Current Climate Change Reports (2021) 7:1–13

[Link]

ADVANCES AND FUTURE DIRECTIONS IN EARTH SYSTEM MODELLING (I SIMPSON, SECTION

EDITOR)

Trait-Based Modeling of Terrestrial Ecosystems: Advances

and Challenges Under Global Change
Xiangtao Xu 1 & Anna T. Trugman 2

Accepted: 24 November 2020 / Published online: 8 January 2021

# The Author(s), under exclusive licence to Springer Nature Switzerland AG part of Springer Nature 2021

Abstract
Purpose of Review We summarize the general structure of modern terrestrial ecosystem models and investigate how advances in
trait-based modeling approaches help to better constrain predictions for ecosystem sensitivity to global change.
Recent Findings In ecosystem models, empirical parameters are increasingly being replaced with plant physiological trait-based
parameters, which can be directly measured in the field. The needs to predict long-term terrestrial ecosystem dynamics under
climate change have spurred novel model developments including the representation of (i) vegetation processes across the critical
zone, (ii) wood and belowground ecophysiology, and (iii) the effects of physiological trait acclimation.
Summary Trait-based modeling of terrestrial ecosystems allows for the direct integration of measured plant ecophysiology with
model processes, increasing the potential to constrain uncertainty and improve predictions under novel climate regimes.
However, such increased model complexity requires careful model design, standardized intercomparisons, and benchmarking
for model responses to both climate extremes and long-term trends.

Keywords Terrestrial ecosystem model . Climate change . Plant functional trait . Trait acclimation . Optimality theory

Introduction a rapidly changing climate and increased anthropogenic dis-

turbance [5]. Despite the known importance of terrestrial pro-
Terrestrial ecosystems host enormous biological diversity and cesses in Earth system feedbacks, internal model uncertainty
provide vital ecosystem services to human and natural sys- associated with terrestrial ecosystem processes exceeds the
tems [1]. From a biogeochemical perspective, terrestrial eco- uncertainty associated with different climate scenarios in the
systems sequester approximately a quarter of anthropogenic current generation of ESMs, even at the decadal and longer
CO2 emission annually [2], contribute disproportionately to time scale [5], implying the paramount need to improve ter-
inter-annual variability in the global carbon cycle [3], and restrial ecosystem models.
exert diverse and complex climate feedbacks [4]. Therefore, Vegetation is the foundation of all terrestrial ecosystems
terrestrial ecosystem dynamics are a central component of and regulates the carbon-water-energy nexus as well as other
Earth system models (ESMs), which are a primary tool for biogeochemical cycles from local to global scales. Because of
understanding the dynamic responses of the Earth system to its foundational role as primary producer, vegetation is the
focus, and in many cases the only life form incorporated in
the ecosystem component of ESMs. Historically, ecosystem
This article is part of the Topical Collection on Advances and Future
Directions in Earth System Modelling
model development was motivated to represent biological and
Earth system processes at two contrasting scales. At the global
* Xiangtao Xu scale, the need to explain biogeographic patterns, biogeo-
xx286@[Link] chemical cycles, and their interactions with climate spurred
development of the very first generation of dynamic global
1
Department of Ecology and Evolutionary Biology, Cornell
vegetation models (DGVMs) [6]. DGVMs prognostically
University, Ithaca, NY 14850, USA simulate the distribution of vegetation plant functional types
2
Department of Geography, University of California, Santa Barbara,
(PFTs) instead of merely prescribing vegetation types from
Santa Barbara, CA 93106, USA land cover maps. In these early DGVMs, vegetation was
2 Curr Clim Change Rep (2021) 7:1–13

simplified as multiple “green blankets” (usually referred to as ecosystem models targeted at a wide-ranging audience from
big leaf models) with no explicit representation of ecosystem field-based ecophysiologists to Earth system scientists and
structures within a model grid cell (usually > 1° in spatial ecosystem modelers.
resolution). In contrast, at the stand scale, individual-based
models (IBMs or stochastic gap models) aim to predict forest
dynamics by explicitly simulating the fate of individual trees Key Process-Based Modules That Comprise
with consideration of stochasticity within a finite population Modern Terrestrial Ecosystem Models
[7, 8]. In recent decades, cohort-based models, such as the
ecosystem demography (ED) model, integrate the global Modern terrestrial ecosystem models consist of interacting
scope of big leaf models with a more detailed representation modules (here defined as relatively self-contained sub-
of vegetation demography drawn from IBMs into vegetation models) that reflect distinct ecophysiological processes and
classes or “cohorts” within terrestrial biosphere models [8, 9•]. their integral to capturing ecosystem dynamics at time scales
Such refinement of ecosystem structure in models benefits ranging from minutes to decades. Inevitably, these ecophysi-
from a more mechanistic representation of ecophysiological ological processes are tightly coupled with land surface pro-
processes, such as competition for light within a canopy. As a cesses such as carbon, water, nutrient, and energy balances,
result, model parameters that used to be highly empirical and forming an integrated terrestrial biosphere model (TBM) that
require calibration and/or tuning can now be directly connect- simulates terrestrial ecosystem dynamics and the associated
ed with field observations of plant physiological traits. physical, chemical, and biological processes at regional to
The development of ecosystem models toward trait-based global scales for coupling with ESMs. Here, we focus only
approaches over empirical functions has several advantages. on the plant physiological processes that occur at the individ-
First, trait-based models have the potential to be more accurate ual scale and the ecological interactions that occur between
when predicting ecosystem responses to unprecedented novel individuals. These ecosystem processes are nested within oth-
climate regimes because they better reflect the ecophysiolog- er land surface processes in the Earth system. In this section,
ical processes underlying ecosystem responses to climate var- we summarize the general module structure shared by most
iations [10•]. Thus, trait-based models can reduce model pre- mainstream ecosystem models (Fig. 1), as well as common
dictive error associated with physiological processes [11], as implementations of these modules. Subsequently, we discuss
well as the need for empirical model parameters or constraints novel developments and future research needs for trait-based
[12]. Second, trait-based models can better account for plant modeling.
functional diversity compared with the traditional
physiognomy-based PFTs (e.g., evergreen broad-leaf forest). Individual-Level Modules
Third, trait-based models allow for assimilation of the wealth
of global functional trait data to reduce model predictive error Resource acquisition is the foundational process in terrestrial
associated with parameter uncertainty [11, 13•]. Lastly, trait- ecosystems. The representation of resource acquisition in ter-
based approaches empower ecosystem models to become the restrial ecosystem models begins at the leaf-level with the
platform for numerical experiments that can help to generate photosynthesis-stomata module. The purpose of the photosyn-
and evaluate hypotheses in an iterative feedback loop between thesis module is to predict leaf-level carbon assimilation from
modelers and eco-physiologists. measurable biochemical parameters (or physiological traits)
Recent reviews on ecosystem modeling span vegetation and abiotic environment (such as intercellular CO2 concentra-
demographics [9•], the evolution of gap models [8], hydrody- tion, light, and temperature). The stomatal module bridges the
namics [14], and agricultural ecosystem modeling [15]. leaf intercellular space with canopy air spaces and
However, despite the rapid evolution and model development biophysically constrains carbon and water fluxes from the
in trait-based modeling, a review on advances and challenges aspect of gas diffusion. The most commonly used framework
of trait-based approaches in mechanistic terrestrial ecosystem in ecosystem models is the Faqhuar, von Caemmerer, and
modeling is lacking. Here, we aim to summarize (1) the key Berry system [29]. Collectively, the photosynthesis-stomata
ecophysiological processes in modern trait-based ecosystem modules form the foundation for mechanistic ecosystem pre-
models, drawing on developments from approximately the dictions under climate change. In addition to leaf-level photo-
last decade up to 2020, (2) the incorporation of new model synthesis, the rhythmic seasonal changes, or phenology, of
processes, physiological traits, and the associated predictive plant leaf area are also important for characterizing vegetation
benefits, particularly in the context of model predictions dur- carbon gain [30]. Generally, ecosystem models simulate sea-
ing a time of rapid global change, and (3) the trade-offs asso- sonal dynamics of leaf area using abiotic cues such as light,
ciated with more mechanistic, but also more complex ecosys- temperature, and water for deciduous trees while assuming a
tem models. Finally, we make recommendations as to the constant leaf area for evergreen trees [31]. In contrast to the
usage, interpretation, and future development of trait-based aboveground, ecosystem model representation of
Curr Clim Change Rep (2021) 7:1–13 3

Fig. 1 Key process-based modules (rectangles) within modern ecosystem belowground resource acquisition will directly exert limitation on
models, separated into individual-level physiological processes (green) aboveground resource acquisition using semi-empirical parameters.
that operate from minute to seasonal scales and community to Novel model developments of resource storage and circulation (dashed
ecosystem-level ecological processes (orange) that operate beyond green rectangle) can represent resource flows within plants in a more
seasonal times scales. Life history events bridge the two scales and are mechanistic way and interfaces with both resource acquisition and
thus colored by both green and orange. Arrows denote the dominant allocation. Select trait-based parameters in each module as summarized
direction of interactions between modules. In most ecosystem models, in Table 1 are shown in red circles and ellipses

belowground resource usage is relatively simple due to limited defense investment, including bark growth [33] and VOC
data availability and mechanistic knowledge on belowground emissions [34].
processes. Key belowground processes, including water and Resource acquisition and allocation ultimately determine
nutrient acquisition (if applicable), are usually represented key plant life history events such as recruitment and mortality,
using empirical functions dictated by belowground plant root bridging individual- and ecosystem-scale processes (Fig. 1)
biomass [32]. and enabling the prediction of long-term ecosystem dynamics
Resource allocation is coupled to resource acquisition and [35]. In traditional big leaf ecosystem models without vegeta-
describes how plants distribute resources, particularly carbon tion demographics, seedling recruitment is absent, and new
derived from photosynthates, among different tissues, includ- plant individuals are added to a common biomass pool of a
ing leaves, stems, roots, and nonstructural carbohydrate re- given PFT within a given grid cell. Thus, recruitment is anal-
serves (NSC). In ecosystem models, the carbon gained by an ogous to growth in that both increase the PFT biomass pool.
individual through photosynthesis, or the gross primary pro- The introduction of vegetation demographics into modern
duction (GPP), is allocated based on a distinct hierarchy of ecosystem models [9•] requires explicit consideration of veg-
priorities for different physiological processes. First, carbon is etation recruitment and its sensitivity to different environmen-
used to satisfy processes which maintain plant activity includ- tal factors because the states of the seemingly insignificant
ing autotrophic respiratory costs of different organs (leaves, understory seedlings have the potential to determine ecosys-
stems, and roots), and necessary tissue turnover from pheno- tem structure and composition decades in the future [36].
logical processes or disturbance. Extra carbon is then allocated Mortality, and especially mortality of large individuals, can
to structural and reproductive growth following allometric re- strongly affect vegetation carbon storage in the Earth system
lationships [10•]. At the cutting edge of ecosystem model [37]. However, the mortality module in ecosystem models has
development, some models have begun to incorporate plant been recognized as one of the most mechanistically uncertain
4 Curr Clim Change Rep (2021) 7:1–13

processes, the impacts of which are especially apparent for across life forms, few ecosystem models mechanistically in-
predictions over long (decades to centuries) time scales clude these interactions (but see [48] for representing parasitic
[38•]. Most ecosystem models include a background mortality lianas in the ED2 model).
rate, which is insensitive to environmental stressors and stand Heterogeneity due to variations in ecosystem structure rep-
density and is usually constrained by observed maximum tree resents another layer of complexity that affects ecosystem
size or age (in models with vegetation demography) [38•, 39]. dynamics in conjunction with biodiversity. Variations in abi-
Stress-driven mortality is usually limited to mortality driven otic and biotic conditions like tree size distribution, soil type,
by carbon starvation based on a variety of metrics ranging and topography regulate the micro-environment for individ-
from net primary productivity, growth rates, or plant carbon uals [49, 50]. It is thus critical to represent within-ecosystem
balance [38•]. In practice, carbon starvation mortality in eco- heterogeneity because some ecological and ecophysiological
system models has mostly been optimized for systems where processes, like photosynthesis and plant hydraulics, scale non-
the stress of light competition dominates, rather than other linearly within the canopy or among individuals [9•, 51].
stressors such as water or nutrient stress [40]. Ecosystem mod- Failing to capture these heterogeneous effects on ecosystem
el mortality modules may also represent mortality from other dynamics in an ecosystem model can result in cascading ef-
biotic stresses such as low temperature, hydraulic failure, and fects and diverging predictions for ecosystem structure and
insect attack [41], which can be more dominant processes function over decadal scales [52, 53].
during extreme events [42]. To account for sub-grid scale heterogeneity, modern eco-
system models are increasingly incorporating vegetation de-
Ecosystem-Level Modules mography (i.e., explicit representation of tree size distribu-
tions and the associated competition for resources, see [9•]),
Ecosystem-level processes can be organized into two inter- enabling for the prediction of nonlinear dynamics associated
connected ecological themes: biodiversity and heterogeneity with light competition within a forest canopy and the physio-
(Fig. 1). Biodiversity across multiple scales strongly shapes logical constraints of vegetation hydraulics that scale with tree
ecosystem function and sensitivity to environment [43]. In size (to name a few processes). Vegetation demography,
fact, ecosystem diversity is both a key motivation and a grand which is controlled by the self-organized individual-level re-
challenge for improving ecosystem dynamics within Earth source competition [54], is also strongly shaped by natural and
system models [5]. The most common method for anthropogenic disturbances. We subsequently review three
representing biodiversity in ecosystem models is through mul- major disturbance types included in ecosystem models, in-
tiple co-occurring PFTs, a conceptual group of species that cluding fire, biotic disturbance, and anthropogenic
perform similar functions in the ecosystem. The level of eco- disturbance.
system complexity can range from relatively simplistic, such Among disturbance types, fire is the most commonly rep-
as a tree PFT shading a grass PFT, to more sophisticated, such resented in ecosystem models [41, 55–57] due to its large
as competition between trees of different size of the same PFT impact on not only vegetation but also other Earth system
and between different PFTs in demography-enabled models. processes (e.g., carbon, water, and energy fluxes). The level
Depending on the model, competition for resources may in- of detail and complexity varies widely across fire models, but
clude light, water, nutrients, or multiple co-limiting resources. some models have even included the impact of functional
PFTs are often defined by life forms and physiognomy (e.g., traits such as bark thickness on plant survival with fire [33,
deciduous broadleaf trees and evergreen needleleaf trees) with 58]. In fact, the importance and complexity of fire dynamics
biome boundaries constrained by empirical climate envelopes have led to the development of a number of stand-alone fire
[12]. This approach has the advantage of easy parameteriza- models [57]. Though the inclusion of fire modules within
tion and is readily able to be integrated with landscape-scale ecosystem models is by no means a new development, it is
observations. However, the lack of (i) mechanistic represen- still an active area of research and validation. For example,
tations of biome shifts [12] and (ii) functional diversity within none of the ecosystem models used in climate change predic-
each biome makes it impossible to robustly simulate ecosys- tions were able to capture the extent of the severe fire season
tems under novel climate regimes and thus difficult to predict in Australia in 2019 [59], and a fire model intercomparison
ecosystem resilience to global change [44]. Therefore, recent has not yet been included in the Coupled Model
modeling studies tend to define multiple competing PFTs fol- Intercomparison Project (CMIP) as of Phase 6 [60].
lowing key plant functional and demographic trait trade-offs, Other natural disturbance types, such as tree windthrow,
especially in hyper-diverse tropical forests [45–47]. In addi- are usually modeled as prescribed events or constant distur-
tion to competition among plants, there are other ecological bance rates (e.g., [41]) with increasing efforts toward more
interactions among diverse life forms such as symbiosis, pre- mechanistic representations (e.g., wind [61] and lightning
dation, and parasitism. Due to limited observations and mech- [62]). Meanwhile, few modeling studies explicitly include bi-
anistic understanding of complex ecological interactions otic disturbances from herbivores and insect attacks other than
Curr Clim Change Rep (2021) 7:1–13 5

by treating them as prescribed events that modify ecosystem is tightly coupled to stem and belowground processes.
states empirically [63, 64]. Finally, anthropogenic distur- Moving forward, it is widely acknowledged that a better rep-
bances such as logging and land use changes are usually in- resentation of the integrated critical zone, and in particular
troduced as external forcing to ecosystem models and are belowground processes, should be one priority in model de-
based on historical land use data sets and stand-alone land velopment [32, 79, 80].
cover and land use models [65]. Wood biology has also been long under-represented in
A common trend of model development is to replace em- ecosystem models [79]. Woody stems regulate the transport
pirical parameters and functions with mechanistic trait-driven of water, sugar, and nutrients through formation and mainte-
approaches (see Table 1). Though this trait-based approach nance of xylem and phloem, thus comprising a fundamental
has the potential to improve ecosystem model predictions, in part of organismal metabolism. In contrast, in ecosystem
practice, biological systems are extremely complex and bio- models, stems are generally represented as a passive biomass
logical traits are more variable and dynamic than their physi- pool dictated by empirical allometric equations [81].
cal parameter counterparts used in ecosystem models. In the However, an increasing realistic representation of wood biol-
next two sections, we summarize the progress and challenges ogy is becoming a more standard module in ecosystem models
toward representing biological complexity, and specifically (Fig. 1). For example, xylem vessels in the stem transport
hone in on temporal variations of trait-based parameters, to water from roots to canopy and thus link resource acquisition
attain realistic predictions at decadal or longer time scales. from both the above- and belowground. Water transport in the
xylem can become the major bottleneck for plant water use,
carbon assimilation, and survival even with adequate soil wa-
Beyond the Canopy: Integrating Plant ter supply [82]. While plant water stress functions and param-
Function Across the Critical Zone eters are highly uncertain among ecosystem models [40, 83],
incorporating hydrodynamics along the soil-plant-atmosphere
Vegetation function connects natural processes across the crit- continuum allows for a more mechanistic simulation of plant
ical zone (Earth’s permeable near-surface layer) from tens of water stress for both acquisition, resource allocation, and plant
meters above the ground at the top of forests’ canopies, all the life history (see [14, 84] for excellent reviews). Recently, the
way to the deep groundwater tapped by tree roots. Historical incorporation of plant hydraulics has received significant at-
ecosystem model development focused mainly on canopy tention in the ecosystem modeling community [46, 85–87],
processes [6, 31, 66–68] for several reasons. First, canopy fueled by the increasing availability for key hydraulic traits
leaves exert the first-order control on plant resource acquisi- such as xylem hydraulic conductivity and vulnerability [88].
tion and ecosystem carbon and water dynamics. Second, can- In hydraulically enabled ecosystem models, the downregula-
opy states and dynamics such as leaf area index (LAI), leaf tion or catastrophic failure of plant physiological activities
carbon assimilation rates, and evapotranspiration are relatively under water stressed conditions where plant water potentials
easier to measure across scales with the increasing prevalence (a diagnostic of plant water status) become critically negative
of high resolution environmental remote sensing products [69, (see Table 1) is explicitly represented. Importantly, the phys-
70] compared to belowground processes. iological traits governing plant responses to water stress can
Recent advances in canopy ecophysiology have led to be directly measured and mapped spatially with increasing
more mechanistic and trait-based canopy models, although capacity [85, 88–90].
many of them are still experimental and have not been inte- Wood biology also affects tree carbon sink capacity (respi-
grated into ecosystem models as standard components. Some ration costs and growth activity, in contrast carbon source
examples of these experimental new developments include capacity, which is regulated by photosynthetic carbon supply).
the representation of triose phosphate use (TPU) limitation As described in the last section, resource allocation in ecosys-
to photosynthesis [29], optimality-based stomatal conduc- tem models is tightly coupled to GPP (carbon source capaci-
tance models [71], and optimality-based ecohydrological ty), and thus variability in growth and other investments al-
equilibrium models [72]. Another avenue of active research ways tracks variability in GPP at all time scales.
pertains scientists’ physiological understanding of leaf func- However, this modeling framework contradicts field ob-
tion (such as maximum photosynthetic rate), and how leaf servations on the seasonal and inter-annual variability in
function varies with phenological stage due to ontogenetic woody growth rates [91, 92]. Therefore, ecophysiologists
changes [73], especially in tropical evergreen forests with have begun experimenting with sink-driven frameworks
mild climatic seasonality [74]. Modeling experiments suggest by representing growth needs as a function of environ-
that these phenological processes can substantially change mental factors [93]. A sink-centric representation may
ecosystem carbon budgets [75–78]. change the sensitivity of ecosystem carbon drawdown
The canopy-centric view in ecosystem models has increas- and storage to climate change. To better integrate the pro-
ingly been recognized as myopic, given that canopy function cess of sink-driving growth, ecosystem models need to
6 Curr Clim Change Rep (2021) 7:1–13

Table 1 Summary of select trait-based parameters for ecophysiological processes in ecosystem models

Trait-based parameter groups Label in Parameter Abiotic drivers of Notes

Fig. 1 uncertainty* trait acclimation**

Photosynthetic capacity A1 Low Temperature e.g., Vcmax, Jmax, TPUmax

Light Can vary due to leaf aging and leaf position in the canopy
[16, 17]
Temperature sensitivity of A2 Medium Temperature e.g., Q10 or activation energy for Vcmax and Jmax
photosynthetic capacity [18]
Stomatal sensitivity A3 Medium CO2 [19] e.g., stomatal slope in traditional stomatal models and marginal
water use efficiency in stomatal optimality models
Thresholds for leaf flushing and A4 Medium Unknown e.g., thresholds for growing degree days, photoperiod, and soil
shedding moisture
Leaf structural traits A5 Medium Light [17, 20] e.g., leaf mass per area, leaf size
Root biomass spatial distribution B1 Medium-high Soil moisture and e.g., maximum depth, lateral and vertical distribution
nutrient supply
[21]
Root uptake potential for nutrients B2 High Unknown Mainly for nitrogen and phosphorus
Root structural traits B3 High Unknown e.g., root specific length/area
Root mycorrhizae association B4 Medium Unknown e.g., arbuscular mycorrhizae, ecto-mycorrhizae, and nitrogen
fixation
Root/shoot ratio C1 Medium-high Moisture [22]
Stoichiometry C2 Medium-high Unknown e.g., C:N, C:P ratios
Biomass allometry C3 Medium-high Light [23] Aboveground allometry is relatively well characterized but
Moisture [24] belowground allometry has high uncertainty
Wood density C4 Low-medium Unknown Determines carbon density of a given stem volume
Respiration rates and temperature C5 Low-medium Temperature [25] Uncertainty of leaf respiration is relatively low while wood and
sensitivity root respiration have limited data
Tissue turnover rates (or C6 Medium-high Unknown Leaf turnover rates can be constrained by leaf longevity while
maintenance cost) branch turnover is highly uncertain
Carbon partitioning between C7 High Unknown Can be size dependent
reproductive and structural
growth
Hydraulic conductance H1 Medium CO2 [26] Including conductivity and sapwood area
Hydraulic vulnerability H2 Medium Water [27] e.g., Turgor loss point and P50 (water potential with 50% loss of
conductivity)
Hydraulic Capacitance H3 Medium Unknown Include saturated water content and other parameters from
pressure-volume curves
Allocation and turnover rates of S1 High Unknown
non-structural carbohydrates
Seed germination, dispersal, and R1 High Unknown
seedling survival
Background mortality rates M1 Medium Unknown Independent from plant density
Density-dependent mortality rates M2 Medium Unknown Mortality due to carbon stress
Mortality sensitivity to biotic stress M3 High Unknown e.g., mortality threshold for hydraulic failure, flooding, fire, and
and natural disturbance wind

*Check [28] for a more quantitative analysis of parameter uncertainty

**Note that trait acclimation is often not incorporated in ecosystem models
To improve readability, we group closely related parameters and relate them to the relevant modules in Fig. 1. We qualitatively assess the uncertainty of
each parameter group using a three-tier classification: (i) traits with little or well-characterized variability across scales are labeled as low parameter
uncertainty; (ii) traits with a qualitative understanding of variability but that are less well constrained than (i) are labeled as medium parameter
uncertainty; (iii) traits with very limited understanding of variability are labeled as high parameter uncertainty. Finally, we identify the key abiotic
factors that have been found to be important in trait acclimation from the literature
Curr Clim Change Rep (2021) 7:1–13 7

simulate non-structural carbohydrates and other more la- for 10–20% total biomass in moist forests and up to
bile and flexible resource storage pools [94] so that the 50–90% in dry and grassy ecosystems [103].
source and sink capacities are partially decoupled. Overlooking structural root carbon pools or assuming a
Finally, woody respiration is generally overlooked or high- constant root/shoot ratio can lead to biases in carbon
ly empirical in ecosystem models (e.g., implicitly included by storage estimation in ecosystem models.
a growth efficiency parameter [56]). In contrast, physiologists In sum, the expansion of trait-based processes beyond the
and ecologists have found that wood respiration can comprise canopy is shown to improve model performance [46, 85, 102]
one-third of canopy leaf respiration [95]. Thus, partitioning and significantly change long-term model predictions [98,
wood respiration into maintenance and growth-driven compo- 100•] while comprehensively evaluating the effects of incor-
nents with explicit linkages with plant size [96] would be porating these processes remain challenging at regional to
helpful to constrain ecosystem carbon budgets in future eco- global scale because of the increased model complexity and
system model developments. parametric uncertainty (Table 1). Moving forward, under-
Similar to woody stem processes, plant roots and the rhi- standing how trait-based parameters across plant organs are
zosphere have long been represented in a simplistic manner or related remains a critical area of research. Indeed, some stud-
are totally absent in many ecosystem models, largely due to ies have shown strong trait correlations driven by either coor-
serious data limitations for the belowground [32]. However, dination or trade-offs to form resource usage niches within a
there is strong interest across fields including plant ecologists, given plant organ ( [106–108]; but see [88] for a lack of trade-
microbiologists, biogeochemists, and modelers to improve off for hydraulic traits). However, within-species, among-spe-
understanding and modeling of the belowground [32, 97]. cies, and cross-site patterns can differ significantly [109], and
For example, in a recent experimental ecosystem model de- in some cases, coordination along the economic spectra is
velopment, the coupling of vegetation dynamics and rhizo- weak or absent [110]. Since trait assemblages represent a
sphere root-microbe nutrient competition enabled a more re- unique resource use strategy that may underlie distinctive re-
alistic representation of soil nutrient limitation on biomass sponses to climate change, either overstating or understating
storage when compared to observed ecosystem responses such trait correlations can lead to potentially large biases for
[98]. Further, ecosystem models that include a representation future ecosystem predictions. This biological complexity can
of major plant-microbe symbiosis (e.g., facultative N-fixation, result in uncertainty in ecosystem model predictions and sub-
mycorrhizal association) predict higher vegetation carbon sequent compensating errors when validating model predic-
storage compared to models where these processes are absent tions against present-day observations. Although the uncer-
[99, 100•]. These two model examples demonstrate the poten- tainty associated with unknowns in the physiology can be
tial importance of the belowground on terrestrial carbon cy- partly mitigated through new techniques such as integrating
cling and serve as a platform for future research. trait and ecosystem dynamics observations into ecosystem
In addition to nutrients, water uptake is also highly depen- models using Bayesian frameworks [13•], overcoming the
dent on root structure and traits, especially in arid and semi- biological complexity challenge ultimately requires novel data
arid ecosystems. Important physiological attributes include collection [69], open and effective data sharing [111], and
rooting depth, root spatial distribution, and their inter- careful data interpretation drawing on ecological and evolu-
specific variations [101]. Understanding both within- and tionary theory.
between-species variations has been recognized as increasing-
ly valuable given the introduction of vegetation demography
into ecosystem models because water stress responses can be
size-dependent [102]. While most models assume an expo- Acclimation and Adaptation: Trait Plasticity
nential distribution of root biomass with soil depth following to Environmental Changes
ecosystem-level observations [103] and ecohydrological the-
ory [104], the root distribution and its absolute rooting depth Plant traits are normally assimilated into ecosystem models as
can be extremely variable depending on local hydrological fixed parameters for each PFT. However, plant traits can vary
conditions and species composition [102, 105]. Thus widely in space and time, even within the same species [109,
root morphology and root water access can contribute 112–114]. In fact, many key ecosystem model parameters
to the predicted uncertainty in plant water stress across respond to abiotic environmental changes (Table 1) such as
space and time. Experimental model developments using temperature, moisture, light, nutrients, and CO2 through accli-
a dynamic root hydraulic framework where water avail- mation (within a life cycle) and adaptation (across many gen-
ability dictates plant root investment have recently been erations) of plant physiological traits. Representing such trait
tested in some water-limited ecosystems [21]; however, evolution is another priority for trait-based ecosystem models
this is still an area of active research. Finally, structural [115] because trait acclimation and adaptation may greatly
roots can also act as a large biomass storage, accounting modify biosphere-climate feedback [116]. In this section, we
8 Curr Clim Change Rep (2021) 7:1–13

review the recent progress and challenges in modeling key plants are super plastic and adaptive) for ecosystem dynamics
trait plasticity responses to environmental changes. under global change. Another enticing feature of optimality-
The most widely studied trait plasticity response is thermal based models is the compression of trait-based parameters into
acclimation of plant photosynthetic traits, such as the maxi- several key factors controlling coordination and trade-offs
mum rate of carboxylation (a photosynthetic trait) and leaf [121•]. However, understanding the time scale of reaching
dark respiration under warming [114, 116, 117]. In ecosystem optimal equilibrium is a crucial consideration when aspiring
models, the temperature sensitivity of photosynthetic param- to attain realistic future predictions, especially considering the
eters is typically represented through literature-constrained changing disturbance regimes and increasing frequency of
Q10 relationships representing enzymatic sensitivity to tem- climate extremes. As an example, the novel P-model built
perature. However, these Q10 relationships in ecosystem on the optimality rules was able to capture average spatiotem-
models include no acclimation or adaptation to changes in poral ecosystem carbon fluxes but showed substantial biases
temperature with routine exposure. In contrast, meta- during climate anomalies without additional constraints [123].
analyses of spatial variations in plant traits provide evidence Overall, using optimality theory to constrain spatiotemporal
for significant temperature acclimation in the sensitivity of variability in trait-based parameters holds significant potential
photosynthetic [18] and respiratory [118] parameters. In addi- for both improving the next-generation ecosystem models and
tion, warming experiments [114, 119] and optimality theory enhancing our understanding as ecologists and physiologists
[16, 120] further support meta-analysis diagnosed acclimation as to the variations and limits of trait acclimation and adapta-
trends. Recent physiological evidence indicates that such trait- tion. However, such optimality approaches can only be suc-
temperature acclimation can happen as quickly as 2 weeks cessful with careful model benchmarks for both long-term
[114] and counters enzymatic sensitivity such that photosyn- trends and responses to climate extremes [124, 125].
thesis and respiration rates are relatively stable across a gradi-
ent of growth temperatures, in contrast to the strong tempera-
ture sensitivity currently represented in ecosystem models [16,
120]. As a result, if thermal acclimation processes are included Conclusion
in ecosystem models, projections for terrestrial carbon seques-
tration and storage may increase [117], depending on the rel- Ecosystem models serve as our main tool for predicting the
ative change in sensitivity of photosynthesis versus future fates of our terrestrial ecosystems, and for informing
respiration. effective mitigation plans to prevent catastrophic turning
Another well-documented plasticity for plant biochemical, points in ecosystem structure and function under the ongoing
structural, and allometric traits is driven by light availability rapid climate change [5, 115, 126–128]. While near-term var-
[17, 20, 113]. However, the extent to which these plasticities iability in ecosystem dynamics can be successfully forecasted
are represented is generally confined to photosynthetic accli- with minimalistic empirical models, such as data-driven
mation in sunlit versus shade leaves [67], and the full impli- models based on empirical relationships and remote sensing
cations of the light-driven plasticity (e.g., changes in leaf mass products [129], projecting long-term evolution of terrestrial
per area, leaf longevity, and allometry as presented in [17]) ecosystems under novel climate regimes requires ecosystem
under global change are not well constrained. Aside from models that consider ecophysiological and evolutionary pro-
temperature and light, increasing CO2 and changing moisture cesses. Our review summarizes the recent progress in ecosys-
conditions can also influence the slope parameters in stomatal tem models toward a more explicit representation of physio-
conductance modules [19] and plant hydraulic traits [26, 27]. logical processes using plant functional traits and lays out
In contrast to acclimation that can happen in a relatively major challenges for future trait-based model development
short period of time, genetic adaptation takes multiple gener- (summarized below).
ations and thus is hard to measure and model with few at-
tempts to account for mutation and adaptation in ecosystem 1. Development of trait-based models often increases the
models [115]. The physiology of adaptation of these processes complexity of model structures by increasing the number
has not yet been constrained to the point where it is feasible to of physiological and ecological processes (Fig. 1). While
include a representation in ecosystem models. such complexity may help to make more rigorous model
Given our limited understanding of the physiology of ac- predictions of ecosystem functions (such as productivity,
climation and adaptation, an alternate approach is to use opti- evapotranspiration, and biomass), it can hinder effective
mality theory to estimate the best set of traits from an eco- interpretation of model intercomparisons, especially when
evolutionary perspective [121•]. Coupled with evolutionary the comparisons only focus on high-level model outputs.
stable strategy analyses, optimality theory can help to generate As a community, we must carefully consider how we can
the equilibrium sets of traits for an ecosystem [122], which in gain knowledge about ecosystem sensitivity to global
turn can constrain a best-case scenario prediction (where change from a range of complex models with different
Curr Clim Change Rep (2021) 7:1–13 9

assumptions of ecosystem dynamical processes that are biology. These types of measurements need to be priori-
often not apparent to the user of model products. ties for future data collections.
2. Compared with the traditional physiognomy-based PFT 3. Constraining model sensitivity to trait-based parameters
configuration, trait-based models require a more detailed under different scenarios (e.g., long-term climate change
representation of trait diversity for canopy leaves and in- and extreme events) is critical when prioritizing data
creasingly stems and roots (Table 1). As a community, we needs for model parameterization. Such analyses are pos-
must effectively incorporate plant functional diversity in sible using Bayesian data assimilation system such as the
trait-based models while maintaining model parsimony Predictive Ecosystem Analyzer (PEcAN, [11, 13•]) and
and identify priority datasets of traits and observations would benefit significantly from coordination and integra-
on vegetation dynamics for model parameterization and tion with ecosystem experiments [124, 125].
benchmarking.
3. In the context of climate change, the role of trait acclima- Acknowledgments X.X. acknowledges funding from Cornell CALS.
A.T.T acknowledges funding from the USDA National Institute of
tion and adaptation in determining ecosystem sensitivity
Food and Agriculture, Agricultural and Food Research Initiative
to various environmental factors has been identified as Competitive Programme Grant No. 2018-67012-31496, the University
important, but more data is needed to parameterize eco- of California Laboratory Fees Research Program Award No. LFR-20-
system models. As a community, to advance this difficult 652467, and the NSF Grant 2003205. We also thank two anonymous
reviewers for providing insightful suggestions to improve the manuscript.
problem, we need to identify what the key trait plasticities
are to long-term climate regime shifts, at which point we
can move towards determining how to incorporate these Compliance with Ethical Standards
plasticities, while still capturing ecosystem dynamics to
Conflict of Interest On behalf of all authors, the corresponding author
short-term environmental perturbations. states that there is no conflict of interest.

Comprehensively addressing these three challenges re-

quires close collaboration between modelers, ecosystem ecol-
ogists, ecophysiologists, evolutionary biologists, and climate References
scientists. Here, we make the following recommendations for
the usage, future development, and applications of trait-based Papers of particular interest, published recently, have been
ecosystem models. highlighted as:
• Of importance
1. Conducting standardized intercomparison studies for sub-
modules (e.g., [29]) with homogenized inputs in addition 1. Millennium Ecosystem Assessment. Ecosystems and human well-
being: synthesis: Island press United States of America; 2005.
to ecosystem models as a whole is critical to understand- 2. Friedlingstein P, Jones MW, O’Sullivan M, et al. Global carbon
ing the distinctive behaviors in sub-modules and diagnos- budget 2019. Earth Syst Sci Data. 2019;11:1783–838.
ing idiosyncratic model behaviors. Such module inter- 3. Poulter B, Frank D, Ciais P, Myneni RB, Andela N, Bi J, et al.
comparisons will require a more modular design of the Contribution of semi-arid ecosystems to interannual variability of
models within the generalized overarching modeling the global carbon cycle. Nature. 2014;509:600–3.
4. Field CB, Lobell DB, Peters HA, Chiariello NR. Feedbacks of
framework for ecosystem processes. One possible method terrestrial ecosystems to climate change. Annu Rev Environ
of systematic organization could require organizing pro- Resour. 2007;32:1–29.
cesses according to key resource stocks and flows, as well 5. Bonan GB, Doney SC (2018) Climate, ecosystems, and planetary
as the dominant time scales of the associated dynamics (as futures: the challenge to predict life in earth system models.
Science (80- ). [Link]
outlined in Fig. 1) so that each sub-module has a set of
6. Foley JA, Prentice IC, Ramankutty N, Levis S, Pollard D, Sitch S,
well-defined and stable inputs and outputs. However, et al. An integrated biosphere model of land surface processes,
restructuring existing code can require careful planning, terrestrial carbon balance, and vegetation dynamics. Glob
software engineering, and monetary resources. Biogeochem Cycles. 1996;10:603–28.
2. Understanding coordination and trade-offs among physi- 7. Pacala SW, Canham CD, Saponara J, Silander JA, Kobe RK,
Ribbens E. Forest models defined by field measurements: estima-
ological traits that affect plant carbon-, water-, and
tion, error analysis and dynamics. Ecol Monogr. 1996;66:1–43.
nutrient-use strategies is critical to constraining the trait- 8. Shugart HH, Wang B, Fischer R, Ma J, Fang J, Yan X, et al. Gap
based parameter spaces. We recommend approaching this models and their individual-based relatives in the assessment of
problem both through further measurements of trait coor- the consequences of global change. Environ Res Lett. 2018;13:
dination in conjunction with optimality theory and other 033001.
9.• Fisher RA, Koven CD, Anderegg WRL, et al. Vegetation demo-
organizing principles revealed for ecosystem dynamics graphics in earth system models: a review of progress and prior-
[122]. In particular, belowground traits and dynamics ities. Glob Chang Biol. 2018;24:35–54 The review discusses
are still the most uncertain processes followed by wood implementations and challenges of vegetation demography, a
10 Curr Clim Change Rep (2021) 7:1–13

key driver of ecosystem heterogeneity among mainstream eco- elevated CO2 (free-air CO2 enrichment) and N-fertilization.
system models. Plant Cell Environ. 2009;32:1500–12.
10.• Falster DS, Duursma RA, FitzJohn RG. How functional traits 27. Bartlett MK, Zhang Y, Kreidler N, Sun S, Ardy R, Cao K, et al.
influence plant growth and shade tolerance across the life cycle. Global analysis of plasticity in turgor loss point, a key drought
Proc Natl Acad Sci. 2018;115:E6789–98 This study uses an tolerance trait. Ecol Lett. 2014;17:1580–90.
elegant theoretical framework to identify a key set of traits 28. Dietze MC, Serbin SP, Davidson C, Desai AR, Feng X, Kelly R,
that interact to influence plant growth rates under heteroge- et al. A quantitative assessment of a terrestrial biosphere model’s
nous light condition. data needs across north American biomes. J Geophys Res
11. Dietze MC. Prediction in ecology: a first-principles framework: A. Biogeosci. 2014;119:286–300.
Ecol Appl. 2017;27:2048–60. 29. Rogers A, Medlyn BE, Dukes JS, Bonan G, von Caemmerer S,
12. Fisher RA, Muszala S, Verteinstein M, Lawrence P, Xu C, Dietze MC, et al. A roadmap for improving the representation of
McDowell NG, et al. Taking off the training wheels: the properties photosynthesis in earth system models. New Phytol. 2017;213:
of a dynamic vegetation model without climate envelopes, 22–42.
CLM4.5(ED). Geosci Model Dev. 2015;8:3593–619. 30. Richardson AD, Keenan TF, Migliavacca M, Ryu Y, Sonnentag
13.• Fer I, Kelly R, Moorcroft PR, Richardson AD, Cowdery EM, O, Toomey M. Climate change, phenology, and phenological con-
Dietze MC. Linking big models to big data: efficient ecosystem trol of vegetation feedbacks to the climate system. Agric For
model calibration through Bayesian model emulation. Meteorol. 2013;169:156–73.
Biogeosciences. 2018. [Link] 31. Richardson AD, Anderson RS, Arain MA, Barr AG, Bohrer G,
Advances in integrating trait and other observations with Chen G, et al. Terrestrial biosphere models need better represen-
ecosystem models in a Bayesian framework. tation of vegetation phenology: results from the north American
14. Mencuccini M, Manzoni S, Christoffersen B. Modelling water carbon program site synthesis. Glob Chang Biol. 2012;18:566–84.
fluxes in plants: from tissues to biosphere. New Phytol. 32. Warren JM, Hanson PJ, Iversen CM, Kumar J, Walker AP,
2019;222:1207–22. Wullschleger SD. Root structural and functional dynamics in ter-
15. Peng B, Guan K, Tang J, Ainsworth EA, Asseng S, Bernacchi CJ, restrial biosphere models - evaluation and recommendations. New
et al. Towards a multiscale crop modelling framework for climate Phytol. 2015;205:59–78.
change adaptation assessment. Nat plants. 2020;6:338–48. 33. Trugman AT, Medvigy D, Hoffmann WA, Pellegrini FA.
16. Smith NG, Keenan TF, Colin Prentice I, Wang H, Wright IJ, Sensitivity of woody carbon stocks to bark investment strategy
Niinemets Ü, et al. Global photosynthetic capacity is optimized in Neotropical savannas and forests. Biogeosciences. 2018;15:
to the environment. Ecol Lett. 2019;22:506–17. 233–43.
17. Poorter H, Niinemets Ü, Ntagkas N, Siebenkäs A, Mäenpää M, 34. Wang B, Shugart HH, Lerdau MT. An individual-based model of
Matsubara S, et al. A meta-analysis of plant responses to light forest volatile organic compound emissions—UVAFME-VOC
intensity for 70 traits ranging from molecules to whole plant per- v1.0. Ecol Model. 2017;350:69–78.
formance. New Phytol. 2019;223:1073–105. 35. Fisher R, McDowell N, Purves D, Moorcroft P, Sitch S, Cox P,
18. Kattge J, Knorr W. Temperature acclimation in a biochemical et al. Assessing uncertainties in a second-generation dynamic veg-
model of photosynthesis: a reanalysis of data from 36 species. etation model caused by ecological scale limitations. New Phytol.
Plant Cell Environ. 2007;30:1176–90. 2010;187:666–81.
19. Manzoni S, Vico G, Katul G, Fay PA, Polley W, Palmroth S, et al. 36. Trugman AT, Fenton NJ, Bergeron Y, Xu X, Welp LR, Medvigy
Optimizing stomatal conductance for maximum carbon gain under D. Climate, soil organic layer, and nitrogen jointly drive forest
water stress: a meta-analysis across plant functional types and development after fire in the North American boreal zone. J Adv
climates. Funct Ecol. 2011;25:456–67. Model Earth Syst. 2016;8:1180–209.
20. Niinemets Ü, Keenan TF, Hallik L. A worldwide analysis of 37. Körner C (2017) A matter of tree longevity. Science (80- ). https://
within-canopy variations in leaf structural, chemical and physio- [Link]/10.1126/science.aal2449.
logical traits across plant functional types. New Phytol. 2015;205:
38.• Bugmann H, Seidl R, Hartig F, et al. Tree mortality submodels
973–93.
drive simulated long-term forest dynamics: assessing 15 models
21. Mackay DS, Savoy PR, Grossiord C, Tai X, Pleban JR, Wang DR,
from the stand to global scale. Ecosphere. 2019;10:e02616
et al. Conifers depend on established roots during drought: results
Intercomparison of mortality modules among various ecosys-
from a coupled model of carbon allocation and hydraulics. New
tem models.
Phytol. 2020;225:679–92.
39. Johnson DJ, Needham J, Xu C, Massoud EC, Davies SJ,
22. Waring BG, Powers JS. Overlooking what is underground: root:
Anderson-Teixeira KJ, et al. Climate sensitive size-dependent sur-
shoot ratios and coarse root allometric equations for tropical for-
vival in tropical trees. Nat Ecol Evol. 2018;2:1436–42.
ests. For Ecol Manag. 2017;385:10–5.
40. Powell TL, Galbraith DR, Christoffersen BO, Harper A,
23. Holbrook NM, Putz FE. Influence of neighbors on tree form:
Imbuzeiro HMA, Rowland L, et al. Confronting model predic-
effects of lateral shade and prevention of sway on the allometry
tions of carbon fluxes with measurements of Amazon forests sub-
of Liquidambar styraciflua (sweet gum). Am J Bot. 1989;76:
jected to experimental drought. New Phytol. 2013;200:350–65.
1740–9.
41. Longo M, Knox RG, Medvigy DM, Levine NM, Dietze MC, Kim
24. Chave J, Réjou-Méchain M, Búrquez A, Chidumayo E, Colgan
Y, et al. The biophysics, ecology, and biogeochemistry of func-
MS, Delitti WBC, et al. Improved allometric models to estimate
tionally diverse, vertically and horizontally heterogeneous ecosys-
the aboveground biomass of tropical trees. Glob Chang Biol.
tems: the ecosystem demography model, version 2.2-part 1: model
2014;20:3177–90.
description. Geosci Model Dev. 2019;12:4309–46.
25. Atkin OK, Bruhn D, Hurry VM, Tjoelker MG. Evans review no. 2
- the hot and the cold: unravelling the variable response of plant 42. McDowell N, Allen CD, Anderson-Teixeira K, Brando P, Brienen
respiration to temperature. Funct Plant Biol. 2005;32:87–105. R, Chambers J, et al. Drivers and mechanisms of tree mortality in
moist tropical forests. New Phytol. 2018;219:851–69.
26. Domec JC, Palmroth S, Ward E, Maier CA, ThÉrézien M, Oren R.
Acclimation of leaf hydraulic conductance and stomatal conduc- 43. Tilman D, Isbell F, Cowles JM. Biodiversity and ecosystem func-
tance of Pinus taeda (loblolly pine) to long-term growth in tioning. Annu Rev Ecol Evol Syst. 2014;45:471–93.
Curr Clim Change Rep (2021) 7:1–13 11

44. Sakschewski B, Von Bloh W, Boit A, Poorter L, Peña-Claros M, 63. Berzaghi F, Longo M, Ciais P, Blake S, Bretagnolle F, Vieira S,
Heinke J, et al. Resilience of Amazon forests emerges from plant et al. Carbon stocks in central African forests enhanced by ele-
trait diversity. Nat Clim Chang. 2016;6:1032–6. phant disturbance. Nat Geosci. 2019;12:725–9.
45. Sakschewski B, von Bloh W, Boit A, Rammig A, Kattge J, 64. Medvigy D, Clark KL, Skowronski NS, Schäfer KVR. Simulated
Poorter L, et al. Leaf and stem economics spectra drive diversity impacts of insect defoliation on forest carbon dynamics. Environ
of functional plant traits in a dynamic global vegetation model. Res Lett. 2012;7. [Link]
Glob Chang Biol. 2015;21:2711–25. 65. Lawrence DM, Hurtt GC, Arneth A, Brovkin V, Calvin KV, Jones
46. Xu X, Medvigy D, Powers JS, Becknell JM, Guan K. Diversity in AD, et al. The Land Use Model Intercomparison Project (LUMIP)
plant hydraulic traits explains seasonal and inter-annual variations contribution to CMIP6: rationale and experimental design. Geosci
of vegetation dynamics in seasonally dry tropical forests. New Model Dev. 2016;9:2973–98.
Phytol. 2016;212:80–95. 66. Sellers PJ, Mintz Y, Sud YC, Dalcher A. A simple biosphere
47. Rüger N, Condit R, Dent DH, DeWalt SJ, Hubbell SP, Lichstein model (SiB) for use within general circulation models. J Atmos
JW, Lopez OR, Wirth C, Farrior CE (2020) Demographic trade- Sci. 1986;43:505–31.
offs predict tropical forest dynamics. Science (80- ) 368:165–168. 67. Bonan GB, Lawrence PJ, Oleson KW, Levis S, Jung M,
48. di Porcia e Brugnera M, Meunier F, Longo M, Krishna Moorthy Reichstein M, et al. Improving canopy processes in the
SM, De Deurwaerder H, Schnitzer SA, et al. Modeling the impact Community Land Model version 4 (CLM4) using global flux
of liana infestation on the demography and carbon cycle of trop- fields empirically inferred from FLUXNET data. J Geophys Res
ical forests. Glob Chang Biol. 2019;25:3767–80. Biogeosci. 2011;116:G02014.
49. Quesada CA, Phillips OL, Schwarz M, Czimczik CI, Baker TR, 68. Moorcroft PR, Hurtt GC, Pacala SW. A method for scaling veg-
Patiño S, et al. Basin-wide variations in Amazon forest structure etation dynamics: the ecosystem demography model (ED). Ecol
and function are mediated by both soils and climate. Monogr. 2001;71:557–85.
Biogeosciences. 2012;9:2203–46. 69. Schimel D, Schneider FD. Flux towers in the sky: global ecology
50. Steidinger BS, Crowther TW, Liang J, et al. Climatic controls of from space. New Phytol. 2019;224:570–84.
decomposition drive the global biogeography of forest-tree sym- 70. Dubayah R, Blair JB, Goetz S, Fatoyinbo L, Hansen M, Healey S,
bioses. Nature. 2019;569:404–8. et al. The global ecosystem dynamics investigation: high-
51. Bennett AC, Mcdowell NG, Allen CD, Anderson-Teixeira KJ. resolution laser ranging of the earth’s forests and topography.
Larger trees suffer most during drought in forests worldwide. Sci Remote Sens. 2020;1:100002.
Nat Plants. 2015;1:15139. 71. Wang Y, Sperry JS, Anderegg WRL, Venturas MD, Trugman AT.
52. Levine NM, Zhang K, Longo M, Baccini A, Phillips OL, Lewis A theoretical and empirical assessment of stomatal optimization
SL, et al. Ecosystem heterogeneity determines the ecological re- modeling. New Phytol. 2020;227:311–25.
silience of the Amazon to climate change. Proc Natl Acad Sci U S 72. Sperry JS, Venturas MD, Todd HN, Trugman AT, Anderegg
A. 2016;113:793–7. WRL, Wang Y, et al. The impact of rising CO2 and acclimation
53. Longo M, Knox RG, Levine NM, Alves LF, Bonal D, Camargo on the response of US forests to global warming. Proc Natl Acad
PB, et al. Ecosystem heterogeneity and diversity mitigate Amazon Sci U S A. 2019;116:25734–44.
forest resilience to frequent extreme droughts. New Phytol. 73. Detto M, Xu X (2020) Optimal leaf life strategies determine Vc,
2018;219:914–31. max dynamic during ontogeny. New Phytol nph.16712.
54. Farrior CE, Bohlman SA, Hubbell S, Pacala SW (2016) 74. Wu J, Albert LP, Lopes AP, et al. Leaf development and demog-
Dominance of the suppressed: power-law size structure in tropical raphy explain photosynthetic seasonality in Amazon evergreen
forests. Science (80- ) 351:155–157. forests. Science. 2016;(80):351, 972–976.
55. Hantson S, Arneth A, Harrison SP, Kelley DI, Prentice IC, Rabin 75. Medvigy D, Jeong SJ, Clark KL, Skowronski NS, Schäfer KVR.
SS, et al. The status and challenge of global fire modelling. Effects of seasonal variation of photosynthetic capacity on the
Biogeosciences. 2016;13:3359–75. carbon fluxes of a temperate deciduous forest. J Geophys Res
56. Sitch S, Huntingford C, Gedney N, et al. Evaluation of the terres- Biogeosci. 2013;118:1703–14.
trial carbon cycle, future plant geography and climate-carbon cy- 76. Chen X, Maignan F, Viovy N, Bastos A, Goll D, Wu J, et al.
cle feedbacks using five dynamic global vegetation models Novel representation of leaf phenology improves simulation of
(DGVMs). Glob Chang Biol. 2008;14:2015–39. Amazonian evergreen forest photosynthesis in a land surface mod-
57. Rabin SS, Melton JR, Lasslop G, Bachelet D, Forrest M, Hantson el. J Adv Model Earth Syst. 2020;12. [Link]
S, et al. The fire modeling Intercomparison project (FireMIP), 2018MS001565.
phase 1: experimental and analytical protocols with detailed model 77. De Weirdt M, Verbeeck H, Maignan F, Peylin P, Poulter B, Bonal
descriptions. Geosci Model Dev. 2017;10:1175–97. D, et al. Seasonal leaf dynamics for tropical evergreen forests in a
58. Baudena M, Dekker SC, Van Bodegom PM, et al. Forests, sa- process based global ecosystem model. Geosci Model Dev
vannas, and grasslands: bridging the knowledge gap between ecol- Discuss. 2012;5:639–81.
ogy and dynamic global vegetation models. Biogeosciences. 78. Kim Y, Knox RG, Longo M, Medvigy D, Hutyra LR, Pyle EH,
2015;12:1833–48. et al. Seasonal carbon dynamics and water fluxes in an Amazon
59. Sanderson BM, Fisher RA. A fiery wake-up call for climate sci- rainforest. Glob Chang Biol. 2012;18:1322–34.
ence. Nat Clim Chang. 2020;10:175–7. 79. Friend AD, Eckes-Shephard AH, Fonti P, Rademacher TT,
60. Eyring V, Bony S, Meehl GA, Senior CA, Stevens B, Stouffer RJ, Rathgeber CBK, Richardson AD, et al. On the need to consider
et al. Overview of the Coupled Model Intercomparison Project wood formation processes in global vegetation models and a sug-
Phase 6 (CMIP6) experimental design and organization. Geosci gested approach. Ann For Sci. 2019;76:1–13.
Model Dev. 2016;9:1937–58. 80. Fatichi S, Leuzinger S, Körner C (2013) Moving beyond photo-
61. Anyomi KA, Mitchell SJ, Ruel JC. Windthrow modelling in old- synthesis: from carbon source to sink-driven vegetation modeling.
growth and multi-layered boreal forests. Ecol Model. 2016;327: New Phytol n/a-n/a.
105–14. 81. Wolf A, Ciais P, Bellassen V, Delbart N, Field CB, Berry JA.
62. Gora EM, Muller-Landau HC, Burchfield JC, Bitzer PM, Hubbell Forest biomass allometry in global land surface models. Glob
SP, Yanoviak SP. A mechanistic and empirically supported light- Biogeochem Cycles. 2011;25. [Link]
ning risk model for forest trees. J Ecol. 2020;1365-2745:13404. 2010gb003917.
12 Curr Clim Change Rep (2021) 7:1–13

82. Trugman AT, Detto M, Bartlett MK, Medvigy D, Anderegg Cycles. 2019;33:501–23 A comprehensive implementation of
WRL, Schwalm C, et al. Tree carbon allocation explains forest diverse mycorrhizal associations including priming,
drought-kill and recovery patterns. Ecol Lett. 2018;21:1552–60. arbuscular mycorrhizae, ecto-mycorrhizae, and nitrogen
83. Trugman AT, Medvigy D, Mankin JS, Anderegg WRL. Soil fixation.
moisture stress as a major driver of carbon cycle uncertainty. 101. Schenk HJ, Jackson RB. Rooting depths, lateral root spreads and
Geophys Res Lett. 2018;45:6495–503. below-ground/above-ground allometries of plants in water-limited
84. Fatichi S, Pappas C, Ivanov VY. Modeling plant-water interac- ecosystems. J Ecol. 2002;90:480–94.
tions: an ecohydrological overview from the cell to the global 102. Smith-Martin CM, Xu X, Medvigy D, Schnitzer SA, Powers JS.
scale. Wiley Interdiscip Rev Water. 2016;3:327–68. Allometric scaling laws linking biomass and rooting depth vary
85. Christoffersen BO, Gloor M, Fauset S, et al (2016) Linking hy- across ontogeny and functional groups in tropical dry forest lianas
draulic traits to tropical forest function in a size-structured and and trees. New Phytol. 2020;226:714–26.
trait-driven model (TFS v.1-hydro). Geosci Model Dev Discuss 103. Jackson RB, Canadell J, Ehleringer JR, Mooney HA, Sala OE,
0:1–60. Schulze ED. A global analysis of root distributions for terrestrial
86. De Kauwe MG, Medlyn BE, Ukkola AM, et al. Identifying areas biomes. Oecologia. 1996;108:389–411.
at risk of drought-induced tree mortality across South-Eastern 104. Laio F, D’Odorico P, Ridolfi L. An analytical model to relate the
Australia. Glob Chang Biol gcb. 2020:15215. vertical root distribution to climate and soil properties. Geophys
87. Kennedy D, Swenson S, Oleson KW, Lawrence DM, Fisher R, Res Lett. 2006;33:L18401.
Lola da Costa AC, et al. Implementing plant hydraulics in the 105. Fan Y, Miguez-Macho G, Jobbágy EG, Jackson RB, Otero-Casal
community land model, version 5. J Adv Model Earth Syst. C. Hydrologic regulation of plant rooting depth. Proc Natl Acad
2019;11:485–513. Sci U S A. 2017;114:10572–7.
88. Gleason SM, Westoby M, Jansen S, Choat B, Hacke UG, Pratt 106. Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z,
RB, et al. Weak tradeoff between xylem safety and xylem-specific Bongers F, et al. The worldwide leaf economics spectrum.
hydraulic efficiency across the world’s woody plant species. New Nature. 2004;428:821–7.
Phytol. 2016;209:123–36. 107. Chave J, Coomes D, Jansen S, Lewis SL, Swenson NG, Zanne
89. Trugman AT, Anderegg LDL, Shaw JD, Anderegg WRL. Trait AE. Towards a worldwide wood economics spectrum. Ecol Lett.
velocities reveal that mortality has driven widespread coordinated 2009;12:351–66.
shifts in forest hydraulic trait composition. Proc Natl Acad Sci.
108. Reich PB. The world-wide ‘fast–slow’ plant economics spectrum:
2020;117:8532–8. a traits manifesto. J Ecol. 2014;102:275–301.
90. Choat B, Jansen S, Brodribb TJ, Cochard H, Delzon S, Bhaskar R,
109. Anderegg LDL, Berner LT, Badgley G, Sethi ML, Law BE,
et al. Global convergence in the vulnerability of forests to drought.
HilleRisLambers J. Within-species patterns challenge our under-
Nature. 2012;491:752–5.
standing of the leaf economics spectrum. Ecol Lett. 2018;21:734–
91. Delpierre N, Berveiller D, Granda E, Dufrêne E. Wood phenolo-
44.
gy, not carbon input, controls the interannual variability of wood
110. Baraloto C, Paine CET, Poorter L, Beauchene J, Bonal D,
growth in a temperate oak forest. New Phytol. 2016;210:459–70.
Domenach AM, et al. Decoupled leaf and stem economics in rain
92. Wagner F, Rossi V, Stahl C, Bonal D, Hérault B. Asynchronism in
forest trees. Ecol Lett. 2010;13:1338–47.
leaf and wood production in tropical forests: a study combining
satellite and ground-based measurements. Biogeosciences. 111. Kattge J, Díaz S, Lavorel S, et al. TRY - a global database of plant
2013;10:7307–21. traits. Glob Chang Biol. 2011;17:2905–35.
93. Fatichi S, Pappas C, Zscheischler J, Leuzinger S. Modelling car- 112. Reich PB, Rich RL, Lu X, Wang Y-P, Oleksyn J. Biogeographic
bon sources and sinks in terrestrial vegetation. New Phytol. variation in evergreen conifer needle longevity and impacts on
2018;221:652–68. boreal forest carbon cycle projections. Proc Natl Acad Sci.
94. Dietze MC, Sala A, Carbone MS, Czimczik CI, Mantooth JA, 2014;111:13703–8.
Richardson AD, et al. Nonstructural carbon in woody plants. 113. Xu X, Medvigy D, Joseph Wright S, Kitajima K, Wu J, Albert LP,
Annu Rev Plant Biol. 2014;65:667–87. et al. Variations of leaf longevity in tropical moist forests predicted
95. Malhi Y, Doughty C, Galbraith D. The allocation of ecosystem net by a trait-driven carbon optimality model. Ecol Lett. 2017;20:
primary productivity in tropical forests. Philos Trans R Soc B Biol 1097–106.
Sci. 2011;366:3225–45. 114. Reich PB, Sendall KM, Stefanski A, Wei X, Rich RL,
96. Robertson AL, Malhi Y, Farfan-Amezquita F, Aragão LEOC, Montgomery RA. Boreal and temperate trees show strong accli-
Silva Espejo JE, Robertson MA. Stem respiration in tropical for- mation of respiration to warming. Nature. 2016;531:633–6.
ests along an elevation gradient in the Amazon and Andes. Glob [Link]
Chang Biol. 2010;16:3193–204. 115. Scheiter S, Langan L, Higgins SI. Next-generation dynamic global
97. Treseder KK, Balser TC, Bradford MA, Brodie EL, Dubinsky EA, vegetation models: learning from community ecology. New
Eviner VT, et al. Integrating microbial ecology into ecosystem Phytol. 2013;198:957–69.
models: challenges and priorities. Biogeochemistry. 2012;109:7– 116. Lombardozzi DL, Bonan GB, Smith NG, Dukes JS, Fisher RA.
18. Temperature acclimation of photosynthesis and respiration: a key
98. Medvigy D, Wang G, Zhu Q, Riley WJ, Trierweiler AM, Waring uncertainty in the carbon cycle-climate feedback. Geophys Res
BG, et al. Observed variation in soil properties can drive large Lett. 2015;42:8624–31.
variation in modelled forest functioning and composition during 117. Smith NG, Malyshev SL, Shevliakova E, Kattge J, Dukes JS.
tropical forest secondary succession. New Phytol. 2019;223: Foliar temperature acclimation reduces simulated carbon sensitiv-
1820–33. ity to climate. Nat Clim Chang. 2016;6:407–11.
99. Levy-Varon JH, Batterman SA, Medvigy D, Xu X, Hall JS, van 118. Tjoelker MG, Oleksyn J, Reich PB. Modelling respiration of veg-
Breugel M, et al. Tropical carbon sink accelerated by symbiotic etation: evidence for a general temperature-dependent Q10. Glob
dinitrogen fixation. Nat Commun. 2019;10:5637. Chang Biol. 2001;7:223–30.
100.• Sulman BN, Shevliakova E, Brzostek ER, Kivlin SN, Malyshev 119. Slot M, Rey-Sánchez C, Gerber S, Lichstein JW, Winter K,
S, Menge DNL, et al. Diverse mycorrhizal associations enhance Kitajima K. Thermal acclimation of leaf respiration of tropical
terrestrial C storage in a global model. Global Biogeochem trees and lianas: response to experimental canopy warming, and
Curr Clim Change Rep (2021) 7:1–13 13

consequences for tropical forest carbon balance. Glob Chang Biol. 125. Wieder WR, Lawrence DM, Fisher RA, Bonan GB, Cheng SJ,
2014;20:2915–26. Goodale CL, et al. Beyond static benchmarking: using experimen-
120. Wang H, Atkin OK, Keenan TF, Smith NG, Wright IJ, tal manipulations to evaluate land model assumptions. Glob
Bloomfield KJ, et al. Acclimation of leaf respiration consistent Biogeochem Cycles. 2019;33:1289–309.
with optimal photosynthetic capacity. Glob Chang Biol. 2020;26: 126. Zhang K, de Almeida Castanho AD, Galbraith DR, Moghim S,
2573–83. Levine NM, Bras RL, et al. The fate of Amazonian ecosystems
121.• Wang H, Prentice IC, Keenan TF, Davis TW, Wright IJ, Cornwell over the coming century arising from changes in climate, atmo-
WK, et al. Towards a universal model for carbon dioxide uptake spheric CO2, and land use. Glob Chang Biol. 2015;21:2569–87.
by plants /704/47 /704/158/851 letter. Nat Plants. 2017;3:734–41 127. Higgins SI, Scheiter S. Atmospheric CO 2 forces abrupt vegeta-
Theoretical basis of optimality-based photosynthesis model tion shifts locally, but not globally. Nature. 2012;488:209–12.
that accounts for acclimation and adaptation. 128. Purves DW, Lichstein JW, Strigul N, Pacala SW. Predicting and
122. Franklin O, Harrison SP, Dewar R, Farrior CE, Brännström Å, understanding forest dynamics using a simple tractable model.
Dieckmann U, et al. Organizing principles for vegetation dynam- Proc Natl Acad Sci U S A. 2008;105:17018–22.
ics. Nat Plants. 2020;6:444–53. 129. Jiang C, Ryu Y. Multi-scale evaluation of global gross primary
123. Stocker BD, Wang H, Smith NG, Harrison SP, Keenan TF, productivity and evapotranspiration products derived from
Sandoval D, et al. P-model v1.0: an optimality-based light use Breathing Earth System Simulator (BESS). Remote Sens
efficiency model for simulating ecosystem gross primary produc- Environ. 2016;186:528–47.
tion. Geosci Model Dev. 2020;13:1545–81.
124. Medlyn BE, Zaehle S, De Kauwe MG, et al. Using ecosystem
experiments to improve vegetation models. Nat Clim Chang. Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
2015;5:528–34. tional claims in published maps and institutional affiliations.