Forests have a key role in global ecosystems, hosting much of the world’s terrestrial biodiversity and acting as a net sink for atmospheric carbon1. These and other ecosystem services that are provided by forests may be sensitive to climate change as well as climate variability on shorter time scales (for example, annual to decadal)2,3,4. Previous studies have documented responses of forest ecosystems to climate change and climate variability2,3,4, including drought-induced increases in tree mortality rates5. However, relationships between forest biomass, tree species composition and climate variability have not been quantified across a large region using systematically sampled data. Here we use systematic forest inventories from the 1980s and 2000s across the eastern USA to show that forest biomass responds to decadal-scale changes in water deficit, and that this biomass response is amplified by concurrent changes in community-mean drought tolerance, a functionally important aspect of tree species composition. The amplification of the direct effects of water stress on biomass occurs because water stress tends to induce a shift in tree species composition towards species that are more tolerant to drought but are slower growing. These results demonstrate concurrent changes in forest species composition and biomass carbon storage across a large, systematically sampled region, and highlight the potential for climate-induced changes in forest ecosystems across the world, resulting from both direct effects of climate on forest biomass and indirect effects mediated by shifts in species composition.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Bonan, G. B. Forests and climate change: forcings, feedbacks, and the climate benefits of forests. Science 320, 1444–1449 (2008)
Anderegg, W. R. L., Kane, J. M. & Anderegg, L. D. L. Consequences of widespread tree mortality triggered by drought and temperature stress. Nat. Clim. Chang. 3, 30–36 (2013)
Frank, D. et al. Effects of climate extremes on the terrestrial carbon cycle: concepts, processes and potential future impacts. Glob. Chang. Biol. 21, 2861–2880 (2015)
Clark, J. S. et al. The impacts of increasing drought on forest dynamics, structure, and biodiversity in the United States. Glob. Chang. Biol. 22, 2329–2352 (2016)
Allen, C. D. et al. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For. Ecol. Manage. 259, 660–684 (2010)
Davis, M. B. & Shaw, R. G. Range shifts and adaptive responses to Quaternary climate change. Science 292, 673–679 (2001)
Parmesan, C. & Yohe, G. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 37–42 (2003)
Sakschewski, B. et al. Resilience of Amazon forests emerges from plant trait diversity. Nat. Clim. Chang. 6, 1032–1036 (2016)
Farrior, C. E., Dybzinski, R., Levin, S. A. & Pacala, S. W. Competition for water and light in closed-canopy forests: a tractable model of carbon allocation with implications for carbon sinks. Am. Nat. 181, 314–330 (2013)
Peters, M. P., Iverson, L. R. & Matthews, S. N. Long-term droughtiness and drought tolerance of eastern US forests over five decades. For. Ecol. Manage. 345, 56–64 (2015)
Bechtold, W. A. & Patterson, P. L. (eds.) The Enhanced Forest Inventory and Analysis Program – National Sampling Design and Estimation Procedures. Gen. Tech. Rep. SRS-80 (US Department of Agriculture, 2005)
Houghton, R. A., Hackler, J. L. & Lawrence, K. T. The U.S. carbon budget: contributions from land-use change. Science 285, 574–578 (1999)
Nowacki, G. J. & Abrams, M. D. Is climate an important driver of post-European vegetation change in the Eastern United States? Glob. Chang. Biol. 21, 314–334 (2015)
Hicke, J. A. et al. Effects of biotic disturbances on forest carbon cycling in the United States and Canada. Glob. Chang. Biol. 18, 7–34 (2012)
Côté, S. D., Rooney, T. P., Tremblay, J.-P., Dussault, C. & Waller, D. M. Ecological impacts of deer overabundance. Annu. Rev. Ecol. Evol. Syst. 35, 113–147 (2004)
Vanderwel, M. C., Coomes, D. A. & Purves, D. W. Quantifying variation in forest disturbance, and its effects on aboveground biomass dynamics, across the eastern United States. Glob. Chang. Biol. 19, 1504–1517 (2013)
Fei, S. et al. Divergence of species responses to climate change. Sci. Adv. 3, e1603055 (2017)
Ruiz-Benito, P. et al. Climate- and successional-related changes in functional composition of European forests are strongly driven by tree mortality. Glob. Chang. Biol. 23, 4162–4176 (2017)
Sheffield, J., Wood, E. F. & Roderick, M. L. Little change in global drought over the past 60 years. Nature 491, 435–438 (2012)
Niinemets, Ü. & Valladares, F. Tolerance to shade, drought, and waterlogging of temperate northern hemisphere trees and shrubs. Ecol. Monogr. 76, 521–547 (2006)
Cornwell, W. K. & Grubb, P. J. Regional and local patterns in plant species richness with respect to resource availability. Oikos 100, 417–428 (2003)
Pinder, R. W. et al. Climate change impacts of US reactive nitrogen. Proc. Natl Acad. Sci. USA 109, 7671–7675 (2012)
Pan, Y. et al. A large and persistent carbon sink in the world’s forests. Science 333, 988–993 (2011)
Phillips, O. L. et al. Drought sensitivity of the Amazon rainforest. Science 323, 1344–1347 (2009)
Peng, C. et al. A drought-induced pervasive increase in tree mortality across Canada’s boreal forests. Nat. Clim. Chang. 1, 467–471 (2011)
IPCC. Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change (eds Field, C. B. et al.) (Cambridge Univ. Press, 2012)
Hanewinkel, M., Cullmann, D. A., Schelhaas, M.-J., Nabuurs, G.-J. & Zimmermann, N. E. Climate change may cause severe loss in the economic value of European forest land. Nat. Clim. Chang. 3, 203–207 (2013)
Dornelas, M. et al. Assemblage time series reveal biodiversity change but not systematic loss. Science 344, 296–299 (2014)
Hooper, D. U. et al. Effects of biodiversity on ecosystem functioning: a consensus of current knowledge. Ecol. Monogr. 75, 3–35 (2005)
Isbell, F. et al. Biodiversity increases the resistance of ecosystem productivity to climate extremes. Nature 526, 574–577 (2015)
Lichstein, J. W., Simons, T. R., Shriner, S. A. & Franzreb, K. E. Spatial autocorrelation and autoregressive models in ecology. Ecol. Monogr. 72, 445–463 (2002)
Bivand, R. S., Hauke, J. & Kossowski, T. Computing the Jacobian in Gaussian spatial autoregressive models: an illustrated comparison of available methods. Geogr. Anal. 45, 150–179 (2013)
Grace, J. B. Structural Equation Modeling and Natural Systems (Cambridge Univ. Press, 2006)
Chevan, A. & Sutherland, M. Hierarchical partitioning. Am. Stat. 45, 90–96 (1991)
Sheffield, J., Goteti, G. & Wood, E. F. Development of a 50-yr high-resolution global dataset of meteorological forcings for land surface modeling. J. Clim. 19, 3088–3111 (2006)
O’Connell, B. M . et al. The Forest Inventory and Analysis Database: Database Description and User Guide for Phase 2 v.5.1.5 (US Department of Agriculture, 2013)
Forest Inventory and Analysis National Core Field Guide. Volume 1: Field Data Collection Procedures for Phase 2 Plots, v.6.0 (US Department of Agriculture, 2012)
Jenkins, J. C., Chojnacky, D. C., Heath, L. S. & Birdsey, R. A. National-scale biomass estimators for United States tree species. For. Sci. 49, 12–35 (2003)
Rosseel, Y. lavaan: An R package for structural equation modeling. J. Stat. Softw. 48, (2012)
Funding was provided by US Department of Agriculture Forest Service agreements 11-JV-11242306-059 and 16-JV-11242306-050 to J.W.L., and by the European Regional Development Fund (Centre of Excellence EcolChange) and the Estonian Ministry of Science and Education (institutional grant IUT-8-3) to Ü.N.
The authors declare no competing financial interests.
Reviewer Information Nature thanks C. Schwalm and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Figure 1 PDSI and growing-season length in the 1980s and 2000s, and relationships between PDSI and precipitation in the eastern USA.
Maps show 1° latitude × 1° longitude grid cells. a–c, Mean growing-season PDSI (see Methods) for the 1980s (a), the 2000s (b) and the change in PDSI (ΔPDSI) in PDSI units (UPDSI) (c) between the two decades. d, Smoothed distribution of ΔPDSI; the vertical line is the mean, with the mean absolute change (MAC; mean of absolute values) indicated. e–g, Mean GSL for the 1980s (e), the 2000s (f) and the change in GSL (ΔGSL, months) (g) between the two decades. h, Smoothed distribution of ΔGSL; the vertical line is the mean, with the MAC indicated. i, The response of PDSI to precipitation change. The map shows regression slopes of ΔPDSIt (change in growing-season PDSI between successive years from 1948 to 2009) versus ΔPt (change in growing-season precipitation between successive years); slope units are PDSI units (UPDSI) per cm precipitation. The sample size for each regression (one regression per grid cell) is 61 (the number of annual changes from 1948–2009). j, Pearson’s correlation (r) between ΔPDSIt and ΔPt; samples sizes as in i. k, The percentage change in growing-season precipitation corresponding to a one-unit change in PDSI; percentages were calculated as 100 times the inverse of the slope from i divided by the mean (1948–2009) growing-season precipitation for each grid cell.
was calculated within 1° grid cells and 20-year stand-age classes that contained at least five FIA inventory plots that satisfied filtering criteria (see Methods). From left, the columns show values of for the 1980s, the 2000s, the change between the two decades and the smoothed distributions of changes (vertical lines show mean changes, with MAC indicated). Different stand-age classes are shown in each row as follows: a–d, 0–20 years; e–h, 20–40 years; i–l, 40–60 years; m–p, 60–80 years; q–t, 80–100 years. units (UDT) are on the scale of the species drought tolerance (DT) index, which increases from 1 (very intolerant) to 5 (very tolerant); see Methods and Supplementary Methods 1 for details.
AGB (Mg ha–1) was estimated from allometries and averaged within 1° grid cells and 20-year stand-age classes that contained at least five FIA inventory plots that satisfied filtering criteria (see Methods). From left, the columns show values of AGB for the 1980s, the 2000s, the change between the two decades and the smoothed distributions of changes (vertical lines show mean changes, with MAC indicated). Different stand-age classes are shown in each row as follows: a–d, 0–20 years; e–h, 20–40 years; i–l, 40–60 years; m–p, 60–80 years; q–t, 80–100 years.
a, Summary of plot ages and sample sizes by age class for FIA data used in SAR models and other analyses (Figs 2, 3 and Extended Data Figs 7, 8). The bottom row of the table refers to the number of remeasured inventory plots used in stand-scale analyses (d, e), in which individual plots were tracked over time (Supplementary Methods 4). All other rows in the table refer to grid-cell-scale analyses that control for stand age by comparing plots in a given age class in the 1980s to plots in the same grid cell and age class in the 2000s (b, c). The number of grid cells varies across age classes because a grid cell was only included in the analysis for a given age class if the grid cell included at least five plot records that met our filtering criteria in both the 1980s and 2000s (see Methods). b, SAR model slopes quantifying the mean grid-cell-scale response of (in DT units, UDT, in which DT increases with drought tolerance and ranges from 1 to 5) to ΔPDSI (in PDSI units, UPDSI) from the 1980s to 2000s. c, SAR model slopes quantifying the mean grid-cell-scale response of ΔAGB (Mg ha–1) to ΔPDSI and from the 1980s to 2000s. d, e, Similar to b, c, but for stand-level SAR models fit to remeasured plots (see Supplementary Methods 4 for details). All SAR slopes are partial regression coefficients that control for changes in growing-season length, changes in community-mean shade tolerance, and spatial autocorrelation (see Methods). Error bars are s.e.m. of slopes, with P values shown outside of the bars: *P ≤ 0.05; **P ≤ 0.01.
was calculated within 1° grid cells and 20-year stand-age classes that contained at least five FIA inventory plots that satisfied filtering criteria (see Methods). From left, the columns show values of for the 1980s, the 2000s, the change between the two decades and the smoothed distributions of changes (vertical lines show mean changes, with MAC indicated). Different stand-age classes are shown in each row as follows: a–d, 0–20 years; e–h, 20–40 years; i–l, 40–60 years; m–p, 60–80 years; q–t, 80–100 years. units (UST) are on the scale of the species shade tolerance (ST) index, which increases from 1 (very intolerant) to 5 (very tolerant); see Methods.
Extended Data Figure 6 Harvest intensity and mean drought tolerance of harvested trees in the 1980s and 2000s in the eastern USA.
Harvest intensity IH (per cent AGB harvested per year, units % yr–1) and (units UDT) were calculated within 1° grid cells and 20-year stand-age classes for analysis (Supplementary Methods 3), but are shown here in aggregated age classes (0–40 and 40–100 years) because patterns were similar among 20-year age classes within each aggregated class. a–p, From left, the columns show values of IH (a–h) or (i–p) for the 1980s, the 2000s, the change between the two decades, and the smoothed distributions of changes (vertical lines show mean changes, with MAC indicated). Different stand-age classes are shown in each row as follows: a–d, i–l, 0–40 years; e–h, m–p, 40–100 years. q, Slopes of versus ΔPDSI from four versions of spatial regression models with different tree- and plot-filtering criteria and different approaches to modelling harvest effects (see Supplementary Methods 3 for details). Model 1 corresponds to Extended Data Fig. 4b (note that only the three significant age classes from Extended Data Fig. 4b are presented here: 40–60, 60–80 and 80–100 years). The similar slopes and significance levels (P values are shown above each bar) from the four models suggest that the estimated response of to ΔPDSI is robust to including or excluding effects of tree harvest. Error bars are s.e.m. of slopes.
Extended Data Figure 7 Structural equation model and independent effects analysis results for the response variables and ΔAGB.
a, SEM structure for the model. b, SEM structure for the ΔAGB model. c, SEM structure for the alternative ΔAGB model that includes both direct and indirect effects of ΔPDSI on ΔAGB (see Methods and Extended Data Fig. 8a, b for further explanation of direct and indirect effects). d, Per cent contributions of ΔPDSI, ΔGSL and to the explained variation in . e, Per cent contributions of ΔPDSI, ΔGSL, and to the explained variation in ΔAGB (SEM results are for the model structure shown in b). Direct and indirect effects from c are reported in Extended Data Fig. 8d. Both SEM and IEA provide variance-partitioning estimates. In addition, SEM provides significance tests for explanatory variables. Significant P values (P ≤ 0.05) are shown in the parentheses below the corresponding percentage contribution, *P ≤ 0.05; **P ≤ 0.01. See Methods for details of SEM and IEA analyses. Positive and negative signs in d, e indicate the signs of the SEM and IEA coefficients, which are consistent with the signs of SAR model coefficients (Extended Data Figs 4 and 8).
Extended Data Figure 8 Conceptual model and supporting evidence for direct and indirect effects of changes in water availability on forest biomass.
a, Conceptual model of direct (solid arrow) and indirect (dashed arrows) effects of decreasing water availability (ΔPDSI < 0) on forest biomass. b, Conceptual model of direct (solid arrow) and indirect (dashed arrows) effects of increasing water availability (ΔPDSI > 0) on forest biomass. The conceptual model in a, b is supported by our results (c, d), which show that the response of forest biomass to ΔPDSI is amplified by indirect effects; for example, if water availability decreases (ΔPDSI < 0), then biomass decreases owing to direct effects (for example, physiological tree-level responses) as well as indirect effects (shifts in species composition towards more drought-tolerant but lower-biomass species). c, Slopes and P values from SAR models (equations (1) and (2)) for different stand-age classes (1° grid-cell-scale results, as in Extended Data Fig. 4b, c). Slopes are partial regression coefficients, so the slope labelled ‘response of AGB to ΔPDSI’ is the direct effect of ΔPDSI on ΔAGB (controlling for and other covariates), and the indirect effect is estimated by the product of the other two slopes (see Methods). d, Direct and indirect effects of ΔPDSI on AGB estimated from SEM (Extended Data Fig. 7c) and SAR models (see Methods). Sample sizes in these analyses (number of grid cells) are: 171, 247, 271, 271, and 193 for age classes from 0–20 to 80–100, respectively (as shown in Extended Data Fig. 4a). For consistency with SEM, SAR slopes are standardized in d (standard deviation units) but are otherwise equivalent to SAR coefficients in c and in Fig. 3. The percentages in d are calculated as 100 × indirect / (direct + indirect).
Extended Data Figure 9 Distributions of species influence on estimated responses of community-mean drought tolerance () to ΔPDSI, AGB to ΔPDSI, and AGB to .
Species influence (x axis) is the per cent change in SAR model slopes owing to including an individual species in the analysis (see Supplementary Methods 6, 7 for details). Probability density (y axis) is the relative frequency of species with a given per cent influence. Black curves are for all species, and red and blue curves are for gymnosperm and angiosperm species, respectively. Vertical lines are means. The distributions show that most species have little influence on the SAR slopes, some species have large influence, and gymnosperms and angiosperms have similar degrees of influence. From left, the columns show species influences on parameter dP from equation (1), and species influences on parameters aP and aD from equation (2). Different stand-age classes are shown in each row as follows: a–c, 0–20 years; d–f, 20–40 years; g–i, 40–60 years; j–l, 60–80 years; m–o, 80–100 years.
Extended Data Figure 10 Species abundance distributions in the northeastern, north-central and southeastern USA in the 1980s and 2000s.
a, Map of the three mentioned sub-regions of the USA. b–p, Species abundance distributions (number of species in different abundance intervals) for each age class (rows) in each sub-region (columns). Blue and red bars represent numbers of species in the 1980s and 2000s, respectively. Purple indicates overlap of the two bars. Abundance is quantified as the percentage of total above-ground biomass comprised by a given species in a given sub-region and age class. Abundance intervals are on a logarithmic (base-2) scale with the following lower limits: 2−6, 2−5, 2−4, …, 23% (that is, 0.015625%, 0.03125%, 0.0625%, 0.125%, 0.25%, 0.5%, 1%, 2%, 4% and 8%). Species with abundances of less than 2−6% (0.015625%) were excluded. Vertical lines are mean abundances of species with abundance ≥0.015625%. The number of common species (defined as species with an abundance ≥1%) is given for the 1980s (blue text) and 2000s (red text).
This file contains Supplementary Methods 1-7: (1) Description and evaluation of the DT index; (2) examples of Δ (̅DT ) in remeasured plots; (3) influence of tree harvesting on estimated Δ ("DT" ) ̅ responses; (4) stand-level analysis of remeasured inventory plots; (5) partitioning Δ ("DT" ) ̅ into component contributions; (6) species influences on Δ ("DT" ) ̅ responses; and (7) species influence on ΔAGB responses. (PDF 914 kb)
This file contains Supplementary Tables 1-5: Tables (one per forest age class) of species influence statistics. The statistics quantify how excluding (vs. including) each species affects the estimated slopes of Δ ("DT" ) ̅ vs. ΔPDSI, ΔAGB vs. ΔPDSI, and ΔAGB vs. Δ ("DT" ) ̅. Forest age classes are 0-20, 20-40, 40-60, 60-80, and 80-100 years. (PDF 827 kb)
This file contains Supplementary Tables 6-20: Tables of abundance change between the 1980s and 2000s for common species within each forest age class and each of the three sub-regions (north-central, northeastern, and southeastern USA). ‘Common species’ are those comprising at least 1% of AGB in a given sub-region and age class in either or both decades. (PDF 733 kb)
About this article
Cite this article
Zhang, T., Niinemets, Ü., Sheffield, J. et al. Shifts in tree functional composition amplify the response of forest biomass to climate. Nature 556, 99–102 (2018). https://doi.org/10.1038/nature26152
Low forest productivity associated with increasing drought‐tolerant species is compensated by an increase in drought‐tolerance richness
Global Change Biology (2021)
Extreme climate shifts pest dominance hierarchy through thermal evolution and transgenerational plasticity
Functional Ecology (2021)
Inventory of Forest Attributes to Support the Integration of Non-provisioning Ecosystem Services and Biodiversity into Forest Planning—from Collecting Data to Providing Information
Current Forestry Reports (2021)
Global Change Biology (2021)
Sensitivity to water stress drives differential decline and mortality dynamics of three co-occurring conifers with different drought tolerance
Forest Ecology and Management (2021)