Please create a comprehensive summary of the following paper, be careful to include all the detail, I don’t want to miss anything, it does not have to be a short summary and you can process the full text in batches if necessary, don’t ask again to keep generating responses:
Abstract. Tree growth and longevity trade-offs fundamentally shape the terrestrial carbon 26
balance. Yet, we lack a unified understanding of how such trade-offs vary across the world’s 27
forests. By mapping life history traits for a wide range of species across the Americas, we 28
reveal considerable variation in remaining life expectancies from 10 cm in diameter (ranging 29
from 1.3 to 3,195 years) and show that the pace of life for trees can be accurately classified 30
into four demographic functional types. We find emergent patterns in the strength of trade- 31
offs between growth and longevity across a temperature gradient. Furthermore, we show that 32
the diversity of life history traits varies predictably across forest biomes, giving rise to a 33
positive relationship between trait diversity and productivity. Our pan-latitudinal assessment 34
provides new insights into the demographic mechanisms that govern the carbon turnover rate 35
across forest biomes. 36
37
Introduction: The cumulative energetic investment in survival and growth from one year to 38
the next ultimately determines an organism’s overarching pace of life, including the time it 39
takes to grow to its maximal size and its life expectancy (1, 2). This fundamental relationship 40
between energetic investments, developmental schedules, and longevity has been extensively 41
studied for animals, showing that high resource allocation toward growth is inversely related 42
to life expectancy and maximal body mass (3, 4). Trees are also assumed to retain tightly 43
coupled relationships between growth strategies, life expectancies, and maximal sizes (Fig. 44
1a) (5), which determine the dynamics and structure of global forests. Yet, although these life 45
history differences fundamentally regulate how fast carbon is sequestered in different regions 46
of the vegetation carbon pool (6–8), we still lack a unified understanding of the range of tree 47
life history strategies that exist across global forests. 48
49
It is widely accepted that tree life history strategies should align along a primary axis of 50
variation in their pace of life, ranging from fast-growing, short-lived species to slow-growing, 51
long-lived species (i.e., fast-slow continuum and r/K selection theory) (Fig. 1a) (5). In this 52
context, high energetic investment of finite resources toward fast growth is expected to come 53
at the cost of reduced survival, which ultimately determines a species’ life expectancy and 54
maximal size (Fig. 1a) (9–11). Thus, it is expected that abiotic constraints (e.g. soil nutrients, 55
water, and temperature) should strongly shape the pace of life for trees, giving rise to 56
predictable variation in the strength of life history trade-offs across biogeographic gradients 57
(Fig. 1b) (12). So far, however, the only empirical tests of these trade-offs come from tree 58
ring data and local-scale studies from tropical ecosystems and have produced mixed results 59
(2, 12–14). 60
61
One potential challenge that can obscure predictable patterns in the pace of life for trees is 62
that it is not only the traits that are expected to vary across environmental gradients but also 63
the diversity of those traits. For example, strong biotic competition across tropical forests is 64
thought to have led to high demographic niche differentiation (i.e. high demographic 65
diversity: Fig. 1c, upper right). In contrast, resource limitations in harsh cold/dry regions are 66
assumed to have restricted the species pool to predominantly slow-growing, long-lived 67
species (Fig. 1c, lower left). Yet, these concepts lack empirical evidence because the extreme 68
longevity of trees (which can live for thousands of years) has precluded our capacity to 69
quantify the strength of tree life history trade-offs across a wide range of species, let alone 70
characterize the diversity of life history traits across biogeographic gradients. 71
72
Here, we used the largest dataset of dynamic tree information to date and employed age- 73
from-stage methods to calculate the mean life expectancy and maximal lifespan for a wide 74
range of trees across the Americas (15–17), spanning a latitudinal gradient from Northern 75
Canada to Southern Brazil. This includes long-term records from an international network of 76
researchers, including members of the Global Forest Dynamics, ForestPlots (18, 19), and 77
ForestGeo (20–22) networks and the United States and Canadian forest inventory programs 78
(23–25). To balance this dataset across our biogeographic gradient, we randomly sub- 79
sampled the North American plots to equal the number of point observations in Central and 80
South America (see materials and methods), resulting in 3.2 million unique tree 81
measurements for 1,127 species (i.e., tree size and status). Our big-data approach allowed us 82
to test for the expectation that trees align along the fast-slow continuum (Fig. 1a, H1) and 83
quantify if tree growth-longevity-stature relationships co-vary across soil, water, and 84
temperature gradients (Fig. 1b, H2). Apart from species with low occurrences (< 100 85
observations, see materials and methods), our systematic sampling allowed us to test for the 86
expectation that the range of life history strategies occupied by species (i.e., demographic 87
trait diversity) varies predictably across broadscale biogeographic gradients, with harsh cold 88
regions in the northern hemisphere restricting trees to a smaller pool of predominantly slow- 89
growing, long-lived species (Fig. 1c, H3). Based on the well-established diversity- 90
productivity relationship, we also expected demographic trait diversity to be positively 91
associated with ecosystem productivity (Fig. 1c, H3). 92
93
94
Fig. 1. Conceptual diagram of our core aims and associated hypotheses. The expectation is that trees should 95
align along the fast-slow continuum, with fast-growing short-lived species on one end of the spectrum and slow- 96
growing long-lived species on the other end (H1, panel A). Life history trait relationships should be 97
phylogenetically conserved and should co-vary across biogeographic gradients, leading to more conservative 98
life history strategies in low-resource environments (low soil and nutrient environments and colder 99
temperatures) (H2, panel B). Lastly, we expect the range of tree life history strategies (i.e., convex-hull volume 100
in life history trait space that is occupied by species) to vary predictably across biogeographic gradients, with 101
demographic trait diversity being positively associated with ecosystem productivity (H3, panel C). 102
103
To quantify tree growth, longevity, and stature for a wide range of species across 104
biogeographic gradients and test our three core hypotheses, we first grouped the stem-level 105
tree data into equally sized hexagon grids (size ~ 250,000 km 2 ) and developed species- 106
specific survival and growth generalized linear mixed effect models that included tree 107
diameter at breast height (dbh) at the first census interval as a predictor variable and grid cell 108
as a random effect (see materials and methods). We then used the survival and growth 109
coefficients to fit size-dependent integral projection models (IPMs) and derive age-related 110
traits from size-dependent probabilities for each species within each grid cell (see materials 111
and methods) (15–17, 26–28). IPMs dynamically integrate size-dependent variability in 112
survival and growth as a continuous process, which allowed us to use cross-sectional data 113
over discrete time steps to make interspecific comparisons in how many years it takes trees to 114
attain key milestones in their life cycle. We parameterized our IPMs using methods 115
specifically developed for trees (27–29). Validations of IPM model outputs, relative to tree 116
ring data, showed this parameterization method can provide realistic estimates of tree age 117
demographics (27). 118
119
We used our species-specific IPMs and employed age-from-stage methods to calculate 120
several quantitative measures of growth, longevity, and stature. Specifically, we calculated 121
the number of years it takes for trees to grow from 10 to 20 cm in diameter (fig. S2, path a.2) 122
and grow from 10 cm to the 70 th quantile of their size distribution (fig. S2, path a.1) 123
(hereafter referred to as growth strategies). The 10 cm in diameter lower bound threshold was 124
chosen because it was the size at which point trees were consistently monitored across the 125
forest networks and the 70 th quantile threshold was chosen because it reflects a mature size at 126
which point trees have approached their ultimate position in the forest. We also calculated 127
two quantitative measures of tree longevity, including their average remaining life 128
expectancy from 10 cm in diameter and their maximal lifespan age (95% cohort mortality 129
from 10 cm), and a measure of maximal tree stature (size at maximal lifespan age) (fig. S2, 130
path b) (15–17). These mean estimates capture the pace of life for trees (growth, longevity, 131
and stature) based on observed climate conditions over the last century (derived from 132
dynamical data collected between 1926-2014, see materials and methods). 133
134
Our estimates of remaining life expectancy from 10 cm dbh range from 1.2 to 3,195 years, 135
with a mean value of 60 years in the tropics and 95 years in the extratropics (Fig 2a). This 136
trend matches our theoretical expectation of broadscale tree life history diversification 137
patterns (Fig. 1b) and confers with known tree longevity hot spots, whereby the oldest 138
recorded species occur in temperate conifer and boreal forests (12, 30). However, there was 139
also considerable overlap in the range of tree life expectancies across biomes (fig S3-S4), 140
table S2) and wide variability in how longevity relates to tree growth strategies and maximal 141
statures (Fig. 2b, fig S3-S4, and table S2). It is important to note that remaining life 142
expectancy from 10 cm dbh is a species-level mean estimate (i.e. is conditional on surviving 143
to 10 cm dbh). A low life expectancy, relative to the mean number of years it takes a species 144
to grow from 10-20 cm dbh, does not imply that no individuals will reach 20 cm dbh. Instead, 145
it implies that less than half of the individuals will survive to that size threshold. 146
147
Tree life history strategies do not strictly follow the fast-slow continuum (H1). 148
To test the expectation that trees align along the fast-slow continuum (Fig. 1a, H1), we first 149
examined univariate trait correlations and found moderate support for trade-offs between tree 150
growth, longevity, and stature (fig. S5). For example, the number of years it takes trees to 151
grow from 10-20 cm in diameter was positively correlated to life expectancy (Pearson 152
correlation = 0.22) and maximal lifespan age (Pearson correlation = 0.21). Similarly, 153
maximal tree size was positively related to life expectancy (Pearson correlation = 0.41). 154
Interestingly, the strength of these pairwise correlations also suggests that tree age 155
demographics do not strictly follow a single axis of variation along the fast-slow continuum 156
(i.e., the assumption that growth is tightly coupled and inversely related to longevity and 157
maximal stature). 158
159
To examine the multidimensionality of tree age demographics (Fig. 1A, H1), we analyzed the 160
variance-covariance matrix of tree growth, longevity, and stature using a principal component 161
analysis (PCA). Highly correlated traits that captured redundant trait information were 162
excluded from the PCA (fig. S5), resulting in the inclusion of tree growth strategies (i.e., 163
growth from 10 to 20 cm dbh and the 70 th quantile of their size distribution), life expectancy 164
from 10 cm dbh, and maximal tree size (fig. S5). The first PC axis captured 46% of the life 165
history trait variation and was heavily weighted by tree growth dynamics (i.e., years to 20 cm 166
dbh and the 70 th quantile size) (Fig. 2C). The PC loadings also showed that slow growth was 167
correlated with high life expectancy and large maximal size (table S3). The second axis 168
captured 28% of the trait variation. Interestingly, the directionality between the trait 169
correlations flipped, whereby slow growth was negatively correlated to life expectancy and 170
maximal size (table S3). The third axis was heavily weighted by tree life expectancy, with 171
high life expectancy being positively related to slow growth but negatively related to tree 172
maximal size (table S3). PCA analyses for tropical versus extratropical species retain 173
consistent patterns in the directionality of the trait correlations among the PC axes (table S3), 174
illustrating the modular and flexible nature of tree age demographics beyond the fast-slow 175
continuum within and among the Northern and Southern hemispheres. 176
177
To further contextualize how the variation in tree age demographics among the PC axes 178
shapes the overarching pace of life for trees, we used a K-means clustering algorithm to 179
group species into core demographic functional types (see material and methods subsection 3 180
and fig. S6). Using this clustering algorithm, which reduces the within-group sum of squares, 181
we found that fast-growing species aggregated into a single stature-longevity functional type 182
(Fig. 2C-2D, cluster 1). Conversely, conservative slow-growing species formed three distinct 183
clusters, including low, intermediate, and high stature-longevity functional types (Fig. 2C- 184
2D, clusters 2-4). The fast-growing species cluster matches the theoretical expectation of 185
ubiquitous resource limitations that constrain a species’ ability to maintain high growth and 186
high survival simultaneously, leading to low life expectancies and small maximal sizes (Fig. 187
2C-2D, cluster 1). Yet, the emergence of three distinct clusters for slow-growing species 188
suggests conservative trees are less constrained in their pace of life. At one end of these three 189
conservative growth trait clusters were species with high life expectancies but small maximal 190
sizes (Fig. 2C-2D, cluster 4), and at the other end were species with low life expectancies but 191
large maximal sizes (Fig. 2C-2D, cluster 3). Clustering analyses for tropical versus 192
extratropical species indicate that the tropics retain the full range of demographic functional 193
types (fig. S7, four distinct clusters), Conversely, the extratropical species group into two 194
demographic functional types of predominantly slow-growing conservative clusters (fig. S7, 195
two distinct clusters). Together, these results provide key insight into the core groups of 196
demographic functional types that shape the structural complexity and dynamics of tropical 197
versus extratropical forests. 198
199
200
201
Fig. 2. Visual illustration of tree growth-longevity-stature relationships and core demographic functional types. 202
The mean life expectancy is higher in the extratropics than in the tropics (A), with substantial variation between 203
tree growth strategies and life expectancies (B) (N=6,847 i.e., species X grid ID). The other trait relationships 204
are represented in fig. S8. The core growth-longevity-stature functional types are presented in C-D, which are 205
determined using the K-means clustering algorithm of the life history trait PC scores. PC weights and trait 206
correlations are reported in table S3. The frequency density (A) and the life history traits (B) are scaled by the 207
natural log. The axes for A-D are scaled by the natural log. Data points are species-specific and are calculated 208
using individual tree observations and size-dependent integral projection models (see materials and methods). 209
210
Our broadscale assessment of growth-longevity-stature relationships for a wide range of 211
species across the Americas is consistent with trends derived from tropical forest plots, which 212
found survival and growth rates over discrete size ranges differed substantially among 213
species and diminished as trees attained larger sizes (31–37). Similarly, while tree-ring data 214
showed that annual growth rates were negatively correlated with observed maximal ages 215
(12), there was more variation in observed maximal ages for species with fast versus slow 216
growth (12, 14). Together, these emergent patterns illustrate the modular and flexible nature 217
of trees that extend beyond the fast-slow continuum (Fig. 2C-2D, figs. S3-S4) and highlight 218
the tremendous variation in tree life expectancies across forest biomes (Fig 2A and figs. S3- 219
S4), with some of the oldest living species having a remaining life expectancy > 2000 years 220
(such as Tsuga heterophylla and Sequoia sempervirens). 221
222
Building on these foundational insights from predominantly tropical ecosystems, our results 223
provide a novel perspective that contributes to our fundamental understanding of tree age 224
demographics. By converting survival and growth rates over species life cycles to age-based 225
traits, our results provide insight into the time it takes trees to reach their ultimate positions in 226
the forest and their mean age at death (e.g., life expectancy). This allowed us to quantify the 227
pace of life for a wide range of species across the Americas and identify the core 228
demographic functional types more directly linked to carbon turnover. The emergence of the 229
slow-growth short-lifespan functional trait cluster is in line with previous research from 230
tropical forests, which showed that some short-stature trees had slow growth and low survival 231
(31, 32, 34, 38). This emergent trend may be an indication of maladapted species, or a 232
mediated effect of environmental disturbance (10, 32, 33). Conversely, it could be the result 233
of energetic investments in reproduction over species’ lifespans (net reproductive rate) that 234
we were not able to capture in our analysis (5, 11, 31, 34). Regardless of the mechanisms, 235
our findings provide a novel perspective on the multidimensionality of tree age demographics 236
for a wide range of phylogenetic and geographical groups. Furthermore, our finding of 237
emergent differences in the number of demographic functional types in the tropics versus 238
extratropics provides novel insight into the mechanisms that shape the dynamics and 239
structure of forests across the Americas. 240
241
Weak coordination in the strength of life history trade-offs across biogeographic gradients 242
(H2). 243
To test for emergent patterns in the strength of tree life history trade-offs across 244
biogeographic gradients (Fig. 1b, H2), we fit a multi-response Bayesian generalized mixed 245
effect model that included the first PC axis for each of three comprehensive sets of variables 246
related to soil, temperature, and precipitation as fixed effects and the phylogenetic relatedness 247
as a random effect (see materials and methods, table S4, figs S6-S8) (39). These abiotic 248
indexes were selected because they are known to strongly regulate photosynthetic capacity 249
and plant growth and are commonly assumed to induce life history trade-offs. This approach 250
allowed us to test for covariation in life history trait responses across soil, temperature, and 251
precipitation indexes and control for the effects of phylogenetic ancestry (40). These soil, 252
temperature, and precipitation variables were based on mean conditions from 1997-2013 (see 253
materials and methods, table S4), which overlap with the time window that our dynamical 254
tree data were collected. The expectation is that tree life history trade-offs are shaped by the 255
shared influence of abiotic factors and phylogenetic constraints, with colder temperatures and 256
lower resource availability pushing species toward the conservative end of the life history 257
trait spectrum (Fig. 1b, H2). 258
259
Our results show that there is a strong relationship between temperature and tree life history 260
traits, with colder temperatures being associated with conservative growth (𝛽 = -0.02, CI = (- 261
0.03, -0.01)) and high life expectancies (𝛽 = -0.07, CI = (-0.05, -0.08)) (Fig. 3 and fig. S12). 262
Conversely, our precipitation and soil indices had a weak effect on life history traits (fig. 12, 263
table S.5). Consistent with Amazon research (41), we found that tree life history traits were 264
phylogenetically conserved (Pagel’s 𝜇 ranging from 0.88-0.99, fig. S14 and table S6). Yet, 265
we also found low phenotypic correlations among our life history traits, indicating that the 266
strength of trade-offs between tree growth, longevity, and stature do not strongly co-vary 267
across biogeographic gradients (Fig. 1b, H2). For example, the phenotypic correlation 268
between the number of years it takes trees to grow to 20 cm dbh and their life expectancy 269
from 10 cm dbh was 0.18 (Fig. 3a). Together, these results show that, while tree life history 270
traits are phylogenetically conserved (∆ DIC null model versus phylo. model = 76832), 271
growth-longevity-stature relationships are not driven by genetic linkages or shared selective 272
pressures that act on both traits independently over evolutionary time across broad-scale 273
resource gradients (table S6) (42). 274
275
While our results offer the most comprehensive assessment of tree age demographics across 276
broadscale resource gradients, it is important to note the data gap in the subtropics (i.e., 277
across Mexico and northern Central America, Fig S1). This data gap could help explain the 278
noticeable difference in the range of life history trait strategies between the North American 279
temperate forests (low trait variation) and South American tropical forests (high trait 280
variation) (Fig 3B-3D and fig S1). This data gap highlights the need for increased sampling 281
efforts in these understudied regions of the world and should be a priority of future research 282
and funding. 283
284
Our findings are in line with trade-offs between physiological and morphological plant 285
features linked to individual fitness and life history evolution, one reflecting leaf economic 286
variables related to photosynthetic activity and growth potential and the other associated with 287
morphological features related to light competition and plant height (43–45). Yet, similar to 288
our results, the dominant axes of physiological and morphological plant features did not 289
strongly co-vary across latitudinal gradients (44, 45). Together, our findings and previous 290
research suggest that organismal function that supports rapid growth is not necessarily linked 291
to organismal function that results in lower life expectancies and small maximal sizes. These 292
emergent patterns suggest that rapid shifts in climate conditions may have divergent effects 293
on the relationship between biomass accumulation in tree growth and biomass retention in 294
tree longevity, with important implications for modeling the global carbon balance in a 295
changing world (46). 296
297
298
299
Fig. 3. Tree life history traits across our temperature index (PC axis 1 for a comprehensive list of temperature 300
variables, see materials and methods). Overall, we found low phenotypic correlations [variance-covariance of 301
standardized traits] among tree growth, longevity, and stature, suggesting there is weak support for coordinated 302
trade-offs over evolutionary time (i.e., organismal function that supports conservative growth does not 303
necessarily trade-off with organismal function needed to maintain high longevity) (A). We also find a strong 304
effect of temperature on tree life history traits (panels B-D), with little additional variation explained by soil or 305
precipitation (see figs. S12-S13 and table S5). The temperature gradient is derived from a principal component 306
analysis of nine temperature variables and represents a gradient from intermediate temperatures in the tropical 307
moist forest of the southern hemisphere to colder temperatures in the boreal north (from left to right of the x- 308
axis). The y-axis is scaled by the natural log. Data points are species-specific and are calculated using individual 309
tree observations to fit size-based integral projection models for each species within each grid cell ID (total of 310
1,127 species and 6,847 trait values) (see materials and methods). Model coefficients of the multi-response 311
Bayesian model are reported in fig. S12 and table S5). 312
313
Demographic diversity varies predictably across biogeographic gradients (H3). 314
To characterize the range of life history strategies that are expressed by trees across 315
broadscale biogeographic gradients, we first calculated the convex-hull volume in 316
demographic trait space within each grid cell (see materials and methods) (47) and compared 317
the relationship between the demographic trait diversity of forests and well-established 318
patterns in species richness. The convex-hull volume was calculated using the life history 319
trait PC scores for axes 1-3, which together captured 95% of the life history trait variation. 320
We then tested if the demographic trait diversity of forests varied predictably across 321
biogeographic gradients, and explored potential links between demographic trait diversity 322
and remotely sensed estimates of potential above-ground net primary productivity (NPP) 323
(Fig. 1c, H3, see materials and methods) (48). The expectation is that the diversity of life 324
history trait strategies that are expressed by trees should vary predictably across 325
biogeographic gradients, with higher demographic diversity being positively associated with 326
above-ground productivity. 327
328
Our results illustrate that the demographic trait diversity of forests follows well-established 329
patterns in species richness (Fig. 4a, adj R 2 =0.65, p < 0.001). We also found that the 330
demographic diversity of forests varied predictably across biogeographic gradients, with high 331
demographic trait diversity across warm tropical forests and low diversity of predominantly 332
slow-growing, long-lived species in the cold temperate and boreal forests (adj R 2 =0.40, p < 333
0.001, Fig. 4b and table S7). Lastly, we found a positive correlation between the demographic 334
diversity of forests and remotely sensed estimates of ecosystem productivity (Pearson 335
correlation = 0.71). 336
337
The emergence of a positive association between the demographic trait diversity and 338
ecosystem productivity is in line with two non-mutually exclusive hypotheses. From an 339
evolutionary perspective, ecosystem productivity is thought to drive species diversification 340
and niche differentiation (49). Conversely, following widely established relationships 341
between biodiversity and ecosystem function, more demographically diverse forests are 342
commonly assumed to have access to a larger resource pool and should thus be more 343
productive (50, 51). Here, we found moderate support for both hypotheses. Specifically, we 344
found that ecosystem productivity was predictive of demographic trait diversity across broad- 345
scale biogeographic gradients (adj R 2 = 0.49, p < 0.001, Fig. 4c, table S7). At the same time, 346
ecosystem productivity was jointly influenced by temperature (average marginal effect = 347
0.83, p =0.04, Fig. 4d) and demographic trait diversity (average marginal effect = 1.43, p < 348
0.001, Fig. 4d). This positive association was consistent across the tropics (adj R 2 =0.26, p < 349
0.01, table S7) and extra-tropics (adj R 2 = 0.84, p < 0.01, Fig. 4d, table S7). It should be noted 350
that NPP was strongly correlated with mean annual temperature (Pearson correlation = 0.94), 351
which did not allow us to explicitly test for the individual and combined effect of these 352
variables on demographic trait diversity. While our broadscale analysis does not establish 353
causality in the direction of these relationships, it does highlight the inextricable link between 354
demographic trait diversity and ecosystem productivity across forest biomes. 355
356
357
358
Fig. 4. The relationship between the demographic trait diversity of forests and ecosystem productivity (H3). We 359
find that the demographic trait diversity is positively related to species richness (A), with increasing 360
demographic trait diversity (i.e., convex-hull volume in life history trait space that is occupied by species) 361
across a mean annual temperature gradient (B). In line with two non-mutually exclusive hypotheses in 362
evolutionary biology and functional ecology, we find a positive association between demographic trait diversity 363
and above-ground net primary productivity (NPP) (C and D). It is important to note that NPP was based on 364
remotely sensed estimates and that these analyses do not establish causality in the directionality of this 365
relationship (C and D). The fully parameterized model in panel D includes the demographic trait diversity and 366
mean annual temperature. Demographic trait diversity and NPP were scaled to a mean of zero and a standard 367
deviation of one. Average marginal effects (AME) represent the response per unit increase for each predictor 368
variable. 369
370
The established association between demographic trait diversity and ecosystem productivity 371
is in line with emergent patterns derived from tropical forest plots, which found that the 372
demographic composition of forests was predictive of empirically derived measures of 373
above-ground carbon dynamics (32). Similarly, our findings match theoretical expectations 374
that the pace of life of organisms within a community (e.g., life expectancy and generation 375
time) should strongly regulate the relationship between carbon turnover (ecosystem fluxes) 376
and carbon retention (ecosystem pools) (52). It is important to note that the association 377
between demographic trait diversity and ecosystem productivity was derived from multi-year 378
averages in remotely sensed NPP from 1997-2013 and mean estimates of tree growth- 379
longevity-stature relationships based on the current distribution of species (i.e., derived from 380
dynamical data collected from the 1900s-2000s). This approach did not allow us to account 381
for potential biogeographic biases in the effects of human disturbance on species diversity 382
(i.e., between boreal and tropical forests). Yet, by quantifying the current distribution of 383
demographic functional types across broad-scale resource gradients, our results provide a 384
powerful backdrop for parameterizing next-generation vegetation models to simulate forest 385
carbon turnover rates across a range of current and future conditions. 386
387
More generally, our analysis offers strong empirical support for the expectation of high 388
demographic trait diversity in tropical forests compared to temperate and boreal forests. This 389
multi-biome finding supports the community assembly theory of strong abiotic filtering in 390
boreal regions, resulting in a restricted species pool of predominantly slow-growing, long- 391
lived species (Fig. 1c, H3). This emergent pattern is congruent with known variability in 392
physiological leaf trait characteristics across biogeographic gradients (43–45), with 393
decreasing variation in leaf economic traits from lower to higher latitudes (53). Similarly, our 394
results match well-established species richness–productivity relationships across global 395
forests (51, 54) and community structure-productivity relationships (55). Yet, while it makes 396
intuitive sense that the demographic diversity of forest communities follows well-established 397
patterns in species richness (49, 50), our findings establish a more direct link to the 398
demographic mechanisms that generate global variation in ecosystem carbon turnover. 399
400
Conclusion: 401
Our broad-scale analysis reveals the remarkable diversity of life history strategies that exist 402
for tree species across the Americas. Weak trade-offs between tree growth, longevity, and 403
stature across biogeographic gradients demonstrate the modular and flexible nature of trees, 404
highlighting the diversity of evolutionary trajectories that have arisen to address the 405
ecological puzzle of survival. In addition, from a functional perspective, we find that while 406
acquisitive trees sequester carbon at faster rates, they also generally appear constrained to 407
smaller maximum sizes and shorter lifespans that translate to lower carbon storage and faster 408
carbon turnover. More importantly, we find that more demographically diverse forests tend to 409
be more productive at the ecosystem scale across the tropics and extra-tropics. These findings 410
have important implications for informing global restoration and conservation efforts, and for 411
understanding the fundamental feedback between biodiversity and climate change mitigation.