Access to this full-text is provided by Wiley.
Content available from Journal of Advances in Modeling Earth Systems
This content is subject to copyright. Terms and conditions apply.
Perspectives on the Future of Land Surface Models and
the Challenges of Representing Complex
Terrestrial Systems
Rosie A. Fisher
1,2
and Charles D. Koven
3
1
Climate and Global Dynamics Division, National Center for Atmospheric Research, Boulder, CO, USA,
2
Centre
Européen de Recherche et de Formation Avancée en Calcul Scientifique, Toulouse, France,
3
Climate and Ecosystem
Sciences Division, Lawrence Berkeley National Lab, Berkeley, CA, USA
Abstract Land surface models (LSMs) are a vital tool for understanding, projecting, and predicting the
dynamics of the land surface and its role within the Earth system, under global change. Driven by the
need to address a set of key questions, LSMs have grown in complexity from simplified representations of
land surface biophysics to encompass a broad set of interrelated processes spanning the disciplines of
biophysics, biogeochemistry, hydrology, ecosystem ecology, community ecology, human management, and
societal impacts. This vast scope and complexity, while warranted by the problems LSMs are designed to
solve, has led to enormous challenges in understanding and attributing differences between LSM
predictions. Meanwhile, the wide range of spatial scales that govern land surface heterogeneity, and the
broad spectrum of timescales in land surface dynamics, create challenges in tractably representing processes
in LSMs. We identify three “grand challenges”in the development and use of LSMs, based around these
issues: managing process complexity, representing land surface heterogeneity, and understanding
parametric dynamics across the broad set of problems asked of LSMs in a changing world. In this review,
we discuss progress that has been made, as well as promising directions forward, for each of
these challenges.
Plain Language Summary Land surface models (LSMs) are the part of climate models that
simulate processes happening at the Earth's surface. These include reflection of the sunlight, evaporation
from ecosystems, and the amount of carbon from human emissions that the land takes up. LSMs also need to
simulate how human management of the land surface changes the climate both directly (e.g., via the
effect on evaporation) and in the long term (via changing the amount of carbon stored in wood and soil).
Not surprisingly, trying to make a single mathematical representation of all of these different parts of the
Earth system is difficult. Here we discuss themes that repeatedly affect all teams developing LSMs: how to
manage the increasing number of complicated model components, how to represent the high degree of
variability of the land surface, and how to predict how the properties of the surface (particularly those of
plant communities) will change. These are large problems, with no obvious easy solutions. We hope to spark
discussion and investment into their resolution, concomitant with the increasing importance of LSMs as
our best tools for translating possible trajectories of climate change into impacts on humans, ecosystems,
food and water supplies, and river systems.
1. Introduction
The land surface is the only part of the Earth system that is directly experienced by the majority of humans,
terrestrial animals, and plants. Land surface processes mediate the majority of the impacts of climate on
human societies and ecosystems, and accurate representation of land surface processes is critical for our
understanding of how climate and climate change actually affect living systems. Land surface models
(LSMs) are numerical models that solve the coupled fluxes of water, energy, and carbon between the land
surface and atmosphere, within a context of direct and indirect human forcings and ecological dynamics.
LSMs are arguably the most sophisticated tools that society currently possesses for predicting how the con-
ditions for life on the surface of the Earth will change in the coming years, decades, and centuries. The scope
of land surface modeling activities naturally encompasses a huge set of overlapping and interconnected dis-
ciplines relevant to this problem.
©2020. The Authors.
This is an open access article under the
terms of the Creative Commons
Attribution License, which permits use,
distribution and reproduction in any
medium, provided the original work is
properly cited.
COMMISSIONED
MANUSCRIPT
10.1029/2018MS001453
Key Points:
•Land surface models have grown in
complexity, and new methods of
managing this complexity are
required for scientific understanding
•New methods are also needed to
represent, classify, and benchmark
models across the multidimensional
heterogeneity of the land surface
•A further challenge is to constrain
model parameters in ways that are
consistent with allowing long‐term
ecological dynamics to occur
Correspondence to:
C. D. Koven,
cdkoven@lbl.gov
Citation:
Fisher, R. A., & Koven, C. D. (2020).
Perspectives on the future of land
surface models and the challenges of
representing complex terrestrial
systems. Journal of Advances in
Modeling Earth Systems,12,
e2018MS001453. https://doi.org/
10.1029/2018MS001453
Received 31 OCT 2019
Accepted 4 MAR 2020
Accepted article online MAR 10 2020
R. A. Fisher and C. D. Koven contribu-
ted equally to this manuscript.
FISHER AND KOVEN 1of24
In this paper, we attempt to provide a high‐level illustration of a set of different classes of challenges that
arise from such a complex and high‐dimensional activity. We further indicate, where appropriate, promising
approaches around which one might organize the development of tools that can predict the complex and
heterogeneous functioning of the land surface under the radically altered climatic, ecological, and societal
conditions anticipated by Earth system projections.
Land surface models were originally developed (and thus continue to be primarily supported) by
atmospheric/climate modeling and forecasting activities that demand physical boundary conditions in terms
of energy partitioning, surface roughness, and albedo, to represent the influence of the land on meteorolo-
gical processes. As applied to the global climate change problem, two key model results set the LSM commu-
nity on its current trajectory: (1) the prediction that plant biophysical responses to elevated CO
2
could have
an appreciable effect on the global climate itself (Sellers et al., 1996), and (2) that coupling of climate and
carbon cycle could substantially strengthen the rate of global warming (Cox et al., 2000). The need for
LSMs to quantify such biogeophysical and biogeochemical feedbacks (respectively) to the climate system
has formed the basis of their recent development, but increasingly, questions pertaining to the impacts on
the land surface itself have attained a higher profile.
State‐of‐the‐art LSMs (e.g., Decharme et al., 2019; D. M. Lawrence et al., 2019; Wiltshire et al., 2019;
Yokohata et al., 2019) typically provide a set of prognostic variables related to land‐mediated feedbacks on
global biogeochemical cycles. In particular, the terrestrial carbon cycle, by partially controlling what fraction
of CO
2
that humans emit remains in the atmosphere, has a role in determining the transient climate
response to emissions and the remaining carbon emissions budget compatible with a given climate goal
(Matthews et al., 2018). In addition, LSMs predict changes in the biophysical function of the land surface
as climate and ecosystems change and thus how the land interacts with both the atmosphere and with rivers
and downstream ecosystems. Lastly, LSMs provide information on risks to human societies and natural eco-
systems associated with future climate scenarios, including crop productivity, heat waves, urban climates,
the severity and frequency of fire and other disturbances, flooding, ecosystem productivity, permafrost
and land ice status, and health and freshwater availability.
Through time, representations of numerous processes that are known to impact the dynamics of systems
relevant to these questions have been incrementally added to LSMs. As a result, land surface models have
expanded from their initial simple biophysical configurations (Sellers et al., 1986), to include representations
of soil moisture dynamics, stomatal functioning, land surface heterogeneity, surface hydrological processes,
plant and soil carbon cycling, dynamic vegetation distributions, fire, urban environments, land cover and
management, nitrogen cycling and crops (Lawrence et al., 2019, Figure 1), and latterly plant demographic
processes (Fisher et al., 2018; Sato et al., 2007; Weng et al., 2017), phosphorus cycling, (Goll et al., 2017;
Reed et al., 2015; Yang et al., 2014), and plant hydraulics (Joetzjer et al., 2018; Kennedy et al., 2019).
Figure 1. A schematic depiction of the evolution of land surface model process representation through time, representing
the approximate timing of emergence of different model components as commonly employed features of Earth system
models. Note that all modeling groups follow a different pathway and that this diagram is primarily intended to act as an
illustration of increasing complexity through time.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 2of24
This ever‐widening scope of land surface models may be unavoidable, due to the interrelated nature of the
questions being asked of them. For example, the processes that govern carbon cycle feedbacks are highly
affected by both biophysical feedbacks in the Earth system and by land use decisions that are in turn affected
by climate impacts on human societies. Climate change impacts such as drought and fire are mediated by
plant biophysical responses to elevated CO
2
, which are themselves impacted by limitations imposed by
nutrient limitations on growth. Changing ecosystem boundary conditions impacts the composition and thus
biogeophysical and biogeochemical functionality of plant communities, and thus all these processes benefit
from being considered within the context of dynamic and/or demographic vegetation.
Arguably, the inclusion of process representation in land surface models is accelerating, driven by the needs
of various different user communities (hydrologists, biogeochemists, ecologists, atmospheric scientists, and
crop modelers) and by arguments put forward that the overall biospheric feedbacks are themselves impor-
tantly affected by a great number of interacting systems, including, for example (at the time of writing),
insect dynamics and impacts (Dietze & Matthes, 2014; Huang et al., 2019), vegetation sink limitations to
growth (Fatichi et al., 2019), soil microbial dynamics (Wieder et al., 2013), subcanopy turbulence (Bonan
et al., 2018), leaf mesophyll processes (Knauer et al., 2019), and polygonal tundra parameterizations
(Pau et al., 2014).
At the same time, both land surface models and the atmospheric models to which they may be coupled are
refining their spatial resolution, as enabled by new data sets and higher computational capabilities. A decade
ago, Wood et al. (2011) argued that achieving such increases into the 10
2
–10
3
m resolution range was itself a
grand challenge of land surface modeling, requiring increases in both the model capabilities and new data
sets to drive and test such models. In response, Beven and Cloke (2012) argued that, while such increases
in resolution should in principle allow for better simulations, the deeper problem lay with the epistemic
uncertainty of how to represent any given process and how to capture the effects of smaller‐scale unresolved
processes, at any given scale. As the scope of land surface models has increased, and alongside computa-
tional advances that have largely allowed the hyperresolution goal to be attained (Bierkens et al., 2015),
the questions of epistemic uncertainty and unresolved heterogeneity have grown in importance.
Rather than focus our discussion here on the arguments for and against inclusion of specific new processes
in land surface models, or whether increasing spatial resolution by itself will qualitatively change the nature
of LSM simulations, we instead focus on three broader challenges that integrate across model components,
namely:
1. Managing and understanding the process complexity of LSMs
2. Heterogeneity and the dimensionality of the land surface
3. Projecting the temporal and spatial dynamics of model parameters
Within each of these three “grand challenges”we describe the nature of the challenge, illustrate ongoing
developments, and propose pathways within which research and model development might best be struc-
tured to meet the important but comprehensively difficult task of predicting the future of the terrestrial sur-
face and biosphere.
2. Challenge: Managing and Understanding Process Complexity
2.1. Process Complexification
The wide variety of processes that interact to form the terrestrial system, and the depth of complexity present
in every one of these processes, together create a deep obstacle to creating tractable models of the land sur-
face. The increasing complexity of land models reflects both the tendency of scientists to focus on their own
particular areas of interest and expertise, as well as the reality that the Earth is in fact complex and that the
details of a great number of processes do in fact matter. But at the same time, the scope and complexity of
some modern land surface models have reached the point that no individuals are able to comprehensively
understand all facets of any one model. Further, a majority of model development teams (which are typically
situated within and primarily funded by Earth system modeling centers) struggle to meet all of the demands
placed on modern LSMs.
The set of processes required to make long‐term projections of the land surface and biosphere is large, and
their complexification has touched many different areas. The representation of soil hydrology, for example,
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 3of24
has proceeded from simple “bucket”representation (Manabe, 1969), through 1‐D Richards equation
(Bonan, 1996; Cox et al., 1999), to 3‐D variably saturated flow models that span from the soil through plant
tissues (Bisht & Riley, 2019). The representation of biogeochemistry has proceeded from the small set of
equations required to represent photosynthesis at the leaf scale (Dickinson et al., 1991), through full carbon
cycle models (Dickinson et al., 1991), to multiple coupled nutrient models (Fisher et al., 2019; Thornton
et al., 2007; Y.‐P. Wang et al., 2015; Zaehle & Friend, 2010). The representation of plant community ecology
has proceeded from the static plant functional types (Bonan, 1996; D. M. Lawrence et al., 2011; Zeng
et al., 2002), through mean individual dynamic models with simple rules governing competition (Arora &
Boer, 2006; Cox, 2001; Sitch et al., 2003), to models that resolve physiological processes at the canopy level
and implicitly downscale to population demography using self‐thinning or allometric scaling relations
(Argles et al., 2019; Bellassen et al., 2010; Haverd et al., 2013), and to demographic or individual‐based mod-
els with resolved competition between cohorts or individual plants (Fisher et al., 2018; Longo et al., 2019;
Moorcroft et al., 2001; Sakschewski et al., 2015; Weng et al., 2017). The shift toward representing the agents
of change has led groups to represent microbial types and their population dynamics in soil biogeochemical
models as well (Treseder et al., 2012; Wieder et al., 2013). The role of both natural and anthropogenic distur-
bance, missing in early land surface models, has been a major focus of developments in order to represent
the many direct effects that humans have on modifying the land surface (P. J. Lawrence et al., 2012;
Nabel et al., 2019; Pongratz et al., 2018; Shevliakova et al., 2009; Yue et al., 2018). Many further dimensions
of process complexification exist as well including canopy radiative transfer, trace gases, fire, permafrost,
boundary layer turbulence, and rivers.
While the arguments behind all of these process developments are sound, the historical development path-
ways by which process complexification has proceeded in any given land surface model have been largely ad
hoc and based on a collection of institutional, geographic, and individual preferences and interests. As a
result, the representation of any given process across models is extremely heterogeneous: Some models
may represent in great detail a given process that is entirely absent in peer models. This makes the compar-
ison of model predictions and projections difficult and frequently uninformative (Clark et al., 2011), a fea-
ture which was noted in early model intercomparison efforts (Koster & Milly, 1997) and remains true
today. Complexity also creates problems for those wanting to bring the evolving understanding of a given
process into models: How do we weigh the costs and benefits of a given increase in complexity?
A frequently proposed strategy to dealing with the problems that arise through complexification is to pursue
a“hierarchy of complexity”(Claussen et al., 2002) wherein parameters of simple(r) models are diagnosed
from the aggregate behavior of complex models. Such approaches are enormously valuable, and show up
across disciplines, but are generally themselves reflective of a particular perspective, because the specific
“simple model”chosen is dependent on the question being asked and conditional on all the other processes
deemed to be outside the hierarchy of complexity. To a hydrologist, the simple model may be a water balance
model, while to a community ecologist the simple model may be the growth rate of trees as conditional on
their size. How can we approach the complexity problem in a way that maintains sufficient flexibility to
allow multiple different ways of simplifying things across the wide set of processes that comprise land
surface modeling?
2.2. Modular Complexity as a Strategy
As land surface models themselves emerged from the introduction of interfaces between the land and the
atmosphere in early climate models (Polcher et al., 1998), a possible solution to the complexification pro-
blem is to take a more modular approach to the representation of processes in the land surface, in order
to allow the scaling of complexity and process representation across many dimensions (Figure 2). The crucial
requirements of such a modeling system are (1) the ability for it to represent a given process (or cluster of
processes) in multiple ways, recognizing the epistemic uncertainty associated with any choice of representa-
tion as well as the possibility of very different degrees of complexity (from highly resolved process represen-
tations to highly simplified “stub”representations such as representing a given process as having a fixed
value), and (2) to not necessarily assume which among a set of potential processes are the ones to be simpli-
fied or replaced, nor which aspects of a given process are the ones that a simpler configuration would be
dependent on. For example, a simplified configuration focused on vegetation dynamics may want to ask
for growth and mortality rates from a simplified representation of plant physiology, while a
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 4of24
Figure 2. Methods for isolating components of land surface model complexity. (a) A process schematic of a full‐complexity LSM. (b) Possible configurations of sim-
plified LSMs. Processes, and sets of processes, are represented as boxes in the diagram, with information connections represented as arrows. All processes—though
here shown only for stomatal conductance—are intended to allow alternate specifications, including possibly multiple hypothetical process realizations, empirical
or machine learning‐derived formulations, and/or simplified stub or null representations to allow for holding a given process constant while other processes vary.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 5of24
meteorology‐focused simplified configuration may only require stomatal conductance and optical properties
from the representation of vegetation (Laguë et al., 2019). Such efforts are already underway for subsets of
the land surface modeling scope such as basin‐scale (Clark, Nijssen, et al., 2015) or site‐scale (Coon
et al., 2016) water and energy budget models, leaf photosynthesis models (Walker et al., 2018), offline models
of forest structure (Farrior et al., 2016) (Moore et al., 2018), and soil biogeochemical “testbed”models
(Wieder et al., 2018) some of which are included as schematics in Figure 2b, but more effort is required to
generate overarching frameworks that can encompass these various themes.
The difficulty of designing a model architecture with this ability in mind is that the boundary conditions for
any one specific process (or cluster of processes) tend in practice to be very fluid. As representation of say,
fire, tree mortality, or soil respiration evolve over time, new variables need to be passed from one part of
the model to another for each iteration of the process representation (e.g., one fire model might need infor-
mation on the status of a single pool of coarse woody debris, whereas its successor may need several
size‐structured pools). Any such coupling strategy must thus be specifically designed to accommodate a wide
set of specific process representations and their variable boundary conditions at the outset, as well as flexibil-
ity in the numerical approach to creating the coupling. Thus, the design of interfaces that are robust to chan-
ging properties of submodules is a high priority. A further difficulty is in deciding how to aggregate processes
into higher‐level submodels: While it may be straightforward to define alternate hypotheses for, say, models
of stomatal conductance or within‐leaf carbon assimilation (Walker et al., 2018), other sets of processes may
not be as unambiguously delineated.
In principle, such an approach to land surface modeling may be much more powerful than current
approaches that use “ensembles of opportunity”to characterize structural uncertainty across a wide
range of model predictions. The key weakness with contemporary model intercomparison projects such
as C4MIP (Arora et al., 2013), TRENDY (Le Quéré et al., 2018), MSTMIP (Huntzinger et al., 2013;
Schwalm et al., 2019; Zscheischler et al., 2014), ISIMIP (Nishina et al., 2015) and others is the inability
to understand how process and parameter uncertainty maps onto uncertainty in the relevant model pro-
jections. Explanations that attempt to identify the largest variation in model projections in terms of spe-
cific processes such as nutrient or land use dynamics (Friedlingstein et al., 2013) are useful in suggesting
what may be driving the models, but such approaches are currently limited by the poor control on struc-
tural and parametric variation between models. The more detailed assumption‐centered approach of attri-
buting divergences between models and experiments described by Medlyn et al. (2015) allows a better
estimate of how structural differences lead to model divergences (see also De Kauwe et al., 2014;
Walker et al., 2015; Zaehle et al., 2014); however, even in that framework the many model differences
other than the specific assumptions being tested (e.g., as enumerated in Rogers et al. (2017)) add a degree
of ambiguity to the interpretation. Schwalm et al. (2019) attempt a post hoc linkage between various com-
ponents of LSM structure within the MsTMIP archive with model skill scores but still emphasize that
their analysis undersamples the potential range of model configurations. Building intercomparison efforts
around model frameworks that use a modular complexity approach, as has been explored in specific
models around specific aspects of process representation, such as the stomatal conductance example
shown in Figure 2 (Franks et al., 2018; Knauer et al., 2015), but expanded and systematized such that
each model system could explore all aspects of the structural uncertainty questions investigated with a
breadth comparable or greater than current MIPs, would provide a much firmer basis for attributing
and understanding differences in model behaviors. Such an approach would allow the community to
move away from its current dependence on ensembles of opportunity and toward one built upon ensem-
bles of purpose.
One further potential benefit of such an approach is that model components could be developed collabora-
tively. Given that the majority of models in the CMIP6 intercomparison do not at present represent the key
processes relevant to biogeochemical feedbacks (nutrient cycling, fire, and dynamic vegetation) (Arora et al.,
2019), we argue that the current system, with its intrinsic massive duplication of effort, could be improved if
certain components were shared across models, with international teams of the relevant process‐domain
experts contributing to the representation of individual modules. Modern online collaboration and commu-
nication tools should make such “horizontal”division of effort more tenable for a new generation of land
surface modelers. The CICE consortium, an international team of sea ice model developers (Roberts
et al., 2018), provides an excellent example of this modus operandi.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 6of24
A notable barrier to developing a culture where model frameworks are deployed by default using parametric
and/or structural ensembles is the “one‐model‐one‐vote”format of the CMIP exercise (Eyring et al., 2016),
and other MIPs, wherein it is expected that single releases of each Earth system model (ESM) provide either
one integration, or an ensemble of integrations that cover the chaos‐induced natural variability in the cli-
mate system by slightly modifying initial conditions (Kay et al., 2015). Atmospheric and ocean processes,
in particular, are known to be highly dependent on initial condition uncertainty, but this focus is somewhat
misplaced in the context of land surface models, where parametric and structural variation is apparently
dominant over initial conditions at all timescales longer than a few years (Bonan & Doney, 2018). Shifting
some substantial fraction of computational resources away from initial‐condition‐focused approaches, and
toward structural and parametric uncertainty approaches, is thus also required to better represent the total
uncertainty inherent in land surface projections.
A further advantage of such a modular complexity framework may be to embed purely empirical
approaches, such as from machine learning, within a given model process. Such approaches may be useful
in solving two distinct sets of problems. The first is that, because of the large scope of land surface modeling,
several aspects of the models have little theoretical basis and are currently based on empirical or ad hoc
approaches. Some of these processes, such as phenology (Dahlin et al., 2015, 2017; Taylor & White, 2019)
and hydrology (Lapeyre et al., 2019) are the subject of a large number of observations, and thus may be
amenable to machine learning approaches. The second set of problems are ones where we may have a
detailed process‐level understanding, but where solving these equations are computationally too expensive
for a given application. In this case, surrogate or reduced order models, based on machine learning
approaches that have been trained on the full process representation models, may allow for higher fidelity
solutions than current, purely process‐driven approach used across LSMs. Given the increasing emphasis
on machine learning approaches and the successes of machine learning in solving problems in ESM
(Gentine et al., 2018) or offline hydrologic model (Fang et al., 2017; Shen, 2018) behavior, designing models
with an emphasis on modular complexity to allow for such hybrid approaches is a crucial challenge in mod-
eling the land surface.
3. Challenge: Heterogeneity and the Dimensionality of the Land Surface
3.1. Horizontal Heterogeneity
The boundary conditions of the land surface change as a function of the climate, which is typically provided
to LSMs as gridded products, either from Earth system models or climate “reanalysis”data products
(Sheffield et al., 2006). Even, however, at the highest resolutions foreseen using modern climate models
(1–10 km), land surface processes can be notably variable (Fox et al., 2008; Lundquist & Dettinger, 2005;
Tai et al., 2017) within a single “climatic”grid cell. Simulating areas with disparate functionality as a single
homogenous entity can lead to numerous errors in prognosis, particularly on account of strong nonlineari-
ties that are common features of land surface processes (Sellers et al., 2007). One approach to resolving sub-
grid heterogeneity is to further increase the resolution of the model. This approach was advocated from a
hydrological perspective by Wood et al. (2011) who described the implementation of “hyperresolution”mod-
els operating down to a scale of ~100 m as a “Grand Challenge”in hydrology. While the resolution at which
land surface models can be run continues to increase, the majority of LSMs can be run on spatial grids of
arbitrary resolution, and their typical deployment remains at much larger spatial scales (0.5–2°) in the con-
text of simulating global climate feedbacks and impacts. However, such resolutions only solve the heteroge-
neity problem where the length scale of the dimension of variation is of the same order of magnitude as the
grid cell size. In practice, many elements of landscape heterogeneity, including forest gaps and microtopo-
graphy manifest at smaller scales (Aas et al., 2019; Bonan et al., 1993; Thomas et al., 2008). In response to
Wood et al. (2011), Beven and Cloke (2012) noted this point, as well as the concern that “hyperresolution”
does not address the numerous epistemic uncertainties remaining.
To allow for operation across a multitude of scales, most modern land surface models, for example, SURFEX
(Masson et al., 2013), JULES (Burton et al., 2019); CLM5 (Lawrence et al., 2019); CABLE (Haverd et al., 2018),
ORCHIDEE, (Naudts et al., 2014), and JSBACH (Mauritsen et al., 2019), operate using some sort of subgrid
“tiling”system, which disaggregates pools and fluxes of relevance (water, energy, carbon, and nutrients)
along predetermined axes of variation that capture various properties of the physical surface. In modern
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 7of24
land surface models, the elements of heterogeneity most often captured include lakes, rocks, ice and urban
areas, as well as “soil.”Typically, the soil area is divided into plant functional type (PFT)‐based tiles,
potentially also including crop types, as well as bare soil. Tiles are typically defined as spatially implicit
aggregations of all of the area within a grid cell belonging to a particular land surface category.
As process complexity grows, however, the need to represent fine‐scale gradients in land surface heteroge-
neity grows with it. An ongoing theme of land surface model development is the proposal of additional axes
of variation which might be considered necessary to accurately represent particular land surface processes.
For example, Aas et al. (2017) illustrate the importance of representing snow‐covered and snow‐free parts of
an alpine landscape on runoff characteristics, illustrating that the melting of an area‐averaged snowpack can
be delayed by 2–3 months compared to a disaggregated and variable‐depth snowpack. Sellers et al. (2007)
argue that, on account of the nonlinearity between soil moisture, plant water stress and evapotranspiration,
that landscapes might be binned according to soil wetness, and the bulk evaporative stress functions calcu-
lated as an area average across bins. They, and latterly Baker et al. (2017) show that area averaging of soil
moisture (to reflect the patchiness of time since the last rainfall event) substantially reduces model respon-
siveness of evapotranspiration to rainfall events.
Fan et al. (2019) and the hydrology community more generally (Clark, Fan, et al., 2015), have argued that
landscapes might be tiled according to “hydrological response units”(HRU) which attempt to capture the
dynamics of lateral water drainage, and thus the nonlinear impacts on hydrological and ecological processes
downstream from the simulated topographic convergence of moisture. Such schemes define HRUs in terms
of fractions of a grid cell, and thus can represent bulk properties of hydrological variation without increasing
computational costs by orders of magnitude. Subin et al. (2014) illustrate the impacts of subgrid representa-
tion of hillslope hydrology in the GFDL model, noting, in particular, an increase in soil carbon resulting
from saturated lowland areas. Swenson et al. (2019) report the implementation of an HRU approach into
the Community Land Model v5, illustrating that the strongest impact of hydrological tiling occured in semi-
arid areas. HRU tiling efforts are underway in other LSMs, for example, JULES (https://www.ceh.ac.uk/
hydrojules). Fan et al. (2019) further note that as well as lateral drainage from hills to valley, slope aspect
(the difference between sunny and shady slopes) is another first‐order control on water and energy availabil-
ity across the landscape.
A largely orthogonal set of developments pertains to representing the basic principles of community ecology,
wherein the primary axis of variation in productive natural ecosystems is the patch mosaic generated by sec-
ondary succession: the processes of ongoing vegetation mortality and disturbance, gap formation, and the
recovery of vegetation back to a closed‐canopy state. Once again, many processes, in particular recruitment
of young plants, are nonlinear with respect to the surface light environment (which itself is also highly non-
linear with respect to canopy shading). An absence of gaps in “big leaf”ecosystem models leads to an inabil-
ity of trees to regenerate where the grid cell average forest has a closed canopy, leading to substantial
low‐biomass biases in models where forest demography is not resolved (Moorcroft et al., 2001). Similarly,
where natural systems are subject to ongoing disturbance from natural mortality, in any given grid cell,
anthropogenic disturbance also gives rise to a range of ages of secondary forest where biomass recovery also
proceeds in a nonlinear fashion after abandonment. Shevliakova et al. (2009) and Nabel et al. (2019) illus-
trate the importance of capturing the regrowth after disturbance in anthropogenically disturbed ecosystems
for simulating the terrestrial carbon sink. At larger scales of soil heterogeneity, some models also implement
tiling regimes for soil type. The ED2 land surface model (Longo et al., 2019), for example, divides the surface
into components of different soil texture (sand & clay fractions). Later versions of the same model also imple-
ment tiling for soil nutrients but specifically allied to variation in disturbance history (Trugman et al., 2016).
This makes at least seven (snow depth, hydrological regime, aspect, rainfall distribution, soil texture, soil fer-
tility, and time since disturbance plus time since land abandonment) relatively strong arguments for addi-
tional dimensionalities of subgrid‐scale horizontal heterogeneity within land surface models. In addition
to which, land surface processes outside of the “natural vegetation”type have been disaggregated within spe-
cific land use classes, into new crop types (including greater varieties of plant, plus the degree to which those
crops are fertilized or irrigated; D. M. Lawrence et al., 2019) and subcategories of urban environments (roofs,
sunlight and shaded walls, and impervious and pervious ground (Oleson et al., 2008). The majority of such
new dimensions of tiling are typically proposed in isolation, but clearly, when considered collectively, it is
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 8of24
not computationally tractable to divide up each climatic grid cell along all proposed dimensions. What is
missing is an overall strategy via which one might discern the most important axes of variation for a given
time and place. Capturing multiple simultaneous elements of landscape heterogeneity (Figure 3) must
surely be a feature of such a strategy.
In principle, one might reduce the dimensionality of the land surface by means of representing covariance
between different elements of heterogeneity (e.g., between hydrological regime, snow cover, and soil type).
Newman et al. (2014), for example, present a kmeans clustering approach to defining a predetermined tiling
scheme for a specific location, generating a set of 10 tiles that capture the dominant multifactoral regimes
affecting land surface dynamics within a given tile. Identification of functionally similar units is an intui-
tively appealing approach to reducing the dimensionality of the multifactoral tiling regime, but of course
rests on the nature of the questions one will ask of the model, for example, whether those are weighted
toward hydrological, biogeochemical or ecological questions. Further, a priori determination of physical
covariances assumes that the important axes of tiling are fixed in time and space.
3.2. Adaptive Tiling Strategies
While some axes of land surface variation (aspect, altitude, etc.) are indeed fixed on the timescales (tens to
hundreds of years) under consideration, many of the given reasons for subgrid tiling are by definition
dynamic in time and space. Thus, the degree to which tiling is needed along a particular axis varies. By
way of illustration, within the Ecosystem Demography model (Moorcroft et al., 2001) the degree of discreti-
zation of the landscape along the disturbance‐recovery axis is responsive to the current need for the model to
distinguish ecosystems of varying size‐structure. New tiles (or patches, in ED terminology) are formed when
a disturbance event occurs. Subsequently, patches with ecosystem structure that are considered “sufficiently
similar”(a user‐defined characteristic) are fused and become a single model unit, with the physical and bio-
logical characteristics recalculated in the process. In practice, this means that large parts of the world with
low productivity are not tiled for disturbance at all, saving significant computational time in the processes.
It is possible to imagine that areas impacted in a transient fashion by snow, rainfall, large gradients in soil
moisture, and so forth, might be amenable to an “adaptive tiling”approach. Difficulties with generalizing
this concept exist, pertaining in particular to the nontrivial complexity of merging and splitting highly com-
plex model objects, with possibly different timescales of persistence in land surface heterogeneity. Lawrence
et al. (2019) document the introduction of “dynamic”land unit transitions, which also allow fusion and
lumping of, for example, physical and biogeochemical soil states. Limited to a smaller number of specifically
transient dimensions, an extension of the ED approach to adaptive temporally variant tile resolution across
multiple dimensions of heterogeneity (e.g., snow, surface moisture) appears at least theoretically plausible
(Figure 4). This approach could allow the needs of multiple modeling communities to be met simulta-
neously, without expanding computational cost excessively. Such a scheme could operate within the context
of subgrid tiling based on temporally invariant (e.g., topographic) landscape features. Numerous modeling
Figure 3. Illustration of multiple concurrent aspects of surface heterogeneity within a hypothetical model grid cell.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 9of24
groups are implementing both hydrological response unit tiling (Hazenberg et al., 2015; Subin et al., 2014;
Swenson et al., 2019) and also vegetation demographics with disturbance tiling (Fisher et al., 2018; Longo
et al., 2019; Weng et al., 2017). Therefore, the most likely near‐term pathway for the representation of
subgrid horizontal heterogeneity is the nesting of vegetation demographics (VDM) models (with
time‐varying adaptive tiling schemes) inside hydrological response unit tiles (which typically are
determined from topography, and thus fixed in time). This methodology should allow for the prediction
of, say, the responses of vegetation to variation in water stress across landscapes, the variation in drought
mortality risk with differential access to water tables, and more generally allow a closer linkage between
hydrological environments and vegetation quantities, which should in principle lead to more accurate
responses to future change. It is possible to envisage further refinement of these architectures, both to
expand the adaptive elements within each HRU, as well as refinement of how the fixed tiles represent
covariance structures of other time‐invariant structures such as altitude, aspect and soil fertility. The
simultaneous operation of HRU and VDM schemes represents a substantial increase in the complexity
and cost of the representation of the land surface, and thus it is imperative that they are implemented in
ways that are flexible enough that they can either be turned off, and/or that the degree of disaggregation
can be modified in accordance with the nature of the research question. This capacity, to probe the
response of the model to alternative degrees of tiling in ecological and hydrological domains is a highly
novel tool that should both provide more tangible answers to outstanding questions of tiling strategy, and
provide a forum for greater collaboration across ecological and hydrological domains (e.g., Tague and
Dugger (2010)).
3.3. Patch Length Scale and Adjacency
Discretization of the land surface along any particular vector leads to (and indeed, is motivated by) a separa-
tion of state variables into categories which evolve according to variable input and output fluxes. In reality,
however, some diffusion of various quantities (energy, water, nutrients) between tiles existing in different
states is likely, reducing the heterogeneity of the system. The rate of diffusion is dependent on the length
scale (or adjacency) of different elements within the heterogeneous real‐world landscape matrix. Given,
however, that the tiling systems in LSMs are nearly always spatially implicit, and that each “tile”
Figure 4. Illustration of the potential for “dynamic adaptive”tiling regimes, to better capture features of the landscape
that are variable in time and thus require differing degrees of resolution under different circumstances. (a, b, and c)
Conditions under which forest structure, snow water equivalent, and surface moisture (respectively) are sufficiently het-
erogeneous to merit separation into independent tile units. (d, e, and f) Conditions where heterogeneity in these features is
low and would not require resolution. Panels (a) and (d) represent the mechanisms already present in ecosystem
demography‐type models, whereby new tiles are generated for each disturbance event and are then fused back into pre-
existing tiles if biomass structure is not sufficiently different to merit resolution. Thus, the model adapts to the complexity
of the landscape and does not generate tiles where vegetation stature is low. “Event‐based”tile generation and fusion
could thus also form the basis of representing time‐varying hydrological dynamics with new tile generated during snow
and rainfall events, becoming homogenized with melting and/or drying. Other aspects of tiling that are not dynamic on
the timescales in question (topography and aspect) would still require resolution at a higher level.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 10 of 24
represents an aggregation of a set of different elements of a complex spatial mosaic, the rate of diffusion of
quantities between tiles is difficult to ascertain. Typically, diffusion is either assumed to lead to complete
homogeneity, where no tiling exists, or impossible, where it does. Models that capture a degree of diffusion
between tiles would in principle need to be informed of the relevant length scale of their subtile elements
(Jupp & Twiss, 2006). As above, the length scale of time‐invariant features of the landscape might be distin-
guishable from remote sensing (topography, land use history, river and road fragmentation, likely requiring
machine learning methods to identify such features globally), while other time‐varying features might better
be retained from the inferred or simulated size distribution of disturbance events. For example, Koster
et al. (2019) illustrate the utility of soil moisture retrievals for informing the length scale of surface moisture
following rainfall events, illustrating its dependence on the type of rainfall (convective vs. large‐scale
precipitation).
Relatedly, some phenomena (fire, insects) intrinsically “spread”through the landscape via contagion, a pro-
cess which is difficult to model explicitly at the level of LSM grid cells. McCabe and Dietze (2019) propose a
method for estimating the size distribution of contagious disturbance events based on their disturbance,
initiation and spread probabilities as well as retaining through the simulation a metric of the “adjacency”
of tiled elements within grid cells. Their method evolves the spatial adjacency of disturbed patches through
time, and therefore could be generally applicable to the problem of retaining length‐scale information for
time‐varying quantities. An estimate of the initial adjacency (presumably including time‐invariant elements
of landscape patchiness) is required, again, from analysis of remote sensing data. The definition of a “patch”
for the purposes of calculating adjacency is, however, dependent on the target processes of interest.
McCabe and Dietze (2019) further argue that the inclusion of the concept of adjacency (and its dynamics)
would in principle allow for a myriad of additional ecological phenomena to be captured, including edge
effects on forest microclimate (of particular importance for the spread of fires), the dependence of dispersal
limitation on spatial arrangement of forests, simulation of invasive species dynamics, and also as above the
flow of matter and energy between patches. Thus, the extraction and use of both tiling units and their bulk
spatial relationships might also be elements of the “grand challenge”of representing the heterogeneity of the
land surface and the living systems that exist within and upon it.
3.4. Other Dimensions of Heterogeneity
Clark, Nijssen, et al. (2015) note that vertical stratification is much more refined in models that focus more
on vegetation physiology than in models that focus on the hydrological cycle. While early land surface mod-
els were built around a one‐dimensional representation of the terrestrial surface to correspond to a single
grid cell of an atmospheric model, they soon expanded to resolve vertical gradients in soil moisture and tem-
perature to better capture surface energy fluxes and the representation of plant water access. Resolving ver-
tical gradients in soil biogeochemistry, for example, is essential for systems such as permafrost‐affected soils,
where steep gradients in the soil physical climate mean that carbon cycles very differently at depth than at
the surface (Koven et al., 2015). Slater et al. (2017) also note the improved performance of models with ver-
tical profiles of temperature within the snowpack (Chadburn et al., 2015; van Kampenhout et al., 2017).
Vertical gradients in light, water status, temperature, leaf properties, and atmospheric conditions within the
vegetative canopy are typically not resolved in most mainstream land surface models. The “two‐stream”
approximation of Sellers (1985), provides a closed‐form solution for the scattering of direct and diffuse light
through homogenous vegetative canopies, thus collapsing the vertical structure down to one or two (e.g.,
sunlit vs. shaded leaves) states. On account of its computational parsimony, this approach was widely
adopted as standard in LSMs, precluding vertical representation of other quantities. In recent years, how-
ever, the gradual inclusion of increasing vertical detail has been an ongoing feature of land surface model
development, enabling more robust comparisons with field data, which by definition are made on particular
canopy layers. Implementations of the vertical structure of light absorption by leaves (Fisher et al., 2010;
Mercado et al., 2007), for example, provides the capacity to further vary plant physiological traits with
canopy depth, allowing models to represent the observation that plant traits do not in fact appear to scale
consistently with light availability as assumed by the two‐stream model (Lloyd et al., 2010; Meir et al., 2002).
The further introduction of vertical variation in leaf water potential, within the context of plant hydrody-
namic models, gives rise to the possibility of testing plant water status against field observations of leaf water
potential, stem water potential and sap flow, which differ substantially with canopy height and irradiance
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 11 of 24
(Christoffersen et al., 2016; Fisher et al., 2006; Joetzjer et al., 2018; Matheny et al., 2017; Mirfenderesgi
et al., 2016; X. Xu et al., 2016). A full treatment of the vertical structure of vegetative canopies, however,
requires resolution of how light, temperature, CO
2
and water content/humidity vary throughout the leaf
layers and canopy space. In forest ecosystems, in particular, large within‐canopy gradients generate substan-
tial environmental heterogeneity, which, as well as modulating gross canopy fluxes, is potentially an impor-
tant driver of niche separation and the capacity to represent functional diversity. A few land surface models
have recently implemented vertical gradients of irradiance, water content, leaf temperature, and also the
feedback between the evaporation of water into canopy air space and the humidity of the airspace, modulat-
ing by turbulence processes within the canopy and the roughness sublayer (which extends to roughly twice
the height of the canopy) (Bonan et al., 2018; Chen et al., 2016; Longo et al., 2019). These efforts represent the
cutting edge of physical representation of forest‐atmospheric exchange, and further challenge traditional
assumptions about the distinction between the atmospheric and the planetary surface, boundary layer, as
they bring the calculation of atmospheric mixing processes well into the realm traditionally occupied by
LSMs, once again raising issues related to the management of model complexity discussed earlier.
A significant intersection between the resolution of heterogeneity and the prior challenge of complexity
management is that agent‐based models typically require a representation of the relevant gradients of het-
erogeneity that are appropriate to the scale of the agents being modeled. This may apply equally to micro-
bially explicit soil biogeochemical models as to the cohort‐based vegetation models discussed above. In
principle, complex rhizosphere gradients in nutrient density radiating away from root surfaces are required
for appropriate simulation of microbial communities (Sulman et al., 2014; Wieder et al., 2015), and non-
linear dynamics of soil‐root resistance, a principal bottleneck on transpiration (Fisher et al., 2007; Sperry
et al., 1998; Williams et al., 2001). Similarly, soil physical heterogeneity, from mineralogical gradients at
micron scales to soil structural gradients at centimeter scales, may be crucial for governing both biogeo-
chemical gradients that govern soil microbial ecology, as well as macropore flows that determine
large‐scale hydrologic functioning. As model process representation shifts toward representing the agents
responsible for ecosystem function, rather than the aggregate behavior of ecosystem function, the need to
match scales of process with resolved heterogeneity represents one of the more complex edges of model
structural variation.
Reflecting our arguments in the previous section, coherent strategies to define the boundaries between inter-
acting complex systems will be necessary to allow informed and useful deployment of models with this level
of complexity in tandem with increasingly refined depictions of the horizontal domains included in LSM. As
the dimensionality of LSMs increases, it will be imperative to build models that are sufficiently flexible that
we can assess how resolving various gradients matters in the full system.
4. Challenge: Projecting the Temporal and Spatial Dynamics of
Model Parameters
Land surface models tend to have a large number of parameters. Hourdin et al. (2017) argue that atmo-
spheric models, are in general “founded on well understood physics combined with a number of heuristic
process representations.”LSMs, in contrast, combine numerous physical processes (themselves often depen-
dent on the complex heterogeneity of the surface, as previously described) with large numbers of biological
processes that in principle operate at a molecular level and are thus not practical to represent at their native
scale. These processes are encapsulated as parameters, often formulated at the scale of relevant observations
(e.g., individual leaves or trees). These parameters contribute to model uncertainty in a few different ways,
which we describe below, as a set of distinct problems of parametric uncertainty which we refer to here as
“parametric dynamics.”
Ideally, process‐based models should use input parameter values that represent properties of the system that
are static in time and space (Hourdin et al., 2017). For a plant trait or ecosystem property that is observed to
vary in time and space, choosing whether the model parameters should represent either the mean value of
that trait, parameters of observed relationship of the trait with the environment, parameters of a model that
optimizes that trait with respect to the environment, or a whole range of parameters representing alternative
types of plant that can be selected for or against according to the environment, is an open question. Thus,
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 12 of 24
idealized discussion of what aspects of ecosystems can and cannot consid-
ered “parameters”remain largely out of scope for LSMs.
4.1. Parametric Uncertainty and Fitting
The first problem of parametric uncertainty is the simplest: How do we
choose a set of parameters that gives a high agreement between model
predictions and a wide suite of data sets? While simple to state, the large
number of degrees of freedom makes this problem difficult to solve in
practice. Numerous efforts have been made to optimize parameters in
land surface models, using a variety of Bayesian approaches with priors
coming from plant trait or other data (LeBauer et al., 2013), and based
on optimizing to fit many different data sets, including optimizing hydro-
logic models against stream data (K. Beven & Binley, 1992), fitting gas
exchange parameters to eddy covariance data (Mäkelä et al., 2019; Post
et al., 2017) or using emulators (Fer et al., 2018; Sargsyan et al., 2014) or
adjoints to full land surface models (Verbeeck et al., 2011) to optimize
against eddy covariance observations. However, because of the high
dimensionality of parameters, such efforts typically run into the barrier
of equifinality: Running a model with many different sets of parameters
can lead to equally good fit to data, and these equally good models may
lead to widely divergent results under novel conditions (Tang
& Zhuang, 2008).
One possible solution to this is to optimize models parameters against
multiple types of data simultaneously, to allow separation by processes
acting on different timescales or on different aspects of model predictions,
as was done by MacBean et al. (2016). Extending such approaches to cover
the large set of processes and parameters relevant to LSM predictions is
itself an enormous challenge. Further, the direct assimilation of data for
calibration (Kaminski et al., 2013) also leads to philosophical questions
related to the interpretation of benchmarking and performance metrics
(Collier et al., 2018; D. M. Lawrence et al., 2019), which the LSM commu-
nity is yet to confront systematically.
A primary issue with LSMs is that biases in one part of a complex coupled model can undermine effective
calibration of other components. In principle, embracing a more comprehensive modularity framework
(section 2 and Figure 2) might allow for some individual processes to be calibrated in isolation with bound-
ary conditions prescribed from observations or data products (Kemp et al., 2014). Many existing calibration
studies have implicitly used low‐complexity versions of carbon cycle models, for example, Bloom et al. (2016).
Extension of this concept might facilitate the necessary dimensional reductions required to make this pro-
blem more tractable.
4.2. The Challenge of Living Systems: Predicting Changes in Ecosystem Properties
Beyond the problem of parameter optimization lies a deeper challenge: Many of the key canopy‐scale prop-
erties of the land surface are determined by the traits of plants or other organisms (Kattge et al., 2020), which
may vary enormously in their functional diversity across otherwise relatively homogeneous patches of
ground. Because plants are constantly growing, dying, reproducing, and competing for resources, these com-
positional mosaics are also dynamic in time. Thus, under the large changes to the environment currently
underway, we expect complex responses in the plant community composition at any given location that—
in addition to constituting a major class of ecological impacts to be understood in their own right—deter-
mine the distribution of plant traits (as defined at a canopy scale) and the dynamics of the land surface.
Thus, we must decide upon methods to predict how plant function at the community level is likely to shift in
response to global change. Approaches to this problem can be roughly grouped into three types: correlative,
optimizing, and competitive (Figure 5).
Figure 5. Illustration of correlative (a), optimizing (b), and competitive (c)
approaches to plant trait and thus ecosystem parameter dynamics under a
changing climate.
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 13 of 24
Correlative approaches take empirically observed relationships between environmental variables and trait
values, and assert that such relationships are conserved under global change. Many different variants on this
argument exist. Early dynamic vegetation models (Sitch et al., 2003; Woodward & Lomas, 2004) used a dis-
crete PFT version of this logic, where PFT distributions are constrained by bioclimatic indices, and each PFT
defined a set of traits in a land surface model; thus when climate changed, the PFT coverage changed with it,
which in turn changed the parameters representing plant processes of the model at a given grid cell. More
modern versions of this approach isolate specific traits that are clearly observed to vary as a function of envir-
onment within the lifetime of an individual plant, and allow these to vary in time and space and a function of
environmental conditions. Examples include the thermal acclimation of leaf photosynthetic and respiratory
temperature sensitivities (Atkin et al., 2015; Kumarathunge et al., 2019; Lombardozzi et al., 2015; Slot
et al., 2014), models that define allocation patterns (Thornton et al., 2007), N fixation (Thornton et al., 2007),
and stem mortality rates (Delbart et al., 2010) all as functions of NPP, or more general relationships between
plant traits and climate as inferred across multiple traits (Butler et al., 2017; van Bodegom et al., 2014;
Verheijen et al., 2015).
Optimizing approaches work by, in principle, constraining predictions of plant trait values with the principle
that evolution and competitive dynamics should have selected trait values that confer the highest “fitness”in
a given environment. Thus, one can hypothesize that these optimal values are those most likely to be pre-
sent. The crucial requirement for such approaches is to be able to define a functional relationship of costs
and benefits (or fitness criteria) for a given trait value as conditional on the environment, which can then
be optimized. Like correlative approaches, optimizing approaches make an assumption of rapid adjustment
to environmental variation, and thus may be only strictly valid for traits that can be shown to vary over the
lifetime of an individual plant. Examples of the expanding literature on plant optimality theory include the
prediction of leaf nitrogen allocation to colimitation metabolic processes under varying environmental con-
ditions (Smith et al., 2019; C. Xu et al., 2012), the response of canopy nitrogen to CO
2
fertilization
(Franklin, 2007; Franklin et al., 2009), control of stomata to maximize assimilation while avoiding dessica-
tion (Bonan et al., 2014; Eller et al., 2018; Kennedy et al., 2019; Medlyn et al., 2011; Williams et al., 1996; Wolf
et al., 2016; X. Xu et al., 2016). Wang et al. (2017) predict internal leaf CO
2
balance based on a model that
optimizes assimilation while accounting for the costs of water transport and nutrient uptake, and Street
et al. (2012) show that N profiles in arctic canopies are consistent with optimal allocation theory controlled
by diffuse light profiles. All optimality approaches rest on the determination of a proxy of fitness that should
be maximized (which is uncertain, per Caldararu et al. (2019)), the definition of a timescale over which the
optimization is relevant, and an assumption concerning the physiological limits of optimization and the
timescales within which it can be achieved. Optimization is, for different purposes assumed to occur at scales
from the lifetime of a single plant, to the timescale of adaptation of a whole ecosystem to its prevailing
climate. The capacity for whole ecosystems to rapidly change functionality under a changing climate
(particularly those dominated by long lived trees) may be slowed by the rate of demographic change and
migration of better adapted individuals into the system, and therefore optimality approaches should perhaps
be viewed as the likely equilibrium state of a system (with the caveat that the optimal strategy for individuals
does not necessarily represent the evolutionary stable strategy within a competitive framework
(Dybzinski et al., 2011).
Competitive approaches more directly address the need to simulate the demographics of transient ecosystem
states. Instead of optimizing a specific function, competitive approaches attempt to resolve the population
dynamics of individual agents, competing for resources in the context of the environment. Thus, the popula-
tion dynamics themselves act to find optimal values among a set of possible trait values from the pool of com-
peting types. The dynamics of competition may range from Lotka‐Volterra‐type formulations with the
competing agents being canopies comprised of plants from a given PFT (Cox, 2001; Harper et al., 2016) to
demographic or individual based models where the agents are either cohorts of size‐resolved plants or indi-
vidual plants competing for space in the canopy and other resources (Christoffersen et al., 2016; Moorcroft
et al., 2001; Purves et al., 2008;Sakschewski et al., 2015 ; Scheiter et al., 2013). Advantages of this approach
are that it can in principle capture the timescales of community adjustment to global change, as well as that
it does not require an a priori estimation of a fitness function to be optimized, and thus may be applicable to a
wider range of traits, or interactions between traits, than optimizing or correlative approaches. A significant
challenge of this methodology is the maintenance of functional diversity, particularly in models that
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 14 of 24
specifically are not inclusive of many mechanisms known to stabilize competitive exclusion processes
between plants with differential fitness (Chesson, 2000; Gravel et al., 2011), as well as spatial dispersal pro-
cesses, high dimensional resource partitioning and the density‐dependent impact of “natural enemies”on
demographic stability. Investigation of the maintenance of functional diversity in demographic vegetation
models is thus an emerging field of research in this domain (Falster et al., 2017; Fisher et al., 2018; Koven
et al., 2019; Maréchaux & Chave, 2017; Powell et al., 2018).
A further challenge in competitive approaches is to better understand how the definition of PFTs in a given
model relates to model predictions. As model complexity has grown, such definitions have grown more com-
plex, from early approaches that equated PFTs with biomes to newer approaches that define PFTs via multi-
ple axes of trait variation. Key challenges relate to the specific ways in which continuous and
multidimensional trait variation is discretized into the axes of trait coordination that define PFTs, including
(1) the number of axes needed to distinguish a comprehensive set of PFTs needed to solve a given problem,
(2) how these axes are specified from trait observations while taking into account both represented and unre-
presented tradeoffs that may prevent dominance by any one PFT (Sakschewski et al., 2015; Scheiter
et al., 2013), and (3) how finely should a set of possible PFTs resolve any given axis of trait variation?
We outline three alternative, but not necessarily competing, philosophies for addressing the dynamics of
organism traits—and thus ecosystem properties—in time and space. In principle, all of these approaches
(correlative, optimizing, and competitive) may be combined in a given land surface model, but theory for
how to do so is not well developed. For any given trait, the inclusion of a high degree of plasticity through
either correlative or optimizing approaches would reduce the role that such a trait plays in determining com-
petitive outcomes. In principle, to best reflect reality, observed within‐lifetime plastic responses to climate
could in principle be nested within competitive demographic approaches for projecting distributions of traits
where no such individual‐level plasticity is evident.
To capture trait dynamics on timescales of many generations (or to take the optimisation of plant fitness in
the presence of competition properly into account) would require demographic models to be embedded
within representations of trait evolution (including mutation and selection), per (Falster et al., 2017;
Scheiter et al., 2013). This consideration, combined with the need to enhance coexistence of functional types
within competitive models, suggests a specific need to open a greater dialog with other formerly separate dis-
ciplines in ecology. The field of biodiversity and ecosystem function is also motivated, for example, by
improving understanding of the means by which functional variations within extant communities of species
may or may not confer resistance and/or resilience to climate shifts (Hooper et al., 2005; Isbell et al., 2015;
Turnbull et al., 2013; Yachi & Loreau, 1999). Interactions between coexistence theorists and land surface
modelers are rare, a situation which we hope improves as our ecological tools at the intersection of these
fields mature.
5. Further Challenges in Land Surface Model Science
In this discussion of “grand challenges”we have focused on several higher‐order elements of LSM develop-
ment: complexity management, surface heterogeneity, and parametric dynamics. There remain numerous
other aspects of LSM science where substantial progress is necessary, but for which the overall solutions
are perhaps more apparent within contemporary organizational structures. For example, development of
the scientific collaboration and software infrastructure to conduct comprehensive and rapid model bench-
marking will continue to be a major priority of the community, with particular reference to the adoption
of community tools to avoid duplication of effort (Abramowitz et al., 2008; Best et al., 2015; Collier et al., 2018;
Nearing et al., 2018) (and as illustrated at, e.g., www.ilamb.org), and these efforts will need to be extended to
encompass data products and outputs relevant to the emerging capacities of LSMs (hillslopes, vegetation
demographics, etc.).
Relatedly, a major focus is required to generate and apply data products that can be used within land surface
model development, from the ever‐expanding scope of Earth observation remote sensing activities and other
large data sets and data synthesis activities. This is an especially wide and active field in which clearly a great
quantity of effort is already expended (Duncanson et al., 2019; Houborg et al., 2015; Quegan et al., 2019;
Stavros et al., 2017) and the depth and breadth of new satellite observations of, for example, biomass and
canopy structure, carbon dioxide, multispectral and hyperspectral surface reflectance, chlorophyll
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 15 of 24
fluorescence, emissivity, and water content will likely revolutionize our knowledge of the terrestrial bio-
sphere and our capacity for predictive understanding. The development of data products (including those
scaled from site level observations to gridded products, e.g., Beer et al., 2010) should bear in mind, in parti-
cular, the likely future trajectory of land surface model developments in future years. Increasing process
resolution of LSMs (along the dimensions discussed above) should allow significant improvement in the
capacity for Earth observations of the real world to be directly comparable with model states and fluxes,
and both activities should be designed to leverage this potential, in particular, by prioritizing the availability
of data products more closely related to the raw signals than to products aggregated by default to the same
degree as older land surface schemes.
As LSMs have matured to provide more detailed representations of the land surface, another key develop-
ment has been to follow the lead of atmospheric models by providing short‐term forecast cycles for aspects
such as hydrologic prediction (NOAA, 2016). Relatedly, in the context of short‐term forecasting, numerous
“land data assimilation systems”(LDAS) have been implemented in the last two decades, as reviewed by
(Xia et al., 2019). The focus of these efforts is typically on improving the system state for the purposes of bet-
ter short‐to medium‐term predictability. Such efforts are useful for identifying where LSMs do and do not
have predictive skill, but with some exceptions (Fox et al., 2018; Kaminski et al., 2013; Peylin et al., 2016)
efforts are not yet particularly well integrated into climate‐focused land surface modeling activities. To some
extent, short‐term weather forecasting operations are concerned only with a subset of the problems faced by
climate‐oriented modeling activities. Integration of observed leaf area index, snow cover and surface soil
moisture, for example, overrides many of the higher‐order predictive processes in a complex land surface
scheme. The emergence of the concept of ecological forecasting (Dietze et al., 2018) however, aims to probe
and illustrate the degree to which the concepts of data assimilation can help constrain predictions dependent
on accurate representation of ecosystem processes (Fer et al., 2020; Niu et al., 2014).
More practical areas of concern relate to the availability of sufficient computing resources, and to the tech-
nical challenges of implementing modularized code structures and adaptive tiling schemes. Meeting these
challenges—via access to supercomputer infrastructure, and critically via the entrainment of modern profes-
sional software engineering—is intrinsically linked to the need to strengthen funding infrastructures for
land surface modeling activities. LSMs have most typically developed associated with and adjunct to atmo-
spheric modeling activities. Simultaneously, the scope of LSMs has expanded such that their applications
rest firmly within the domains of hydrology, ecology, geography and biogeochemistry. This situation is chal-
lenging for the majority of national‐or agency‐level funding networks, particularly those where funding
streams are aligned with more traditional academic disciplines. Strengthening the connection between
the LSM community and the disciplines on which the evolving model capacity both encroaches and depends
is the most likely means by which changing the status quo can be achieved.
6. Conclusions
Global concern is more deeply focusing on the fate of the terrestrial biosphere and the land surface, as we
accelerate toward rapid changes in climate, atmospheric composition, and land use. With this increased
focus, studies using land surface models regularly make international headlines. LSMs are the primary tools
that we have to simulate conditions for life on the terrestrial surface of planet Earth and play a crucial role in
our ability to estimate the quantity of carbon that humanity can, in principle, emit to limit climate change to
any given international target (e.g., 1.5 and 2 °C).
Despite this extraordinary degree of interest, the number of individual scientists and software engineers
actively developing LSMs could comfortably all be housed in one medium‐sized village, and even the most
active LSM teams struggle to meet multifaceted demands placed upon them. These include predictions of the
terrestrial cycling of carbon, water, energy, nitrogen, methane, and N
2
O, in the context of changing climate,
atmospheric CO
2
, ozone, and N deposition, as well as vegetation cover, land use, fire, and
crop management,.
We argue that new paradigms in complexity management, in the flexible representation of surface hetero-
geneity, and in the representation of parametric and trait dynamics, are needed to meet the overwhelming
challenges that are necessarily imposed upon our community by the questions of society. To modify existing
development practices to encompass the modular complexity and adaptive heterogeneity frameworks, we
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 16 of 24
suggest is a major scientific and software engineering challenge. We emphasize that modern collective, open
approaches to code development, benchmarking, computational methods, data product development, and
publication are necessary to facilitate these paradigm shifts. Further, modularization of model code and
the development of international teams of experts collaborating on advanced process representation of dis-
tinct model elements would have considerable benefits in terms of reduced duplication of effort, while archi-
tectures built on modular complexity approaches may allow differing institutional interests to be
represented via alternate structural configurations and parametric choices within a given model.
Significant effort is required to meet these urgent needs. The status quo investment in land surface model
development is inadequate for the task at hand, and it will not suffice if we seek an ability to make robust
projections of the status of the terrestrial land surface and the living systems which inhabit it over decadal
to century timescales.
Modern LSMs represent a unique and powerful intersection of the fields of physics, biochemistry, physiol-
ogy, ecology, hydrology, geography, statistics, mathematics, and high‐performance computing. To solve
our grand challenges, we must raise the profile and importance of LSMs within all these contributing fields.
Given the overwhelming importance of understanding how our modification of Earth's atmosphere and cli-
mate will affect our direct living conditions and the ecological and hydrological systems on which we
depend, it is imperative that LSMs step out of the shadow of their “atmospheric boundary condition”
beginnings and develop into a science in their own right.
References
Aas, K. S., Gisnås, K., Westermann, S., & Berntsen, T. K. (2017). A tiling approach to represent subgrid snow variability in coupled land
surface–atmosphere models. Journal of Hydrometeorology,18(1), 49–63. https://doi.org/10.1175/jhm‐d‐16‐0026.1
Aas, K. S., Martin, L., Nitzbon, J., Langer, M., Boike, J., Lee, H., et al. (2019). Thaw processes in ice‐rich permafrost landscapes
represented with laterally coupled tiles in a land surface model. The Cryosphere,13(2), 591–609. https://doi.org/10.5194/
tc‐13‐591‐2019
Abramowitz, G., Leuning, R., Clark, M., & Pitman, A. (2008). Evaluating the performance of land surface models. Journal of Climate,
21(21), 5468–5481. https://doi.org/10.1175/2008JCLI2378.1
Argles, A. P. K., Moore, J. R., Huntingford, C., Wiltshire, A. J., Jones, C. D., & Cox, P. M. (2019). Robust Ecosystem Demography (RED): a
parsimonious approach to modelling vegetation dynamics in Earth system models. Geoscientific Model Development Discussion. https://
doi.org/10.5194/gmd‐2019‐300
Arora, V. K., & Boer, G. J. (2006). Simulating competition and coexistence between plant functional types in a dynamic vegetation model.
Earth Interactions,10(10), 1–30. https://doi.org/10.1175/EI170.1
Arora, V. K., Boer, G. J., Friedlingstein, P., Eby, M., Jones, C. D., Christian, J. R., et al. (2013). Carbon‐concentration and carbon‐climate
feedbacks in CMIP5 Earth system models. Journal of Climate,26(15), 5289–5314. https://doi.org/10.1175/JCLI‐D‐12‐00494.1
Arora, V. K., Katavouta, A., Williams, R. G., Jones, C. D., Brovkin, V., Friedlingstein, P., et al. (2019). Carbon‐concentration and
carbon‐climate feedbacks in CMIP6 models, and their comparison to CMIP5 models. Biogeosciences Discuss. https://doi.org/10.5194/
bg‐2019‐473
Atkin, O. K., Bloomfield, K. J., Reich, P. B., Tjoelker, M. G., Asner, G. P., Bonal, D., et al. (2015). Global variability in leaf respiration in
relation to climate, plant functional types and leaf traits. The New Phytologist,206(2), 614–636. https://doi.org/10.1111/nph.13253
Baker, I. T., Sellers, P. J., Denning, A. S., Medina, I., Kraus, P., Haynes, K. D., & Biraud, S. C. (2017). Closing the scale gap between land
surface parameterizations and GCMs with a new scheme, SiB3‐bins. Journal of Advances in Modeling Earth Systems,9, 691–711. https://
doi.org/10.1002/2016ms000764
Beer, C., Reichstein, M., Tomelleri, E., Ciais, P., Jung, M., Carvalhais, N., et al. (2010). Terrestrial gross carbon dioxide uptake: Global
distribution and covariation with climate. Science,329(5993), 834–838. https://doi.org/10.1126/science.1184984
Bellassen, V., Le Maire, G., Dhôte, J. F., Ciais, P., & Viovy , N. (2010). Modelling forest management within a global vegetation model—Part
1: Model structure and general behaviour. Ecological Modelling,221(20), 2458–2474. https://doi.org/10.1016/j.ecolmodel.2010.07.008
Best, M. J., Abramowitz, G., Johnson, H. R., Pitman, A. J., Balsamo, G., Boone, A., et al. (2015). The plumbing of land surface models:
Benchmarking model performance. Journal of Hydrometeorology,16(3), 1425–1442. https://doi.org/10.1175/JHM‐D‐14‐0158.1
Beven, K., & Binley, A. (1992). The future of distributed models: Model calibration and uncertainty prediction. Hydrolog ical Processes,6(3),
279–298. https://doi.org/10.1002/hyp.3360060305
Beven, K. J., & Cloke, H. L. (2012). Comment on “Hyperresolution global land surface modeling: Meeting a grand challenge for monitoring
Earth's terrestrial water”by Eric F. Wood et al.: Commentary. Water Resources Research,48, W01801. https://doi.org/10.1029/
2011WR010982
Bierkens, M. F. P., Bell, V. A., Burek, P., Chaney, N., Condon, L. E., David, C. H., et al. (2015). Hyper‐resolution global hydrological
modelling: what is next?: “Everywhere and locally relevant”.Hydrological Processes,29(2), 310–320. https://doi.org/10.1002/hyp.10391
Bisht, G., & Riley, W. J. (2019). Development and verification of a numerical library for solving global terrestrial multiphysics problems.
Journal of Advances in Modeling Earth Systems,11, 1516–1542. https://doi.org/10.1029/2018MS001560
Bloom, A. A., Exbrayat, J.‐F., van der Velde, I. R., Feng, L., & Williams, M. (2016). The decadal state of the terrestrial carbon cycle: Global
retrievals of terrestrial carbon allocation, pools, and residence times. Proceedings of the National Academy of Sciences of the United States
of America,113(5), 1285–1290. https://doi.org/10.1073/pnas.1515160113
Bonan, G. B. (1996). A land surface model (LSM Version 1.0) for ecological, hydrological, and atmospheric studies: Technical description
and user's guide. UCAR/NCAR. https://doi.org/10.5065/D6DF6P5X
Bonan, G. B., & Doney, S. C. (2018). Climate, ecosystems, and planetary futures: The challenge to predict life in Earth system models.
Science,359(6375), eaam8328. https://doi.org/10.1126/science.aam8328
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 17 of 24
Acknowledgments
R. A. F. acknowledges the support of
the National Center for Atmospheric
Research, which is a major facility
sponsored by the National Science
Foundation under Cooperative
Agreement 1852977. C. D. K
acknowledges support by the Director,
Office of Science, Office of Biological
and Environmental Research of the U.
S. Department of Energy under
Contract DE‐AC02‐05CH11231
through the Early Career Research
Program, the Regional and Global
Model Analysis Program (RUBISCO
SFA), and the Next Generation
Ecosystem Experiment‐Tropics
(NGEE‐Tropics) project. We thank Ben
Sanderson (CERFACS), Ryan Knox
(LBNL), and David Lawrence (NCAR)
for helpful discussion. Data Availability
Statement: No data were used in
writing this article.
Bonan, G. B., Patton, E. G., Harman, I. N., Oleson, K. W., Finnigan, J. J., Lu, Y., & Burakowski, E. A. (2018). Modeling canopy‐induced
turbulence in the Earth system: A unified parameterization of turbulent exchange within plant canopies and the roughness sublayer
(CLM‐ml v0). Geoscientific Model Development. https://doi.org/10.5194/gmd‐11‐1467‐2018
Bonan, G. B., Pollard, D., & Thompson, S. L. (1993). Influence of subgrid‐scale heterogeneity in leaf area index, stomatal resistance, and soil
moisture on grid‐scale land–atmosphere interactions. Journal of Climate,6(10), 1882–1897. https://doi.org/10.1175/1520‐0442(1993)
006<1882:IOSSHI>2.0.CO;2
Bonan, G. B., Williams, M., Fisher, R. A., & Oleson, K. W. (2014). Modeling stomatal conductance in the Earth system: Linking leaf
water‐use efficiency and water transport along the soil–plant–atmosphere continuum. Geoscientific Model Development,7(5), 2193–2222.
https://doi.org/10.5194/gmd‐7‐2193‐2014
Burton, C., Betts, R., Cardoso, M., Feldpausch, T. R., Harper, A., Jones, C. D., et al. (2019). Representation of fire, land‐use change and
vegetation dynamics in the joint UK land environment simulator vn4.9 (JULES). Geoscientific Model Development,12(1), 179–193.
https://doi.org/10.5194/gmd‐12‐179‐2019
Butler, E. E., Datta, A., Flores‐Moreno, H., Chen, M., Wythers, K. R., Fazayeli, F., et al. (2017). Mapping local and global variability in plant
trait distributions. Proceedings of the National Academy of Sciences of the United States of America,114(51), E10937–E10946. https://doi.
org/10.1073/pnas.1708984114
Caldararu, S., Thum, T., Yu, L., & Zaehle, S. (2019). Whole‐plant optimality predicts changes in leaf nitrogen under variable CO
2
and
nutrient availability. Ecology. bioRxiv. https://doi.org/10.1101/785329
Chadburn, S., Burke, E., Essery, R., Boike, J., Langer, M., Heikenfeld, M., et al. (2015). An improved representation of physical per-
mafrost dynamics in the JULES land‐surface model. Geoscientific Model Development,8(5), 1493–1508. https://doi.org/10.5194/
gmd‐8‐1493‐2015
Chen, Y., Ryder, J., Bastrikov, V., McGrath, M. J., Naudts, K., Otto, J., et al. (2016). Evaluating the performance of land surface model
ORCHIDEE‐CAN v1. 0 on water and energy flux estimation with a single‐and multi‐layer energy budget scheme. Geoscientific Model
Development,9(9), 2951–2972. https://doi.org/10.5194/gmd‐9‐2951‐2016
Chesson, P. (2000). Mechanisms of maintenance of species divers ity. Annual Review of Ecology and Systematics,31(1), 343–366. https://doi.
org/10.1146/annurev.ecolsys.31.1.343
Christoffersen, B. O., Gloor, M., Fauset, S., Fyllas, N. M., Galbraith, D. R., Baker, T. R., et al. (2016). Linking hydraulic traits to tropical
forest function in a size‐structured and trait‐driven model (TFS v.1‐hydro). Geoscientific Model Development,9(11), 4227–4255. https://
doi.org/10.5194/gmd‐9‐4227‐2016
Clark, M. P., Fan, Y., Lawrence, D. M., Adam, J. C., Bolster, D., Gochis, D. J., et al. (2015). Improving the representation of hydrologic
processes in Earth system models. Water Resources Research,51, 5929–5956. https://doi.org/10.1002/2015WR017096
Clark, M. P., Kavetski, D., & Fenicia, F. (2011). Pursuing the method of multiple working hypotheses for hydrological modeling: Hypothesis
testing in hydrology. Water Resources Research,47, W09301. https://doi.org/10.1029/2010WR009827
Clark, M. P., Nijssen, B., Lundquist, J. D., Kavetski, D., Rupp, D. E., Woods, R. A., et al. (2015). A unified approach for process‐based
hydrologic modeling: 1. Modeling concept. Water Resources Research,51, 2498–2514. https://doi.org/10.1002/2015WR017198
Claussen, M., Mysak, L., Weaver, A., Crucifix, M., Fichefet, T., Loutre, M.‐F., et al. (2002). Earth system models of intermediate com-
plexity: Closing the gap in the spectrum of climate system models. Climate Dynamics,18(7), 579–586. https://doi.org/10.1007/
s00382‐001‐0200‐1
Collier, N., Hoffman, F. M., Lawrence, D. M., Keppel‐Aleks, G., Koven, C. D., Riley, W. J., et al. (2018). The international land model
benchmarking (ILAMB) system: Design, theory, and implementation. Journal of Advances in Modeling Earth Syste ms,10, 2731–2754.
https://doi.org/10.1029/2018MS001354
Coon, E. T., David Moulton, J., & Painter, S. L. (2016). Managing complexity in simulations of land surface and near‐surface processes.
Environmental Modelling & Software,78, 134–149. https://doi.org/10.1016/j.envsoft.2015.12.017
Cox, P. M. (2001). Description of the TRIFFID dynamic global vegetation model. Hadley Centre technical note 24. Bracknell, UK: Hadley
Centre, Met Office.
Cox, P. M., Betts, R. A., Bunton, C. B., Essery, R. L. H., Rowntree, P. R., & Smith, J. (1999). The impact of new land surface physics on the
GCM simulation of climate and climate sensitivity. Climate Dynamics,15(3), 183–203. https://doi.org/10.1007/s003820050276
Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A., & Totterdell, I. J. (2000). Acceleration of global warming due to carbon‐cycle feedbacks in a
coupled climate model. Nature,408, 184–187.
Dahlin, K. M., Fisher, R. A., & Lawrence, P. J. (2015). Environmental drivers of drought deciduous phenology in the community land
model. Biogeosciences,12(16), 5061–5074. https://doi.org/10.5194/bg‐12‐5061‐2015
Dahlin, K. M., Ponte, D. D., Setlock, E., & Nagelkirk, R. (2017). Global patterns of drought deciduous phenology in semi‐arid and
savanna‐type ecosystems. Ecography,40(2), 314–323. https://doi.org/10.1111/ecog.02443
De Kauwe, M. G., Medlyn, B. E., & Zaehle, S. (2014). Where does the carbon go? A model–data intercomparison of vegetation carbon
allocation and turnover processes at two temperate forest free‐air CO
2
enrichment sites. The New Phytologist. https://doi.org/10.1111/
nph.12847
Decharme, B., Delire, C., Minvielle, M., Colin, J., Vergnes, J., Alias, A., et al. (2019). Recent changes in the ISBA‐CTRIP land surface system
for use in the CNRM‐CM6 climate model and in global off‐line hydrological applications. Journ al of Advances in Modeling Earth Systems,
11, 1207–1252. https://doi.org/10.1029/2018MS001545
Delbart, N., Ciais, P., Chave, J., Viovy, N., Malhi, Y., & Le Toan, T. (2010). Mortality as a key driver of the spatial distribution of above-
ground biomass in Amazonian forest: Results from a dynamic vegetation model. Biogeosciences,7(10), 3027–3039. https://doi.org/
10.5194/bg‐7‐3027‐2010
Dickinson, R. E., Henderson‐Sellers, A., Rosenzweig, C., & Sellers, P. J. (1991). Evapotranspiration models with canopy resistance for
use in climate models, a review. Agricultural and Forest Meteorology,54(2‐4), 373–388. https://doi.org/10.1016/0168‐1923(91)
90014‐H
Dietze, M. C., Fox, A., Beck‐Johnson, L. M., Betancourt, J. L., Hooten, M. B., Jarnevich, C. S., et al. (2018). Iterative near‐term ecological
forecasting: Needs, opportunities, and challenges. Proceedings of the National Academy of Sciences of the United States of America,115(7),
1424–1432. https://doi.org/10.1073/pnas.1710231115
Dietze, M. C., & Matthes, J. H. (2014). A general ecophysiological framework for modelling the impact of pests and pathogens on forest
ecosystems. Ecology Letters,17(11), 1418–1426. https://doi.org/10.1111/ele.12345
Duncanson, L., Armston, J., Disney, M., Avitabile, V., Barbier, N., Calders, K., et al. (2019). The importance of consistent global
forest aboveground biomass product validation. Surveys in Geophysics,40(4), 979–999. https://doi.org/10.1007/s10712‐019‐
09538‐8
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 18 of 24
Dybzinski, R., Farrior, C., Wolf, A., Reich, P. B., & Pacala, S. W. (2011). Evolutionarily stable strategy carbon allocation to foliage, wood,
and fine roots in trees competing for light and nitrogen: An analytically tractable, individual‐based model and qua ntitative comparisons
to data. The American Naturalist,177(2), 153–166. https://doi.org/10.1086/657992
Eller, C. B., Rowland, L., Oliveira, R. S., Bittencourt, P. R. L., Barros, F. V., da Costa, A. C. L., et al. (2018). Modelling tropical forest
responses to drought and El Niño with a stomatal optimization model based on xylem hydr aulics. Philosophical Transactions of the Royal
Society of London. Series B, Biological Sciences,373(1760), 20170315. https://doi.org/10.1098/rstb.2017.0315
Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R. J., & Taylor, K. E. (2016). Overview of the coupled model
Intercomparison project phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development,9(5), 1937–1958.
https://doi.org/10.5194/gmd‐9‐1937‐2016
Falster, D. S., Brännström, Å., Westoby, M., & Dieckmann, U. (2017). Multitrait successional forest dynamics enable diverse competitive
coexistence. Proceedings of the National Academy of Sciences of the United States of America,114(13), E2719–E2728. https://doi.org/
10.1073/pnas.1610206114
Fan, Y., Clark, M., Lawrence, D. M., Swenson, S., Band, L. E., Brantley, S. L., et al. (2019). Hillslope hydrology in global change research and
Earth system modeling. Water Resources Research,55,1737–1772. https://doi.org/10.1029/2018WR023903
Fang, K., Shen, C., Kifer, D., & Yang, X. (2017). Prolongation of SMAP to spatiotemporally seamless coverage of continental U.S. using a
deep learning neural network. Geophysical Research Letters,44(21), 11,030–11,039. https://doi.org/10.1002/2017GL075619
Farrior, C. E., Bohlman, S. A., Hubbell, S., & Pacala, S. W. (2016). Domina nce of the suppressed: Power‐law size structure in tropical forests.
Science,351(6269), 155–157. https://doi.org/10.1126/science.aad0592
Fatichi, S., Pappas, C., Zscheischler, J., & Leuzinger, S. (2019). Modelling carbon sources and sinks in terrestrial vegetation. The New
Phytologist,221(2), 652–668. https://doi.org/10.1111/nph.15451
Fer, I., Gardella, A. K., Shiklomanov, A. N., Serbin, S. P., De Kauwe, M. G., Raiho, A., et al. (2020). Beyond modeling: A roadmap to
community cyberinfrastructure for ecological data‐model integration. Retrieved from https://www.preprints.org/manuscript/
202001.0176
Fer, I., Kelly, R., Moorcroft, P. R., Richardson, A. D., Cowdery, E. M., & Dietze, M. C. (2018). Linking big models to big data: Efficient
ecosystem model calibration through Bayesian model emulation. Biogeosciences ,15(19), 5801–5830. https://doi.org/10.5194/
bg‐15‐5801‐2018
Fisher, R. A., Koven, C. D., Anderegg, W. R. L., Christoffersen, B. O., Dietze, M. C., Farrior, C. E., et al. (2018). Vegetation demo-
graphics in Earth system models: A review of progress and priorities. Global Change Biology,24(1), 35–54. https://doi.org/10.1111/
gcb.13910
Fisher, R. A., McDowell, N., Purves, D., & Moorcroft, P. (2010). Assessing uncertainties in a second‐generation dynamic vegetation model
caused by ecological scale limitations. The New Phytologist. https://doi.org/10.1111/j.1469‐8137.2010.03340.x
Fisher, R. A., Wieder, W. R., Sanderson, B. M., Koven, C. D., Oleson, K. W., Xu, C., et al. (2019). Parametric controls on vegetation responses
to biogeochemical forcing in the CLM5. Journal of Advances in Modeling Earth Systems,11(9), 2879–2895. https://doi.org/10.1029/
2019MS001609
Fisher, R. A., Williams, M., da Costa, A. L., Malhi, Y., da Costa, R. F., Almeida, S., & Meir, P. (2007). The response of an eastern Amazonian
rain forest to drought stress: Results and modelling analyses from a throughfall exclusion experiment. Global Change Biology,13(11),
2361–2378. https://doi.org/10.1111/j.1365‐2486.2007.01417.x
Fisher, R. A., Williams, M., Do Vale, R. L., Da Costa, A. L., & Meir, P. (2006). Evidence from Amazonian forests is consistent with isohydric
control of leaf water potential. Plant, Cell & Environment,29(2), 151–165. https://doi.org/10.1111/j.1365‐3040.2005.01407.x
Fox, A. M., Hoar, T. J., Anderson, J. L., Arellano, A. F., Smith, W. K., Litvak, M. E., et al. (2018). Evaluation of a data assimilation system for
land surface models using CLM4.5. Journal of Advances in Modeling Earth Systems,10, 2471–2494. https://doi.org/10.1029/
2018MS001362
Fox, A. M., Huntley, B., Lloyd, C. R., Williams, M., & Baxter, R. (2008). Net ecosystem exchange over heterogeneous Arctic tundra: Scaling
between chamber and eddy covariance measurements. Global Biogeochemical Cycles,22, GB2027. https://doi.org/10.1029/
2007GB003027
Franklin, O. (2007). Optimal nitrogen allocation controls tree responses to elevated CO
2
.The New Phytologist,174(4), 811–822. https://doi.
org/10.1111/j.1469‐8137.2007.02063.x
Franklin, O., McMURTRIE, R. E., Iversen, C. M., Crous, K. Y., Finzi, A. C., Tissue, D. T., et al. (2009). Forest fine‐root production and
nitrogen use under elevated CO
2
: Contrasting responses in evergreen and deciduous trees explained by a common principle. Global
Change Biology,15(1), 132–144. https://doi.org/10.1111/j.1365‐2486.2008.01710.x
Franks, P. J., Bonan, G. B., Berry, J. A., Lombardozzi, D. L., Holbrook, N. M., Herold, N., & Oleson, K. W. (2018). Comparing optimal and
empirical stomatal conductance models for application in Earth system models. Global Change Biology,24(12), 5708–5723. https://doi.
org/10.1111/gcb.14445
Friedlingstein, P., Meinshausen, M., Arora, V. K., Jones, C. D., Anav, A., Liddicoat, S. K., & Knutti, R. (2013). Uncertainties in CMIP5
climate projections due to carbon cycle feedbacks. Journal of Climate,27(2), 511–526. https://doi.org/10.1175/JCLI‐D‐12‐00579.1
Gentine, P., Pritchard, M., Rasp, S., Reinaudi, G., & Yacalis, G. (2018). Could machine learning break the convection parameterization
deadlock? Geophysical Research Letters,45(11), 5742–5751. https://doi.org/10.1029/2018GL078202
Goll, D., Vuichard, N., Maignan, F., Jornet‐Puig, A., Sardans, J., Violette, A., et al. (2017). A representation of the phosphorus cycle for
ORCHIDEE (Revision 4520). Geoscientific Model Development,10(10), 3745–3770. https://doi.org/10.5194/gmd‐10‐3745‐2017
Gravel, D., Guichard, F., & Hochberg, M. E. (2011). Species coexistence in a variable world. Ecology Letters. https://doi.org/10.1111/
j.1461‐0248.2011.01643.x
Harper, A. B., Cox, P. M., Friedlingstein, P., Wiltshire, A. J., Jones, C. D., Sitch, S., et al. (2016). Improved representation of plant functional
types and physiology in the joint UK land environment simulator (JULES v4.2) using plant trait information. Geoscientific Model
Development,9(7), 2415–2440. https://doi.org/10.5194/gmd‐9‐2415‐2016
Haverd, V., Smith, B., Cook, G. D., Briggs, P. R., Nieradzik, L., Roxburgh, S. H., et al. (2013). A stand‐alone tree demography and landsc ape
structure module for Earth system models. Geophysical Research Letters,40,5234–5239. https://doi.org/10.1002/grl.50972
Haverd, V., Smith, B., Nieradzik, L., Briggs, P. R., Woodgate, W., Trudinger, C. M., et al. (2018). A new version of the CABLE land surface
model (subversion revision r4601) incorporating land use and land cover change, woody vegetation demography, and a novel
optimisation‐based approach to plant coordination of photosynthesis. Geoscientific Model Development,11(7), 2995–3026. https://doi.
org/10.5194/gmd‐11‐2995‐2018
Hazenberg, P., Fang, Y., & Broxton, P. (2015). A hybrid‐3D hillslope hydrological model for use in E arth system models. Water Resources
Research,51, 8218–8239. https://doi.org/10.1002/2014WR016842
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 19 of 24
Hooper, D. U., Chapin, F. S. III, Ewel, J. J., Hector, A., Inchausti, P., Lavorel, S., et al. (2005). Effects of biodiversity on ecosystem func-
tioning: A consensus of current knowledge. Ecological Monographs,75(1), 3–35. https://doi.org/10.1890/04‐0922
Houborg, R., Fisher, J. B., & Skidmore, A. K. (2015). Advances in remote sensing of vegetation function and traits. International Journal of
Applied Earth Observation and Geoinformation,43,1–6. https://doi.org/10.1016/j.jag.2015.06.001
Hourdin, F., Mauritsen, T., Gettelman, A., Golaz, J.‐C., Balaji, V., Duan, Q., et al. (2017). The art and science of climate model tuning [data
set]. Bulletin of the American Meteorological Society,98(3), 589–602. https://doi.org/10.1175/BAMS‐D‐15‐00135.1
Huang, J., Kautz, M., Trowbridge, A. M., Hammerbacher, A., Raffa, K. F., Adams, H. D., et al. (2019). Tree defence and bark beetles in a
drying world: Carbon partitioning, functioning and modelling. New Phytologist,225(1), 26–36. https://doi.org/10.1111/nph.
16173
Huntzinger, D. N., Schwalm, C., Michalak, A. M., Schaefer, K., King, A. W., Wei, Y., et al. (2013). The North American Carbon Program
Multi‐scale Synthesis and Terrestrial Model Intercomparison Project—Part 1: Overview and experimental design. Geoscientific Model
Development Discussion,6(3), 3977–4008. https://doi.org/10.5194/gmdd‐6‐3977‐2013
Isbell, F., Craven, D., Connolly, J., Loreau, M., Schmid, B., Beierkuhnlein, C., et al. (2015). Biodiversity increases the resistance of eco-
system productivity to climate extremes. Nature,526(7574), 574–577. https://doi.org/10.1038/nature15374
Joetzjer, E., Maignan, F., Chave, J., Goll, D., Poulter, B., Barichivich, J., et al. (2018). Effect of the importance of tree demography and
flexible root water uptake for modelling the carbon and water cycles of Amazonia. Biogeosciences Discussions,1–33. https://doi.org/
10.5194/bg‐2018‐308
Jupp, T. E., & Twiss, S. D. (2006). A physically motivated index of subgrid‐scale pattern. Journal of Geophysical Research,111, D19112.
https://doi.org/10.1029/2006JD007343
Kaminski, T., Knorr, W., Schürmann, G., Scholze, M., Rayner, P. J., Zaehle, S., et al. (2013). The BETHY/JSBACH carbon cycle
data assimilation system: Experiences and challenges. Journal of Geophysical Research –Biogeosciences,118, 1414–1426.
Retrieved from https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1002/jgrg.20118%4010.1002/%28ISSN%292169‐8961.
EMITES1
Kattge, J., Bönisch, G., Díaz, S., Lavorel, S., Prentice, I. C., Leadley , P., et al. (2020). TRY plant trait database—Enhanced coverage and open
access. Global Change Biology,26(1), 119–188. https://doi.org/10.1111/gcb.14904
Kay, J. E., Deser, C., Phillips, A., Mai, A., Hannay, C., Strand, G., et al. (2015). The commu nity Earth system model (CESM) large ensemble
project: A community resource for studying climate change in the presence of internal climate variability. Bulletin of the American
Meteorological Society,96(8), 1333–1349. https://doi.org/10.1175/BAMS‐D‐13‐00255.1
Kemp, S., Scholze, M., Ziehn, T., & Kaminski, T. (2014). Limiting the parameter space in the Carbon Cycle Data Assimilation System
(CCDAS). Geoscientific Model Development,7(4), 1609–1619. https://doi.org/10.5194/gmd‐7‐1609‐2014
Kennedy, D., Swenson, S., Oleson, K. W., Lawrence, D. M., Fisher, R., Lola da Costa, A. C., & Gentine, P. (2019). Implementing plant
hydraulics in the Community Land Model, Version 5. Journal of Advances in Modeling Earth Systems,11, 485–513. https://doi.org/
10.1029/2018MS001500
Knauer, J., Werner, C., & Zaehle, S. (2015). Evaluating stomatal models and their atmospheric drought response in a land surface scheme:
A multibiome analysis: Multibiome stomatal model evaluation. Journal of Geophysical Research –Biogeosciences,120, 1894–1911.
https://doi.org/10.1002/2015JG003114
Knauer, J., Zaehle, S., De Kauwe, M. G., Haverd, V., Reichstein, M., & Sun, Y. (2019). Mesophyll conductance in land surface models:
Effects on photosynthesis and transpiration. The Plant Journal. https://doi.org/10.1111/tpj.14587
Koster, R. D., & Milly, P. C. D. (1997). The interplay between transpiration and runoff formulations in land surface schemes used with
atmospheric models. Journal of Climate,10(7), 1578–1591.
Koster, R. D., Reichle, R. H., Schubert, S. D., & Mahanama, S. P. (2019). Length scales of hydrological variability as inferred from SMAP soil
moisture retrievals. Journal of Hydrometeorology. https://doi.org/10.1175/JHM‐D‐19‐0070.1
Koven, C. D., Knox, R. G., Fisher, R. A., Chambers, J., Christoffersen, B. O., Davies, S. J., et al. (2019). Benchmarking and parameter
sensitivity of physiological and vegetation dynamics using the Functionally Assembled Terrestrial Ecosystem Simulator (FATES) at
Barro Colorado Island, Panama. Biogeosciences Discussions. https://doi.org/10.5194/bg‐2019‐409
Koven, C. D., Lawrence, D. M., & Riley, W. J. (2015). Permafrost carbon‐climate feedback is sensitive to deep soil carbon decomposability
but not deep soil nitrogen dynamics. Proceedings of the National Academy of Sciences,112(12), 3752–3757. https://doi.org/10.1073/
pnas.1415123112
Kumarathunge, D. P., Medlyn, B. E., Drake, J. E., Tjoelker, M. G., Aspinwall, M. J., Battaglia, M., et al. (2019). Acclimation and adaptation
components of the temperature dependence of plant photosynthesis at the global scale. The New Phytologist,222(2), 768–784. https://doi.
org/10.1111/nph.15668
Laguë, M. M., Bonan, G. B., & Swann, A. L. S. (2019). Separating the impact of individual land surface properties on the terrestrial surface
energy budget in both the coupled and uncoupled land–atmosphere system. Journal of Climate,32(18), 5725–5744. https://doi.org/
10.1175/JCLI‐D‐18‐0812.1
Lapeyre, C. J., Cazard, N., Roy, P. T., Ricci, S., & Zaoui, F. (2019). Reconstruction of hydraulic data by machine learning. arXiv [cs.CE].
Retrieved from http://arxiv.org/abs/1903.01123
Lawrence, D. M., Fisher, R. A., Koven, C. D., Oleson, K. W., Swenson, S. C., Bonan, G., et al. (2019). The Community Land Model Version 5:
Description of new features, benchmarking, and impact of forcing uncertainty. Journal of Advances in Modeling Earth Systems,11,
4245–4287. https://doi.org/10.1029/2018MS001583
Lawrence, D. M., Oleson, K. W., Flanner, M. G., Thornton, P. E., Swenson, S. C., Lawrence, P. J., et al. (2011). Parameterization
improvements and functional and structural advances in Version 4 of the Community Land Model. Journal of Advances in Modeling
Earth Systems,3, M03001. https://doi.org/10.1029/2011MS00045
Lawrence, P. J., Feddema, J. J., Bonan, G. B., Meehl, G. A., O'Neill, B. C., Oleson, K. W., et al. (2012). Simulating the biogeochemical and
biogeophysical impacts of transient land cover change and Wood harvest in the community climate system model (CCSM4) from 1850 to
2100. Journal of Climate,25(9), 3071–3095. https://doi.org/10.1175/jcli‐d‐11‐00256.1
Le Quéré, C., Andrew, R. M., Friedlingstein, P., Sitch, S., Hauck, J., Pongratz, J., et al. (2018). Global carbon budget 2018. Earth System
Science Data,10(4), 2141–2194. https://doi.org/10.5194/essd‐10‐2141‐2018
LeBauer, D. S., Wang, D., Richter, K. T., Davidson, C. C., & Dietze, M. C. (2013). Facilitating feedbacks between field measurements and
ecosystem models. Ecological Monographs,83(2), 133–154. https://doi.org/10.1890/12‐0137.1
Lloyd, J., Patiño, S., Paiva, R. Q., Nardoto, G. B., Quesada, C. A., Santos, A. J. B., et al. (2010). Optimisation of photosynthetic carbon gain
and within‐canopy gradients of associated foliar traits for Amazon forest trees. Biogeosciences,7(6), 1833–1859. https://doi.org/10.5194/
bg‐7‐1833‐2010
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 20 of 24
Lombardozzi, D. L., Bonan, G. B., Smith, N. G., Dukes, J. S., & Fisher, R. A. (2015). Temperature acclimation of photosynthesis and
respiration: A key uncertainty in the carbon cycle‐climate feedback. Geophysical Research Letters,42, 8624–8631. https://doi.org/
10.1002/2015GL065934
Longo, M., Knox, R. G., Medvigy, D. M., Levine, N. M., Dietze, M. C., Kim, Y., et al. (2019). The biophy sics, ecology, and biogeochemistry of
functionally diverse, vertically and horizontally heterogeneous ecosystems: the Ecosystem Demography Model, Version 2.2—Part 1:
Model description. Geoscientific Model Development,12(10), 4309–4346. https://doi.org/10.5194/gmd‐12‐4309‐2019
Lundquist, J. D., & Dettinger, M. D. (2005). How snowpack heterogeneity affects diurnal streamflow timing. Water Resources Research,41,
W05007. https://doi.org/10.1029/2004WR003649
MacBean, N., Peylin, P., Chevallier, F., Scholze, M., & Schürmann, G. (2016). Consistent assimilation of multiple data streams in a carbon
cycle data assimilation system. Geoscientific Model Development,9(10), 3569–3588. https://doi.org/10.5194/gmd‐9‐3569‐2016
Mäkelä, J., Knauer, J., Aurela, M., Black, A., Heimann, M., Kobayashi, H., et al. (2019). Parameter calibration and stomatal conductance
formulation comparison for boreal forests with adaptive population importance sampler in the land surface model JSBACH.
Geoscientific Model Development,12(9), 4075–4098. https://doi.org/10.5194/gmd‐12‐4075‐2019
Manabe, S. (1969). Climate and the ocean circulation: I. the atmospheric circulation and the hydrology of the Earth's surface. Monthly
Weather Review,97(11), 739–774. https://doi.org/10.1175/1520‐0493(1969)097<0739:CATOC>2.3.CO;2
Maréchaux, I., & Chave, J. (2017). An individual‐based forest model to jointly simulate carbon and tree diversity in Amazonia: Description
and applications. Ecological Monographs,87(4), 632–664. https://doi.org/10.1002/ecm.1271
Masson, V., Le Moigne, P., Martin, E., Faroux, S., Alias, A., Alkama, R., et al. (2013). The SURFEXv7. 2 land and ocean surface platform for
coupled or offline simulation of Earth surface variables and fluxes. Geoscientific Model Development,6(4), 929–960. https://doi.org/
10.5194/gmd‐6‐929‐2013
Matheny, A. M., Mirfenderesgi, G., & Bohrer, G. (2017). Trait‐based representation of hydrological functional properties of plants in
weather and ecosystem models. Plant Diversity,39(1), 1–12. https://doi.org/10.1016/j.pld.2016.10.001
Matthews, H. D., Zickfeld, K., Knutti, R., & Allen, M. R. (2018). Focus on cumulative emissions, global carbon budgets and the implications
for climate mitigation targets. Environmental Research Letters,13(1), 010201. https://doi.org/10.1088/1748‐9326/aa98c9
Mauritsen, T., Bader, J., Becker, T., Behrens, J., Bittner, M., Brokopf, R., et al. (2019). Developments in the MPI‐M Earth System Model
Version 1.2 (MPI‐ESM1. 2) and its response to increasing CO
2
.Journal of Advances in Modeling Earth Systems,11, 998–1038. https://doi.
org/10.1029/2018MS001400
McCabe, T. D., & Dietze, M. C. (2019). Scaling contagious disturbance: A spatially‐implicit dynamic model. Frontiers in Ecology and
Evolution,7, 64. https://doi.org/10.3389/fevo.2019.00064
Medlyn, B. E., Duursma, R. A., Eamus, D., Ellsworth, D. S., Prentice, I. C., Barton, C. V. M., et al. (2011). Reconciling the optimal and
empirical approaches to modelling stomatal conductance. Global Change Biology,17(6), 2134–2144. https://doi.org/10.1111/
j.1365‐2486.2010.02375.x
Medlyn, B. E., Zaehle, S., De Kauwe, M. G., Walker, A. P., Dietze, M. C., Hanson, P. J., et al. (2015). Using ecosystem experiments to
improve vegetation models. Nature Climate Change,5(6), 528–534. https://doi.org/10.1038/nclimate2621
Meir, P., Kruijt, B., & Broadmeadow, M. (2002). Acclimation of photosynthetic capacity to irradiance in tree canopies in relation to leaf
nitrogen concentration and leaf mass per unit area. Plant, Cell & Environment. https://doi.org/10.1046/j.0016‐8025.2001.00811.x
Mercado, L. M., Huntingford, C., Gash, J. H. C., Cox, P. M., & Jogireddy, V. (2007). Improving the representation of radiation interception
and photosynthesis for climate model applications. Tellus Series B: Chemical and Physical Meteorology,59(3), 553–565. https://doi.org/
10.1111/j.1600‐0889.2007.00256.x
Mirfenderesgi, G., Bohrer, G., Matheny, A. M., Fatichi, S., de Moraes Frasson, R. P., & Schäfer, K. V. R. (2016). Tree level hydrodynamic
approach for resolving aboveground water storage and stomatal conductance and modeling the effects of tree hydraulic strategy. Journal
of Geophysical Research –Biogeosciences,121, 1792–1813. https://doi.org/10.1002/2016JG003467
Moorcroft, P. R., Hurtt, G. C., & Pacala, S. W. (2001). A method for scaling vegetation dynamics: The ecosystem demography model (ED).
Ecological Monographs,71(4), 557–586. https://doi.org/10.1890/0012‐9615(2001)071[0557:AMFSVD]2.0.CO;2
Moore, J. R., Zhu, K., Huntingford, C., & Cox, P. M. (2018). Equilibrium forest demography explains the distribution of tree sizes across
North America. Environmental Research Letters,13(8), 084019. https://doi.org/10.1088/1748‐9326/aad6d1
Nabel, J. E. M. S., Julia, E. M., Naudts, K., & Pongratz, J. (2019). Accounting for forest age in the tile‐based dynamic global vegetation model
JSBACH4 (4.20p7; git feature/forests)—A land surface model for the ICON‐ESM. Geoscientific Model Development Discussion. https://
doi.org/10.5194/gmd‐2019‐68
Naudts, K., Ryder, J., McGrath, M. J., Otto, J., Chen, Y., Valade, A., et al. (2014). A vertically discretised canopy description for ORCHIDEE
(SVN r2290) and the modifications to the energy, water and carbon fluxes. Geoscientific Model Devel opment,8(7), 2035–2065. https://doi.
org/10.5194/gmd‐8‐2035‐2015
Nearing, G. S., Ruddell, B. L., Clark, M. P., Nijssen, B., & Peters‐Lidard, C. (2018). Benchmarking and process diagnostics of land models.
Journal of Hydrometeorology,19(11), 1835–1852. https://doi.org/10.1175/JHM‐D‐17‐0209.1
Newman, A. J., Clark, M. P., Winstral, A., Marks, D., & Seyfried, M. (2014). The use of similarity concepts to represent subgrid variability in
land surface models: Case study in a snowmelt‐dominated watershed. Journal of Hydrometeorology,15(5), 1717–1738. https://doi.org/
10.1175/JHM‐D‐13‐038.1
Nishina, K., Ito, A., Falloon, P., Friend, A. D., Beerling, D. J., Ciais, P., et al. (2015). Decomposing uncertainties in the future terrestrial
carbon budget associated with emission scenarios, climate projections, and ecosystem simulations using the ISI‐MIP results. Earth
System Dynamics,6(2), 435–445. https://doi.org/10.5194/esd‐6‐435‐2015
Niu, S., Luo, Y., Dietze, M. C., Keenan, T. F., Shi, Z., Li, J., & Stuart Chapin, F. III (2014). The role of data assimilation in predictive ecology.
Ecosphere. https://doi.org/10.1890/es13‐00273.1
NOAA. (2016). The National Water Model. Retrieved October 29, 2019, from https://water.noaa.gov/about/nwm
Oleson, K. W., Bonan, G. B., Feddema, J., Vertenstein, M., & Grimmond, C. S. B. (2008). An urban parameterization for a global climate
model. Part I: Formulation and evaluation for two cities. Journal of Applied Meteorology and Climatology,47(4), 1038–1060. https://doi.
org/10.1175/2007JAMC1597.1
Pau, G. S. H., Bisht, G., & Riley, W. J. (2014). A reduced‐order modeling approach to represent subgrid‐scale hydrological dynamics for
land‐surface simulations: Application in a polygonal tundra landscape. Geoscientific Model Development,7(5), 2091–2105. https://doi.
org/10.5194/gmd‐7‐2091‐2014
Peylin, P., Bacour, C., MacBean, N., Leonard, S., Rayner, P., Kuppel, S., et al. (2016). A new stepwise carbon cycle data assimilation system
using multiple data streams to constrain the simulated land surface carbon cycle. Geoscientific Model Development,9(9). 3321‐3346
Retrieved from https://www.osti.gov/biblio/1361538
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 21 of 24
Polcher, J., McAvaney, B., Viterbo, P., Gaertner, M.‐A., Hahmann, A., Mahfouf, J.‐F., et al. (1998). A proposal for a general interface
between land surface schemes and general circulation models. Global and Planetary Chan ge,19(1‐4), 261–276. https://doi.org/10.1016/
S0921‐8181(98)00052‐6
Pongratz, J., Dolman, H., Don, A., Erb, K.‐H., Fuchs, R., Herold, M., et al. (2018). Models meet data: Challenges and opportunities in
implementing land management in Earth system models. Global Chang e Biology,24(4), 1470–1487. https://doi.org/10.1111/gcb.13988
Post, H., Vrugt, J. A., Fox, A., Vereecken, H., & Hendricks Franssen, H.‐J. (2017). Estimation of community land model parameters for an
improved assessment of net carbon fluxes at European sites. Journal of Geophysical Research –Biogeosciences,122, 661–689. https://doi.
org/10.1002/2015JG003297
Powell, T. L., Koven, C. D., Johnson, D. J., Faybishenko, B., Fisher, R. A., Knox, R. G., et al. (2018). Variation in hydroclimate sus-
tains tropical forest biomass and promotes functional diversity. The New Phytologist,219(3), 932–946. https://doi.org/10.1111/
nph.15271
Purves, D. W., Lichstein, J. W., Strigul, N., & Pacala, S. W. (2008). Predicting and understanding forest dynamics using a simple tractable
model. Proceedings of the National Academy of Sciences of the United States of America,105(44), 17,018–17,022. https://doi.org/10.1073/
pnas.0807754105
Quegan, S., Le Toan, T., Chave, J., Dall, J., Exbrayat, J.‐F., Minh, D. H. T., et al. (2019). The European Space Agency BIOMASS mission:
Measuring forest above‐ground biomass from space. Remote Sensing of Environment,227,44–60. https://doi.org/10.1016/j.
rse.2019.03.032
Reed, S. C., Yang, X., & Thornton, P. E. (2015). Incorporating phosphorus cycling into global modeling efforts: A worthwhile, tractable
endeavor. The New Phytologist,208(2), 324–329. https://doi.org/10.1111/nph.13521
Roberts, A. F., Hunke, E. C., Allard, R., Bailey, D. A., Craig, A. P., Lemieux, J.‐F., & Turner, M. D. (2018). Quality control for
community‐based sea‐ice model development. Philosophical Transactions. Series A, Mathematical, Physical, and Engineering Sciences,
376(2129). https://doi.org/10.1098/rsta.2017.0344
Rogers, A., Medlyn, B. E., Dukes, J. S., Bonan, G., von Caemmerer, S., Dietze, M. C., et al. (2017). A roadmap for improving the repre-
sentation of photosynthesis in Earth system models. The New Phytologist,213(1), 22–42. https://doi.org/10.1111/nph.14283
Sakschewski, B., von Bloh, W., Boit, A., Rammig, A., Kattge, J., Poorter, L., et al. (2015). Leaf and stem economics spectra drive diversity of
functional plant traits in a dynamic global vegetation model. Global Change Biology,21(7), 2711 –2725. https://doi.org/10.1111/gcb.12870
Sargsyan, K., Safta, C., Najm, H. N., Debusschere, B. J., Ricciuto, D., & Thornton, P. (2014). Dimensionality reduction for complex models
via Bayesian compressive sensing. International Journal for Uncertainty Quantification,4(1), 63–93. https://doi.org/10.1615/Int.J.
UncertaintyQuantification.2013006821
Sato, H., Itoh, A., & Kohyama, T. (2007). SEIB–DGVM: A new dynamic global vegetation model using a spatially explicit individual‐based
approach. Ecological Modelling,200(3‐4), 279–307. https://doi.org/10.1016/j.ecolmodel.2006.09.006
Scheiter, S., Langan, L., & Higgins, S. I. (2013). Next‐generation dynamic global vegetation models: Learning from community ecology. The
New Phytologist,198(3), 957–969. https://doi.org/10.1111/nph.12210
Schwalm, C. R., Schaefer, K., Fisher, J. B., Huntzinger, D., Elshorbany, Y., Fang, Y., et al. (2019). Divergence in land surface modeling:
Linking spread to structure. Environmental Research Communications,1(11), 111004. https://doi.org/10.1088/2515‐7620/ab4a8a
Sellers, P. J. (1985). Canopy reflectance, photosynthesis and transpiration. International Journ al of Remote Sensing,6(8), 1335–1372.
https://doi.org/10.1080/01431168508948283
Sellers, P. J., Bounoua, L., Collatz, G. J., Randall, D. A., Dazlich, D. A., Los, S. O., et al. (1996). Comparison of radiative and physiological
effects of doubled atmospheric CO
2
on climate. Science,271(5254), 1402–1406. https://doi.org/10.1126/science.271.5254.1402
Sellers, P. J., Fennessy, M. J., & Dickinson, R. E. (2007). A numerical approach to calculating soil wetness and evapotranspiration over large
grid areas. Journal of Geophysical Research,112(D18), D18106. https://doi.org/10.1029/2007JD008781
Sellers, P. J., Mintz, Y., Sud, Y. C., & Dalcher, A. (1986). A simple biosphere model (SIB) for use within general circulation models. Journal
of the Atmospheric Sciences,43(6), 505–531. https://doi.org/10.1175/1520‐0469(1986)043<0505:ASBMFU>2.0.CO;2
Sheffield, J., Goteti, G., & Wood, E. F. (2006). Development of a 50‐year high‐resolution global dataset of meteorological forcings for land
surface modeling. Journal of Climate,19(13), 3088–3111. https://doi.org/10.1175/JCLI3790.1
Shen, C. (2018). A transdisciplinary review of deep learning research and its relevance for water resources scientists. Water Resources
Research,54, 8558–8593. https://doi.org/10.1029/2018WR022643
Shevliakova, E., Pacala, S. W., Malyshev, S., Hurtt, G. C., Milly, P. C. D., Caspersen, J. P., et al. (2009). Carbon cycling under 300 years of
land use change: Importance of the secondary vegetation sink. Global Biogeochemical Cycles,23, GB2022. https://doi.org/10.1029/
2007GB003176
Sitch, S., Smith, B., Prentice, I. C., Arneth, A., Bondeau, A., Cramer, W., et al. (2003). Evaluation of ecosystem dynamics, plant geography
and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology,9(2), 161–185. https://doi.org/
10.1046/j.1365‐2486.2003.00569.x
Slater, A. G., Lawrence, D. M., & Koven, C. D. (2017). Process‐level model evaluation: A snow and heat transfer metric. The Cryosphere,
11(2), 989–996. https://doi.org/10.5194/tc‐11‐989‐2017
Slot, M., Rey‐Sánchez, C., Gerber, S., Lichstein, J. W., Winter, K., & Kitajima, K. (2014). Thermal acclimation of leaf respiration of tropical
trees and lianas: Response to experimental canopy warming, and consequences for tropical forest carbon balance. Global Change
Biology,20(9), 2915–2926. https://doi.org/10.1111/gcb.12563
Smith, N. G., Keenan, T. F., Colin Prentice, I., Wang, H., Wright, I. J., Niinemets, Ü., et al. (2019). Global photosynthetic capacity is opti-
mized to the environment. Ecology Letters,22(3), 506–517. https://doi.org/10.1111/ele.13210
Sperry, J. S., Adler, F. R., Campbell, G. S., & Comstock, J. P. (1998). Limitation of plant water use by rhizosphere and xylem conductance:
Results from a model. Plant, Cell & Environment,21(4), 347–359. https://doi.org/10.1046/j.1365‐3040.1998.00287.x
Stavros, E. N., Natasha Stavros, E., Schimel, D., Pavlick, R., Serbin, S., Swann, A., et al. (2017). ISS observations offer insights into plant
function. Nature Ecology & Evolution,1(7), 0194. https://doi.org/10.1038/s41559‐017‐0194
Street, L. E., Shaver, G. R., Rastetter, E. B., van Wijk, M. T., Kaye, B. A., & Williams, M. (2012). Incident radiation and the allocation of
nitrogen within Arctic plant canopies: Implications for predicting gross primary productivity. Global Change Biology. https://doi.org/
10.1111/j.1365‐2486.2012.02754.x
Subin, Z. M., Milly, P. C. D., Sulman, B. N., Malyshev, S., & Shevl iakova, E. (2014). Resolving terrestrial ecosystem processes along a subgrid
topographic gradient for an Earth‐system model. Hydrology and Earth System Sciences Discussions,11(7), 8443–8492. https://doi.org/
10.5194/hessd‐11‐8443‐2014
Sulman, B. N., Phillips, R. P., Oishi, A. C., Shevliakova, E., & Pacala, S. W. (2014). Microbe‐driven turnover offsets mineral‐mediated
storage of soil carbon under elevated CO
2
.Nature Climate Change,4(12), 1099–1102. https://doi.org/10.1038/nclimate2436
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 22 of 24
Swenson, S. C., Clark, M., Fan, Y., Lawrence, D. M., & Perket, J. (2019). Representing Intrahillslope lateral subsurface flow in the com-
munity land model. Journal of Advances in Modeling Earth Systems,11, 4044–4065. https://doi.org/10.1029/2019MS001833
Tague, C., & Dugger, A. L. (2010). Ecohydrology and climate change in the mountains of the Western USA—A review of research and
opportunities. Geography Compass. https://doi.org/10.1111/j.1749‐8198.2010.00400.x
Tai, X., Mackay, D. S., Anderegg, W. R. L., Sperry, J. S., & Brooks, P. D. (2017). Plant hydraulics improves and topography mediates pre-
diction of aspen mortality in southwestern USA. The New Phytologist,213(1), 113–127. https://doi.org/10.1111/nph.14098
Tang, J., & Zhuang, Q. (2008). Equifinality in parameterization of process‐based biogeochemistry models: A significant uncertainty source
to the estimation of regional carbon dynamics. Journal of Geophysical Research,113, G04010. https://doi.org/10.1029/2008JG000757
Taylor, S. D., & White, E. P. (2019). Automated data‐intensive forecasting of plant phenology throughout the United States. Ecological
Applications. https://doi.org/10.1002/eap.2025
Thomas, R. Q., Quinn Thomas, R., Hurtt, G. C., Dubayah, R., & Schilz, M. H. (2008). Using lidar data and a height‐structured ecosystem
model to estimate forest carbon stocks and fluxes over mountainous terrain. Canadian Journal of Remote Sensing. https://doi.org/
10.5589/m08‐036
Thornton, P. E., Lamarque, J.‐F. C., Rosenbloom, N. A., & Mahowald, N. M. (2007). Influence of carbon‐nitrogen cycle coupling on land
model response to CO
2
fertilization and climate variability. Global Biogeochemical Cycles,21, GB4028. https://doi.org/10.1029/
2006GB002868
Treseder, K. K., Balser, T. C., Bradford, M. A., Brodie, E. L., Dubinsky, E. A., Eviner, V. T., et al. (2012). Integrating microbial ecology into
ecosystem models: Challenges and priorities. Biogeochemistry,109(1‐3), 7–18. https://doi.org/10.1007/s10533‐011‐9636‐5
Trugman, A. T., Fenton, N. J., Bergeron, Y., Xu, X., Welp, L. R., & Medvigy, D. (2016). Climate, soil organic layer, and nitrogen jointly drive
forest development after fire in the North American boreal zone. Journal of Advances in Modeling Earth Systems,8, 1180–1209. https://
doi.org/10.1002/2015MS000576
Turnbull, L. A., Levine, J. M., Loreau, M., & Hector, A. (2013). Coexistence, niches and biodiversity effects on ecosystem functioning.
Ecology Letters,16(Suppl 1), 116–127. https://doi.org/10.1111/ele.12056
van Bodegom, P. M., Douma, J. C., & Verheijen, L. M. (2014). A fully traits‐based approach to modeling global vegetation distribution.
Proceedings of the National Academy of Sciences of the United States of America,111(38), 13733–13738. https://doi.org/10.1073/
pnas.1304551110
van Kampenhout, L., Lenaerts, J. T. M., Lipscomb, W. H., Sacks, W. J., Lawrence, D. M., Slater, A. G., & van den Broeke, M. R. (2017).
Improving the representation of polar snow and firn in the community Earth system model. Journal of Advances in Modeling Earth
Systems,9, 2583–2600. https://doi.org/10.1002/2017MS000988
Verbeeck, H., Peylin, P., Bacour, C., Bonal, D., Steppe, K., & Ciais, P. (2011). Seasonal patterns of CO
2
fluxes in Amazon forests: Fusion of
eddy covariance data and the ORCHIDEE model. Journal of Geophysical Research,116, G02018. https://doi.org/10.1029/2010JG001544
Verheijen, L. M., Aerts, R., Brovkin, V., Cavender‐Bares, J., Cornelissen, J. H. C., Kattge, J., & van Bodegom, P. M. (2015). Inclusion of
ecologically based trait variation in plant functional types reduces the projected land carbon sink in an Earth system model. Global
Change Biology,21(8), 3074–3086. https://doi.org/10.1111/gcb.12871
Walker, A. P., Ye, M., Lu, D., Kauwe, M. G. D., Gu, L., Medlyn, B. E., et al. (2018). The multi‐assumption architecture and testbed (MAAT
v1.0): R code for generating ensembles with dynamic model structure and analysis of epistemic uncertainty from multiple sources.
Geoscientific Model Development,11(8), 3159–3185. https://doi.org/10.5194/gmd‐11‐3159‐2018
Walker, A. P., Zaehle, S., Medlyn, B. E., De Kauwe, M. G., Asao, S., Hickler, T., et al. (2015). Predicting long‐term carbon sequestration in
response to CO
2
enrichment: How and why do current ecosystem models differ? Global Biogeochemical Cycles,29, 476–495. https://doi.
org/10.1002/2014GB004995
Wang, H., Colin Prentice, I., Keenan, T. F., Davis, T. W., Wright, I. J., Cornwell, W. K., et al. (2017). Towards a universal model for carbon
dioxide uptake by plants. Nature Plants,3(9), 734–741. https://doi.org/10.1038/s41477‐017‐0006‐8
Wang, Y.‐P., Zhang, Q., Pitman, A. J., & Dai, Y. (2015). Nitrogen and phosphorous limitation reduces the effects of land use change on land
carbon uptake or emission. Environmental Research Letters,10(1), 014001. https://doi.org/10.1088/1748‐9326/10/1/014001
Weng, E., Farrior, C. E., Dybzinski, R., & Pacala, S. W. (2017). Predicting vegetation type through physiological and environmental
interactions with leaf traits: Evergreen and deciduous forests in an Earth system modeling framework. Global Change Biology,23(6),
2482–2498. https://doi.org/10.1111/gcb.13542
Wieder, W. R., Allison, S. D., Davidson, E. A., Georgiou, K., Hararuk, O., He, Y., et al. (2015). Explicitly representing soil microbial pro-
cesses in Earth system models: Soil microbes in Earth system models. Global Biogeochemical Cycles,29,1782–1800. https://doi.org/
10.1002/2015GB005188
Wieder, W. R., Bonan, G. B., & Allison, S. D. (2013). Global soil carbon projections are improved by modelling microbial processes. Nature
Climate Change,3(10), 909–912. https://doi.org/10.1038/nclimate1951
Wieder, W. R., Hartman, M. D., Sulman, B. N., Wang, Y.‐P., Koven, C. D., & Bonan, G. B. (2018). Carbon cycle confidence and uncertainty:
Exploring variation among soil biogeochemical models. Global Change Biology,24(4), 1563–1579. https://doi.org/10.1111/gcb.13979
Williams, M., Law, B. E., Anthoni, P. M., & Unsworth, M. H. (2001). Use of a simulation model and ecosystem flux data to examine carbon–
water interactions in ponderosa pine. Tree Physiology,21(5), 287–298. https://doi.org/10.1093/treephys/21.5.287
Williams, M., Rastetter, E. B., Fernandes, D. N., Goulden, M. L., Wofsy, S. C., Shaver, G. R., et al. (1996). Modelling the
soil‐plant‐atmosphere continuum in a Quercus–Acer stand at Harvard Fore st: The regulation of stomatal conductance by light, nitrogen
and soil/plant hydraulic properties. Plant, Cell & Environment,19(8), 911–927. https://doi.org/10.1111/j.1365‐3040.1996.tb00456.x
Wiltshire, A. J., Rojas, C. D., Edwards, J., Gedney, N., Harper, A. B., Hartley, A., et al. (2019). JULES‐GL7: The global land configuration of
the joint UK land environment simulation version 7.0. Geoscientific Model Development Discussion. https://doi.org/10.5194/
gmd‐2019‐152
Wolf, A., Anderegg, W. R. L., & Pacala, S. W. (2016). Optimal stomatal behavior with competition for water and risk of hydraulic impair-
ment. Proceedings of the National Academy of Sciences of the United States of America,113(46), E7222–E7230. https://doi.org/10.1073/
pnas.1615144113
Wood, E. F., Roundy, J. K., Troy, T. J., van Beek, L. P. H., Bierkens, M. F. P., Blyth, E., et al. (2011). Hyperresolution global land surface
modeling: Meeting a grand challenge for monitoring Earth's terrestrial water: OPINION. Water Resources Research,47, W05301. https://
doi.org/10.1029/2010WR010090
Woodward, F. I., & Lomas, M. R. (2004). Vegetation dynamics—Simulating responses to climatic change. Biological Reviews. https://doi.
org/10.1017/s1464793103006419
Xia, Y., Hao, Z., Shi, C., Li, Y., Meng, J., Xu, T., et al. (2019). Regional and global land data assimilation systems: Innovations, challenges,
and prospects. Journal of Meteorological Research,33(2), 159–189. https://doi.org/10.1007/s13351‐019‐8172‐4
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 23 of 24
Xu, C., Fisher, R., Wullschleger, S. D., Wilson, C. J., Cai, M., & McDowell, N. G. (2012). Toward a mechanistic modeling of nitrogen lim-
itation on vegetation dynamics. PLoS ONE,7(5), e37914. https://doi.org/10.1371/journal.pone.0037914
Xu, X., Medvigy, D., Powers, J. S., Becknell, J. M., & Guan, K. (2016). Diversity in plant hydraulic traits explains seasonal and inter‐annual
variations of vegetation dynamics in seasonally dry tropical forests. New Phytologist. https://doi.org/10.1111/nph.14009
Yachi, S., & Loreau, M. (1999). Biodiversity and ecosystem productivity in a fluctuating environment: The insurance hypothesis.
Proceedings of the National Academy of Sciences of the United States of America,96(4), 1463–1468. htt ps://doi.org/10.1073/pnas.96.4.1463
Yang, X., Thornton, P. E., Ricciuto, D. M., & Post, W. M. (2014). The role of phosphorus dynamics in tropical forests—A modeling study
using CLM‐CNP. Biogeosciences,11(6), 1667–1681. https://doi.org/10.5194/bg‐11‐1667‐2014
Yokohata, T., Kinoshita, T., Sakurai, G., Pokhrel, Y., Ito, A., Okada, M., et al. (2019). MIROC‐INTEG1: A global bio‐geochemical land
surface model with human water management, crop growth, and land‐use change. Geoscientific Model Development Discussion. https://
doi.org/10.5194/gmd‐2019‐184
Yue, C., Ciais, P., Luyssaert, S., Li, W., McGrath, M. J., Chang, J., & Peng, S. (2018). Representing anthropogenic gross land use change,
wood harvest, and forest age dynamics in a global vegetation model ORCHIDEE‐MICT v8.4.2. Geoscientific Model Development. https://
doi.org/10.5194/gmd‐11‐409‐2018
Zaehle, S., & Friend, A. D. (2010). Carbon and nitrogen cycle dynamics in the O‐CN land surface model: 1. Model description, site‐scale
evaluation, and sensitivity to parameter estimates. Global Biogeochemical Cycles,24, GB1005. https://doi.org/10.1029/2009GB003521
Zaehle, S., Medlyn, B. E., de Kauwe, M. G., Walker, A. P., Dietze, M. C., Hickler, T., et al. (2014). Evaluation of 11 terrestrial carbon–
nitrogen cycle models against observations from two temperate Free‐Air CO
2
Enrichment studies. The New Phytologist,202(3), 803–822.
https://doi.org/10.1111/nph.12697
Zeng, X., Shaikh, M., Dai, Y., Dickinson, R. E., & Myneni, R. (2002). Coupling of the common land model to the NCAR Community climate
model. Journal of Climate,15(14), 1832–1854. https://doi.org/10.1175/1520‐0442(2002)015<1832:COTCLM>2.0.CO;2
Zscheischler, J., Michalak, A. M., Schwalm, C., Mahecha, M. D., Huntzinger, D. N., Reichstein, M., et al. (2014). Impact of large‐scale
climate extremes on biospheric carbon fluxes: An intercomparison based on MsTMIP data. Global Biogeochemical Cycles,28, 585–600.
https://doi.org/10.1002/2014gb004826
10.1029/2018MS001453
Journal of Advances in Modeling Earth Systems
FISHER AND KOVEN 24 of 24