (daily means) as pF-values, log(|hPa|) for three labelled (black, grey, light grey), two unlabelled (green, light green), and one unplanted control (light blue) mesocosms. Solid vertical line: time of deuterium labelling; dotted vertical line: re-watering of beech compartments. Very high pF values of some replicates from 10 April onwards were likely related to volumetric-water-content (VWC) measurement bias in very dry soil conditions.
These data showed that the beech trees experienced severe drought stress during the experiment that was not significantly counteracted by the silver-fir seedlings. Previous studies also reported a similar development for mature trees in the field under severe drought conditions at mixed fir and beech cultivation [66].
In unlabeled mesocosms and the labelled control mesocosm, δ2H values of soil moisture in the beech compartment remained between –120‰ and –80‰ VSMOW throughout the entire incubation period (Figure 4a). Labelled mesocosms showed similar or slightly enriched values for the first three weeks after labelling. In addition, 2H excess recovery in the beech compartments was negligible during that time. This suggested that no detectable HR occurred for three weeks after label injection.
Forests 11 00162 g004 550
Figure 4. (a) Time series of δ2H in soil moisture in beech compartment of 2H-labeled mesocosms (grey), planted unlabelled mesocosms (green), and labelled control mesocosm without plants (light blue). Dashed lines: time series of each respective replicate; solid lines: their mean values. Re-watering was with unlabelled water. (b) Time series of 2H excess recovery in beech compartments expressed as percentage of 2H excess added by deuterated water to fir compartment. Isotopic composition of control mesocosms’ beech compartment used as baseline for excess calculation. Grid was spaced at 14 day intervals.
During this initial three-week period, soil pF further increased in the beech compartments (Figure 3c). Values for δ2H in the beech compartments started to notably rise afterwards (Figure 4a). On the basis of 2H excess recovery, we could assume that traceable amounts of water were reallocated to the beech compartment through the fir roots (Figure 4b). Hence, HR occurred in this experiment with a delay of ca. 2–3 weeks after label application. However, this delayed HR occurred at pF values above the permanent wilting point (i.e., pF = 4.2 according to Amelung et al. [73]; Figure 3c), and partially resulted in the wilting of the beech seedlings. Seedling wilting is also represented by the concurrent drop of photosynthetic activity and stomatal conductance to values close to zero (Figure 3a,b).
Prior to re-watering, 2H excess recovered in the beech compartments amounted to an average of 0.2% of 2H excess added with the 2 L of deuterated water (Figure 4b). Re-watering of the beech compartments with 1.5 L of water at the end of October clearly diluted the δ2H signal (Figure 4a). Furthermore, pF values increased above the permanent wilting point. However, 2H excess recovery was initially doubled by re-watering, followed by rapid decline (Figure 4b). We assumed that the unexpected doubling of 2H excess upon re-watering was caused by overestimated VWC measurements directly after re-watering the beech compartments (water amount was part of the calculation of 2H excess recovery; see Equation 2). As indicated by δ2H dynamics, re-watering seemed to eliminate the large gradient in water potential, which was needed to sustain HR through water loss from the fir roots.
Overall, δ2H was correlated to pF values in the beech compartments (Figure 5). This suggests that water-potential gradients largely drove HR by the silver-fir roots, and HR was stronger at high pF values. A similar observation was made in a loblolly pine stand, where hydraulic redistribution was determined by reverse root flow and was shown to increase with soil drought [74]. However, in split-root experiments for Norway spruce, European beech, and English oak, Hafner et al. [27] observed hydraulic redistribution already under moderate drought (water potential of ca. –0.5 MPa/ –5000 hPa/pF value of 3.7) after only several days. In these conditions, we could not find evidence to support HR at similar time scales. Still, our results were consistent with the observation that, in a mature mixed beech–fir forest, the presence of fir slightly improved the xylem-sap flow density of beech compared to a beech monoculture [66]. However, this improvement was not enough to counteract the negative effect of drought on xylem-sap flow density. In addition, the positive effect of the presence of fir on xylem-sap flow density in beech upon drought was reversed when the forest stand was subjected to precipitation after drought. These data indicated that cocultivation of beech and fir may not enhance the resilience of beech to drought events through HR.
Forests 11 00162 g005 550
Figure 5. Relationship between δ2H and pF values in beech compartments of labelled mesocosms. Only incubation period from labelling until re-watering was considered.
- Conclusions
We set up a split-root experiment with silver-fir and European beech seedlings using 2H labelled water to track HR from wet- to dry-soil zones through silver-fir roots. In situ measurements of the isotopic composition of soil moisture showed detectable HR at 2–3 weeks after label application. A positive correlation between soil pF values and the isotopic composition of soil moisture confirmed that changes in water potential drive HR. Considering that only 0.2% of the added 2H could be recovered at the permanent wilting point, we concluded that the admixing of silver fir to European beech stands likely does not increase the resilience of beech monocultures to severe drought conditions. The delayed onset of HR of 2–3 weeks in this study led us to recommend that future field studies allow for a longer monitoring time to fully detect hydraulic redistribution.
SECOND ONE:
Synergies and trade-offs in the management objectives forest health and flood risk reduction
Fabian Rackelmann,Fabian Rackelmann1,2Zita SebesvariZita Sebesvari2Rainer BellRainer Bell1
1Department of Geography, University of Bonn, Bonn, Germany
2Institute for Environment and Human Security, United Nations University, Bonn, Germany
While healthy forest ecosystems deliver various services that can reduce flood risk, they can also contribute to flooding by providing wood that potentially contributes to the clogging of waterways and associated backwater effects. In this regard, deadwood, as a key aspect of healthy forests, is often in focus of post-flood disaster discourses. This research reflects on this ambiguity in the different forest management goals when it comes to managing forests for flood risk reduction versus forest health. A working definition of forest health will be presented and an overview of the different aspects of how a forest potentially can affect the flood hazard will be provided. This will refer to the ways forests influence (1) the discharge of water from the landscape into channels and (2) the characteristics of the channel and its riparian area and their respective influence on the transport of water, sediment, and debris. Often these two determining factors for the development of the flood peak are discussed separately and by different academic fields. This paper aims to connect the existing knowledge spheres and discusses the synergies and trade-offs. The review shows that the two objectives of forest health and flood risk reduction are largely synergetic. However, in direct proximity to watercourses trade-offs might occur. This is especially due to the ambivalent relation of living vegetation and deadwood to flood hazard. In places without susceptible infrastructures to clogging, deadwood and diverse vegetation structures should be supported due to their beneficial effects on water retention and channel characteristics. In places where susceptible infrastructures exist, trade-offs between the two objectives arise. Here the potential of freshly uprooted vegetation to cause damages should be reduced while maintaining the vegetation’s supportive characteristics, for example, concerning bank and slope stability. Where the risk of clogging is assessed as too high, also the selective removal or shortening of dead in-channel Large Wood can be considered. However, based on the literature review the risk deriving from dead Large Wood is evaluated as comparably low. This is related to its generally lower proportions and its smaller and less stable characteristics compared to freshly uprooted vegetation.
- Introduction
In mid-July 2021 Western Europe experienced severe floods. More than 220 people died in Germany and Belgium and huge public and private economic damages occurred amounting to €46 billion in Europe of which €33 billion accounted to Germany. This makes it the costliest extreme weather event in Germany and Europe to date (Munich RE, 2022). One particularly affected area is the Ahr valley in the low-ranging mountain area of Eifel, Germany. Within a few hours, the water level rose several meters. At 7 p.m. on the 14th of July, the flood wave exceeded the historic mark of 3.2 m meters in the small city of Altenahr, Rhineland-Palatinate (RLP). Just 1 h later it overflowed the gauge at a water level of more than 5 m. The water level reached its highest level of about 10 m in the early morning of the 15th of July (Landesamt für Umwelt Rheinland-Pfalz, 2022). Ultimately, more than 130 humans lost their lives and more than 42,000 people are affected in the Ahr valley (Deutsches Komitee Katastrophenvorsorge e.V, 2022). Despite recovery and aid initiatives many people in the area continue to suffer from the effects of the flood due to, i.e., property loss, livelihood disruption, and mental health effects (Welsch, 2022).
The flood was triggered by the low-pressure system “Bernd,” which brought regionally pronounced heavy rainfall in large parts of river basins. In West Germany, 60–180 mm of rain fell in only 22 h on the 14th of July. These large quantities of water quickly ran off due to the prevailing orographic conditions and already saturated soils. In the partly narrow valleys, the water got channeled and led to quickly rising streams (Dietze and Ozturk, 2021; Junghänel et al., 2021; Deutsches Komitee Katastrophenvorsorge e.V, 2022). This factor in interaction with mobilized sediment and debris led to not yet commonly observed effects in the region. Bridges were clogged which resulted in extending floods to areas hundreds of meters away from the initial riverbed (Dietze and Ozturk, 2021). One major factor that has been in the focus for this effect is deadwood (Dietze and Ozturk, 2021; tagesschau, 2021; ZDF Frontal, 2021). This discourse was also taken up locally. Similar to the discussions observed by Borga et al. (2019) after a flash flood in the Veneto Region of Italy, statements ranged from “exaggerated” nature conservation that did not allow to clear riparian forests from deadwood (ZDF Frontal, 2021) or to calls to remove trees along the streams (SWR Aktuell, 2022) to a not sufficient protection of ecosystems, as, e.g., healthy forest ecosystems allow rainfall to infiltrate and drain better in contrast to monocultures or heavily managed forests (BÜNDNIS 90/DIE GRÜNEN Ahrweiler, 2022). Indeed, this has been acknowledged in alpine areas for long under the concept of protective forests,1 however, has not received much attention in low and medium-ranging mountain areas (Markart et al., 2021; Welle, 2021). Against this background, the question of the adequate management of forests arises, in particular regarding the treatment of deadwood as a key aspect of healthy forest ecosystems, to strengthen its function in providing flood protection, while taking other interests into account.
This article aims to shed light on this issue and to support future developments in the complex forest-flood nexus by elaborating on the potential effects of forest ecosystem health aspects on flood risk. For this purpose, the research aims to address the following question: What are potential trade-offs and synergies between the forest management objectives of forest ecosystem health and of flood risk reduction?
To answer this question, firstly a theoretical background on forest health and flood risk will be provided. This is followed by an investigation of the ways a forest and its condition can influence the development of the flood wave after a heavy precipitation event occurs (Figure 1). Due to the large public attention on the discourse on dead and freshly uprooted wood, we will reflect on this aspect more in detail. Lastly, we will discuss to what extent synergies between the objectives of forest health and of flood risk reduction exist and where likely trade-offs appear.
FIGURE 1
www.frontiersin.org
Figure 1. The green frame depicts the thematic foci of the review to identify synergies and trade-offs between the forest management objectives forest health and flood risk reduction. It covers the influences of forest conditions on water discharge into channels and on flood relevant channel processes.
- Theoretical and analytical background
2.1. Defining forests and forest health
The origin of the term forest health is strongly connected with the concerns over acidic precipitation and the dying-off of forests, the “Waldsterben,” in the 1970s and 1980s in North America and Europe (ICP Forests, 2018; Forest Information System for Europe, 2022). However, there is no consensus on the definition of forest health or of forests to date.
Over time, different definitions of forests have been developed that reflect different management objectives and which shape the development of environmental policies (Chazdon et al., 2016). The most widely applied forest definition is from the Food and Agriculture Organization of the United Nations (FAO) which broadly frames forests as
Land spanning more than 0.5 hectares with trees higher than 5 meters and a canopy cover of more than 10 percent, or trees able to reach these thresholds in situ. It does not include land that is predominantly under agricultural or urban land use (FAO, 2020, p. 10).
Further, it differentiates them into naturally regenerating forests, planted forests, and plantation forests. The FAO’s generic definition is partly heavily criticized for not reflecting on the ecology and complexity of the natural biotope forest (Naturwald Akademie, 2020; Persson, 2020). In this regard, for example, Buettel et al. (2017) propose that deadwood should be an integrated part of the definition of forests.
As there is no agreed-upon definition of forests, also the definition of forest health varies widely. The FAO provides no official definition of forest health. However, its AGROVOC multilingual thesaurus, a tool to classify data homogeneously and facilitate interoperability, defines forest health as “the perceived condition of a forest derived from concerns about such factors as its age, structure, composition, function, vigor, presence of unusual levels of insects or disease, and resilience to disturbance” (FAO, 2022). This highlights the term’s subjectivity and dependence on forest management objectives. Accordingly, definitions of forest health range broadly, from strictly utilitarian to ecological perspectives (Trumbore et al., 2015). For instance, from a utilitarian perspective, factors indicating forest health include diseases, growth rates, and damages but also water quality. Many of these indicators are also included from an ecological perspective. However, it expands to indicators that do not have an immediate use, such as existing deadwood, or which could even be considered unhealthy from a utilitarian perspective, such as the patchiness of forests with different successional stages. For example, patches with a high ratio of old and dying trees would not count as healthy in a strictly utilitarian sense as no immediate use or even loss, e.g., of income would result from it. However, these patches facilitate various ecological processes, such as regeneration or habitat creation on a wider scale (Trumbore et al., 2015). Therefore, according to Trumbore et al. (2015) forest health is characterized by “mosaic(s) of successional patches representing all stages of the natural range of disturbance and recovery” (Trumbore et al., 2015, p. 815). Whereas, the frequency, strength, and spatial extent of disturbances should not exceed the natural variability nor affect the trajectory of recovery at the landscape to regional scale (Trumbore et al., 2015). This characterization of forest health is close to the structures and processes of old-growth forests where in absence of large disturbances, the forest dynamics tend to be small scale and homogenous stand structures rarely reach an extent of more than 0.5–1 ha (Korpel’, 1995). Old-growth forests are characterized by a multi-layered structure with high growing stocks and large deadwood volumes (Burrascano et al., 2013; Commarmot et al., 2013).
For the purpose of the article’s research objective, we adapt a definition of forest health that draws upon the structure of old-growth forests. Therefore, forest health is characterized by (1) mosaics of successional patches that represent all stages of forest dynamics, (2) a multi-layered structure, and (3) high amounts of growing stock and deadwood.
Forest health depends on various factors and is influenced by natural and anthropogenic stressors. Anthropogenic climate change is seen as one of the major challenges for the forestry sector and many different management approaches are discussed and explored to adapt the forest to climate change. These comprise, among others, the use of other or newly bred tree species or more diverse forest structures and deadwood (Bundesministerium für Ernährung und Landwirtschaft, 2021; European Commission, 2021b; Wissenschaftlicher Beirat für Waldpolitik beim Bundesministerium für Ernährung und Landwirtschaft, 2021). However, many of the stressors are related directly to the trade-offs that come along with the forest management objectives. This is particularly the case when the ecosystem management’s target is to maximize the provision of one ecosystem service, such as the provision of timber, which often leads to a considerable decline in the ecosystem’s ability to provide other ecosystem services, such as water retention (Bennett et al., 2009; Cavender-Bares et al., 2015; Watson et al., 2018).
2.2. Ecosystem (dis)services, nature-based solutions, ecosystem-based adaptation, and disaster risk reduction
As touched upon before, intact forest ecosystems provide various services that are beneficial to people (Watson et al., 2018), so-called ecosystem services (ES). The Millennium Ecosystem Assessment Framework categorizes ES into provisioning, regulating, cultural, and supporting services (Millennium Ecosystem Assessment, 2003). These services have the potential to reduce disaster risk by addressing one or several of its three dimensions: hazard, vulnerability, and exposure (Walz et al., 2021a). Their potential in providing a multi-purpose and sustainable approach to Disaster Risk Reduction (DRR) and Climate Change Adaptation (CCA) is widely acknowledged (Sudmeier-Rieux et al., 2021; IPCC, 2022) and also highlighted in the latest Adaptation Strategy of the European Commission (2021a). However, to what extent an ecosystem can support DRR is largely determined by its condition (Walz et al., 2021a).
Approaches that make use of these services to mitigate climate change, adapt to its impacts, and reduce existing disaster risks are clustered under the umbrella term of “Nature-based Solutions” (NbS). Depending on the target of the approach, it is further differentiated into “Ecosystem-based Adaptation” (EbA) and “Ecosystem-based Disaster Risk Reduction” (Eco-DRR). While EbA addresses climate-related hazards with a focus on long-term changes and future uncertainties due to climate change, Eco-DRR addresses all types of natural hazards and, in contrast to EbA, rather focuses on existing risks (European Environment Agency, 2017a; Walz et al., 2021b). In addition, in the sphere of flood risk management terms like “green infrastructure” (European Environment Agency, 2017b) or “bioengineering” (Bischetti et al., 2014) are frequently used.
While functional ecosystems are widely acknowledged as a promising solution to reduce both current and future risk, they may as well contribute to risk by means of “ecosystem disservices” (Sudmeier-Rieux et al., 2021). An example happened during the 2021 Eifel flood with the provision of woody material which contributed to blocking bridges and similar structures staunching the water and leading to backwater effects (Dietze et al., 2022).
Therefore, comprehensive DRR and CCA management schemes need to reflect on potential ecosystem services and disservices (Sudmeier-Rieux et al., 2021). An example of an established Eco-DRR approach to protect against several natural hazards is the concept of protective forests.
2.3. The concept of protective forests
In low-ranging mountains, which usually have absolute heights between 500 and 1,500 m above sea level and overtop their surrounding countryside by only 300–1,000 m, the protective forest concept has not received much attention so far even though low mountain ranges can already cause considerable orographic precipitation (Spektrum Akademischer Verlag, 2001; Welle, 2021).
In alpine areas protective forests have a long tradition to protect against natural hazards. For example, 42% of Switzerland’s area profits from protective forests, whereas one-third of the Swiss protective forests are located in altitudes below 1,000 m (Brändli et al., 2020). Protective forests, as a “Forest-based Solution” (Teich et al., 2021), target to prevent and mitigate natural hazards. They are generally implemented at the slope scale and are comparably inexpensive and feasible to protect against several hazards at the same time (Huber et al., 2015; Teich et al., 2021). Together with spatial planning, they are used to enhance the effectiveness of existing and to reduce the costs of new technical protection structures downstream (Markart et al., 2021; Teich et al., 2021).
Various definitions of protective forests exist, also in the German-speaking alpine area. According to the current Swiss definition from 2013, protective forests protect against the hazard categories of channel process associated with flowing water, avalanche, and rock- and icefall (Losey and Wehrli, 2013; Bundesamt für Umwelt, 2021). However, forests that exclusively reduce runoff but do not mitigate hazardous channel processes, such as debris flow, do not qualify as protection forests (Bundesamt für Umwelt, 2021). The Austrian Forest Act, amended in 2002, distinguishes protection forests into site-protecting forests (Standortschutzwald) (§ 21[1]) focusing on soil eroding processes, object-protecting forests (Objektschutzwald) protecting certain settlements, infrastructures, and other objects (§ 21[2]), and so-called “ban forests” (Bannwald). Here, the economic or other public interest to be protected proves to be more important than the disadvantages associated with the restriction of forest management and therefore declared by official notice a protection forest (§ 27). In comparison to the Swiss definition, the three protection forest types cover a wider range of services, including runoff reduction and also the protection of drinking water (Bundesministerium für Finanzen, 2022). Similarly broad is the Bavarian concept of protective forest which, however, does not distinguish between different types. It was defined as early as 1852 in the first forestry law for Bavaria and has not changed fundamentally until today (Bayerisches Staatsministerium für Ernährung, Landwirtschaft und Forsten, 2016). According to it, protective forests are forests in the high and ridge regions of the Alps and the low mountain ranges, on sites that tend to karstification or are highly susceptible to erosion, and serve to prevent avalanches, rock falls, rockslides, landslides, floods, soil drifts or similar hazards, or to preserve riverbanks (Art. 10 [1]). Further, forests that protect neighboring forest stands from storm damage are also defined as protection forests (Art. 10 [2]) (Bayerische Staatskanzlei, 2022). Also in Bavaria, runoff reduction is an explicitly mentioned task of the protection forest and its increasing importance is highlighted in the context of climate change to reduce flood risk (Bayerisches Staatsministerium für Ernährung, Landwirtschaft und Forsten, 2016).
2.4. Flood risk
Flood risk is, according to the EU directive on the assessment and management of flood risks, “the combination of the probability of a flood event and of the potential adverse consequences for human health, the environment, cultural heritage and economic activity associated with a flood event” (European Parliament and Council of the European Union, 2007, p. 29). The definition is touching upon the general understanding of risk as the product of hazard, exposure, and vulnerability (IPCC, 2014) which is nowadays the agreed understanding in the DRR and CCA communities. Transferring this understanding to the EU directives definition, the flood event is the hazard and the potential adverse consequences for a given subject constitute of its exposure and its vulnerability to being negatively affected by the flood event. Therefore, a hazard alone does not constitute a risk or a disaster (IPCC, 2014). While acknowledging that forests can affect all three factors of risk–e.g., riparian forests could prevent people to settle in flood plains, or could contribute via timber sale to vulnerability reduction of a municipality when used to better equip emergency teams– this review focuses primarily on the flood hazard factor.
-
Methodology
This study is based on a comprehensive review of English and German literature. For this, peer-reviewed journal articles and book chapters were retrieved from the platforms “Google Scholar,” “Web of Science,” and “Scopus.” The initial applied Boolean operators were based on the keywords “forest*” and “flood*” to receive a broad overview. They have then been further concretized by adding keywords reflecting on forest health, i.e., “deadwood,” “drought,” and “health*,” and flood risk aspects, i.e., “erosion,” “runoff,” “driftwood,” “large wood*,” and “debris.” Relevant papers were selected by screening their titles and abstracts. Papers that largely covered the interests of this review during full-text review, e.g., connecting forest management with runoff or with large wood, were inserted in “connectedpapers.com” to identify other related articles and topics that might not have yet been reflected on. The yielded articles were supplemented with relevant gray literature, such as management guidelines and policy documents to capture current developments and practices in the management of forests and flood risk. During the research and writing process, articles have been continuously added to the initial literature bank via cross-references and specific searches for the applied thematic foci. The information for this review has then been extracted from the full articles, chapters, or their sections and summarized, compared, and analyzed according to the presented structure.
-
Results: the influence of the forest on the flood hazard
The severity of the flood hazard depends, besides the duration and intensity of the precipitation event, largely on two major components which determine the flood runoff volume and the development of its peak: (1) The discharge of water from the landscape into channels and (2) the characteristics of the channel and surrounding green and gray infrastructure and their influence on the channel and the transport of water, sediment, and debris.
4.1. The forests’ influence on water discharge into channels
The discharge is to a certain degree influenced by the landscape characteristics. Forests and their soils provide various supporting and regulating ES that influence the potential discharge rate and velocity of surface and sub-surface runoff. In general, forests are characterized by lower run-offs than, for instance, grasslands (e.g., Chen et al., 2021; Scheidl et al., 2021). This is due to their generally higher evapotranspiration [evaporation from soils and water bodies, and vegetation interception and transpiration (Brooks et al., 2013)]) and infiltration rates, higher soil water storage capacities, and higher surface roughness leading to lower runoff velocities (Schüler, 2006; Eisenbies et al., 2007; Hümann et al., 2011; Markart et al., 2021).
While forests may reduce flood peaks for moderate precipitation events, the effect of the forest and its condition decreases with increasing severity (Schüler, 2006; Bathurst et al., 2022; Xiao et al., 2022). These characteristics are, however, strongly influenced by complex interactions of geographical, edaphic, vegetative and climatic factors and the current and historical management (Schüler, 2006; Eisenbies et al., 2007). Optimal are multilayered forests with rich ground vegetation. In well-structured forest stands, up to 6 mm of precipitation in the canopy and depending on the ground vegetation a further 1–4 mm can be retained per rain event (Markart, 2000; Markart et al., 2021). However, it is remarked that this interception capacity is comparatively quickly reached in case of an extreme precipitation event (Markart et al., 2021). While evaporation and transpiration effects during a heavy rain event are neglectable (Markart et al., 2021), the higher evapotranspiration of forests, especially during the growing season in summer, can lead to relatively low antecedent soil moisture contents. In effect, forest soils generally obtain higher degree of available storage volume before a rain event (Brooks et al., 2013; Archer et al., 2016).
To reduce the runoff amount and velocity, the surface roughness of the forest ground plays an essential role. Hereby, the ground vegetation is important, but also structures like dead branches or laying trunks are essential to increase the roughness and eventually help to decrease erosion (Figure 6A; Lachat et al., 2019; Markart et al., 2020).
4.1.1. Deadwood and other organic material
Deadwood and other organic material, such as litter, act as a sponge for water and increase the water absorption capacity. As such, they can create wet microhabitats that obtain an important role for natural regeneration in dry forest types (Frehner et al., 2005). Decaying wood also serves as a seedbed. In some humid mountain forests, more than half of all spruce trees grow on decaying wood (Lachat et al., 2019). Further, deadwood supports soil structuring micro fauna and is therefore an important factor for the development of forest soils characterized by high water infiltration and retention potential (Schüler, 2006; Bundesamt für Naturschutz, 2020).
While high organic contents of forest soils are important for a high water absorption capacity under normal weather conditions, it can lead to stronger water repellency when long dry periods occur. The drying out of organic material leads to hydrophobic conditions increasing the soil water repellency (Mao et al., 2016; Hewelke et al., 2018) and potentially leading to surface runoff (Hümann et al., 2011). Especially dense spruce forests with no ground vegetation (Piceetum nudum) are disadvantageous as they have a pure needle litter which has a strong water-repellent effect (Markart et al., 2020). This repellent “straw roof” (“Strohdach”) effect is also observable in grasslands with a very dense root system, e.g., some Festuca species, leading to a generally higher runoff as compared to forests (Markart et al., 2011). In low-ranging mountains, peat and swamp areas of swamp forests play an important role due to their high water retention capacities. They are able to buffer sudden heavy rainfall after dry periods and mitigate it until their water storage capacity is fully replenished. A hydrophobic phase of organic material, as previously described, is not observed in swamp forests because of permanent moisture penetration (Schüler, 2007).
4.1.2. Soil structure
The forest soil is one of the key factors influencing water retention and stands in strong relation to the forest type and its tree types. While the soil influences the type of forest that is growing on it, it is also shaped by the forest. Besides the provision of organic material, the forest’s root system is a major factor in the development of soils. The tree roots of forests create a complex structure that acts as a drainage system, as water can more easily infiltrate along roots and macro pore structures left by decaying roots (Figure 6B; Wu et al., 2021). In this way, the root system increases preferential flow, partitions, and transports water within catchments (Archer et al., 2016; Luo et al., 2023). The older the forest, the higher the infiltration of water generally is (Lange et al., 2009; Neary et al., 2009; Karl et al., 2012; Archer et al., 2016). Further, forests with deep-rooting tree species, foremost deciduous species, show generally higher water retention rates than shallow-rooted tree species (e.g., Nordmann et al., 2009).
In a study by Archer et al. (2016), the effect of different forest types in the Cairngorm Mountains, Scotland, has been investigated for in situ field-saturated hydraulic conductivity, root fraction, and proportions and connectivity of macropores. The measured values indicate that with increasing tree and forest age the infiltration rates increase due to larger root fractions and higher proportions and connectivity of macropores. Remarkably, an old-growth Caledonian forest had a 7–15 times larger hydraulic conductivity than a 48-year-old pine plantation, while it was evaluated that a 6-year-old pine plantation likely does not allow for heavy precipitation infiltration. The latter is brought into connection with the treatment of the area before the new trees have been planted. The felling of the previous plantation and the preparation of the soil for the new plantation with heavy machinery resulted in amorphous soil structures with a reduced macropore proportion and connectivity (Archer et al., 2016). However, as the 48-year-old plantation has higher values than the young plantation, it is suggested that the pine plantation was able to relatively quickly recover a root system capable of enhancing water infiltration (Archer et al., 2016).
The results of a study by Hümann et al. (2011), however, draw a different picture. They conducted a series of sprinkling experiments on hillslopes in low-mountain ranges of Rhineland-Palatinate (RLP), Germany, in order to investigate the effect of different forest types and soil properties on runoff. In line with Archer et al. (2016), they found that the soils under established forests are porous with relatively high infiltration and water conductivity values. However, in a 40-year-old Douglas fir plantation and in a 30-year-old deciduous afforestation surface runoff has been observed, indicating a limited infiltration. This was associated with the prevalent soil properties. In the Douglas fir stand, the surface runoff has been associated with local soil compaction and dry weather conditions resulting in higher water repellency of the humus layer. In the 30-year-old afforestation, it is related to the compacted soil layer at 20 cm depth that was derived from previous agricultural practices. In effect, the afforestation still reacts similarly to arable land. This indicates that the potential of forests to reduce runoff generation and enhance water retention mainly depends on the physical soil properties and conditions (Lüscher and Zürcher, 2002; Hümann et al., 2011).
The limiting effect of the soil is also acknowledged in the guidance document “Sustainability and Success Control in Protection Forests” [Nachhaltigkeit und Erfolgskontrolle im Schutzwald -NaiS (Frehner et al., 2005)], which describes the official standards and principles of protection forest management in Switzerland. It states that silvicultural management should focus on sites with deep but inhibited permeable soils as here deep-rooting tree species can significantly increase the water storage effect and, therefore, changes in the forest condition have the highest influence. In contrast, the forest condition has only a limited effect on water storage on sites with heavily waterlogged, very shallow, or excessively permeable soils or, on the contrary, on sites with profound and permeable soils that naturally already have a high soil water storage effect (Frehner et al., 2005). For afforestation on soils that are heavily waterlogged, e.g., due to agricultural use, it is suggested to consider the use of special deep-loosening machinery prior to the planting to break up the compacted soil layers to accelerate the establishment of a complex root system (Hümann et al., 2011).
4.1.3. Disturbance and regeneration
As previously touched upon, discharge reduction is impaired by human management practices. This refers to the age and structure of forests, prevalent trees, and existing forest infrastructures. Naturally, in wide parts of central Europe the beech would be dominant (Ellenberg and Leuschner, 2010). The natural forest types were, however, often replaced by plantation forests. This does not solely negatively affect the discharge due to the different afore-described factors but also the resistance of the forest stand to disturbances which is important for the continuous delivery of runoff-reducing services.
In general, forests with a multi-layered, stepped, or routed structure are considered less susceptible to disturbance than single-layered stands (Hanewinkel et al., 2014). Especially, single-layered needle forests own a high risk to be largely destabilized by wind throw or bark beetle infestations resulting in a wide loss of their protective services (Huber et al., 2015; Sebald et al., 2019). In case areas are disturbed by wind throw, lying deadwood initially provides high surface roughness, but as it decomposes, the protective effect may be temporarily reduced if regeneration is slow to emerge (Lachat et al., 2019). According to NaiS natural regeneration should be present on at least 3–6% of the area to guarantee stable protective forest stands (Frehner et al., 2005). Generally, natural regeneration does not present any difficulties in the lower montane, sub-montane, and colline stages. Here beech dominated forests would naturally occur with different ratios of other tree species (Korpel’, 1995; Schwitter et al., 2019). However, the high browsing pressure of ungulate game species is considered a major challenge to natural regeneration (Frehner et al., 2005; Huber et al., 2015; Brändli et al., 2020). Other challenges for natural regeneration pose the quick spreading of blackberry (Rubus silvaticus) on disturbed or cleared sites (Figure 2), and droughts (Frehner et al., 2005).
FIGURE 2
www.frontiersin.org
Figure 2. Blackberry (Rubus silvaticus) dominated area (Fabian Rackelmann).
In case there is missing natural regeneration in undisturbed forest stands, e.g., due to missing light evoked by the dense canopy of one-layered, single-aged forest stands, the cutting out of gaps can help to promote its development. Further, the development of stable trees can be supported by cutting down neighboring trees to give them enough space to develop (Huber et al., 2015). A respective selective thinning approach applied in protective forests in lower altitudes is the “Z-tree care” (Z-Baum Pflege) presented by Schwitter et al. (2019) which targets the support of single trees to transform single-layered forest stands toward multilayered forests.
As drought is one of the main challenges for natural regeneration, it is important to reflect on the risk of excess sun and heat development when opening the forest canopy which could lead to the drying out of the young trees and the organic matter (Schwitter et al., 2019; Ibisch et al., 2021). Similarly, this should be the case when considering the clearing of dead spruce stands triggered by bark beetle infestations. With the dying of the spruces the temperature within the stand increases. It is, however, considerably lower compared to cleared areas which facilitates the upcoming of natural vegetation (Ibisch et al., 2021).
Clearing of dead trees leads to losses of organic matter and reduces the existing structures and thus the surface roughness (Markart et al., 2021). Moreover, for the harvest often large machines are used that compress the soil, as, e.g., mentioned in the study by Hümann et al. (2011). To reduce the impact of machines on forest soils logging trails are established to avoid areal driving. However, the runoff reducing potential of soil in these trails is lost (Figure 3) and therefore they should have minimum density. Further, in the line structures of logging trails and forest roads and respective trenches, it comes to a concentration of water potentially leading to a quick runoff into forest canals (Figure 4; Schüler, 2007).
FIGURE 3
www.frontiersin.org
Figure 3. Logging trail with standing water during a hot summer day in 2022 indicating the destroyed water infiltration capacity of the soil due to the compaction of the vehicles, Ahrhütte, Blankenheim, NRW (Fabian Rackelmann).
FIGURE 4
www.frontiersin.org
Figure 4. Forest road design tilted toward hillslope favoring the concentration and quick discharge of water, Altenahr, RLP (Fabian Rackelmann).
While the influence of the forest and their conditions in small catchments is largely acknowledged, with increasing catchment size the influence of forest vegetation on flood peaks becomes more contested (Markart et al., 2021; Bathurst et al., 2022). Calder and Aylward (2006) present three possible explanations for it: First, flood peaks from smaller catchments will unlikely come together in the bigger catchment at the same time and therefore will not add up to each other. Second, storms with a spatial scale able to affect a large basin area are potentially also of high severity. As previously described, the higher the severity the smaller the effect of the forest and its condition on the discharge is. And third, the proportional change in water retention by measurements is likely to be higher within a smaller catchment than within a larger catchment.
4.2. The forests’ influence on flood relevant channel processes
The previously described higher roughness of forests also applies to forests in the floodplain and along the channel. Riparian forests increase the flow resistance contributing to the reduction of the flood peak (Bölscher et al., 2010; Reinhardt et al., 2011). They are also a major source of deadwood in channels with a high ecological value (Neuhaus and Mende, 2021) and provide various ES, inter alia, positively affecting channel processes (compare following chapter) (Linstead and Gurnell, 1999; Thomas and Nisbet, 2012). Further, a number of studies have shown the importance of vegetation on streambank stability (Simon and Collison, 2002; Docker and Hubble, 2008; Gasser et al., 2020) and for the reduction of mass movements, such as landslides, by increasing slope stability due to, e.g., root reinforcement and reduced soil’s pore-water pressure (Moos et al., 2016; Vergani et al., 2017; Gasser et al., 2019). Therefore, forests help to reduce the flood peak by decreasing the mobilization of different forms of debris from the riparian stripes, including wood. Further, they also help to retain already mobilized in-channel debris (Figures 5, 6C; Gasser et al., 2019).
FIGURE 5
www.frontiersin.org
Figure 5. Wood retained by a yew tree (Taxus baccata) close to Brück (Ahr), Altenahr, RLP (Fabian Rackelmann).
FIGURE 6
www.frontiersin.org
Figure 6. Overview of different forest influences on flood hazard with a particular focus on Large Wood processes. Deadwood is highlighted in yellow. Fresh Large Wood is depicted in green.
However, the services provided by trees and deadwood hold the potential to evolve to flood accelerating disservices during flood events due to mobilized material contributing to blockages of infrastructures, such as bridges.
Dead and fresh woody materials which are located within the floodplain are widely defined as “Large Wood (LW),” “instream wood,” or “(large) woody debris.” The connection with the term “debris” is, however, not frequently used anymore due to its negative connotation (Bundesamt für Umwelt, 2019). The dimensions of LW are generally defined as being at least 1 m in length and 10 cm in diameter. Various LW aspects are increasingly comprehensively researched. The research foci vary, inter alia, from investigating its role for flora and fauna, the origins of the woody pieces that are mobilized during a flood event, to the transport in the watercourse and its interactions with different hydro-geological channel processes, and its potential in contributing to clogging of infrastructures.
As the issue of LW has been very prominent regionally in the discussions on the causes of the Eifel flood disaster of 2021, the following sections will give an overview of what the distribution of dead and fresh LW in previous floods has been and where they originated from to better evaluate the potential risk from different types of LW. Further, the interaction of dead LW with channel processes relevant for the development of floods, will be described. The sections will, moreover, inform on suggested management practices concerning flood risk reduction.
4.2.1. The role of dead and fresh Large Wood during previous flood events
The type of LW, its size and form, the type of vegetation, and whether it is deadwood (woody vegetation, or pieces thereof, that have been dead before the event) or fresh wood (woody vegetation that has been uprooted or broken apart and washed away during the event) determine its potential to contribute to the clogging of bridges and other infrastructure within the floodplain. Long pieces with branches or adjunct root systems pose a higher risk for wood jams than smaller less complex LW forms as further LW and debris can be more easily retained by the more complex structures. Fresh LW is generally longer and more complex than dead LW (Schalko, 2018). For its potential of clogging also the stability of the LW is an important factor. This is determined by the size, especially the diameter, but also by the type and vitality as it determines the stability of the wood. Fresh LW derived from deciduous trees is generally more stable than dead LW originating from coniferous trees and, therefore, potentially owns a larger clogging potential to block bridges (Figure 6D; Rickli et al., 2018; Schalko, 2018).
Besides the type of LW, also its transport form influences its potential to block infrastructure, e.g., single pieces of LW have a lower potential to block bridges than congested LW (Mazzorana et al., 2018). Models to estimate the potential of different types of LW to contribute to clogging are important for an adequate LW management. They have become more complex in recent years. For example, from initially only accounting for straight LW pieces options to including different forms, such as the roots, have been incorporated (Schalko, 2018). For more information on different methods and models to assess LW-related hazards, e.g., Mazzorana et al. (2018) and Friedrich et al. (2022) can be consulted.
Large Wood that has been transported during flood events, has different proportions of deadwood and fresh wood. The proportions are heavily influenced by the management of the watercourse and the floodplain, e.g., the age and management of the forest but will also be influenced by the methodology of the assessment. Even though areas around LW are increasingly investigated, studies on the proportions of dead and fresh LW (Table 1) remain rare. Bänziger (1990) assessed the proportions of dead and fresh LW for the summer 1987 floods in Switzerland. Based on interviews and regional aerial photographs, he estimated that 35% of the total LW has been in-channel deadwood, 19% was derived from eroded alluvial forests, 12% from the streamside vegetation, 17% was mobilized by landslides, and another 17% was derived from lumberyards and timber. Interestingly, the interviewees suggested that the alluvial forests have been responsible for 52% of the total LW mobilization, which was far higher than according to the 19% of his final analysis.
TABLE 1
www.frontiersin.org
Table 1. Average proportions of dead and fresh Large Wood (LW) per event.
For the 2005 flood in Switzerland, Waldner et al. (2007) assessed the proportion of different LW types with standardized visual assessments of the bark of full piles of LW that either naturally accumulated or artificially were piled in the aftermath of the flood in 4 torrent catchments. The collected data were reevaluated by Rickli et al. (2018) applying some filters to reflect on human influences such as the mechanic piling and cutting of wood (Figure 7). Large variations in the distribution among the different locations were observed. For instance, the fresh LW proportion varied from 35 to 79%. However, as the LW has often completely been debarked during the mobilization process, this may have led to an overestimation of the contribution of deadwood (Waldner et al., 2007; Steeb et al., 2017). Steeb et al. (2017) have modeled, based upon parameters that reflect the recruitment and deposition processes of a catchment, the LW budget during the 2005 flood in Switzerland for four tributaries to the Kander catchment. The proportion of deadwood varied within the 4 rivers from 9.3 to 22% with an average of 12.1%, whereas approx. two-thirds of it, or 8% of the total LW amount, has been located in the channel, and one-third, 4% of the total LW, has been deadwood located in the riparian forests which were then eroded during the flood event.
FIGURE 7
www.frontiersin.org
Figure 7. Mechanically pilled Large Wood as one factor exacerbating potential endeavors to determine the source and proportion of Large Wood, Sinzig, RLP (Fabian Rackelmann).
Waldner et al. (2007) particularly paid attention in their investigation on the role of LW to the question of the proportion of storm-damaged wood, since after the event, the assumption was made on various occasions that the large amount of LW had been caused to a substantial extent by storm-damaged areas that had not been cleared. Some of the study regions were severely affected by the storm “Lothar” in 1999. These showed proportions of fresh LW of 53%, while in areas that were barely affected by the storm a fresh LW proportion of 78% was found. In the regions severely affected by the storm, bark beetle traces were found in about 17% of the total volume of LW recorded, in the others it was about 2%. Since bark beetles proliferate especially in spruce stands after storm damage, this is an indication that relatively little LW originated from storm areas. Especially in the areas downstream of the alpine rim lakes, a high proportion of softwood tree species typical of the riparian vegetation of valley rivers was found. This is an indication that a large part of the LW was recruited by lateral erosion during the flood along the larger water courses (Waldner et al., 2007).
As touched upon earlier, the type of LW also depends on the recruitment processes which vary from tree mortality (Figure 6E) and wind-throw to landslides (Figure 6F), debris flows (Figure 6G), and bank erosion (Figure 6H). Depending on the climatic conditions, topography, and geology of the river catchment, and the channel and landscape management the input sources and processes vary. In areas that are not frequently affected by extreme precipitation events, LW recruitment occurs mostly by tree mortality, wind throw, and bank erosion. Conversely, in areas where heavy precipitation events are frequently occurring, LW recruitment is increasingly shaped by mass wasting processes, such as landslides and debris flows (Seo et al., 2010). In mountain catchments, the recruitment processes are quite variable. This is, i.e., related to the varying precipitation distribution which leads to different discharge rates within catchments (Rickli et al., 2018). For example, Rickli et al. (2018) observed no drastic LW transport in four headwater streams during the 2005 flood in Switzerland, which stands in contrast to the high amounts of LW observed in larger rivers further downstream (Steeb et al., 2017; Rickli et al., 2018). Further, the proportions of LW recruitment from the valley bottom by alluvial processes and from the hillslopes by foremost gravitational processes vary considerably, even within small mountain catchments (Comiti et al., 2016a). For the 2005 flood in Switzerland, it has been shown that in the steep headwater torrent catchments few mass wasting events such as landslides and debris flows have been the dominant LW recruitment processes contributing to much of the LW input (Waldner et al., 2007; Rickli et al., 2018). In the lower reaches of the mountains, where catchment areas reached around 100 km2, lateral bank erosion has become the major process for LW recruitment (Steeb et al., 2017). Similar relations have been obtained by other studies, as well (Seo et al., 2010; Lucía et al., 2015). Click or tap here to enter text. However, for example, during the 2011 floods in the Magra River basin in northwestern Italy, severe bank erosion has been already the dominant LW recruitment process for much smaller catchments. For the tributaries and their catchments Gravegnola (34.3 km2) and Pogliaschina (25.1 km2), it has been estimated by Lucía et al. (2015) that 70–80% of the LW were derived from the valley bottom and its flood plains through severe channel widening. The rest of the LW came from a few hillslope processes. While flood plain erosion has been the prevailing process in most of the observed areas, in some upstream sub-basins hillslope processes, foremost landslides, have been majorly responsible for LW recruitment (Lucía et al., 2015).
Based on these observations, it is assumed that with increasing catchment area and decreasing channel slope, the proportion of LW recruited by mass wasting events decreases and the proportion of LW mobilized by erosion increases (Rickli et al., 2018).
As channel widening has been a major contributor to LW recruitment and a direct hazard to infrastructure along the channels during the 2011 flood in Italy, studies by Nardi and Rinaldi (2015), Surian et al. (2016), and Comiti et al. (2016b) helped to better grasp this phenomenon by analyzing geomorphological and hydraulic variables. Despite that, the studies provided a good understanding of the influence of channel slope, channel confinement, upstream sediment supply, and unit stream power, the quantitative prediction of the channel widening during an extreme flood event remains subject to large uncertainties. This in turn limits the capability to precisely evaluate the LW volumes that need to be expected during flood events from vegetated floodplains (Comiti et al., 2016a). However, to estimate LW supply to rivers different models have been developed which follow empirical, deterministic, stochastic, and GIS-based approaches. A review of the different models and their characteristics is provided by Steeb et al. (2023), Preprint who further evaluated the usefulness of two recent GIS-based models for flood hazard assessments.
4.2.2. Suggested multi-goal management of vegetation along watercourses
To meet these uncertainties, water managers and many foresters tend to prevent vegetation growth along the watercourses to avoid the risk of clogging (Gasser et al., 2019). This is also suggested partly by recent practices in which it is recommended to clear the channel along one to one and half times the tree’s height (Pichler and Stöhr, 2018; Markart et al., 2021). This might reduce the input of LW but also reduce the before described benefits of riparian vegetation and thus tends to, inter alia, favor the creation of main LW recruitment processes such as bank erosion (Simon and Collison, 2002; Docker and Hubble, 2008; Gasser et al., 2020) and landslides (Moos et al., 2016; Vergani et al., 2017; Scheidl et al., 2021). According to Markart et al. (2021), the effects of such a clearance management approach became apparent after the 2005 flood in Austria where a lot of debris was mobilized but nearly no LW.
Therefore, considerations between the stabilizing effect of vegetation and the potential recruitment during floods need to be made. To help practitioners in their decision-making, Gasser et al. (2019) provide guidance on this issue for two main LW recruitment processes during heavy precipitation events: Hydraulic bank erosion and landslides. They created respective graphs which evaluate the positive stabilizing effect of the vegetation in high, variable, and low depending on the main characteristics of the slope and channel.
Regarding hydraulic bank erosion, the positive effect of vegetation compared to its potential to act as a LW source tends to decrease with increasing channel width and channel gradient. In areas where the positive effect is variable or low, large trees should be removed and lower-growing vegetation such as different shrubs and grasses should be planted (Gasser et al., 2019).
Regarding landslides, the positive effect of vegetation tends to decrease with increasing slope steepness and soil thickness. Nevertheless, also in areas where the positive effect of vegetation is variable or comparatively low, vegetation is important for overall slope stability. However, here tree heights and densities should be adapted to ensure continuous regeneration to guarantee a continually existing root structure. Further, a mixture of tree species should be adapted to initiate higher root reinforcement (Gasser et al., 2019).
4.2.3. The effect of in-channel dead Large Wood on watercourse processes
The previous section has shown that the risk deriving from dead in-channel LW is likely to be comparatively low. Nevertheless, its widespread removal from the watercourses is widely practiced to reduce flood risk (Thomas and Nisbet, 2012). As with the removal of living vegetation this, however, can have detrimental effects on the development of the flood peak. This section will take a closer look at the ways dead in-channel LW interacts with the watercourse processes and its effect on the flood hazard.
4.2.3.1. Presence, storage, and transport of in-channel dead Large Wood
The presence, storage, and mobilization of in-channel dead LW depend on the characteristics of the LW, and of the stream and its hydro-geomorphic conditions and management. Generally, the lower the channel widths and depths are, the higher the storage of LW in the channel and its interaction with the channel bed and bank. LW pieces that are longer than the channel width are considered largely stable (Figure 8; Lienkaemper and Swanson, 1987; Ruiz-Villanueva and Stoffel, 2017). In this regard, Gurnell et al. (2002) categorize channels in small, medium, and large. Small channels have widths smaller than most of the LW lengths. Medium channels are wider than most of the LW pieces and large channels have widths larger than all of the supplied LW pieces.
FIGURE 8
www.frontiersin.org
Figure 8. Dead Large Wood with very low potential of being mobilized during a flood event as their lengths are greater than the stream widths. (Left) Stream Aulbach close to Ahrhütte, Blankenheim, NRW. (Right) Stream Vischelbach close to Kreuzberg, RLP (Fabian Rackelmann).
Further, the stream flow power influences the stream’s capacity to transport LW. In small headwater catchments, the mobilized LW generally deposits not far away from its source areas due to narrower valley channels and lower stream power compared to larger rivers. Even during extreme flood events in a larger catchment, the LW in the smaller contributing headwater catchments remains often stable as the precipitation and discharge peaks locally are generally not as severe as compared to the larger catchment (Seo et al., 2010; Rickli et al., 2018). An exception is debris flows as they often transport large amounts of LW, boulders, and sediment into headwater streams and their adjunct watercourses, where, however, the LW recruited generally breaks into smaller pieces (Seo et al., 2010; Steeb et al., 2017). As with the recruitment of fresh LW, in larger streams fluvial processes play the major role in the transportation of LW downstream or on the floodplains (Figure 9). Here, LW has less influence on the channel characteristics than in small streams (Lienkaemper and Swanson, 1987; Seo et al., 2010).
FIGURE 9
www.frontiersin.org
Figure 9. Large Wood with a high potential of being mobilized during a flood event. River Ahr close to Müsch, RLP (Fabian Rackelmann).
The dominant watercourse morphologies in mountainous regions are step-pool channels (Chin, 1989). In mountain streams, the natural accumulation form of LW are dams, also called logjams. Logjams are frequently defined as more than two LW pieces that exert combined effects on hydraulics and sediment storage of a channel (Rickli et al., 2018; Wohl and Scamardo, 2020). A simple and widely applied categorization of LW accumulations is the one from Gregory et al. (1985) in “partial dams,” “complete dams” and “active dams,” which have been further elaborated on by Linstead and Gurnell (1999). Partial dams present only an incomplete barrier to the water flow as it does not cover the full width of the channel. Complete dams extend across the full channel width, however, consisting of a rather loose structure that allows the water flow to continue and not to influence the water profile at regular flows. Active dams also extend across the complete width, however, induce a step in the water profile by (partially) blocking the water flow. Active dams need the longest time in their generation but are also the most stable type, persisting often many years at the same location (Linstead and Gurnell, 1999). The porosity of the LW dams depends on the structure and size of the LW and the accumulating material. A matrix out of small and big wood pieces will create denser dams than big wood pieces alone. The porosity of the dams also depends on the seasons, being the lowest in autumn when leaf litter in the stream accumulates behind the dam (Thomas and Nisbet, 2012). Over time, the characteristics and hydro-geomorphic effect of stable LW dams vary by differing amounts of wood pieces and sediment trapped by the key pieces of the LW dam (Wohl and Iskin, 2022). The size of the dams increases downstream, whereas the frequency of LW accumulations decreases (Swanson et al., 1982; Wohl and Scamardo, 2020).
Under largely unmanaged conditions it was found for British headwater rivers that approximately every 7–10 times of the channel width LW dams occur naturally whereas the forms and characteristics of dams vary (Linstead and Gurnell, 1999). Most of the LW dams are relatively transient. Only a few dams persist for several years and remain at the same site, however, the density over time remains relatively equal under unmanaged conditions. Despite the relative transience of LW dams, persistent effects in channel morphology and backwater storage are observed (Wohl and Scamardo, 2020; Wohl and Iskin, 2022).
4.2.3.2. The effect of in-channel Large Wood on morphology and water retention
Several studies have highlighted the role of LW in dissipating excess energy that otherwise would contribute to slope erosion (Chin, 1989; Chin and Wohl, 2005). This is, inter alia, because LW accumulations cause an increase in the flow resistance, its effect, however, decreases with increasing water discharge rates (Gregory et al., 1985; Linstead and Gurnell, 1999). They exhibit a higher dispersive fraction than streams without LW and it has been shown that the flow resistance is far greater in step-pool channels with LW than without. This, however, is largely influenced by the location of the LW in the watercourse. LW located close to the lip of the step rises the step height and increases the flow resistance more than compared to LW which is located in the pool of the channel. In this way, LW is an essential factor for the stabilization of these channels (Abbe and Montgomery, 1996; Curran and Wohl, 2003; Chin and Wohl, 2005; Wilcox and Wohl, 2006).
Further, LW creates greater variability in flow depth and flow velocity. Different types of pools are created in the channel which plays an integral role in the retention of water and sediments and provides habitats for fauna and flora (Linstead and Gurnell, 1999; Thomas and Nisbet, 2012). In streams in which LW is unmanaged pools can occur approximately every two channel-widths (Linstead and Gurnell, 1999).
When water is facing LW dams (Figure 6D), the flow is concentrated when it flows over, under, or through the dam’s matrix where it gets concentrated leading to higher velocities after the dam (Thomas and Nisbet, 2012). During high flows, dammed pools can act as locations of flow avulsion. Further, if active dams persist for long, they can also lead to a change in the position of the channel itself if the bank is erodible (Linstead and Gurnell, 1999; Thomas and Nisbet, 2012; Wohl, 2013). These effects lead to an increased capacity to hold back water in the channel and in the floodplain, which decreases the total runoff and stretches the runoff time. Thomas and Nisbet (2012) have reintroduced LW in the tributaries of the Welsh river Fenni and have modeled based on the observed changes that each dam has the potential to delay the flood peak by 2–3 min. This effect might be low for a single dam, however, on a catchment scale it may be considerably high (Linstead and Gurnell, 1999; Thomas and Nisbet, 2012). Linstead and Gurnell (1999) argued that this might also help to desynchronize the runoff of the different tributaries leading to a decrease in the flood peak of the downstream river. This, however, could not be proven by Thomas and Nisbet (2012).
Moreover, particularly active dams are important for trapping sediment and LW and keeping it within the headwater system (Linstead and Gurnell, 1999). In turn, the removal of LW dams would trigger the downstream mobilization of sediment and the remaining LW, incise the channel bed, and reduce habitat diversity (Linstead and Gurnell, 1999). In this regard, the downstream accumulation of LW can help restore previously eroded and incised channels by forming a renewed step and therefore contributing to the momentary channel-hillslope stability reducing the probability of landslides (Golly et al., 2017).
4.2.3.3. The influence of water and forest management on Large Wood and suggested approaches for Large Wood management
As previously described, the presence and mobilization of in-channel LW are dependent on the characteristics of the stream which is strongly influenced by existing gray infrastructures, such as bridges, and green infrastructures, such as living vegetation and existing LW dams. These are largely shaped by river and forest management (Seo et al., 2010; Lucía et al., 2015; Comiti et al., 2016a).
Forest management substantially affects the input of the species and sizes of the wood. The removal of trees from the riparian forest impairs the development of stable dams as they generally form the basis of them (Linstead and Gurnell, 1999). Along afforested riverbanks, LW densities (number of pieces per km of shoreline) are much higher than in open landscapes such as fields (Angradi et al., 2004). Also, river management largely affects the presence of LW. It often interferes with the flow regime and targets an increased flow conveyance. This includes the direct removal of LW but also targets to reduce the LW retention capacity of the channel. At stabilized riverbanks considerably less LW is found than on natural riverbanks consisting of sand or silt. Also, the normally high effect of riparian forests on LW density is diminished by stabilized riverbanks (Angradi et al., 2004). Combined this leads to a reduced possibility that stable LW dams can form which in turn means that potentially more LW and sediment are transported downstream and the blockage of infrastructure becomes more likely (Linstead and Gurnell, 1999). Therefore, Linstead and Gurnell (1999) come to the same conclusion as Benke et al. (1985):
Although there are certain situations that may require wood removal to eliminate stream blockage, the wisest management practice is usually no “management” other than protection of the adjacent flood plain (Benke et al., 1985, p. 12).
However, while indiscriminate removal of LW should be avoided, a (partly) removal of LW should be considered if the channel conveyance receives a high priority, or when extensive amounts of LW have entered the channel, e.g., by forest operations. Excess input of small materials into rivers might lead to fully closing the matrix of dams and thus hinder the migration of fish and other organisms. If the riparian forest is not managed, this problem is unlikely to occur (Linstead and Gurnell, 1999). Further, as it is with living vegetation, dead LW should be removed when less stable LW accumulations are close to infrastructure susceptible to clogging. However, in case the estimated flooding is of a lower extent, the partial removal of looser LW should be considered to reduce negative ecological and morphological impacts, such as the incising of riverbanks. The most stable LW pieces should remain at their place, therefore primarily LW should be removed that is not longer than the channel width, unstable and not fixed within the stream bed on one or two ends, by either being buried, anchored, or braced by riparian trees, boulders, bedrock outcrops, or by LW that does not have just listed characteristics. The removal should only happen at a certain length of the watercourse and not include the removal of all dams in the river, as especially active dams hold back LW in the upper stream system (Gurnell et al., 1995; Linstead and Gurnell, 1999). Besides the removal of LW, the systematic management of LW should also be supplemented by engineering approaches, such as bridge design and retention structures, to protect areas susceptible to LW (Schmocker and Weitbrecht, 2013; Mazzorana et al., 2018).
The various benefits of LW are increasingly acknowledged. However, often the watercourses and the floodplains are heavily managed, especially in Europe, and managed forested areas along rivers can often only insufficiently provide LW. Therefore, the reintroduction of LW into streams is increasingly considered for river restorations after years of active removal of LW from streams (Neuhaus and Mende, 2021; Swanson et al., 2021; Wohl and Iskin, 2022) and even described as state of the art for river restoration projects in Switzerland (Neuhaus and Mende, 2021). A further approach to river restoration is covered under the rewilding approach which for example includes the support of beaver populations. A study from Bavaria indicates that their dams have a similar effects on the flood hazard as active dams (Neumayer et al., 2020). Further insights on the role of beavers on, for example, river morphology are presented by, e.g., Levine and Meyer (2019) and Larsen et al. (2021). Besides beavers also the management of riparian forests influences the LW supply into streams (Linstead and Gurnell, 1999; Vaz et al., 2011; Neuhaus and Mende, 2021). Missing forests and the clearance of trees along the stream have the consequence that LW dams are more difficult to sustain. Therefore, forest management should include buffer strips along the watercourses including trees of different ages and sizes that can act as LW sources, where the risk for infrastructures is evaluated as low. The buffer stripe should optimally be 20 meters, as this reflects on the natural height of mature native tree species in forests, which defines the range in which wood can reach the watercourse, e.g., through wind throw. If a riparian forest only consists of a single-age structure that is not able to act as a source of LW the input will consequently be low and a (continuous) supply of LW to the watercourse might be necessary to sustain LW dams. Therefore, the development of a more structurally diverse forest should be supported, if the development of LW dams is envisaged (Gurnell et al., 1995; Linstead and Gurnell, 1999).
- Discussion: synergies and trade-offs between forest ecosystem health and flood risk reduction
In general, the objectives of forest ecosystem health and flood risk reduction are largely synergetic. A forest management that is focused on forest health by developing or allowing the development of successional patches and a multilayered structure and increasing growing and deadwood volumes stands in line with forest structures that support water retention in the landscape. High volumes of growing stock, with a multilayered structure, and deadwood support soil structures that have high water holding and infiltration rates. Further, it increases surface roughness reducing runoff. The presence of small successional patches enhances a continuous regeneration which in combination with the lower risk of healthy forests for large-scale disturbances from, e.g., wind throw or bark beetle calamities supports the constant provision of its flood hazard-reducing services. Existing forest disturbances reduce the forest’s ability to retain water. In accordance with the forest health objective for high deadwood volumes, the disturbed areas, however, should not be cleared as this would go along with a reduced soil roughness and, depending on the clearing operation, with a high and long-lasting loss of the soil’s capacity to retain water due to mechanical compaction. Further, deadwood supports the upcoming of new seedlings by acting as a seedbed with high water storage which gains special relevance on soils with low soil thickness or soils with a low water holding capacity. Further, disturbed forests with standing deadwood have lower temperatures compared to cleared areas which are also important for the upcoming and survival of new tree seedlings. Its importance is likely to increase with progressing global warming. Therefore, the restructuring of monocultures and single layered forest stands toward the objectives of forest health benefits in general also the reduction of the flood hazard. Due to its short-term effects, e.g., by not clearing dead spruce stands, and its long term effects by developing diverse and resilient forest structures, the objectives of forest health contribute to both, Eco-DRR and EbA.
Further, removing or adapting man-made infrastructures in the forest, like logging trails, will reduce the water runoff and benefit both the vitality of drought-affected forest stands and the reduction of the flood hazard. Moreover, these infrastructures might trigger landslides during heavy precipitation events that potentially mobilize large amounts of LW into channels (Figure 6E; Seo et al., 2010).
Along water streams, the relationship between forest health and flood risk is more complex. While the afore-discussed benefits of healthy forest structures also hold true along the streams, the close interactions with the water dynamics lead to more aspects that need to be accounted for in the management of riparian forests. This is particularly the case due to the ambivalent role of existing in-channel LW and vegetation that could become LW during a flood event. As described, dead in-channel LW and vegetation in the floodplain or on aligning slopes have the potential to deliver valuable services for the reduction of flood hazards. The objectives of forest health stand in line with it as the diverse structures and high volumes of deadwood and growing stock increase the flow resistance. Further, the multilayered structures and different successional patches enhance the root penetration which positively affects streambank (Simon and Collison, 2002; Docker and Hubble, 2008; Gasser et al., 2020) and slope stability (Moos et al., 2016; Vergani et al., 2017; Gasser et al., 2019). Moreover, these structures, translating in more and thicker stems, are likely to better retain already mobilized in-channel debris and to better keep deadwood than forest stands with fewer volumes of growing stock and less complex structures on the ground that could benefit the locking.
Further, a forest management that follows the objectives of forest health benefits the continuous LW supply. The different layers and successional patches support the regular supply of LW with different size distributions. In contrast to timber-focused forest management, this management would also facilitate the supply of long and stable LW, such as full tree stems. These can act especially in small and medium channels as key pieces for the development of persistent LW dams with their beneficial effects on water retention.
However, the in-channel LW and riparian vegetation might exacerbate the flood hazard in case they are mobilized during the flood event. Therefore, if located close to infrastructure susceptible to LW, the forest management objective of forest health might collide with the interest in flood risk reduction. Here, however, deadwood plays only a subordinate role for hazard creation. This is, as described, related to its lower proportions and its shorter and less stable characteristics compared to fresh LW (Rickli et al., 2018; Schalko, 2018). Higher volumes of deadwood in healthy forests might increase the proportions but as the overall contribution of deadwood from riparian forests is low (Steeb et al., 2017), this effect is likely to be negligible.
Another concern that potentially affects the management toward forest health objectives, is that disturbed forest areas could largely contribute to the mobilization of LW and thus should be cleared, as it has been discussed for example for the 2005 floods in Switzerland over wind throw forests. While the investigations of Waldner et al. (2007) in this regard found that storm-damaged wood was present, its overall contribution to the LW amounts was relatively small. The LW amounts were shaped majorly by the downstream erosion of deciduous trees. Similarly, it is likely the case for forest areas affected by bark beetle calamities and wildfires. Bark beetle calamities, which are widely present in many German low mountain ranges, result in large areas with standing deadwood. Here the potential to exacerbate the clogging is likely even lower compared to storm-hit areas, as once the dead tree stems fall, they will show lower stabilities and sizes due to different decomposition processes. Likewise, this is potentially the case for forest areas that were affected by wildfires. It was found that fire-affected LW is less complex (Vaz et al., 2011) and, thus, has a lower clogging potential. Different forest disturbances, such as wildfires,2 will increasingly come along with climate change and therefore also influence LW dynamics. This will further exacerbate the predictability of how and from where most of the LW is mobilized (Swanson et al., 2021). However, so far there are no indications that deadwood from disturbed areas, and consequently the objectives of forest health, contributes over proportionally to the overall LW volumes mobilized during a flood event and, therefore, should be cleared. On the contrary, a clearance could lead to higher erosion rates and trigger mass movements potentially transporting LW.
The major risk of healthy forests in areas that are susceptible to LW is therefore not stemming from their contribution to deadwood supply but from the mobilization of high volumes of uprooted vegetation during a flood event, especially in medium to large channels. Here particularly late successional patches with outgrown trees pose a risk due to their high potential in contributing to clogging of bridges. As such, in line with the guiding framework by Gasser et al. (2019), this work suggests to remove big trees with a high LW potential in riparian areas close to susceptible infrastructure. While Gasser et al. (2019) provide guidance under which conditions of the channel and adjunct slope the positive services of forests could be outweighed by the potential of being transformed to hazardous LW and suggestions on how vegetation should be (re)structured, no information on deadwood management within these areas is provided. The findings of this work suggest that deadwood from riparian forests play only a negligible role in the severity of the flood hazard. Therefore, in line with the objectives of forest health, we argue that deadwood in riparian forests and adjunct slopes should remain. The existence of high deadwood volumes might also partly compensate for the preventive removal of big trees to reduce the LW potential as suggested by Gasser et al. (2019). The preventive removal will negatively affect the inner forest climate and therefore the success of natural regeneration that is essential for continuous root penetration. The existence of deadwood will support the upcoming of natural regeneration due to its water-storing properties. Further, during flood events, it can provide higher flow resistance and direct the water into the forest stand. In case the risk of large deadwood sizes is considered too high due to potential high erosion rates or the probability of landslides engineering approaches should be considered, especially, downstream of areas that are not accessible for LW removal, e.g., due to difficult terrain or environmental regulations. Further, the concision of deadwood might be an approach to decrease the risk of clogging, as already suggested by Bänziger (1990) three decades ago, while maintaining most of its beneficial properties inside the riparian forest.
- Conclusion
This study has elaborated on the relations between the two forest management objectives of forest health and flood risk reduction. For this, we have provided a theoretical background on related concepts and have presented a working-definition for forest health. The different aspects of how a forest potentially can affect the flood hazard have been reviewed. This included the forests’ influences (1) on the discharge of water from the landscape into channels, as well as (2) its influences on the characteristics of the channel and its riparian area and their respective influence on the transport of water, sediment, and debris. The review’s results and discussion suggest that the two objectives are largely synergetic. However, in direct proximity to watercourses trade-offs might occur. This is especially due to the ambivalent relation of living vegetation and deadwood to flood hazard. In places where no susceptible infrastructures to clogging exist, diversely structured vegetation and deadwood should be supported due to its beneficial effects on water retention and channel characteristics. Here, the objectives of forest health and flood risk reduction are aligned. In places where susceptible infrastructures exist, trade-offs between the two objectives arise. Here the potential of freshly uprooted vegetation to cause damages should be reduced while maintaining the vegetation’s supportive characteristics, e.g., concerning bank and slope stability. For this, primarily the adjustment of existing vegetation structures toward a better root penetration should be targeted while reducing the prevalence of high growing stocks that could result in large and stable LW sizes. Where the risk of clogging is perceived as too high, also the selective removal or shortening of dead in-channel wood can be considered. However, based on the literature review the risk deriving from dead LW is evaluated as comparability low. This is related to its generally lower proportions and its smaller and less stable characteristics compared to freshly uprooted vegetation. In this regard more research needs to be conducted. More insights on the proportions and sources of dead and fresh LW after flood events are necessary to better evaluate the respective roles and, therefore, to deliver a better foundation for the connections and suggestions made in this review. Furthermore, more studies are needed focusing on low-ranging mountains as well as floods spanning several orders of magnitudes (ranging up to 500–1,000 years events like the flood in the Ahr valley). Lastly, we want to acknowledge that forest health and flood risk are only two of the various objectives shaping forest management. Therefore, the findings of this review should be seen as a contribution toward a comprehensive forest management taking different interests into account.
THIRD:
Contrasting Regeneration Patterns in Abies alba-Dominated Stands: Insights from Structurally Diverse Mountain Forests across Europe
by Bohdan Kolisnyk 1,*ORCID,Camilla Wellstein 2,3,*ORCID,Marcin Czacharowski 1ORCID,Stanisław Drozdowski 1ORCID andKamil Bielak 1ORCID
1
Department of Silviculture, Institute of Forest Sciences, Warsaw University of Life Sciences, Nowoursynowska 159, 02-776 Warsaw, Poland
2
Faculty of Agricultural, Environmental, and Food Sciences, Free University of Bolzano-Bozen, Piazza Università 5, 39100 Bolzano, Italy
3
Competence Center for Economic, Ecological and Social Sustainability, Free University of Bolzano-Bozen, Piazza Università 1, 39100 Bolzano, Italy
*
Authors to whom correspondence should be addressed.
Forests 2024, 15(7), 1182; https://doi.org/10.3390/f15071182
Submission received: 24 May 2024 / Revised: 1 July 2024 / Accepted: 3 July 2024 / Published: 8 July 2024
(This article belongs to the Special Issue Ecosystem-Disturbance Interactions in Forests)
Downloadkeyboard_arrow_down Browse Figures Versions Notes
Abstract
To maintain the ecosystem resilience to large-scale disturbances in managed forests, it is essential to adhere to the principles of close-to-nature silviculture, adapt practices to the traits of natural forest types, and utilize natural processes, including natural regeneration. This study examines the natural regeneration patterns in silver fir (Abies alba Mill.)-dominated forests, analyzing how the stand structure—tree size diversity, species composition, and stand density—affects the regeneration. We analyze the data from four sites in Poland, Germany, and Italy, employing generalized linear and zero-inflated models to evaluate the impact of the management strategies (even- vs. uneven-aged) and forester-controlled stand characteristics (structural diversity, broadleaf species admixture, and stand density) on the probability of regeneration, its density, and the developmental stages (seedling, small sapling, and tall sapling) across a climatic gradient. Our results indicate a significantly higher probability of regeneration in uneven-aged stands, particularly in areas with lower temperatures and lower overall regeneration density. The tree size diversity in the uneven-aged stands favors advancement from juveniles to more developed stages (seedling to sapling) in places with higher aridity. A denser stand layer (higher stand total basal area) leads to a lower density of natural regeneration for all the present species, except silver fir if considered separately, signifying that, by regulating the stand growing stock, we can selectively promote silver fir. A higher admixture of broadleaf species generally decreases the regeneration density across all the species, except in a water-rich site in the Bavarian Alps, where it had a strong positive impact. These findings underscore the complex interactions of forest ecosystems and provide a better understanding required for promoting silver fir regeneration, which is essential for a close-to-nature silviculture under climate change.
Keywords: uneven-aged silviculture; mixed stands; ecosystem resilience; silver fir; sustainable forest management
- Introduction
The continuity of forest cover is a key element of sustainable forest management and strongly depends on the effective tree regeneration resulting from the successful completion of a series of interconnected events. Any disruption in the sequence of these events can lead to the failure of the entire process, which is highly undesirable in places where the socio-ecological role of forests is incredibly high [1,2]. Mountain forests serve as primary buffers on slopes, playing a crucial role in protecting against soil erosion, water runoff, avalanches, landslides, and torrential floods. The integrity of mountain forests is crucial for their ecological functioning as habitats for wildlife and biodiversity hotspots. Mountain forests also play an increasingly important role in the provisioning of social functions like places for tourism and connectivity with nature in today’s fast-paced world [3]. While the idea of protecting these forests through strict protection or minimal interventions is widely recognized by society, the reality is ominously different.
The current state of the European mountain forests is largely shaped by human activity and timber demands. Global warming-driven changes will further alter this landscape significantly, yet the demand for timber will persist, making the strict protection of large areas unfeasible. A notable example is the widespread Norway spruce (Picea abies (L.) Karst.) dieback in monodominant, monolayered plantations, which have replaced the mostly naturally diverse mixed mountain forests. There is a pressing need to restore these ecosystems to their “pre-management” diversity and support the species adversely affected by human activities. It is essential to develop sustainable management practices based on uneven-aged silviculture principles to promote a continuously high level of ecosystem functioning to satisfy the growing demand for timber and non-timber ecosystem services [4,5,6,7].
Silver fir (Abies alba Mill.), one of the species impacted by the artificial expansion of Norway spruce, could reclaim its place in the European mountain forests. More resilient to warm climates and drought than spruce [8], silver fir thrives in the montane zone and is found even in the flatlands in Poland and Ukraine [9,10]. Alongside European beech (Fagus sylvatica L.), it is extremely shade-tolerant and can be classified as a competitive stress tolerator, making it well-suited for uneven-aged silviculture [11,12,13]. The continuity and the demographic stability of silver fir-dominated forests hinge on a robust bank of natural regeneration. Recent decades have witnessed a decline in mature silver fir stands, but also in the natural regeneration layer at some places, even often despite the presence of sufficient seed production [14]. The key issues include suboptimal microsite conditions and stand characteristics, but also browsing by ungulates [15,16].
Paluch and Jastrzębski [14] observed the highest regeneration in the pure fir stands, with the regeneration reducing in the mixed stands as the proportion of silver fir decreased. Similarly, Dobrowolska [17] emphasized that the increase in the fir percentage in a stand correlates with enhanced regeneration quantity and height increments. While a higher percentage of silver fir in a stand composition appears to be beneficial for the regeneration quantity in the short term, long-term issues like allelopathic auto-intoxication in pure silver fir stands are well-documented [18]. Furthermore, Paluch and Jastrzębski [14] found that the survival rate of the fir regeneration in nearly mono-specific stands (90% of silver fir) with an admixture of Norway spruce, European beech, and Scots pine (Pinus sylvestris L.) was higher than in pure fir stands. However, the benefits offered by the admixture of other tree species might not be sufficient to offset the reduced seed availability.
The impact of the stand-level tree size diversification on the success of forest stand regeneration remains ambiguous. It is posited that uneven-aged stands exhibit greater resilience to disturbances such as windthrows at the stand level, enabling rapid regeneration. This resilience is attributed partly to the presence of multiple layers, particularly the understory, as well as to the existing regeneration [19]. Hence, as the introductory step to our research, we would like to confirm whether uneven-aged silver fir-dominated stands show a higher self-regeneration capacity compared to even-aged stands, which is gauged by the likelihood of seedling and sapling emergence to ensure forest continuity and maintain the demographic balance of silver fir under varying climatic conditions. Thus, our first hypothesis (H1) is that the probability of regeneration (for all the species and specifically for silver fir) is higher in uneven-aged stands than in even-aged stands (Figure 1, H1). Following the verification of the first hypothesis, we aim to explore in detail how the structural (tree size diversity) and compositional (admixture of broadleaved tree species) characteristics of uneven-aged stands (i.e., stands with at least two age classes or layers) influence the density of regeneration, focusing particularly on silver fir. Hence, our second and third hypotheses (H2) posit that improved stand characteristics (increased tree size diversity, admixture of broadleaf tree species, and reduced total basal area) positively influence the total density of natural regeneration (Figure 1, H2) and (H3) increase the dominant developmental stage of natural regeneration (seedlings, small saplings, and tall saplings) of silver fir (Figure 1, H3). The presence of large overstory trees is considered to be a good forestry practice, leading to increased biodiversity and other socio-ecological functions. For instance, in Poland, a minimum of five large trees per hectare should remain [20]. Considering that large trees can also produce an excessive amount of seeds and are not inherent in the tree size diversity, we are interested in whether the (H4) presence of large overstory silver fir trees contributes to a higher density of silver fir regeneration (Figure 1, H4). Understanding how silver fir at the youngest age responds to different management strategies and stand characteristics, which foresters can influence, is vital for active protection and sustainable management.
Forests 15 01182 g001
Figure 1. Schematic representation of formulated hypotheses (H1–H4).
- Materials and Methods
2.1. Research Area
The study was conducted across four sites in Poland, Germany, and Italy, covering a wide climatic gradient and a large part of the silver fir natural range (Figure 2; Table 1).
Forests 15 01182 g002
Figure 2. Map showing the locations of the study sites, natural range of silver fir (source: European Forest Genetic Resources Programme, EUFORGEN), and average annual precipitation for the period 2012–2022 (source: European Climate Assessment & Dataset, ECAD [21]).
Table 1. The general characteristics of the research sites.
Table 1 is located in the Zagnańsk Forest District, within the Małopolska Upland, on the outskirts of the Świętokrzyskie Mountains range. This district marks the northeastern boundary of the silver fir’s distribution, presenting a unique opportunity to study the marginal populations facing environmental stress and explore regeneration dynamics at the species’ range limit [22]. The second site (PL2) is in the Nawojowa Forest District, situated in the western part of the Low Beskid Mountains. This area lies on the east of the Carpathian flysch, featuring a geological formation of alternating sandstone and shale layers, and is a part of the Sądecki Beskids [23]. Contrary to Zagnańsk, Nawojowa provides optimal conditions for fir growth and development, serving as an uninterrupted habitat for silver fir.
IT in the Tisens-Laurein region is nestled within the Southern Alps in South Tyrol, Italy. IT1 experiences heavy snowfall during the winter, leading to prolonged snow cover that significantly affects local hydrology and ecosystem dynamics. Plots in IT1 are located in the mountain mixed forest zone, close to the upper boundary, as evidenced by the limited vertical growth of broadleaf species (Figure S1). Lastly, the site in Inzel (GE1), situated in the Bavarian Alps, Germany, is an exceptionally water-rich area (Table 1) with pre-Alpine climatic conditions marked by distinct seasonal changes. Cold winters with steady snowfall and mild, moist summers define the climate, with snowpack playing a crucial role in the regional water cycle. The research plots in GE1 are in the middle of the mountain mixed forest zone. All four study sites were severely affected by ungulate browsing, with silver fir being the most heavily impacted among silviculturally important tree species on our plots.
2.2. Data Collection
At each site, 34–36 circular plots were established, with an area of 0.05 hectares, using Field-Map technology [24], to have at least 3-4 replications for each stand type (Figure 3). By having 3-4 replications per stand type, we ensure that our dataset is robust, captures the variability within each type, and meets the minimum requirement for regression analysis. These plots span a gradient from simple, pure, even-aged to complex, mixed, uneven-aged silver fir-dominated forest stands, with the fir constituting ≥ 40% of the stand’s basal area. For even-aged stands, a pair of pre-mature (40–80 years old) and mature (80–120 years old) were included per site. The admixture tree species predominantly include European beech and Norway spruce. The research plots are distributed across stands with varying tree size diversity and admixtures of broadleaf tree species (Figure 3; Figure S2). The positioning of plots within each site is random and confined within a maximum 10 km radius.
Forests 15 01182 g003
Figure 3. Plot selection matrix, including tree size diversity (vertical axis) and admixture of broadleaf tree species (horizontal axis). This design was used for plot selection purposes in the field to ensure that, at each site, plots equally cover compositional and structural gradients. All plots were used to answer H1; plots classified as transition stage and uneven-aged were used to answer H2–H4.
All trees larger than 7 cm in diameter at the breast height (DBH) within the plots were mapped in a 3D local coordinate system. Following, standard dendrometric measurements were performed, including DBH in mm, species identification, and documentation of any significant damages (such as wind breakage, decay, and beetle damage). For selected trees, additional measurements were taken, including total height, height to crown base—the height to the lowest living branch forming the continuous crown—and crown projection in four directions. Height curves were fitted to estimate the heights of the non-measured trees (Figure S1).
Natural regeneration was measured within concentric subplots. In the first subplot, with a radius of 1.26 m, seedlings aged 2+ years and under 50 cm in height (after seedlings) were recorded, with the total count of seedlings per species noted. The second concentric plot, with a radius of 2.52 m, focused on small saplings taller than 50 cm but with a DBH of less than 2 cm (after small saplings). The third subplot, with a radius of 3.99 m, included measurements for specimens with a DBH between 2 and 7 cm (after tall saplings) (Figure 4).
Forests 15 01182 g004
Figure 4. Schematic representation of concentric subplots for natural regeneration, where r1, r2, and r3 are the radiuses of the subplots.
2.3. Tree Size Diversity and Admixture of Broadleaf Tree Species
To address H1, the initial classification of forest stands into even-aged and uneven-aged was performed directly in the field and later validated by measuring age at breast height (from increment cores) to mitigate bias. The number of even-aged plots per site was around 10–12. To answer hypotheses H2–H4 and to numerically describe the continuous tree size diversity, we used the Shannon diversity index (ShD) with the 4-m height classes and the sum of the basal area (BA) of all trees larger than 7 cm per class as a proxy of the class share. For instance, in class (4,8], we summed up the basal area of all trees taller than 4 m but equal to or shorter than 8 m and used it as a proportion of trees in the class (4,8]. The generalized formula is based on the proportion of observed objects in the selected class to the total across all classes (Equation (1)).
𝑆ℎ𝐷=−∑𝑖=1𝑁𝑝𝑖×ln(𝑝𝑖)
(1)
where
N is the number of classes;
𝑝𝑖
is the proportion of trees in the i-th class.
Subsequently, to facilitate comparison across different locations, the Shannon diversity index values were normalized to a 0–1 scale, with each site’s diversity score divided by the maximum observed value for the site. The share of broadleaf tree species within each plot was determined by calculating the proportion of the total BA represented by broadleaf species.
2.4. Developmental Stage of Fir Regeneration
To estimate the dominant developmental stage (DS) of silver fir regeneration (seedling, small sapling, and tall sapling) within each plot, we employed a straightforward approach (Equation (2)) based on the adjusted formula of the weighted mean [25].
𝐷𝑆=1×seedling+3×small saplings+5×tall saplingsdensity all stages
(2)
where
seedlings, small saplings, and tall saplings represent the per hectare density of silver fir regeneration in each respective developmental stage;
density of all stages denotes the total density per hectare of silver fir regeneration across all developmental stages;
1, 3, 5—weights that signify the increasing importance of saplings with increasing developmental stage.
This approach was selected to provide a simple yet effective representation of the predominant developmental stage of silver fir regeneration across the plots. The weighting scheme (1, 3, 5) was empirically tested against an alternative conventional scheme (1, 2, 3) that is typically the first choice for such purposes. The results showed that the DS calculated with weights 1, 3, 5 produced residuals that were more favorable for the fitted model’s performance. Specifically, the residuals with the 1, 3, 5 weighting scheme were smaller and more uniformly distributed, indicating a better fit and more accurate representation of the predominant developmental stages.
2.5. Climatic Data
To account for climatic differences between regions, we utilized different data sources. For IT1 and GE1, climate grid data from ECAD E-OBS were used, featuring a high resolution of 0.1 degrees with daily resolution [21]. However, for PL1 and PL2, data from the Bartków and Nowy Sącz climatic stations were used due to the observed tendency of ECAD E-OBS data to underestimate precipitation in central Poland.
To more accurately assess the impact of precipitation and temperature on natural regeneration processes, we calculated the mean de Martonne aridity index for the last 20 years [26] as a product of these two variables (Equation (3)). The lower the Martonne index, the more arid the climate is.
𝑀𝑎𝑟𝑡𝑜𝑛𝑛𝑒=𝑃𝑇+10
(3)
where
P—annual sum of precipitation in millimeters;
T—average annual temperature in degrees.
2.6. Density of Ungulates
Ungulates, particularly red deer (Cervus elaphus L.) and roe deer (Capreolus capreolus L.), play a significant role in shaping forest ecosystems through browsing habits, often favoring more palatable tree species such as silver fir. This interaction can lead to changes in forest composition. Despite similar average densities of red deer across Poland, Italy, and Germany, their spatial distribution within these countries shows distinct patterns. In Poland and Germany, red deer are relatively uniformly distributed, with some areas of concentration, while, in Italy, about 75% are found in the central and eastern Alps [27,28,29].
Data on ungulate populations were sourced locally due to the lack of uniform data across different regions. For site IT1, red deer density data were obtained from the Autonomous Province of Bolzano—South Tyrol, Department of Forestry Services, Office for Wildlife Management. The mean density of red deer per hectare over the past 10 years was calculated from this source. Similarly, for sites PL1 and PL2 in Poland, data on red and roe deer densities over the previous 10 years were provided by the Zagnansk and Nawojowa forestry districts, respectively. In Germany, for site GE1, official red deer density data were unavailable; instead, data systemized by Suzanne T. S. van Beeck Calkoen [30] were used. To estimate the red deer density in GE1, two points from [30] nearest to the GE1 study site were selected, and the distance-weighted mean was calculated.
2.7. Statistical Analysis
2.7.1. Probability of the Regeneration
To test the first hypothesis, a logistic regression analysis was performed using the glm function from the “stats” package as part of the core R [31], with the binomial distribution family and logit link, focusing on binary outcomes of the probability of regeneration influenced by two predictors, namely structure type (even/uneven-aged) and site (Equation (4)). Due to the presence of regeneration on all the plots, and thus the insignificance of the data from GE1 for this regression, they were not used. Variables explaining climatic differences were not found to be significant.
𝑙𝑜𝑔(𝑝1−𝑝)=𝛽0+𝛽1×𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒+𝛽2×𝑆𝑖𝑡𝑒
(4)
where
𝑝
denotes the probability of the regeneration (all species);
coefficients 𝛽0
, 𝛽1
, and 𝛽2
correspond to the intercept and the effects of the two predictors on the log odds of the event, respectively;
Structure—structure type (even/uneven-aged);
Site—study site.
Following, a second logistic regression was fitted, with the probability of silver fir regeneration as a dependent variable, mean annual temperature and stand structure (even- and uneven-aged) as dependent variables, and site as a random effect using the data from all sites.
𝑙𝑜𝑔(𝑝1−𝑝)=𝛽0+𝛽1×𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒+𝛽2×𝑇𝑒𝑚𝑝+α𝑆𝑖𝑡𝑒
(5)
where
𝑝
denotes the probability of the regeneration (silver fir);
coefficients 𝛽0
, 𝛽1
, and 𝛽2
correspond to the intercept and the effects of the two predictors on the log odds of the event, respectively;
Structure—structure type (even/uneven-aged);
Temp—mean annual temperature for the last 20 years;
α𝑆𝑖𝑡𝑒
—random intercept for each site.
2.7.2. Regeneration Density
To answer the second hypothesis of our research using the “glmmTMB” R package version 1.1.9 [32], we fitted two zero-inflated (ZI) models with the truncated negative binomial distribution (TNB), chosen to address overdispersion and an abundance of zero observations (Table 2). Despite the final similarity in our case, in contrast to two-step models, ZI models assume that there are two kinds of zeros in the data: structural zeros and zeros that occur as the result of the count process. The ZI models simultaneously calculate the probability of an excessive amount of structural zeros (a logit model for the ZI part) and, in the case of the TNB, only the count model for non-zero data [32]. A simultaneous ZI model was selected over the classic hurdle model to check whether count zeros are crucial for the model or not without drastically changing the computation behind it (negative binomial distribution vs. truncated negative binomial).
Table 2. Overview of ZI models’ components and variables used and tested for the regeneration density; those tested and not included were found to be statistically insignificant.
2.7.3. Dominant Developmental Stage of Silver Fir Regeneration
To answer the third hypothesis, we evaluated the Gamma and Inverse-Gaussian distributions within the framework of generalized linear models (GLMs) to accommodate the right-skewed nature of our response variable using the same “stats” base R package [31]. The choice of an appropriate link function was crucial for linearizing the relationship between predictors and the response variable. To this end, we explored a variety of link functions, including the identity, logarithmic, inverse, square, square root, and beta distribution functions. The selection of the optimal model was based on minimizing the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), and also considering the normality of the simulated residuals. This process led us to select the GLM with a Gamma distribution paired with a log link function as the best-fitting model (Table 3).
Table 3. Table of GLM components and variables used and tested for the regeneration-dominant developmental stage.
2.7.4. Presence of Large Overstory Trees
To answer our last hypothesis (H4), we initially defined a criterion for what constitutes an overstory tree, describing an overstory tree as any tree reaching at least 80% of the height of the tallest tree within a given site. The influence of these overstory trees on regeneration density was then analyzed using a binary variable (0 for absence; 1 for presence).
3. Results
3.1. Probability of the Regeneration Appearance
The regeneration density of all the species exhibits considerable variation across the sites, with silver fir consistently emerging as the dominant species in the regeneration layer. In GE1 and IT1, the mountain environment and higher elevations contribute to a greater share of spruce within the regeneration layer. Conversely, at lower altitudes, beech—and hornbeam only in PL1—predominantly complement the regeneration layer (Figure 5).
Forests 15 01182 g005
Figure 5. Comparison of the mean natural regeneration densities by species across four sites.
The likelihood of regeneration across various plots, regardless of the developmental stages (seedlings, small saplings, and tall saplings) and species involved, was significantly higher in the plots managed under uneven-aged forestry practices (Table 4; Figure 6). We observed a consistent trend where the probability of fir regeneration decreased, as well as a widening disparity between the even- and uneven-aged stands with a decrease in the average annual temperature (Figure 6). This trend was especially notable in IT1, where the average annual temperature and the overall density of regeneration were significantly lower compared to the other sites (Figure 5; Figure 6). Similarly, a difference in the regeneration of all the species between the even- and uneven-aged stands was observed. However, for all the species, the influence of the average annual temperature on the regeneration appeared to be insignificant.
Forests 15 01182 g006
Figure 6. The probability of the regeneration for silver fir (a) and all species (b) in even- vs. uneven-aged stands across four sites and temperature gradients, with horizontal lines representing predicted values and boxes confidence intervals. Two solid lines connecting predicted values depict the trend of increased regeneration probability with increasing temperature.
Table 4. Logistic regression summary on the influence of forest structure on the likelihood of regeneration (all species and fir) across four sites.
3.2. Regeneration Density
Across all the sites, higher total basal area (TBA) is associated with a decrease in the density of natural regeneration of all the species regardless of the tree size diversity and the presence of a broadleaf species admixture in the stand layer (Figure 7). Interestingly, while the inclusion of broadleaf species generally tended to slightly (but statistically significantly, Tables S1 and S2) decrease the density of the natural regeneration in places with a higher aridity index (PL1, PL2, and IT1), it notably increased the regeneration density in water-rich GE1 (Figure 7).
Forests 15 01182 g007
Figure 7. Relationship between broadleaf species admixture (mainly European beech) and the natural regeneration density of all present species, considering a climatic gradient (Martonne index) and varying levels of normalized total basal area (TBA). Four colored lines represent modeled trends for study site-specific Martonne index values, while shaded areas indicate 95% confidence intervals (CI). The admixture of broadleaf tree species is expressed as the proportion of broadleaf species in the forest stand’s total basal area (a value of 0 indicates pure silver fir stands). The TBA is a continuous variable, with values close to 1 indicating maximum stand density in our plots. The selected TBA levels were chosen to cover the range of the parameter. Note that factors contributing to the relatively wide CI include high inherent variability in natural regeneration data, sample size, model complexity, and data range. Overlapping CIs may appear as a different color.
When considering solely silver fir regeneration, the findings diverge slightly. The density of silver fir regeneration does not appear to be influenced by TBA. Yet, similar to the overall trend, an increasing admixture of broadleaf species is associated with a decrease in the silver fir regeneration density in PL1, PL2, and IT1, whereas, in GE1, with a lower aridity index (high water sufficiency), it is associated with a significant increase (Figure 8).
Forests 15 01182 g008
Figure 8. Relationship between broadleaf species admixture (mainly European beech) and the natural regeneration density of silver fir, considering a climatic gradient (Martonne index). Four colored lines represent modeled trends for study site-specific Martonne index values, while shaded areas indicate 95% confidence intervals (CI). The admixture of broadleaf tree species is expressed as the proportion of broadleaf species in the forest stand’s total basal area (a value of 0 indicates pure silver fir stands). Note that factors contributing to the relatively wide CI include high inherent variability in natural regeneration data, sample size, model complexity, and data range. Overlapping CIs may appear as a different color.
3.3. Dominant Developmental Stage of Silver Fir Regeneration
The tree size diversification has demonstrated a clear positive impact on the advancement of silver fir in the regeneration layer, indicating that vertical diversification significantly contributes to the progression of silver fir from the early growth stages (seedlings) to more developed stages (saplings) (Table 5; Figure 9). Notably, neither the TBA nor the admixture of broadleaf tree species within the stand layer has a discernible effect on the developmental stages of silver fir regeneration.
Forests 15 01182 g009
Figure 9. Relationship between tree size diversification in uneven-aged stands (only stands with two or more age classes), expressed as the Shannon diversity index (showing how diverse trees are in terms of size in the stand layer) and the dominant developmental stage of silver fir natural regeneration across a climatic gradient (Martonne index). The dominant developmental stages of natural regeneration range from 1 (predominantly seedlings) to 5 (predominantly tall saplings), indicating the advancement of natural regeneration. Four colored lines represent modeled trends for study site-specific Martonne index values, while shaded areas indicate 95% confidence intervals (CI). Note that factors contributing to the relatively wide CI include high inherent variability in natural regeneration data, sample size, model complexity, and data range. Overlapping CIs may appear as a different color.
Table 5. GLM summary on the influence of forest tree size diversity on the DS of silver fir.
- Discussion
4.1. The Likelihood of Regeneration Success—To Be or Not to Be?
The principle of fostering a robust bank of natural regeneration is anchored in the objective of establishing a resilient ecosystem capable of maintaining a stable level of ecosystem services provisioning, species demographic equilibrium, and prioritizing naturally favored species [33,34,35]. This is especially critical in the face of large-scale disturbances, which have become more frequent during the last decade [36]. Creating favorable conditions for the rapid restoration of ecosystem functioning by carefully selecting a forest management approach is paramount, even more so under the challenges posed by climate change [37,38].
Our findings support the implicit idea, resonating in numerous publications, e.g., [39,40,41], that uneven-aged stands are more likely to undergo natural regeneration than even-aged stands with a simplified vertical structure and closed canopy. This holds for all the species collectively and specifically for silver fir. Moreover, the disparity in the regeneration probability between the even-aged and uneven-aged stands is more pronounced for all the species than for silver fir alone (Figure 6).
For the germination of most forest tree species’ seeds, three fundamental conditions are required: sufficient water, oxygen, and an optimal temperature range [42]. However, for the successful establishment and survival of regenerations beyond their first years, the substrate and the level of light availability become critical factors [43,44] in addition to the possible disturbance by browsing. Different species employ diverse life strategies and follow distinct recruitment patterns, resulting in varying resource requirements. Consequently, the environmental heterogeneity at the stand level, especially concerning the light availability, microsites, and soil conditions that partly arise from small-scale disturbances of uneven-aged forestry, is creating multiple niches for varied species regeneration [45,46]. The uneven-aged silviculture primarily involves modulating the competition for light across the forest’s vertical profile, including the natural regeneration layer [19]. The increased light availability facilitates more optimal conditions through enhanced solar radiation reaching the forest floor, creating a favorable environment for regeneration development in general. For instance, Scheller and Mladenoff [47] compared the even-, uneven-aged, and old-growth forests in terms of the understory plant communities, including the regeneration of wooden plants. Scheller and Mladenoff [47] found that the understory species richness was lower in old-growth forests compared to even-aged forests, and, most importantly for us, lower in even-aged stands compared to uneven-aged stands, and attributed the difference to the available light and deadwood debris.
Certain tree species, including silver fir, can derive benefits from the limited light under the horizontally closed canopy of even-aged stands and degraded microsite conditions. Such trees experience reduced competition from other tree species in the regeneration layer, shrubs, and grasses [48]. Particularly in their early developmental stages, seedlings of extremely shade-tolerant species exhibit a remarkable degree of tolerance. Despite the minimal photosynthesizing area, which restricts growth, the small silver fir seedlings under the canopy can maintain their current stage for an extended period of time [49]. A study conducted in Comelico (Italian eastern Alps) in Norway spruce–silver fir stands showed that, in the gap system, the fir saplings were more abundant in the understory and less in the gaps, as compared with spruce [50]. Furthermore, even though the age structure of the regeneration in the gap showed that most of it appeared after the formation of the gap, saplings taller than 2 m were predominantly already present at the moment of the gap harvest [50], underscoring the importance of the understory presence for a rapid post-disturbance recovery. However, the advantage gained from the reduced competition does not outweigh the benefits derived from light availability. In even-aged stands, natural regeneration predominantly remains at the seedling stage, with the saplings that manage to develop often exhibiting a suboptimal “silvicultural” quality mainly due to insufficient light (Figure S2).
In addressing the rhetorical question of whether natural regeneration is more viable in uneven-aged stands, our research affirms its feasibility. Following this, we focused on identifying the forester-controlled stand characteristics only within the class of uneven-aged stands that could further enhance natural regeneration.
4.2. The Complexity of Choice: The Impact of Stand Density and Tree Size Diversity on the Natural Regeneration
We found that the tree size diversity does not affect the regeneration density, suggesting that any level of diversification in the vertical profile, which enables increased light penetration to the forest floor and/or improved microsite conditions, is sufficient to create an optimal environment for seed germination and the initial establishment of regeneration. Further diversification, even with improved microsite conditions and the availability of light, is not needed at the early developmental stages. However, as seedlings grow and progress and the demand for resources increases, a higher stratification in the vertical profile could theoretically partition the competition not only for light but also for other resources [46,51]. This provides a window of opportunity for seedlings to grow and develop, thus enhancing the average developmental stage of fir regeneration. This is supported by our findings that, in IT1, PL1, and PL2, where the main resource of competition is most likely water, the average developmental stage of regeneration increases with increasing tree size diversity, whereas, in the water-rich GE1, the tree size diversity appears to have no significant effect. While the quantity of regeneration may not increase, its quality, in terms of being in a more advanced developmental stage, improves, potentially enabling faster post-disturbance recovery. Moreover, reducing the TBA significantly boosted the density of all the species, although it did not notably influence the silver fir density alone. Thus, when the goal is to promote the natural regeneration of silver fir, selecting the locally optimal stand density emerges as an effective strategy.
The process of formation and further management of stands with high tree size diversity, such as those managed under the renowned “Plenterwald” system, is recognized for its complexity, time-intensive nature, and associated initial costs [52,53,54,55]. The maintenance and stable functioning of these systems necessitate frequent interventions [19,56]. While on a long-term scale such practices have proven to be economically viable in regions with well-developed infrastructure and accessible terrain, the feasibility of frequent interventions becomes questionable in remote areas with less developed infrastructure or challenging terrain [57,58,59]. In these contexts, in practice, foresters often resort to the less frequent single-tree selection system or shelterwood, creating simpler stand structures like two-layered stands [60]. Nevertheless, the critical question persists: do the advantages of an improved regeneration developmental stage, which acts as a form of ecological insurance for the quick recovery of ecosystem functions, outweigh the extra costs in areas where frequent management actions are economically demanding? This question, while beyond the scope of our current research, opens the field for further research.
4.3. Influence of Broadleaf and Conifer Interplay on the Natural Regeneration
The impact of species mixtures on the natural regeneration patterns of tree species is a topic of ongoing research within forest ecology, reflecting a limited understanding of this area. The high level of uncertainty in this field is often linked to the complexity of forested ecosystems, especially those characterized by a high level of diversity, which leads to increased entropy in terms of microsite conditions [61,62,63]. Generally, the introduction of broadleaved species at the stand level is observed to decrease the density of silver fir regeneration, with a belief that low levels of admixture might positively influence the fir regeneration density [14,17]. However, the threshold at which the benefits of improved microsite conditions offset the drawbacks of a reduced seed source remains ambiguous. Our findings in places with a comparatively higher aridity index (PL1, PL2, and IT1) align with the commonly observed trend: a small but overall reduction in the density of silver fir regeneration within stands dominated by fir. Contrary to expectations, the situation significantly diverged with increasing water availability (GE1). The GE1 is characterized by the presence of large overstory fir trees, which can produce a substantial amount of seeds and potentially offset the drawbacks of a reduced number of seed sources. Yet, our models failed to confirm our H4 and such a relationship in general. The possibility of allelopathic effects or soil acidification influencing the regeneration patterns was considered unlikely given the long history of mixed stand management in this region. We assume that the distinct outcomes observed in GE1 are predominantly the result of unique local climatic conditions, signifying that the climate plays a crucial role in shaping the impact of species mixtures on regeneration. This underlines the complexity inherent in forest ecosystem dynamics and emphasizes the need for continued research to elucidate these complex interactions with greater clarity and precision [64,65,66].
4.4. Influence of Ungulates on the Regeneration of Silver Fir
Our study found no significant impact regarding the differences in ungulate density and thus browsing pressure on the regeneration success, density, and developmental stage of silver fir and other tree species. This is not in line with the expectations based on previous case studies that have consistently demonstrated that ungulates preferably browse silver fir trees [51,67]. The presence of red deer has been specifically linked to changes in the regeneration composition, heavily affecting silver fir regeneration.
The role of the ungulates in silver fir regeneration dynamics is complex and influenced by multiple other factors [68]. The differences in data sources and methodologies to estimate the ungulate density and the resulting browsing pressure contribute to high uncertainty, potentially explaining the discrepancies with previous studies. The broad spatial and temporal scale has inherent high variability; thus, more precise long-term local observations of ungulate density and direct measures of ungulate browsing (including control plots without browsing) are needed to fully understand the natural regeneration and wild game interplay.
- Conclusions
Keeping our initial goal in mind, it is fitting to conclude with recommendations for foresters from the perspective of natural regeneration. (1) Adopt uneven-aged silviculture: silver fir-dominated forests, when managed under an uneven-aged system, possess an advanced self-regeneration capacity expressed as a higher probability of regeneration compared to even-aged stands. (2) Balance the tree size diversity: while adding more tree size diversity to these uneven-aged stands helps silver fir progress to more advanced developmental stages, it does not boost the overall density of regeneration, which leads us to an important consideration: the need to balance the ecological benefits of enhanced regeneration quality against the required frequent interventions that are crucial to increasing and maintaining a high level of tree size diversity. (3) Moderate the broadleaf species admixture considering specific growing conditions: including an admixture of broadleaved species into uneven-aged stands in places with a higher aridity index (PL1, PL2, and IT1) results in the reduced density of the natural regeneration of all the species and silver fir specifically. This reduction is statistically significant, although it is negligible in absolute terms. To avoid soil degradation and allelopathy, reaching an equilibrium between the negative and positive effects of the admixture of broadleaf tree species and maintaining a sensible proportion of broadleaf species to favor silver fir regeneration are crucial. This is also important when we consider places with water abundance (GE1), where we noted a significant increase in the natural regeneration density with increasing admixture. (4) Retain large overstory trees: large overstory fir trees did not significantly enhance the density of the fir regeneration. However, given the critical role these trees play in ecosystem functioning and biodiversity conservation, the fact that their presence does not detract from the quality and density of the natural regeneration underlines the importance of retaining them in some places. (5) Use the stand density as a tool to promote desired species: the total BA played a crucial role in boosting the density of all the species, even though it was not significantly influential for silver fir separately. In scenarios where promoting silver fir natural regeneration is a key objective, regulating and carefully selecting the locally appropriate stand density appears to be a suitable tool. Furthermore, considering the importance of the findings, we would like to underline that there is room for further research that incorporates direct measures of light availability and ungulate browsing.
FORTH:
Tree-root control of shallow landslides
Denis Cohen 1 and Massimiliano Schwarz 2,3
1 Department of Earth and Environmental Science, New Mexico Tech, Socorro, NM 87801, USA
2 School of Agricultural, Forest, and Food Sciences, Bern University of Applied Science,
3052 Zollikofen, Switzerland
3 EcorisQ, 1205 Geneva, Switzerland
Correspondence to: Denis Cohen (denis.cohen@gmail.com)
Received: 22 February 2017 – Discussion started: 24 February 2017
Revised: 23 June 2017 – Accepted: 14 July 2017 – Published: 17 August 2017
Abstract. Tree roots have long been recognized to increase slope stability by reinforcing the strength of soils.
Slope stability models usually include the effects of roots by adding an apparent cohesion to the soil to simulate
root strength. No model includes the combined effects of root distribution heterogeneity, stress-strain behavior
of root reinforcement, or root strength in compression. Recent field observations, however, indicate that shallow
landslide triggering mechanisms are characterized by differential deformation that indicates localized activation
of zones in tension, compression, and shear in the soil. Here we describe a new model for slope stability that
specifically considers these effects. The model is a strain-step discrete element model that reproduces the self-
organized redistribution of forces on a slope during rainfall-triggered shallow landslides. We use a conceptual
sigmoidal-shaped hillslope with a clearing in its center to explore the effects of tree size, spacing, weak zones,
maximum root-size diameter, and different root strength configurations. Simulation results indicate that tree
roots can stabilize slopes that would otherwise fail without them and, in general, higher root density with higher
root reinforcement results in a more stable slope. The variation in root stiffness with diameter can, in some cases,
invertthisrelationship.Roottensionprovidesmoreresistancetofailurethanrootcompressionbutrootswithboth
tension and compression offer the best resistance to failure. Lateral (slope-parallel) tension can be important in
cases when the magnitude of this force is comparable to the slope-perpendicular tensile force. In this case, lateral
forces can bring to failure tree-covered areas with high root reinforcement. Slope failure occurs when downslope
soil compression reaches the soil maximum strength. When this occurs depends on the amount of root tension
upslope in both the slope-perpendicular and slope-parallel directions. Roots in tension can prevent failure by
reducing soil compressive forces downslope. When root reinforcement is limited, a crack parallel to the slope
forms near the top of the hillslope. Simulations with roots that fail across this crack always resulted in a landslide.
Slopes that did not form a crack could either fail or remain stable, depending on root reinforcement. Tree spacing
is important for the location of weak zones but tree location on the slope (with respect to where a crack opens) is
as important. Finally, for the specific cases tested here, intermediate-sized roots (5 to 20mm in diameter) appear
to contribute most to root reinforcement. Our results show more complex behaviors than can be obtained with
the traditional slope-uniform, apparent-cohesion approach. A full understanding of the mechanisms of shallow
landslide triggering requires a complete re-evaluation of this traditional approach that cannot predict where and
how forces are mobilized and distributed in roots and soils, and how these control shallow landslides shape, size,
location, and timing.
Published by Copernicus Publications on behalf of the European Geosciences Union.
452 D. Cohen and M. Schwarz: Tree-root control
1 Introduction
Shallow landslides are hillslope processes that play a key
role in shaping landscapes in forested catchments (Istanbul-
luogluandBras,2005;SidleandOchiai,2006).Manystudies
have highlighted the importance of roots and their mechani-
cal properties for the stabilization of hillslopes (e.g., Schwarz
et al., 2015), but usually only basal root reinforcement is
considered. When considering how roots reinforce soil, how-
ever, three different mechanisms of root reinforcement must
be recognized.
- Basal root reinforcement acting on the basal shear sur-
face of the landslide. This is the most efficient mech-
anism, if present. In many cases, however, this mecha-
nism is absent because the position of the failure surface
is deeper than the rooting zone.
- Lateral root reinforcement acting on lateral surfaces
of the landslide. The magnitude of the contribution of
this mechanism depends on the type of deformation of
the landslide mass. If the landslide behaves as a rigid
mass, lateral reinforcement may act almost simultane-
ously along all the edges of the sliding mass (in tension,
shear, and compression). In cases where there is differ-
ential deformation of the soil mass, this leads to the pro-
gressive activation of lateral reinforcement, first in ten-
sion at the top of the landslide, and then in compression
at the toe at the end of the triggering. The magnitude of
lateral root reinforcement depends on the spatial distri-
bution of the root network.
- Roots stiffening the soil mass. The presence of roots in
the soil increases the macroscopic stiffness of the rooted
soil mass, leading to a larger redistribution of forces at
the scale of the hillslope through small deformations.
This mechanism increases the effects of the previous
two (basal and lateral root reinforcements).
On top of these considerations on root reinforcement mech-
anisms acting on a single landslide, it is important to empha-
size that those mechanisms assume different meaning when
considering the more global context of landslide processes
at the catchment scale. Specifically, the effects of root rein-
forcement on landslide processes are considered limited by
the following:
i. The magnitude of root reinforcement (a function of for-
est structure and tree species composition). Root rein-
forcement needs to reach values of the order of a few
kilopascal in order to be significant (Schwarz et al.,
2016).
ii. The heterogeneity of root distribution (tree species, to-
pography, local soil condition, etc.). Root reinforcement
must be active in specific places and at specific times to
have any effect on slope stability: mean values of appar-
ent cohesion across the entire hillslope are not represen-
tative and not sufficient for considering the specifics of
actual root reinforcement effects.
iii. The depth of the landslide shear surface (effects of basal
root reinforcement). The deeper the shear surface is, the
less important the effect of basal root reinforcement is.
iv. The length and volume of the landslide (lateral root
reinforcement and buttressing/arching mechanisms and
stiffening effects). The larger the landslide is, the lower
are the effects of lateral root reinforcement. In order to
characterize the efficacy of roots for slope stabilization,
a spatiotemporal quantification of root reinforcement is
needed.
In view of the importance of root reinforcement and of shal-
low landslides to landscape evolution and to human societies,
mechanistic models that include the processes linked to the
triggering of shallow landslide and the influence of root re-
inforcement on it are needed. In the large majority of cases,
slope stability models add apparent cohesion to the soil to
simulate root reinforcement (e.g., Milledge et al., 2014; Bel-
lugi et al., 2015; Hwang et al., 2015). Few models include the
effectsofrootdistributionheterogeneity(Stokesetal.,2014),
andnoneconsiderthestress-strainbehaviorofrootreinforce-
ment and the strength of roots in compression. Recent field
observations show that shallow landslide triggering mecha-
nisms are characterized by differential deformation that indi-
cates localized loading of soils in tension, compression, and
shear (Schwarz et al., 2012a). These observations contradict
common assumptions used in models until now, yet the direct
coupling of these different root reinforcement mechanisms,
and their mobilization during the triggering of shallow land-
slides, has not yet been made.
Here we present a new model for shallow slope stability
calculations that specifically considers these important ef-
fects. To fully understand the mechanisms of shallow land-
slide triggering, a complete re-evaluation of the traditional
apparent cohesion approach is required. To do so, it is im-
portant to consider the forces held by roots in a way that
is entirely different than done thus far. Moreover, measure-
ments and models indicate that the assumptions of constant
elasticityandhomogeneousrootproperties,asappliedintyp-
ical finite element geotechnical model, cannot reproduce the
mechanisms leading to the triggering of forested slope fail-
ures (Schwarz et al., 2013).
The SOSlope (for Self-Organized Slope) model presented
here fills this gap by developing a mechanistic model for pre-
dicting shallow landslide sizes across landscapes, consider-
ing the effects of root reinforcement in a detailed quantita-
tive manner (spatiotemporal heterogeneity of root reinforce-
ment). The SOSlope model allows for exploring the activa-
tion of root reinforcement during the triggering process and
helps to shed light on the contribution of roots to the slope
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 453
stability. The SOSlope model is used in this work to test the
following main hypotheses:
– Both tensional and compressional forces resulting from
mobilization of forces in the roots and the soil are effi-
cient in stabilizing slopes but have higher effectiveness
when occurring simultaneously.
– Weak zones in the root network (Schwarz et al., 2010b,
2012a) determine the effectiveness of root reinforce-
ment at the slope scale if no basal reinforcement is
present.
– Coarse roots dominate reinforcement and its efficacy,
when present.
In what follows we first present a general background on
the importance of vegetation for geomorphic processes in the
context of hillslopes and landslides (Sect. 2). We then de-
scribe the SOSlope model in detail (Sect. 3), present the data
set for roots and soil used in simulations (Sect. 4), show and
discuss results (Sect. 5), and synthesize a typical force redis-
tribution process during landslide triggering (Sect. 6). Con-
clusions are given in Sect. 7.
2 Background and motivation
Understanding the role of shallow landslides in the geomor-
phic evolution of landscapes is of prime importance and mo-
tivates the present work. In some regions, shallow landslides
are the dominant regulating mechanisms by which soil is de-
livered from the hillslope to steep channels or fluvial sys-
tems (Jakob et al., 2005). The magnitude and intensity of
these phenomena also has important societal impacts both
in the long (landscape evolution and soil resource availabil-
ity Istanbulluoglu and Bras, 2005; Montgomery, 2007) and
short term (risks due to landslides, debris flows and sediment
transport, water quality, soil productivity; Wehrli et al., 2007;
Hamilton, 2008).
On long timescales, shallow landslides are important geo-
morphic processes shaping landscapes of both vegetated and
non-vegetated basins. For vegetated basins, the spatiotempo-
ral distribution of root reinforcement has a major impact on
the dynamic of sediment transport at the catchment scale (Si-
dle and Ochiai, 2006) and on the availability of productive
soil, a key resource for human needs. At the hillslope scale,
the presence of vegetation generally increases soil thickness,
lowering the frequency of landsliding events but increasing
their magnitudes (Amundson et al., 2015). At the catchment
scale, vegetation causes slopes to steepen and sediment mo-
bilization is then often dominated by deep landslides driven
by fluvial incision (Larsen and Montgomery, 2012). The in-
fluence of shallow landslides on shaping the landscape on
long timescales is, in part, masked by continuously changing
factors influenced by human activities, climate change, and
other disturbances such as storms and fires. Under these con-
stant disturbances soils never reach an equilibrium state that
would otherwise require between 10 and 1000 years (Blume
et al., 2010; Bebi et al., 2017). Nevertheless, the presence of
soils on steep slopes is a necessary condition for preserving
important functions of mountain environments, such as water
supply, nutrient production, biodiversity, landscape aesthet-
ics, and cultural heritage.
While soil as a resource is gaining increasing attention in
the context of global sustainable development (Nature Edi-
torial, 2015), risks related to shallow landslides and to pro-
cesses linked to them (debris flows, bedload transport, large
wood transport during floods) as well as the availability of
quality water are issues that impact human societies in the
short term (Miura et al., 2015), particularly in mountainous
regions. Water quality is linked to shallow landslides because
sediments mobilized by landslides are transported as sus-
pended sediments in streams.
While sustainable resource management in forestry and in
agriculture aims to keep the frequency of shallow landslide
events to pseudo-equilibrium conditions at the catchment
scale and to reduce the overall erosion rate (Li et al., 2016),
disturbances such as those due to human activities may lead
to a rapid and dramatic increase in shallow landslide fre-
quency and magnitude. For instance, deforestation and inten-
sive agriculture may lead to an increase in the overall erosion
rate by 1 order of magnitude. Marden (2012) reports that in
the 17km 2 catchment of Waipaoa (New Zealand), erosion
rate increased from 2.7 to 15Mtyear −1 after deforestation
and conversion of slopes to pasture land. In this new environ-
ment, shallow landslides contribute ∼60% of the sediment
yield of the Waipaoa river during floods and 10 to 20% of to-
tal erosion. Similar conditions occurred in the European Alps
until the first half of the 20th century, which led to a con-
siderable increase in erosion rates (Mariotta, 2004). Meus-
burger and Alewell (2008) reported that, in a catchment in
the centralAlps, theincrease inlandslide areaby 92% within
45 years was likely due to dynamic factors like climate and
land-use changes and had a decisive influence on landslide
patterns observed today.
Risks due to shallow landslides are associated with dif-
ferent types of phenomena ranging from hillslope debris
flows (example of process causing a direct risk to infrastruc-
tures and individuals) to various channel processes such as
large sediment transport during floods, wood debris trans-
port, channelized debris flows, etc. (examples of processes
causing an indirect risk to infrastructures and individuals). It
is estimated that landslides triggered by heavy rainfall cause
damages upwards of several billions each year and more than
600 fatalities per year (Sidle and Ochiai, 2006).
Nexttotheconstellationoffactorswellknowntoinfluence
the triggering of shallow landslides, vegetation has been rec-
ognized to play an important role (Sidle and Ochiai, 2006;
Schwarz et al., 2010c; McGuire et al., 2016) and its func-
tion is considered an important component of ecosystem ser-
vices provided in mountain regions. The importance of the
effects of vegetation is, in some cases, recognized at a po-
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
454 D. Cohen and M. Schwarz: Tree-root control
litical level. For instance, the global forest area managed for
protection of soil and water is 25% of all global forested ar-
eas (Miura et al., 2015). In Switzerland, protection forests
occupy more than 50% of all forested areas (Wehrli et al.,
2007). Moreover, bio-engineering measures are often con-
sidered an important part of integrated risk management and
disaster risk reduction strategies. The management of such
protection forests and bio-engineering measures needs quan-
titative tools to optimize the effectiveness of such important
ecosystem services for society. The formulation of such tools
needs to be based on quantitative methods applicable to a
large range of situations. Moreover, these methods need to
consider different time and spatial scales at which vegetation
influences processes. To put the motivation for the present
work in the appropriate context, we briefly summarize the
effects of vegetation on long and short term geomorphic pro-
cesses.
In the long term, the presence of vegetation (i) increases
soil production rates through mechanical and chemical pro-
cesses (Wilkinson et al., 2005; Phillips et al., 2008) (100–
1000 years); (ii) increases soil residence time on hillslopes
due to root reinforcement and protects against runoff ero-
sion (Istanbulluoglu and Bras, 2005) (10–100 years; note that
in the case of natural or human driven disturbances, the re-
sponse time of the system (i.e., root decay) is of the order
of a few years (Vergani et al., 2016)); and (iii) enhances soil
diffusion rates on hillslopes due to tree wind throw (Paw-
lik, 2013; Roering et al., 2010), root mounds (Hoffman and
Anderson, 2014), and biological activity (Gabet and Mudd,
- (100–1000 years).
In the short term, vegetation mainly influences root rein-
forcement and regulates water fluxes. At the hillslope scale,
the hydrological effects of vegetation are assumed to play
a small role on slope stability compared to the contribution
of root reinforcement (Sidle and Bogaard, 2016; Sidle and
Ziegler, 2017). At the catchment scale, however, the regula-
tion of water fluxes may have important implications for the
stability of those slopes that drain large areas, particularly for
short and intense rainfall events.
Root are considered the hidden half of plants due to the
difficulties in characterizing and quantifying their distribu-
tion and mechanical properties. In slope stability, the process
of root reinforcement remains hidden because direct obser-
vations have not yet been made on steep hillslopes. Field and
laboratory experiments (e.g., Zhou et al., 1998; Ekanayake
and Phillips, 1999; Roering et al., 2003; Docker and Hubble,
- generally explore only a small part of the complex root
reinforcement mechanisms.
Methods for the quantification of different types of root
reinforcement mechanisms have been through a succession
of models in the last few decades, starting with the assump-
tion of the simultaneous breakage of all roots (Wu et al.,
1979; Waldron and Dakessian, 1981) to the application of
fiber bundle models that consider the progressive failures of
roots (Pollen and Simon, 2005; Schwarz et al., 2010a; Co-
hen et al., 2011). Fiber bundle models may be differentiated
on the basis of the type of loading, whether it is by stress
(Pollen and Simon, 2005) which does not allow for the cal-
culation of displacement, or by strain (Schwarz et al., 2013;
Cohen et al., 2011), which does. We enumerate below some
aspects of root reinforcement models important for slope sta-
bility.
- Breakage versus slip-out. Field observations show that
in tree-root bundles, the dominant failure mechanism of
roots is by breakage (Schwarz et al., 2012a). Slippage
is limited to small roots that usually contribute only a
small fraction of the total root reinforcement. For this
reason, numerical models usually assume that all roots
fail by breaking (Schwarz et al., 2013; Cohen et al.,
2011).
- Thecontributionofrootreinforcementmustbedifferen-
tiated between different types of stress conditions: ten-
sion, compression, and shearing. While most of the lit-
erature has focused on the shear behavior of rooted soils
(e.g., Docker and Hubble, 2008), some works have in-
vestigated the contribution of root reinforcement under
tension (Zhou et al., 1998; Schwarz et al., 2010a, 2011)
and compression (Schwarz et al., 2015). In general the
contribution of maximum root reinforcement under ten-
sion and shearing is of the same order of magnitude,
whereas under compression the contribution of roots is
about1orderofmagnitudesmaller.However,rootscon-
tribute significantly to increase the stiffness of soil un-
der compression. This may play an important role in the
re-distribution of forces during the triggering of a shal-
low landslide (Schwarz et al., 2015).
- The mechanical interactions of neighboring roots in a
bundle are usually neglected. Giadrossich et al. (2013)
showed with laboratory experiments that the failure
mechanisms of single roots are influenced by neigh-
boring roots only at high root density that are usually
reached only near tree stems (0–0.5m).
- The mechanical and geometrical variability in roots was
recently considered using survival functions (Schwarz
et al., 2013) that represent the complexity of several fac-
tors contributing to the variable stress-strain behavior
of roots. Specifically, these factors are root tortuosity
(Schwarz et al., 2010a), root–soil mechanical interac-
tions (Schwarz et al., 2011), and position of root break-
age along the root. Pulled roots break at different dis-
tances from the point of force application because of
branching, root geometry, changes in root diameter due
to soil properties, presence of stones, etc.
- The spatial and temporal heterogeneity of root rein-
forcement is related to several factors such as topog-
raphy, soil water content, soil disturbances, resistance
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 455
and resilience of forest cover to disturbances, and an-
imal browsing (Schwarz et al., 2012a; Vergani et al.,
2016).
3 The SOSlope model
3.1 General framework
SOSlope is a hydro-mechanical model of slope stability that
computes the factor of safety on a hillslope discretized into a
two-dimensional array of blocks connected by bonds. Bonds
between adjacent blocks represent mechanical forces acting
across the blocks due to roots and soil (Cohen et al., 2009).
These forces can either be tensile or compressive depending
on the relative displacements of the blocks. A digital eleva-
tion model (DEM) is used to divide the hillslope into squares
in plan view, where the centers of the squares are points of
the DEM (Fig. 1). Three-dimensional blocks are created by
extruding the squares to the bottom of the soil layer along
the vertical. The center of mass of a block is connected to
the four lateral blocks by four force bonds (Fig. 1). Initially,
bond forces between blocks are set to zero. Rainfall onto
the slope will increase the mass and decrease the soil shear
strength of the blocks. At each time step, the factor of safety
is calculated for each block using a force balance (resistive
force over active force; see equations below). If the factor of
safety of one or more blocks is less than one, those blocks are
moved in the direction of the local active force (defined be-
low) by a predefined amount (usually 0.1mm) and the factor
of safety is recalculated for all blocks. Because of the relative
motion between blocks that have moved and blocks that re-
main stationary, mechanical bond forces between blocks are
no longer zero and the force balance changes. This relative
motion triggers instantaneous force redistributions across the
entire hillslope similar to a self-organized critical (SOC) sys-
tem of which the spring-block model (Bak et al., 1988; Her-
garten and Neugebauer, 1998; Cohen et al., 2009) is a sub-
set. Looping over blocks and moving those that are unstable
is repeated until all blocks are either stable (factor of safety
greater than or equal to 1) and the system reaches a new equi-
librium or some blocks have failed (their displacements are
greater than some set value, usually a few meters), triggering
a landslide.
3.2 Factor of safety
The factor of safety for each block is calculated as the ratio
of resistive to active forces. Resistive forces include the soil
basal shear strength and the strength of roots that cross the
basal slip surface, assumed to be located at the bottom of the
soil layer. The active forces include the gravitational driving
force due to the soil mass and the push or pull forces between
blocks that include the effects of soil and root tension and
compression. These later forces are the bond forces between
the blocks described above. Including all these forces in a
β
h
D
F 1
F 2
F 3
F 4
(b)
(a)
Figure 1. (a) Plan view of discretized cell with its four neighbors
showing bond forces. (b) Vertical section across neighboring cells
showing the center of mass of cells and the location of the connect-
ing bond. β is the surface slope and h and D are the thicknesses of
soil down to the basal surface, measured vertically and perpendicu-
lar to the surface, respectively.
force balance yields the factor of safety
FOS =
F s +F r
?
? F
d +
4
?
j=1
F j
?
?
, (1)
where F s is the soil basal resistive force that includes soil
cohesionandfriction,F r isthebasalrootresistance,F d isthe
driving force vector due to gravity, and F j , j = 1,…,4, are
the four bond vector forces that quantify soil and root tension
orcompressionbetweentheblockanditsfourneighbors.The
vertical bars in the denominator denote the norm of a vector.
This factor of safety is calculated for each block but an index
for the block number is not included so as not to clutter the
equations.
Soil basal resistance is
F s = Aτ b , (2)
where A is the surface area of the block along the failure sur-
face and τ b is the basal shear stress (described below). In the
present model, we set F r = 0, focusing on lateral root rein-
forcement. This is justified in many cases where the depth
of the slip surface is 1m or greater and very few roots are
present (e.g., Bischetti et al., 2005; Tron et al., 2014). Basal
root reinforcement can easily be added using a formulation
similar to lateral root reinforcement (discussed below) with
values of root reinforcement a function of the shear displace-
ment and the density of roots crossing the slip surface.
The driving force is
F d = γDA ˆ t, (3)
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
456 D. Cohen and M. Schwarz: Tree-root control
where γ is the specific weight of the wet soil, D is the depth
to the shearing surface, perpendicular to slope, and ˆ t is the
unit tangent to the slope in the direction of the maximum
slope. The specific weight of the wet soil is calculated based
on water content and solid fraction, i.e.,
γ =
? ρ
s φ s +ρ w θ
? g,
(4)
where ρ s and ρ w are the solid (grain) and water densities,
respectively, φ s is the solid volumetric fraction, θ the volu-
metric water content, and g is gravity.
Bond forces are given by
F j =
?
F soil
j
+F roots
j
?
ˆ
b j , j = 1,…,4, (5)
where F soil
j
and F roots
j
are the soil and root components of the
four bond forces, respectively, and
ˆ
b j are unit vectors along
the bondaxes pointingoutward ofthe block.These quantities
are detailed below.
3.3 Bond forces due to roots
The force in bond j between a block and its neighbor due
to roots (F root
j
) depends on four factors: the root density and
the root-diameter distribution at the bond center; the strength
of roots, which depends on root diameter; and the change
in length (elongation) of the bond with respect to its initial
length. Changes in root density with depth (e.g., Bischetti
et al., 2005) are not taken into account. This force is com-
puted using the Root Bundle Model (RBM) of Schwarz et al.
(2013) with Weibull statistics, called RBMw. For the sake of
completeness, the full details of the model are given below.
3.3.1 Root density and root-diameter distribution
Roots are binned according to their diameters in 1mm size
bins from 0.5mm to an upper limit given by data. A bin is
usually referred to as a root-diameter class, with φ i denoting
the mean root diameter of class i, i = 1,…,i max . At each of
the four faces of a block, the total number of roots for each
root-diameter class i that crosses a face j is the sum of the
number of roots for that root-diameter class from each sur-
rounding tree in the stand. Summing roots from each tree
implies no competition for resources. Following the empiri-
cal model of Schwarz et al. (2010a) in its version described
by Giadrossich et al. (2016), the number of roots depends
on the distance of the face center to the tree trunks, the tree
trunks diameters, and the tree species. For simplicity all trees
in the stand are assumed to belong to the same species. The
modelassumesalinearallometricrelationbetweentrunksize
and root density, a power-law decay of root density with dis-
tance from the tree trunk, and a logarithmic decrease in root
density with root-diameter size. The number of roots of class
diameter φ i crossing face j is
N j
φ i
= A j
T
?
k=1
ρ j
k
?
1−
ln ? 1+min ? φ i ,φ max
k
? /φ
o
?
ln ? 1+φ max
k
/φ o ?
? ?
φ i
φ o
? γ
, (6)
where A j is the surface area of face j, T is the number
of trees in the stand (more specifically the number of trees
whose roots reach face j of the cell), and ρ j
k
is the density of
fine roots of tree k for face j. This later quantity is given by
ρ j
k
=
N k
d max
k
2πd j
k
?
max(0,d max
k
−d j
k )
d max
k
?
, (7)
where N k , the total number of fine roots of tree k, is
N k = µπ
? φ
k
2
? 2
, (8)
d max
k
, the maximum rooting distance for tree k, is
d max
k
= ψ φ k , (9)
and φ max
k
, the maximum root diameter class of tree k, is
φ max
k
= max
?
0,
d max
k
−d j
k
η
?
. (10)
In these equations, φ o = 1 mm is the size of the smallest root
diameter class, d j
k
is the distance between face j and tree k,
and φ k is the tree diameter (usually diameter at breast height
or simply DBH). This model contains four fitting parame-
ters (µ, η, ψ, and γ) that must be determined from data (Gi-
adrossich et al., 2016; Schwarz et al., 2016).
3.3.2 Root mechanical forces
Roots are assumed elastic in both tension (Schwarz et al.,
- and compression (Schwarz et al., 2015). The linear
elastic force in a root is expressed using a spring constant
(i.e., Hooke’s law) that depends on the root diameter class.
For a root in diameter class i on bond j, that elastic force is
F E
i,j (φ i ,x j ) = k
E
i
x j , (11)
where the superscript E indicates either tension (E = T) or
compression (E = C) and x j is the elongation of the bond
from its initial length (positive for tension, negative for com-
pression). Based on data (e.g., Schwarz et al., 2013, 2015)
we assume the spring constant depends linearly on root di-
ameter, i.e.,
k E
i
= k E
0
+k E
1
φ i , (12)
with k E
0
and k E
1
constants to be determined from data. Other
formulations based on a power-law relation can also be used
(Giadrossich et al., 2016).
The variability in root bio-mechanical properties (e.g.,
maximum tensile or compressive strength, elastic moduli in
tension or compression) due to the presence of biological
or geometrical weak spots is handled probabilistically. The
probability of failure of a root in tension (or in compres-
sion) is captured by multiplying the elastic force by a Weibull
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 457
survival function (S) that depends on a dimensionless bond
elongation. Then, the total root-bond force is obtained by
summing over all roots of each diameter class, i.e.,
F roots
j
(x j ) =
i max
?
i=1
N j
φ i F
E
i,j (φ i ,x j )S
E
i,j (ξ i,j ),
(13)
where N j
φ i
is given by Eq. (6), F E
i,j
by Eq. (11),
S E
i,j (ξ i,j ) = exp
?
−
? ξ
i,j
λ E
? ω E ?
, (14)
and
ξ i,j =
k E
i
x j
F E
i,max
, (15)
where λ E and ω E (E = T or C) are two scale and two shape
parameters to be determined from field or laboratory exper-
iments (see Schwarz et al., 2013, 2015, for details). F E
i,max
is the maximum force held in a root at breakage (in ten-
sion) or at the critical buckling condition (in compression;
see Schwarz et al., 2015) for a root of diameter φ i and is
given by the commonly used power-law equation
F E
i,max = F
E
o
?
φ i
φ o
? α E
, (16)
with α E the power-law exponent and F E
o
a pre-exponential
factor for tension or compression (E = T or C). The scal-
ing of the displacement with the maximum strength of a root
eliminates the effect of root diameter on maximum displace-
ment. Similarly, the parameter λ E scales the root strength
variability to the root diameter. Equation (13) has a maxi-
mum (F roots
j,max ) called the maximum root reinforcement and
occurs at a bond elongation x j,max .
3.4 Bond forces due to soil
The soil bond force (F soil
j
, Eq. 5) depends on whether the soil
is in tension or in compression. For tension, we assume that
resistance scales with soil apparent cohesion (including the
effects of suction stress for unsaturated soils) as a function of
displacement using a logarithmic function (Win, 2006):
F soil, T
j
=
c a W D
?
1−
log ? 1+ε j L j
?
log ? 1+ε T
max L j
?
?
, ε j < ε T
max ,
0, ε j ≥ ε T
max ,
(17)
where c a is the apparent cohesion, ε T
max
is a strain thresh-
old above which soil loses any tensional resistance, and L j
is the length of bond j. In compression, following the work
of Schwarz et al. (2015) we assume that the soil compres-
sional resistance is mobilized across the shear plane that
forms during the failure of a downslope wedge, similar to
the earth pressure force in the geotechnical engineering lit-
erature that develops during the passive state when a retain-
ing wall moves downslope toward the adjacent backfill (e.g.,
Milledge et al., 2014). According to Schwarz et al. (2015),
the mobilized force on the downslope wedge scales with the
maximum passive earth pressure force F p and with the dis-
placement, i.e.,
F soil, C
j
(x j ) = −F p W P w 1 (x j )S w 2 (x j ), (18)
where
F p = K pγ ρg
D 2
2
+K pc c ? D, (19)
and K pγ and K pc are the passive earth pressure coeffi-
cients due to soil weight and to cohesion, respectively, ob-
tained from a fitting of equations given in Soubra and Macuh
(2002); c ? is effective soil cohesion; and P w 1 and S w 2 are the
Weibull cumulative density and the Weibull survival func-
tions, respectively, given by
P w 1 (x j ) = 1−exp
?
−
?
x j
µ 1
? κ 1 ?
(20)
and
S w 2 (x j ) = exp
?
−
?
x j
µ 2
? κ 2 ?
, (21)
with µ 1 , κ 1 , µ 2 , and κ 2 four parameters determined from
compression experiments. The first Weibull function, P w 1 ,
serves to scale the maximum passive earth pressure force
with displacement during initial block motion, while the sec-
ond one, S w 2 , reduces that same force as the wedge is over-
ridden by the block and the failure surface area of the slip
plane decreases (see Schwarz et al., 2015, for details). We
neglect the active earth pressure force on upstream faces of
cells because the magnitude of the active force is small in
comparison to other forces.
3.5 Hydrological triggering
Rainfall-triggered shallow landslides can fail under saturated
conditions during increases of pore-water pressure and/or
loss of suction under unsaturated conditions (Lu and Godt,
2013). Our objective here is not to reproduce the detailed
physical mechanisms by which changes in subsurface hy-
drology trigger a landslide but to develop a simple empiri-
cal model that realistically mimics observed changes in pore-
water pressure and water content during rainfall infiltration.
Althoughdiverse hydrologictriggers havebeen observedand
described(e.g.,Reidetal.,1997;Iverson,2000),hereweuse,
as a representative example for the hydrological conditions
triggering a shallow landslide in our model, pore-pressure
measurements during the artificial triggering of the Rüdlin-
gen shallow landslide experiment in Switzerland (Askarine-
jad et al., 2012; Lehmann et al., 2013). Data from Lehmann
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
458 D. Cohen and M. Schwarz: Tree-root control
et al. (2013) indicate that high pore-water pressures were
attained relatively quickly and remained steady across the
slope long before failure occurred, and that the decrease in
the standard deviation of the water saturation prior to failure
indicated an increase in the connectivity of water-saturated
regions that reduced soil shear strength across the full length
of the slip surface leading to failure. Other data in different
localities (e.g., Matsushi et al., 2006; Bordoni et al., 2015)
have also shown high, steady pore-water pressure prior to
failure. Because our model focuses on the effects of roots
and soil strength on slope stability rather than on the details
of hydrologic triggering, we choose a simplified, empirical,
dual-porosity model for our slope hydrology. Our objective
is only to reproduce reasonable pore-water pressure distri-
bution and water content evolution in both the matrix and
the preferential flow domains, but not to model the physics
of evolving subsurface hydrology. The model embodies the
rapid increase in positive pore pressure in a preferential flow
domain (representing macropores) and the slow decrease in
suction in the soil matrix caused by slow water transfer from
the macropores to the matrix. This decrease in suction is the
equivalent of the increasing connections of water-saturated
regions represented by the decrease in the standard deviation
of water saturation observed by Lehmann et al. (2013) that
eventually caused slope failure in the Rüdlingen experiment.
We assume that water flow in soils during a rainfall event
is a combination of slow matrix flow (also called immobile
water with capillary number lower than 1) and fast prefer-
ential flow (mobile water, capillary number higher than 1)
(Sidle and Ochiai, 2006; Beven and Germann, 2013). While
slow matrix flow influences the change in suction stress,
the fast preferential flow directly influences pore-water pres-
sure in the macropores. Our formulation of this concept
is empirical and is a simplification of the more common
dual-porosity models that employ two flow equations (e.g.,
Richards’ equation) that exchange moisture between the two
domains, and mixture equations for water content, hydraulic
conductivity, rainfall partitioning based on the volumetric ra-
tio of the fast and slow flow domains (e.g., Gerke and van
Genuchten, 1993; Shao et al., 2015). In accord with con-
tinuum mixture theory for effective stress (e.g., Borja and
Koliji, 2009), we write the mean pore-water pressure of the
soil (matrix+macropores), p, as
p = ψ 1 p 1 +ψ 2 p 2 , (22)
where ψ 1 and ψ 2 are the pore fractions along the potential
failure surface of the landslide of the matrix and the macro-
pores, respectively (volume of pore in matrix or macropores
over total pore volume, with indices 1 for matrix and 2 for
macropores)withψ 1 +ψ 2 = 1,andwherep i ,i = 1,2arethe
matrix and macropores intrinsic mean pore pressures. Pore
fractions ψ 1 and ψ 2 s are related to the partial porosities of
the matrix and the macropores, φ 1 and φ 2 , respectively, by
ψ i =
φ i
n
, i = 1,2, (23)
where φ 1 +φ 2 = n, with n being the total porosity of the soil.
The solid volume fraction of the matrix (macropores have
only pore space) is φ s = 1−φ 1 −φ 2 = 1−n. The superscripts
and subscripts in these equations and in equations below re-
fer to partial and intrinsic quantities, respectively. Partial and
intrinsic water content of the matrix and macropores are re-
lated as follows:
θ 1 =
?
φ s +φ 1
?
θ 1 , (24)
θ 2 = φ 2 θ 2 , (25)
where θ 1 and θ 2 are the partial water contents of phase 1 and
2 (volumetric water content of phase 1 or 2 over total soil
volume) and θ 1 and θ 2 are the intrinsic water contents of each
phases (volumetric water content of phase i over volume of
phase i, i = 1,2). At saturation θ 2 = φ 2 since the macropore
phase contains only void space and thus θ 2 = 1. The total
water content of the soil is
θ = θ 1 +θ 2 , (26)
and is used in Eq. (4) to compute the soil-specific weight.
Equations similar to Eq. (26) can be written for saturated and
residual water contents.
We assume that the time evolution of the intrinsic pore-
water pressure in the macropores, p 2 , and of the partial water
contentinboththemacropore(θ 2 )andthematrixphases(θ 1 )
can be modeled using cumulative distribution functions. For
the macropore phase, we write
p 2 (t) = p max F
? t ∗ ,µ
p ,σ p
? ,
(27)
and
θ 2 (t) = φ 2
? θ r
2 +(1−θ
r
2 )F
? t ∗ ,µ
p ,σ p
?? ,
(28)
where p max is a constant here but ultimately depends on
rainfall infiltration rate and upstream contributing area (e.g.,
Montgomery and Dietrich, 1994), t ∗ is a dimensionless time,
F is the normal cumulative distribution function with mean
µ p and standard deviation σ p , and θ r
2
is the intrinsic residual
water content for the macropores (we have used the fact that
the intrinsic saturated water content θ s
2
= 1 since macropores
have no solid fraction). For the water content in the matrix
we assume that
θ 1 (t) =
?
θ o −φ 2
?
+(θ s −θ o ) F fold
? t ∗ ,µ
θ ,σ θ
? ,
(29)
where θ o and θ s are the soil initial and saturated water con-
tents, respectively, and F fold is the folded normal cumula-
tive distribution with mean µ θ and standard deviation σ θ .
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 459
The pore-water pressure in the matrix is given by (Borja and
Koliji, 2009)
p 1 (t) = S 1
e p 1 ,
(30)
where p 1 is the intrinsic pore-water pressure in the matrix
and S e is the equivalent degree of saturation (also called ef-
fective saturation) in the matrix. Following Lu et al. (2010),
we have used the equivalent degree of saturation (S e ) in
Eq. (30) instead of the more commonly used degree of satu-
ration. Under unsaturated conditions, p 1 is a matrix suction
stress (Lu et al., 2010). The equivalent degree of saturation
in the matrix is defined as
S 1
e
=
θ 1 −θ r
1
θ s
1 −θ
r
1
, (31)
where θ r
1
and θ s
1
are the intrinsic residual and saturated water
content of the matrix phase with θ s
1
= φ 1 . Using Eqs. (24)
and (25), and equations for the residual and saturated water
content equivalent to Eq. (26), Eq. (31) can be rewritten as
S 1
e
=
θ 1 −θ r
θ s −φ 2 −θ r
, (32)
where θ 1 is given by Eq. (29). Using van Genuchten formu-
lation (Van Genuchten, 1980), we can write the suction stress
as (Lu et al., 2010)
p 1 (t) = −
S 1
e
α vg
? ?
S 1
e
?
n vg
1−n vg
−1
? 1
n vg
, (33)
where and α vg and n vg are the soil parameters.
Pore-water pressure in the macropores (Eq. 27), matrix
water content (Eq. 29), matrix suction (Eq. 33), and mean
pore-water pressure (Eq. 22) are computed at each block of
the domain at each time step. The dimensionless time t ∗ in
these equations is time scaled with the characteristic time
for reaching steady state (t ∗ = t/t ss ). Figure 2 illustrates the
model behavior for parameters shown in Table 1. The stan-
dard deviations are chosen so that macropore water pressure
reaches its maximum before matrix water content, to mimic,
but not reproduce, the behavior observed by Lehmann et al.
(2013).
3.6 Basal shear stress
Basal shear resistance along the slip surface is calculated us-
ing the Mohr–Coulomb failure criterion including contribu-
tions from both the suction stress and the pore-water pressure
using the mean pore-water pressure p of Eq. (22), i.e.,
τ b = c ? +
? σ
n −p
? tanφ,
(34)
where σ n is the normal stress and φ is the soil friction angle.
Water pressure (kPa)
-1 -1
0 0
1 1
2 2
3 3
4 4
5 5
Matrix suction
Macropore pressure
Mean pressure
Time (min)
Total water content
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0.2
0.3
0.4
0.5
Water pressure (kPa)
0 0
Matrix suction
Macropore pressure
Mean pressure
Figure 2. Time evolution of pore-water pressures and water content
for the dual-porosity model.
Table 1. Hydrological parameters used in all simulations.
Variable Value
t ss 720min
µ p 0.5
σ p 0.125
µ θ 0.0
σ θ 0.6
4 Data
4.1 Soil
Mechanical soil parameters from Schwarz et al. (2013, 2015)
andotherparametersusedinsimulationsarelistedinTable2.
Figure 3 shows the soil strength in tension and compression
(positive and negative values of displacement, respectively)
for different soil thicknesses.
4.2 Roots
Model parameters for roots (Table 3) are taken from field
and laboratory data of Schwarz et al. (2010a, 2012b, 2013,
- for Picea abies (Norway spruce). Figure 4 shows root
reinforcement as a function of bond elongation (both in ten-
sion and compression) for fourvalues of tree diameter (DBH,
diameter at breast height) and for three distances (d) from
the tree trunk (0.5, 1.5, and at 2.5m). The maximum root
reinforcement in tension occurs within the first 5cm of dis-
placement in both tension and compression. The magnitude
is about 5 times higher in tension than in compression and
depends strongly on the size of the tree. Small trees (i.e.,
DBH = 0.1m) provide negligible reinforcement at all dis-
placements. For large trees (DBH > 0.3m) lateral root rein-
forcement upwards of tens of kilopascal is typical (Schwarz
et al., 2012b). In tension, root reinforcement becomes neg-
ligible once the bond has stretched over 0.1m, regardless of
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
460 D. Cohen and M. Schwarz: Tree-root control
Table 2. Soil parameters used in all simulations.
Variable Value Units
ρ s 1700 kgm −3
ρ w 1000 kgm −3
c ? 500 Pa
φ 31
◦
D 1 m
ε T
max
0.003
µ 1 0.58 m
κ 1 0.07
µ 2 2.00 m
κ 2 0.25
θ s = n 0.46
θ r 0.082
θ o 0.26
ψ 1 0.4
ψ 2 0.6
p max 3800 Pa
n vg 3.3
α vg 0.00086 Pa −1
Displacement (m)
Soil strength (kPa)
-0.6 -0.4 -0.2 0 0.2
-25
-20
-15
-10
-5
0
0.5 m
1 m
1.25 m
1.5 m
2 m
Figure 3. Soil strength as a function of displacement for different
soil depths. Values of passive earth pressure coefficients for esti-
mating soil compressional strength are calculated using a surface
slope of 40 ◦ . Other parameters needed for the calculation are given
in Table 2. Negative values of displacement indicate compression.
the distance from the tree trunk. In compression, the bond
elongation over which reinforcement is active depends on the
distance from tree and range from 0.15 m close to the tree
trunk to about 0.05m at 2.5m distance from the tree trunk.
5 Results and discussion
To illustrate the capabilities of SOSlope to reproduce the
triggering of shallow landslides influenced by the presence
of tree roots, we first present simulations of a 70m×70m
conceptual sigmoidal forested hillslope with a 20m×50m
Table 3. Root parameters used in simulations.
Variable Value Units
µ 72453 No. rootsm −2
η 243
ψ 18.5
γ −1.30
k T
0
356 Nm −1
k T
1
2.70×10 5
k C
0
480 Nm −1
k C
1
1.02×10 6
λ T 1.17
ω T 2.33
λ C 1.0
ω C 1.0
α T 1.04
F T
o
1.5×10 5 N
α C 1.67
F C
o
6.5×10 5 N
clearing in its center. The slope is discretized into 1m square
blocks in the horizontal plane. The hillslope has a maxi-
mum slope angle of 40 ◦ and 32m of vertical drop (Fig. 5a).
Soil depth D, perpendicular to the slope surface, is 1m and
uniform across the entire slope. Trees, 50cm in diameter
(DBH), are arranged on a square lattice, 3m apart (horizon-
tal distance). For the base case, the clearing has no tree and
no roots. Other simulations shown later include trees in the
clearing. Figure 5b–d show the spatial distribution of root
density for the base case for roots of three different diam-
eters: 1, 10, and 100mm. The hydrologic behavior of the
slope, identical for all simulations, is shown in Fig. 2. Sim-
ulations are run for 2200min (36.67h) with a time step in-
terval of 1min. A landslide occurs when one or more cells
reach a total displacement of 4m. Soil and root parameters
used for all simulations are those given in Tables 2 and 3.
5.1 Displacement and force redistribution
Figure 6 illustrates the evolution of slope displacement and
soil and root bond forces during loading (the rainfall event) at
four different time steps, 900, 1200, 1358, and 1359min after
the start of loading. The last time step (1359min) is when the
slope (clearing) fails. Time step 1358 shows the slope at the
time step immediately before failure. Until failure, all slope
configurations are stable (factor of safety greater than 1 for
all cells of the slope).
During loading, cells in the clearing move downhill more
than cells in the stand (Fig. 6a–d). A discontinuity in dis-
placement appears near the top of the clearing. This gap,
12m long and slope parallel, occurs where the surface slope
is about 0.62 (ca. 32 ◦ ). This gap represents the formation
of a vertical tension crack at the upper edge of a soil slip
that has yet to fail completely. With increasing loading, dis-
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 461
Displacement (m)
Root reinforcement (kPa)
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
-20
-10
0
10
20
30
40
50
60
70
80
DBH = d = 0.5 m
DBH = d = 0.4 m
DBH = d = 0.3 m
DBH = d = 0.1 m
d = 0.5 m
Displacement (m)
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
-20
-10
0
10
20
30
40
50
60
70
80
d = 1.5 m
Displacement (m)
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
-20
-10
0
10
20
30
40
50
60
70
80
d = 2.5 m
Figure 4. Root reinforcement as a function of bond elongation for different tree diameters (DBH) and different distances from the tree trunk
(d). Positive displacement indicates tension; negative compression.
X
Y
Z
0.9
0.72
0.54
0.36
0.18
0
(a)
Slope
800
640
480
320
160
0
(b)
Roots m
2 -
=1 mm
10
8
6
4
2
0
(c)
Roots m
2 -
=10 mm
1
0.8
0.6
0.4
0.2
0
(d)
Roots m
2 -
=100 mm
Figure 5. Tree-covered sigmoid slope, 70m×70m, with a 20m×50m clearing in its center. (a) Slope (unitless) with cell discretization
(1m). Density of roots crossing a vertical plane in units of roots per square meter for roots of diameter (b) 1mm, (c) 10mm, and (d) 100mm.
placement across the crack grows to exceed 1m prior to
failure (Fig. 6c). Although this crack is in the clearing in a
zone devoid of trees, a few small roots from trees above the
crack are present and extend across this vertical tension crack
(see Fig. 5b). Cells above the crack show barely percepti-
ble displacements (< 0.1m). The situation is different in the
forested area, where, up to failure, displacement is signifi-
cantly smaller (about 10 times smaller), uniform (no discon-
tinuity), and highest in the steepest portion of the slope (not
visible in Fig. 6), with no evidence of a crack forming in the
upper part of the slope. The slope in the stand remains sta-
ble after the clearing fails for the remaining of the simulation
(2200min). In the forested area, cells that have undergone
displacement extend further uphill than in the clearing. We
attribute this effect to the connected root system of trees that
activates tensional forces uphill and pulls rooted cells down-
hill. These tensional forces are absent in the clearing due to
lack of roots and negligible soil tensional strength.
Figure 6e–h and i–l show the downslope (y axis) bond soil
and root forces, respectively. During loading (Fig. 6e–g, i–k),
soil compression forces increase near the bottom of the hills-
lope with significantly higher values in the clearing area (up
to −30kN in Fig. 6g, negative sign for compression). Soil
tension is negligible owing to the soil minimal tensional re-
sistance. In the forested area, roots of trees near the top of
the hillslope are in tension with the tensional force increas-
ing with increasing loading as the slope slowly slips downhill
(Fig. 6i–k). Root tension perpendicular to slope is highest on
both edges of the vertical crack. This is where the largest
displacements are observed generating the highest tensional
forces in the roots. In that zone, tension in roots reaches al-
most 20kN just before the clearing fails (Fig. 6k). Simultane-
ously, some roots of trees in the lower part of the slope are in
compression, relieving some of the compression in the soil.
Across-slope (also referred to as lateral or slope-parallel)
root forces are shown in Fig. 6m–p. Downward motion of
soil in the clearing causes a lateral tension in roots that span
the transition zone from clearing to forested area. This zone
is about 6–10m wide. It is across this boundary that displace-
ment gradients are high and across-slope root forces highest.
The lateral tension increases up to about 6kN with increasing
downhill motion of the clearing and stays high after failure
because the relative downslope displacement of cells across
the slope remains.
Figure 7 yields additional insights into the dynamics and
transfer of forces during loading. In that figure, values of
displacement (Fig. 7a–d), downslope bond force (root+soil,
Fig. 7e–h), and across-slope bond force (Fig. 7i–l) are shown
for three sections perpendicular to slope, at the center line
(x = 0) that passes through the clearing, at x = −9m near
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
462 D. Cohen and M. Schwarz: Tree-root control
Figure 6. Time evolution of (a–d) total displacement, (e–h) downslope (parallel to steepest slope) soil force, (i–l) downslope root force, and
(m–o) across-slope (lateral, also referred to as slope-parallel) root force shown at four time steps (left to right) for the slope shown in Fig. 5.
Failure occurs at t = 1359min (last column). t = 1358 min is the time step immediately preceding slope failure. Black curves in panel (a)
indicate locations of downslope cross sections at x = 0, x = −9m, and x = −12m shown in Fig. 7.
the left edge of the clearing, and at x = −12, which inter-
sects the first row of trees next to the clearing (see black
curves in Fig. 6a for location). Figure 7a–d clearly shows the
formation of the vertical crack with discontinuous displace-
ments across it at about y = 14m, initially only for the center
line (black symbols), but with increasing time (or load) also
at x = −9m (pink symbols). The forested area (x = −12m)
never develops such a crack and the displacement there is
always continuous. The bond force perpendicular to slope
shown in Fig. 7e–h indicates how the main resistive forces
holding the slope are redistributed during loading. Initially,
except for the clearing, which cannot hold much tension be-
cause of a lack of roots, forces on the slope are in tension
in the upper half and in compression in the lower part. The
transition occurs halfway down the slope in the forested area
(red symbols, y ∼ 0), a little uphill at the edge of the clearing
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 463
Y
Displacement (m)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
(a)
x = 0 m
x = -9 m
x = -12 m
Y
Displacement (m)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
(b)
Y
Displacement (m)
-30 -20 -10 0 10 20 30
0
1
2
3
4
(c)
Y
Displacement (m)
-30 -20 -10 0 10 20 30
0
1
2
3
4
(d)
Y
Downslope bond force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(e)
x = 0 m
x = 9 m
x = 12 m
Y
Downslope bond force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(f)
Y
Downslope bond force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(g)
Y
Downslope bond force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(h)
Y
Across-slope bond force (kN)
-30 -20 -10 0 10 20 30
-1
0
1
2
3
4
5
6
(i)
x = 0 m
x = 9 m
x = 12 m
Y
Across-slope bond force (kN)
-30 -20 -10 0 10 20 30
-1
0
1
2
3
4
5
6
(j)
Y
Across-slope bond force (kN)
-30 -20 -10 0 10 20 30
-1
0
1
2
3
4
5
6
(k)
Y
Across-slope bond force (kN)
-30 -20 -10 0 10 20 30
-1
0
1
2
3
4
5
6
(l)
Figure 7. Time evolution of (a–d) displacement, (e–h) downslope bond force, and (i–l) across-slope bond force along three downslope
cross sections at different distances from the center line (0, −9, and −12m; see Fig. 6) at four different times. Note the different scale for
displacement in panels (c) and (d).
(pink symbols, y = 5m). With increasing load, both tension
and compression in the slope increase. Tension is highest
where root density is highest (x = −12) and slightly lower
at the edge of the clearing (x = −9m). At t = 1200min, the
edge of the clearing has formed a crack and forces down-
hill of that crack are now in compression. Roots that cross
the crack at the edge of the clearing are now broken and
no longer provide any tensional resistance (pink symbols in
Fig. 7f).
Bonds that were in tension in the upper part of the slope at
t = 900min are now in compression owing to the failure of
roots across the widening crack near the edges of the clear-
ing. The clearing is now entirely held by compressive forces
and by lateral (across-slope) tensile forces shown in Fig. 7i–
l. These lateral forces are due to root tensile strength and
are highest near the transition from forest to clearing (pink
symbols, x = −9m), where the relative downslope displace-
ment between adjacent cells is highest. Along the first row of
trees (red symbols), cells that host a tree have larger values
of across-slope tensional forces than cells that do not giving
rise to a saw-tooth pattern of tensional force. In the clear-
ing (black symbols), positive lateral tensional forces are en-
tirely due to the soil apparent cohesion, which reaches val-
ues of almost 1kN. With increasing load and decreasing soil
shear strength due to increasing mean pore-water pressure,
the clearing eventually fails at t = 1359min but the forested
area remains stable for the remainder of the simulation (up to
2200min).
Results from this simulation demonstrate that maximum
tensional and compressive forces in rooted slopes do not con-
tributesimultaneouslyandequallytothestabilityoftheslope
during the initiation of a shallow landslide. Roots provide re-
inforcement in tension. This tensional root force can disap-
pears once displacement across a vertical crack becomes suf-
ficiently large. In our example, this occurs when the crack
grows to about 0.1m (see Fig. 4). Compression is higher
in the clearing (no roots) than in the vegetated area. Where
present, when slope-perpendicular root tensional reinforce-
ment is eliminated, soil stability is entirely accommodated
by soil compressive resistance and by lateral tension held
by roots. Lateral root forces provide additional stability to
the clearing by redistributing slope-perpendicular forces lat-
erally across the slope. The clearing fails when soil strength
at the base can no longer be held by the combination of the
lateral root bond forces and downslope soil compression, and
compression in the soil exceeds the maximum strength.
We can summarize the redistribution of forces during the
loading of a rooted hillslope into three distinct phases:
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
464 D. Cohen and M. Schwarz: Tree-root control
- Increasingloadandweakeningofsoilstrengthalongthe
basal failure plane (not shown) without any soil motion
(factor of safety above 1).
- Initiation of downward motion after some cells reach
critical condition (factor of safety equal to 1). Force
redistributions (compression in soil, tension and com-
pression in roots) prevent the slope from failing.
These forces increase with increasing load and in-
creasing mean pore-water pressure (e.g., Fig. 7, t =
900min). The culmination of slope-perpendicular ten-
sional forces across the crack (t = 900min, Fig. 7e) oc-
curs with (1) less-than-maximum compressive forces
in the lower-half of the slope and (2) lateral tensional
forces activated at the edge of the forested area (Fig. 7i).
- Culmination of compressive forces leading to failure
whenexceeded(t = 1359min,Fig.7,lastcolumn).This
occurs after tensile, slope-perpendicular forces due to
roots are lost across the vertical crack and when lateral
root tensile forces reach their maximum values.
The timing and duration of these three phases will vary
with soil mechanical properties, slope inclination, slope mor-
phology, root distribution, and hydrology, resulting in an in-
crease or decrease in the stability of the slope. These three
phases of force redistribution are used as criteria to define
the triggering of a landslide. In civil engineering, calcula-
tions using infinite slope analysis, for example, must yield
a factor of safety greater than 1 for the slope to be deemed
stable. Any values below 1 imply an unstable slope with the
possibility of a landslide, even if slope motion subsequently
stops with no occurrence of a runout. This definition of a
landslide corresponds to the second phase of force redistri-
bution where motion has initiated but complete failure has
not yet occurred. Many such occurrences of a failed land-
slide (at least temporarily) exist; one is shown in Fig. 8. In
risk analyses, or when studying geomorphological processes,
a landslide occurs by definition only when the soil mass fails
completely and is followed by a runout, corresponding to the
third phase of our force redistribution process. In that case,
the transition from phase 2 to phase 3 and the accompanying
redistribution of forces, is the critical process.
Changes in the values of the factor of safety (FOS) over
time help understand the processes of landslide triggering
and illustrates the three phases of landslide initiation and
forceredistribution.Figure9showstheevolutionofdisplace-
ment, the factor of safety, and the mean pore-water pres-
sure with time at the center of the clearing. Initially, FOS
is larger than 1 and decreases with increasing mean pore-
water pressure up until about 400min. This corresponds to
the phase 1 described above. Beyond 400min, the value of
FOS oscillates rapidly just above the value of 1. These os-
cillations, which last until failure, correspond to the critical
state of a self-organized system before global failure (e.g.,
Bak et al., 1988). Here, these oscillations correspond to our
Figure 8. Initiation of slip at Castel Vecchio, Italy, that did not re-
sult in a landslide in the geomorphic sense, but is considered as one
in the engineering sense. See text for details.
Time (min)
FOS
Displacement (m)
p (kPa)
0 200 400 600 800 1000 1200 1400 1600
1
1.02
1.04
1.06
1.08
0
1
2
3
4
-25
-20
-15
-10
-5
0
5
Figure 9. Time series of the factor of safety (red), displacement
(blue), and mean pore-water pressure (p, black) at the center of the
clearing (x = y = 0) for the simulation shown in Figs. 6 and 7.
phases 2 and 3. During this critical period, the number of cell
moves (not shown) increases dramatically as a result of force
redistribution between bonds of connected cells and as the
number of cells with factor of safety less than 1 increases.
The increase in the number of redistributions with loading is
similar to the process of avalanching in load-controlled self-
organized systems like fiber bundle models (Cohen et al.,
2009; Lehmann and Or, 2012). The increase in force re-
distribution across the slope corresponds to the progressive
slope failure stage of coalescence of local failure surfaces
that eventually leads to global failure (Petley et al., 2005;
Cohen et al., 2009). This is equivalent to our phase 3.
The decrease in the factor of safety is linked to the increase
in mean pore-water pressure in the soil (Fig. 9). A detailed
analysis of how hydrology impacts slope stability is beyond
the aim of this paper. Here we wish to point out that our sim-
ple dual-porosity model, with the coexistence of pore-water
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 465
pressure in the macropores and suction stress in the matrix
(see Fig. 2), is realistic and can model a wide range of hydro-
logical situations that can lead to shallow landslide trigger-
ing. In the simulation shown in Fig. 9, there is an impercepti-
ble increase in the factor of safety during the first phase of the
simulation until about 100min. This increase is due to the in-
crease (in absolute value) of the suction stress that increases
the soil apparent cohesion (see Fig. 2). The increase in pore
pressure after about 200min causes the soil to weaken with
an associated decrease in the factor of safety eventually lead-
ing to the critical state (FOS close to 1). A decrease in matrix
suction linked to flow of water from the macropores to the
matrix increases the mean pore-water pressure (Fig. 2) and
eventually causes soil to weaken sufficiently for a landslide
to occur. Depending on the application of the model and on
the local hydrological properties, choices of different values
of hydrological parameters than those used in this example
could lead to different hydrological triggering. For example,
triggering could be due to the rapid increase in macropore
water pressure and the saturation of the soil from top to bot-
tom with little time for changes in matrix pore pressure to
occur. In our example, preferential flow paths lead to local
increases of pore-water pressure that, in combination with a
lossofsuctionstressinthesoilmatrix,resultinacriticaldrop
of soil shear strength typical of forested soils on compacted
bedrock(Lehmannetal.,2013).Yetinanothersituation,high
pore-water pressure can originate from ephemeral springs or
water exfiltration from fractured bedrock (Montgomery and
Dietrich, 1994).
5.2 Effects of root tensile and compressive strength
Our results show that force mobilization and redistribution in
the soil and in the root system during the triggering of a shal-
low landslide is a complex process. Our model can be used to
investigate the effects of the various components of the bond
force system (roots and soil) on the dominant reinforcement
mechanisms (tension or compression, lateral or downslope)
and how these forces control the stability of the slope. Under-
standing which of these forces control slope stability under
certain conditions is important for making appropriate sim-
plifications when the full level of details is not needed or not
known.
Figure 10 shows the displacement for nine hillslope simu-
lations where trees, spaced 3m apart, cover the entire slope.
In Fig. 10a, tree diameter (DBH) is 50cm; in Fig. 10b, it is
40cm;inFig.10c,itis30cm.Ineachofthethreecases,three
simulations are shown: trees with roots that have both tensile
and compressive strength (the standard behavior), roots that
only have tensile strength, and roots with only compressive
strength. All simulations are run with the same hydrologic
loading used in earlier simulations (see Fig. 2).
The slope behaves differently depending on the tree size
and the type of root reinforcement. Root reinforcement for
the 50cm diameter trees is sufficiently large that the slope
Displacement (m)
0
0.05
0.1
0.15
0.2 (a)
Compression only
Tension + compression
DBH = 50 cm
Tension only
Displacement (m)
0
0.1
0.2
0.3
0.4
0.5
0.6 (b)
Compression only
Tension + compression
DBH = 40 cm
Tension only
Failure
Time (min)
Displacement (m)
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0
1
2
3
4
5 (c)
DBH = 30 cm
Roots fail across
tension crack
Tension +
compression
Tension only
Compression only
Figure 10. Effects of tensile and compressive strength of roots on
slope displacement and stability for trees of different diameters.
Displacement at the slope center (x = y = 0) as a function of time
for (a) a stand of trees 50cm in diameter, (b) 40cm in diameter,
and (c) 30cm in diameter. Trees are spaced 3m apart in each cases.
Each graph shows three curves for roots with both compressive and
tensile strength, roots with only tensile strength, and roots with only
compressive strength.
does not fail regardless of the type of root reinforcement
(tensile, compressive, or both). For the 40cm diameter trees,
there is a threshold: the tensile strength of roots is needed to
keep the slope stable. Without root tensile strength (compres-
sion only), the slope fails (Fig. 10b). Finally, for the 30cm
diameter trees, all root reinforcement configurations lead to
slope failure, but at different times, with compression-only
roots failing first and roots with both compression and ten-
sion last.
Results indicate that roots with only tensile strength limit
downward slope slip under loading and delay slope failure
more than roots that have only compressive strength. Roots
that have both tensile and compressive strength offer the best
protection against slope motion and slope failure. Neglecting
root compression in the simulations results in only a couple
of centimeters’ difference in slope displacement or less than
1h in the timing of the landslide. Neglecting root tension,
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
466 D. Cohen and M. Schwarz: Tree-root control
Y
DBH = 50 cm
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(a)
T + C
T only
C only
Y
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(b)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
(c)
Y
DBH = 40 cm
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(d)
Y
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(e)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
(f)
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(g)
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
DBH = 30 cm
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(g)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
(i)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.2
0.4
0.6
0.8
1
Y
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(h)
Y
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Figure 11. Effects of tensile (T) and compressive (C) strength of roots on root and soil bond force distribution along a downslope section at
the centerline (x = 0) along a row of trees for a tree-covered slope with (a–c) DBH = 50cm, (d–f) DBH = 40cm, and (g–i) DBH = 30cm.
(a, d, g) Downslope root force. (b, e, h) Downslope soil force. (c, f, i) Across-slope root force. Sections are shown either at the end of the
run for simulations that did not fail or at the time step just prior to failure for those that did (see Fig. 10).
however, can result in predicting a false slope failure. Also,
neglecting tension misses the jump in displacement during
theearly initiationof thelandslide,when rootsacross theten-
sion gap in the upper part of the slope fail under tension (see
Fig. 10c). Note also that when roots across the vertical crack
fail in tension (as is the case for the 30cm diameter trees),
the slope eventually fails. This appears to be the case for all
simulations we tested. However, simulations with roots that
do not break across the widening crack do not necessarily
remain stable over the duration of the simulation (2200min).
Figure 11 illustrates the conditions under which the slope
fails for the different tree-size diameters and root-strength
configurations. Each graph in Fig. 11 shows a bond force
along the downslope section at the center of the slope (x =
0). Downslope root bond force along that section (Fig. 11a,
d, g) indicates that when roots have no tensile strength (C
only), roots in the lower section of the slope bear a higher
compressive load. Similarly, roots that have no compressive
strength (T only) bear higher tensile loads in the upper part
of the slope, but only slightly higher than roots that have both
tensile and compressive strength (T +C). As expected, roots
of larger trees can bear higher tensile and compressive forces
owing to higher root densities and more roots of larger di-
ameters. Downslope soil bond forces (Fig. 11b, e, h) indicate
thatsoilsinslopescoveredbysmallertreesmusttakemoreof
the compressive force caused by the slope downhill motion.
For the 30cm diameter trees (Fig. 11h), the soil compressive
force eventually reaches its maximum value and the slope
fails, regardless of the root configuration, because roots hold
only a small fraction of the tensile or compressive resistance
that helps maintain the slope stable: roots are too few and too
small for this tree size. This is also the case for the 40cm
diameter trees when roots have only compressive strength
(Fig. 11d, e). Because roots do not hold any tension in the
upper part of the slope, and root compression is insufficient
to support much load, soil bond in compression eventually
reaches a maximum and the slope fails at t = 1960min. Fig-
ure 11c, f, and i show the lateral root force across the slope.
This force is 1 order of magnitude smaller than the downs-
lope force and has only a limited role in the slope stability for
the cases shown here. These simulations indicate that downs-
lope root and soil forces control slope stability, which is reg-
ulated by the maximum soil compression. Roots can reduce
soil compression by taking up some of the force in tension
in the upper part of the slope, preventing or delaying failure.
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 467
Root compression alone is insufficient to offset soil compres-
sion in the lower part of the slope.
5.3 Effects of weak zones
The structure of the stand (dimension, density, and rela-
tive position of trees) plays an important role on root rein-
forcement and slope stability. Moos et al. (2016) found that
susceptibility to landslide was higher in plots with longer
downslope gaps in the tree stand and in locations where the
distance to nearby trees was higher. Conversely, Moos et al.
(2016) also found that susceptibility to landslide was smaller
where root reinforcement, based on tree diameter and dis-
tancefromtree,washigh.Weakzones,zoneswithlowvalues
of root reinforcement, can serve as initiation points for slope
movement and control the location and size of a landslide
(Schwarz et al., 2010b, 2012a). An example of a weak zone
where a soil slip initiated is shown in Fig. 12. Roots around
a tree provide sufficient stiffness to make the soil around the
tree behave as a rigid body. The zone in between the tree and
its neighbors does not provide sufficient root reinforcement
and a gap opens as a result of loading (here rainfall). Here,
we explore independently how tree size (diameter) and tree
spacing can affect landslide initiation and hillslope stability.
5.3.1 Tree diameter
Our base scenario is the simulation presented earlier with
trees 50cm in diameter spaced 3m apart on a sigmoid hill-
slope with a 20m×50m clearing in the center. Five other
simulations were run with the clearing area planted with trees
of diameter 10, 20, 30, 40, and 50cm, all spaced 3m apart as
in the forested area surrounding the clearing. These six sim-
ulations are referred to as 50/0, 50/10, 50/20, 50/30, 50/40,
and 50/50, where the first and second numbers indicate the
stand tree diameter and the tree clearing diameter, respec-
tively. Figure 13 shows the computed factor of safety and
displacement at the center of the slope for these six sim-
ulations. The 50/10 simulation fails earlier (t = 1266min)
than the 50/0 simulation (t = 1359min). For larger trees in
the clearing, time to failure increases, from 1419min for the
50/20 to 1793min for 50/30. Slopes with trees in the clearing
greater than or equal to 40cm do not fail.
The time evolution of the factor of safety depends on the
tree size inside the clearing. Simulations 50/40 and 50/50
have values of factor of safety that remain significantly
higher than the remaining simulations, although their val-
ues sometimes oscillate very close to 1. Although these two
configurations have undergone some downhill motion, it is
limited to a few centimeters, significantly less than the other
cases. These two slopes with large trees are in critical con-
dition because their factors of safety is nearly equal to 1
(< 1.1). Slope motion is limited to a small area near the cen-
ter of the clearing and to very few cell moves owing to the
Figure 12. Example of a weak zone in a forested area showing
isolated tree stumps with a root system that behaved as a stiff island
during the opening of a gap in a weak zone in between root systems
of adjacent trees.
FOS
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1 (a)
50/40
50/10
50/0
50/20
50/30
50/50
Time (min)
Displacement (m)
Displacement (m)
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0
1
2
3
4
5
0
0.01
0.02
0.03
0.04
0.05
(b)
Figure 13. Time series of (a) factor of safety (FOS) and (b) dis-
placement at the center of the slope (x = 0, y = 0) for six simula-
tions with different tree sizes inside the clearing. The sets of two
numbers shown in panel (a) indicate the stand DBH and the clear-
ing DBH in centimeters. For example, 50/10 means a stand of trees
50cm in diameter with a clearing filled with trees 10cm in diame-
ter. Spacing is identical in the clearing and in the stand (3m). Color
code is identical in panels (a) and (b). In panel (b) the simulations
50/40 and 50/50 do not fail and slope displacement plots on the
vertical axis on the right side (indicated by arrows).
largetensional resistance ofrootsthat limit downslopemove-
ment.
Figure 14 shows the distribution of displacement, root and
soil bond forces across the slope just before failure (or at the
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
468 D. Cohen and M. Schwarz: Tree-root control
Figure 14. Effect of clearing tree size diameter on slope displacement and soil and root bond forces. From left to right, slope displacement
(d), soil downslope compression (F y
soil ), downslope root force (F
y
root ), and across-slope root force (F
x
root ), at the time step just before failure
and displacement at failure (d fail ) for the six simulations with different tree size diameters inside the clearing shown in Fig. 13. (a–e) Empty
clearing (50/0), (f–j) DBH = 10cm (50/10), (k–o) DBH = 20cm (50/20), (p–t) DBH = 30cm (50/30), (u–y) DBH = 40cm (50/40), and
(z–ad) DBH = 50cm (50/50). Outside the clearing, DBH = 50cm. All trees are spaced 3m apart. Scale is given in the first row except when
noted.
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 469
Y
x = 0
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(a)
50/30
50/40
50/50
50/0
50/10
50/20
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
(c)
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
Y
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(b)
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
Y
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(e)
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
Y
x = -9 m
Downslope soil force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(d)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
7
8
9
10
(f)
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
7
8
9
10
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
7
8
9
10
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
7
8
9
10
Y
-30 -20 -10 0 10 20 30
0
1
2
3
4
5
6
7
8
9
10
Figure 15. Effects of clearing tree size on (a, d) downslope soil force, (b, e) downslope root force, and (c, f) across-slope root force along
two downslope sections at (a–c) x = 0 and (d–f) x = −9m near the clearing–stand transition just before failure (simulations 50/0, 50/10,
50/20, and 50/30) and at the end of the simulations (50/40 and 50/50).
last time step of the run for simulations that did not fail), and
displacement at failure for all six simulations (one simulation
per row). For cases where a landslide occurred (0, 10, 20,
and 30cm), only the clearing area fails except for the 50/30
case, where the entire slope fails (see Fig. 14, last column).
The clearing with the 30cm trees pulls down the slope with
the stand of 50cm diameter trees. Lateral root forces in the
clearing and across the stand–clearing transition, and downs-
lopetensileforcesinthestand,aresignificantlyhigherforthe
30cm simulation than for any other simulations (see Fig. 14r,
s). Despite smaller displacement before failure (Fig. 14p),
30cm diameter tree roots mobilize more force than the sim-
ulation with smaller trees owing to higher root density and
sizes and larger root stiffness. This causes high downslope
root forces at the upper edges of the clearing. Also, lateral
force in the clearing are higher and extend across the full
width of the clearing (Fig. 14s). As a result, unlike simula-
tions with smaller trees inside the clearing, tensile root fail-
ure does not occur inside the clearing but outside in the stand,
resulting in the collapse of the stand, pulled down by lateral
forces originating from the 30cm diameter trees in the clear-
ing. This is a case where lateral tensile forces plays a crucial
role: by extending spatially across a larger area, lateral forces
of the 30cm trees eventually pull roots of 50cm trees when
theclearingfails.Thisnumericalresulthelpsexplainthefield
observations of Rickli and Graf (2009), who found that the
mean landslide area was greater for forested slopes than for
non-forested slope in the same catchment.
This behavior is also illustrated in Fig. 15, which shows
downslope soil and root forces as well as across-slope root
force along two downslope sections (x = 0 and x = −9m)
for the six simulations at the time step just prior failure (for
simulations that resulted in a landslide, i.e., 10, 20 and 30cm
diameter trees in clearing) or at the end of the run (40 and
50cm diameter trees).
Results in that figure clearly show that soil bond forces
along the slope center (Fig. 15a) for small trees (0 to 30cm
in diameter) reach significantly higher values than for large
trees (factor of 5 to 6), eventually reaching the soil max-
imum compressive strength just before failure. Downslope
root-bond force is smaller for these smaller trees owing to
smaller densities and smaller root sizes (Fig. 15b). For larger
trees (40 and 50cm), roots take up some of the load on the
soil, reducing compressive forces in the soil downslope. The
situation is nearly similar along the clearing–stand transition
(Fig. 15d–f) except that the 30cm diameter trees have the
highest downslope root bond force (Fig. 15e, cyan symbols).
Despite the smaller displacements of soil for the 30cm trees
than for smaller trees (see Fig. 14a, f, k, p), yet significantly
larger than for the 40 and 50cm trees (Fig. 14p, u, z; note
the different scale), downslope root force is maximized for
the 30cm trees owing to the combination of displacement
and root-diameter sizes that are mobilized. This is also ob-
servable on the across-slope root bond force which is highest
for that tree size (Fig. 15f). The across-slope root force is
nearly zero for the large trees, all the load being handled via
the downslope bond forces (Fig. 15c, f). For the smaller trees,
the across-slope root force is significant at the clearing–stand
transition (Fig. 15f) but small or close to zero at the center
of the clearing (Fig. 15c). The 30cm diameter configuration
stands out from the others in having the largest across-slope
root bond forces which eventually fail outside the clearing
area, entraining the large trees in the stand during the col-
lapse.
Figure 15 also helps understand why the 50/10 simulation
fails before the 50/0 simulation. This is counterintuitive but
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
470 D. Cohen and M. Schwarz: Tree-root control
is the result of the force balance and the effects of higher
root stiffness with increasing root diameter. For the 50/10
simulation, slope-parallel (across-slope) root reinforcement
is slightly smaller (of the order of 100N) than for the 50/0
simulation and limited to areas around the large, 50cm, tree
trunks at the edge of the clearing (Fig. 15e and also com-
pare Fig. 13d to i). Reinforcement values are smaller be-
cause displacement is smaller (compare Fig. 13a to f) and
root stiffness small. As a result, slightly more load is taken
into downslope soil compression in the 50/10 case than in
the 50/0 case (again of the order of 100N; see Fig. 15a)
and the 50/10 case reaches maximum soil compression be-
fore the 50/0 case, failing first despite a higher root density
and a higher potential for root reinforcement. For the larger
trees (simulations 50/20 and 50/30), although displacement
is less prior to failure (e.g., Fig. 13k, p), the larger root stiff-
ness associated with these larger tree roots produces larger
root forces at smaller displacements resulting in less downs-
lope soil compression for the same time and thus delaying
the time to failure.
5.3.2 Tree spacing
Effects of tree spacing on slope stability also yielded some
unexpectedresults.Treeswerespacedevenlyontheslopeus-
ing the center of the slope (x = y = 0) as the reference point
for a tree. All other trees are located at equal intervals along
the x and y axes from this central tree. Figure 16 shows dis-
placement as a function of time at the slope center for five
simulations with tree spacing of 3, 5, 7 (two simulations),
and 10m. Intuitively, one would expect that increasing tree
spacing would decrease root reinforcement away from trees
and increase the likelihood of a weak zone to fail. Results,
however, show a different behavior. Slopes with trees spaced
3 or 7m (no offset, simply called 7m spacing in Fig. 16)
apart were stable, but the slope with tree spacing of 5m was
not. Despite having higher tree density than the 7m spacing
simulation, and thus having higher root density and root re-
inforcement values, the slope with the 5m tree spacing failed
at t = 1781min while the 7m spacing did not fail. Figure 17
shows the slope displacement and tree density (a–e), downs-
loperootforce(f–j),anddownslopesoilforce(k–o).Because
trees are spaced at regular intervals around y = 0, tree posi-
tions for the 5m spacing are at y = 0, 5, 10, and 15m. For the
7m spacing, trees are positioned at y = 0, 7, and 14m. The
vertical crack that forms upslope occurs at the smallest root
reinforcement location in between two rows of trees. This is
at about y = 13m for the 5 and 10m spacing, and at about
y = 11m for the 7m spacing (see Fig. 17f–j). The 3m spac-
ing had sufficiently high root reinforcement that a crack did
not form. The vertical crack is 2m higher up the slope for
the 5m tree spacing than for the 7m spacing. Because the
crack is higher upslope, the number of cells along the y axis
that move downhill due to loading is larger for the 5m spac-
ing than for the 7m spacing. As a result, near the bottom of
Time (min)
Displacement (m)
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0
1
2
3
4
5
3 m
5 m
7 m
7 m (2 m offset)
10 m
Zoom 400 to 600 min
400 450 500 550 600
0
0.1
0.2
0.3
Figure 16. Effects of tree spacing on slope displacement at the cen-
ter (x = y = 0) for five different tree spacings and spatial configu-
rations. Inset shows details at early stage of displacement and the
failure of roots across the tension crack at 580min for the 5m tree
spacing.
the hill, compression is significantly higher for the 5m spac-
ing (Fig. 17l). With increasing load, the 5m spacing slope
reaches its ultimate value of compression and fails while the
7m spacing never reaches that point.
To explore the effect of crack location on slope stability,
trees in the 7m spacing slope were offset 2m uphill, so that
a vertical crack would form at a higher elevation than with-
out the offset. This simulation is shown with a dashed curve
in Fig. 16. Figure 17d, i, n show the hillslope for this sim-
ulation and indicate that a crack forms at y = 13, like in
the 5m simulation, thus resulting in high soil compression
forces downslope. In this configuration, the slope eventually
fails. High values of soil compression forces that lead to fail-
ure (5m, 7m with offset, and 10m spacing) are clearly vis-
ible in Fig 17l–o, and contrast with lower soil compression
forces in simulations that did not fail (3m, 7m without off-
set, Fig. 17k, m).
5.4 Effects of maximum root diameter
SOSlopewasusedtotesttheinfluenceoftherangeofrootdi-
ameter classes on the stability of a slope. Figure 18 shows the
displacement at the center of the slope (x = y = 0) as a func-
tion of time for six simulations with different maximum root
diameter: 5, 7, 8, 10, 20, and 100mm. The simulations with
20 and 100mm maximum root size diameter have no land-
slide and are practically indistinguishable. This is because
the number of roots larger than 20mm is insignificant and
contributes little additional strength to the root bundle. The 8
and 10mm simulations also do not fail and have only slightly
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 471
Figure 17. Effect of tree spacing on hillslope behavior. (a–e) Tree location (black circles) on the hillslope over slope displacement at the last
stable time step or last time step for simulations where no landslide occurs (see Fig. 16). (f–j) Downslope root bond force. (k–o) Downslope
soil bond force. Vertical crack position on slope shown in the column at center.
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
472 D. Cohen and M. Schwarz: Tree-root control
Time (min)
Displacement (m)
Displacement (m)
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0
1
2
3
4
5
5 mm
7 mm
8 mm
10 mm
20 mm
100 mm
Figure 18. Effect of maximum root-size diameter (5 to 100mm)
on displacement at the slope center (x = y = 0) for a 3m spaced,
DBH = 40cm, tree-covered slope. The 5 and 7mm simulations
both yield a landslide and their displacement curves plot on the ver-
tical axis to the right (indicated by arrows).
larger displacements (6 to 7cm instead of 5cm for the 20 and
100mm simulations). The two simulations with a maximum
root diameter class of 5 and 7mm, however, fail at 1400 and
1500min, respectively. The threshold for stability is thus ob-
tained by including root size up to 8mm in diameter. Root re-
inforcement that includes only smaller roots is significantly
smaller than if the entire bundle is included. Not including
large roots can yield incorrect predictions of slope behavior.
Figure 19 shows the downslope and across-slope forces at
the center line for several simulations with different maxi-
mum root-size diameter. The 5mm simulation has the small-
est amount of downslope root force but the highest across-
slope root force and downslope soil compression, explaining
why this simulation fails while others (10, 20, and 100mm
maximum root diameter) do not. Insufficient root density and
lack of large roots compromises the stability of the slope
by offering little resistance to loading and declining shear
strength of soil. Lateral root forces are small for all cases and
has a negligible impact here on slope stability (Fig. 19d).
6 Synthesis of force redistributions during
triggering of shallow landslides
Figure 20 summarizes the typical evolution of forces during
landslide initiation of a forested slope for the 50/30 case de-
scribed in Sect. 4.3 (see Figs. 14–16). In this simulation, a
clearing is planted with trees 30cm in diameter, while the
rest of the slope has trees 50cm in diameter.
The largest force that contributes to slope stability is soil
compression in the area above the landslide toe. There, soil
compression increases initially rapidly until it plateaus at
about 700min. During this increase, root tension across a
growing crack increases and also plateaus. Root compression
downslope similarly increases and then plateaus but is signif-
icantly smaller than either root tension upslope or soil com-
pression downslope. This time period is defined as phase 2 of
our landslide initiation process, which starts when many ar-
eas of the slope have a factor of safety that has decreased to
1 (Fig. 20b). Phase 1 of the initiation was the decrease in the
factor of safety due to loading and soil weakening without
any slope motion.
At t = 720min, the roots across the tension crack fail and
that tensional resisting force goes to zero. Instantaneously,
the slope moves downhill and the force lost by tree roots is
taken up by both soil and root compression downslope with
the soil taking up most of the increase. This is the begin-
ning of phase 3. With continued loading, soil compression
increases but root compression slowly decreases. Lateral root
forces at the edge of the clearing begin to take some of the
load to resist downslope movement. Eventually the soil max-
imum compressive strength is reached and the clearing fails
just before 1800min.
The time span of the three phases varies with tree size, tree
spacing, maximum root diameter, and of course soil and hy-
drological properties (here fixed for all simulations). Look-
ing back at Figs. 13, 16, and 18, phase 2 can last from several
hourstolessthanone.Sometimes,nocrackforms,thereisno
crack-root failure, and phase 2 and 3 overlap. When the slope
has no clearing (as in simulations shown in Figs. 17 and 18),
these same three phases exist but lateral forces play no role.
Force redistribution and force balance is dominated by soil
compression, adjusted by root tension in the upslope area and
to a lesser extent root compression downslope. Root forces
modify the force balance significantly but soil compression,
due to its magnitude, dominates and controls the slope sta-
bility and its time to failure. Simulations with smaller soil
depth will change this balance: smaller depth will decrease
the absolute values of soil compression (see Fig. 3) and tree
roots will then support tensile and compressive forces equal
or greater to soil compression. In such a situation, roots may
be the main factor controlling slope stability.
7 Conclusions
There are growing evidences that the effects of root rein-
forcement on slope stability are the results of complex in-
teractions of different factors in which individual contribu-
tions are difficult to isolate using classical methods (e.g., in-
finite slope calculations). The model presented here, SOS-
lope, is the final element of a series of related studies aiming
to quantitatively upscale the stress-strain behavior of rooted
soils under tension, compression, and shearing. In this frame-
work, SOSlope represents the final module where previously
investigated aspects of root reinforcements are combined to
quantify the macroscopic influences of root reinforcement on
slope stability considering spatial heterogeneities of root dis-
tribution. The model can produce a systematic analysis of
the factors influencing the contribution of root reinforcement
on slope stability, yielding a quantitative basis for discussion
of root reinforcement mechanisms for slope stabilization and
support for the assumptions or simplifications needed to im-
plement such effects in simpler approaches for slope stability
Earth Surf. Dynam., 5, 451–477, 2017 www.earth-surf-dynam.net/5/451/2017/
D. Cohen and M. Schwarz: Tree-root control 473
Y
Across-slope soil force (kN)
-30 -20 -10 0 10 20 30
0
0.5
1
1.5
(d)
Y
Across-slope soil force (kN)
-30 -20 -10 0 10 20 30
0
0.5
1
1.5
(d)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.5
1
1.5
(c)
Y
Across-slope root force (kN)
-30 -20 -10 0 10 20 30
0
0.5
1
1.5
(c)
Y
Downslope soil force (kN)
-20 0 20
-30
-20
-10
0
10
20
(b)
Y
Downslope soil force (kN)
-20 0 20
-30
-20
-10
0
10
20
(b)
Y
Downslope root force (kN)
-30 -20 -10 0 10 20 30
-30
-20
-10
0
10
20
(a)
100 mm
20 mm
10 mm
5 mm
(a)
t = 1423 min (just prior to failure)
t = 2200 min (end of run)
Figure 19. Effect of maximum root diameter on (a) downslope root force, (b) downslope soil force, (c) across-slope root force, and
(d) across-slope soil force at the center line for the times specified in panel (a).
Bond force (kN)
-25
-20
-15
-10
-5
0
5
10
15
(a)
Phase 1 Phase 2 Phase3
Lateral root force at edge of clearing
Soil compression downslope
Root force across crack
Root force downslope
Clearing failure
Crack root failure
Solution time
FOS
Displacement (m)
0 200 400 600 800 1000 1200 1400 1600 1800
1
1.02
1.04
1.06
1.08
0
1
2
3
4
(b)
Figure 20. Evolution of (a) soil and root bond force, and (b) factor
of safety and displacement during the initiation of a landslide in a
forested hillslope (simulation 50/30 described earlier) at the center
of the slope.
calculations (Dorren and Schwarz, 2016). Specifically, sim-
ulation results obtained with SOSlope highlight the potential
of the model to investigate fundamental questions such as
the role of forest structure (e.g., tree size, tree spacing), root
distribution, and root mechanical properties on the triggering
mechanisms of shallow landslides. Based on the results pre-
sented here the following general statements can be made:
– Maximum root reinforcement under tension and com-
pression does not take place simultaneously.
– Root tensile strength is more effective than root com-
pressive strength in preventing or delaying a landslide.
– The stabilization effect of roots depends on their spa-
tial distribution: the presence of a “weak zone” leads
to behavior similar to bare soils. With little or no root
reinforcement, slope failure is more likely and occurs
earlier.
– Root reinforcement at the macroscopic scale is domi-
nated by intermediate to coarse roots when present. For
the species considered here and based on available data,
roots between 5 and 20mm contribute the most to root
reinforcement.
– Tree positions in the tension zone of a potential land-
slide influence the stability of the slope. In general,
the effect of lateral root reinforcement in tension con-
tributes most to stability along the transition between
stable and unstable zones of the hillslope where a crack
can form.
These observations indicate that the standard, slope-
uniform, constant apparent cohesion approach for rooted soil
is often inappropriate, especially for forested slopes, where
roots contribute significantly to the balance of forces. For ex-
ample, our model shows that the specific locations of trees on
a slope (Fig. 17) are important for predicting slope failure, a
conclusion that cannot be reached with the apparent cohe-
sion model. Also, root force distribution on the slope may
result in a larger landslide for trees with higher root densi-
ties (Fig. 14), a result impossible to predict with the appar-
ent cohesion model. Also, root stiffness can modify the time
to failure of rooted soils by either increasing or decreasing
www.earth-surf-dynam.net/5/451/2017/ Earth Surf. Dynam., 5, 451–477, 2017
474 D. Cohen and M. Schwarz: Tree-root control
forces mobilized in roots at different displacement (Fig. 13).
Finally, our simulations quantify the importance of consid-
ering the heterogeneous distribution of tensional as well as
compressional root and soil forces, an element that is entirely
missing from traditional infinite slope stability models.
To our knowledge, SOSlope is the first model to im-
plement a new approach that characterizes the force-
displacement behavior of rooted soils under both tension
and compression. Including this fundamental behavior is key
for understanding and modeling shallow landslide triggering.
Further work is needed to extend the applicability of stan-
dard geotechnical methods (e.g., Schwarz et al., 2015) for
the quantification of those soil and root forces.
The SOSlope model can be applied at the hillslope scale
to investigate the effect of single factors such as root distri-
bution and root mechanical properties (species specific) on
slope stability, and quantification of bio-engineering mea-
sures and protective effects of forests. An important appli-
cation at the hillslope scale is the testing of hypotheses that
would support the simplification of calculations in problem-
specific applications, e.g., for slope stability model at a re-
gional scale.
The use of the SOSlope model at the catchment scale will
be useful for studying the effects of vegetation on slope sta-
bility processes in the short and long term. In the long term,
root strength can vary by orders of magnitude (Vergani et al.,
2016), and estimation of slope stability and landslide initi-
ation is necessary for an integrated management of moun-
tain catchments for risk reduction and control of sediment
balance. In the short term, estimations of safety factors for
rooted slopes provide important data for risk assessment in
forested mountain catchments. Future work will focus on
both these short and long timescales.