STEPWAT2: an individual-based model for exploring the impact of climate and disturbance on dryland plant communities

Abstract

2018. STEPWAT2: an individual-based model for exploring the impact of climate and disturbance on dryland plant communities. Ecosphere 9(8): Abstract. The combination of climate change and altered disturbance regimes is directly and indirectly affecting plant communities by mediating competitive interactions, resulting in shifts in species composition and abundance. Dryland plant communities, defined by low soil water availability and highly variable climatic regimes, are particularly vulnerable to climatic changes that exceed their historical range of variability. Individual-based simulation models can be important tools to quantify the impacts of climate change, altered disturbance regimes, and their interaction on demographic and community-level responses because they represent competitive interactions between individuals and individual responses to fluctuating environmental conditions. Here, we introduce STEPWAT2, an individual plant-based simulation model for exploring the joint influence of climate change and disturbance regimes on dryland ecohydrology and plant community composition. STEPWAT2 utilizes a process-based soil water model (SOILWAT2) to simulate available soil water in multiple soil layers, which plant individuals compete for based on the temporal matching of water and active root distributions with depth. This representation of resource utilization makes STEPWAT2 particularly useful for understanding how changes in soil moisture and altered disturbance regimes will concurrently impact demographic and community-level responses in drylands. Our goals are threefold: (1) to describe the core modules and functions within STEPWAT2 (model description), (2) to validate STEPWAT2 model output using field data from big sagebrush plant communities (model validation), and (3) to highlight the usefulness of STEPWAT2 as a modeling framework for examining the impacts of climate change and disturbance regimes on dryland plant communities under future conditions (model application). To address goals 2 and 3, we focus on 15 sites that span the spatial extent of big sage-brush plant communities in the western United States. For goal 3, we quantify how climate change, fire, and grazing can interact to influence plant functional type biomass and composition. We use big sage-brush-dominated plant communities to demonstrate the functionality of STEPWAT2, as these communities are among the most widespread dryland ecosystems in North America.

Full text

EMERGING TECHNOLOGIES
STEPWAT2: an individual-based model for exploring the impact of
climate and disturbance on dryland plant communities
KYLE A. PALMQUIST ,
1,

JOHN B. BRADFORD ,
2
TRACE E. MARTYN ,
3
DANIEL R. SCHLAEPFER ,
4
AND WILLIAM K. LAUENROTH
1,4
1
Department of Botany, University of Wyoming, Laramie, Wyoming 82071 USA
2
U.S. Geological Survey, Southwest Biological Science Center, Flagstaff, Arizona 86001 USA
3
School of Biological Sciences, The University of Queensland, St. Lucia, Queensland 4072 Australia
4
School of Forestry and Environmental Studies, Yale University, New Haven, Connecticut 06511 USA
Citation: Palmquist, K. A., J. B. Bradford, T. E. Martyn, D. R. Schlaepfer, and W. K. Lauenroth. 2018. STEPWAT2: an
individual-based model for exploring the impact of climate and disturbance on dryland plant communities. Ecosphere
9(8):e02394. 10.1002/ecs2.2394
Abstract. The combination of climate change and altered disturbance regimes is directly and indirectly
affecting plant communities by mediating competitive interactions, resulting in shifts in species composi-
tion and abundance. Dryland plant communities, defined by low soil water availability and highly variable
climatic regimes, are particularly vulnerable to climatic changes that exceed their historical range of vari-
ability. Individual-based simulation models can be important tools to quantify the impacts of climate
change, altered disturbance regimes, and their interaction on demographic and community-level responses
because they represent competitive interactions between individuals and individual responses to fluctuat-
ing environmental conditions. Here, we introduce STEPWAT2, an individual plant-based simulation model
for exploring the joint influence of climate change and disturbance regimes on dryland ecohydrology and
plant community composition. STEPWAT2 utilizes a process-based soil water model (SOILWAT2) to simu-
late available soil water in multiple soil layers, which plant individuals compete for based on the temporal
matching of water and active root distributions with depth. This representation of resource utilization
makes STEPWAT2 particularly useful for understanding how changes in soil moisture and altered distur-
bance regimes will concurrently impact demographic and community-level responses in drylands. Our
goals are threefold: (1) to describe the core modules and functions within STEPWAT2 (model description),
(2) to validate STEPWAT2 model output using field data from big sagebrush plant communities (model
validation), and (3) to highlight the usefulness of STEPWAT2 as a modeling framework for examining the
impacts of climate change and disturbance regimes on dryland plant communities under future conditions
(model application). To address goals 2 and 3, we focus on 15 sites that span the spatial extent of big sage-
brush plant communities in the western United States. For goal 3, we quantify how climate change, fire,
and grazing can interact to influence plant functional type biomass and composition. We use big sage-
brush-dominated plant communities to demonstrate the functionality of STEPWAT2, as these communities
are among the most widespread dryland ecosystems in North America.
Key words: Artemisia tridentata; climate change; disturbance; dryland; ecohydrology; fire; gap dynamics; grazing;
individual-based model; plant simulation model; sagebrush; semiarid.
Received 29 January 2018; revised 15 May 2018; accepted 25 June 2018. Corresponding Editor: Robert A.
Washington-Allen.
Copyright: ©2018 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.
E-mail: kpalmqu1@uwyo.edu
❖www.esajournals.org 1August 2018 ❖Volume 9(8) ❖Article e02394

INTRODUCTION
Climate change has important consequences
for plant communities from the leaf to ecosystem
scale (Gornish and Prather 2014). Shifting precipi-
tation and temperature regimes can result in
changes in functional type or species composition
and abundance and cause shifts in species distri-
butions (Lin et al. 2010, Chen et al. 2011, Thom
et al. 2017). Changes in disturbance regimes are
accelerating (Seidl et al. 2014) and act simultane-
ously with climate to impact plant community
structure. The direct effects of climate change and
altered disturbance precipitate indirect effects on
demographic processes and community-level
responses by altering intraspecific and interspeci-
fic competition between plant individuals (Tylia-
nakis et al. 2008). As such, quantifying individual
plant responses and changes in competitive inter-
actions in response to climate change and altered
disturbance regimes is a key challenge.
These issues are particularly pertinent for dry-
land ecosystems, as soil water availability and
temperature are often key limiting factors for
plant growth, influence plant species composi-
tion and richness (Butterfield and Munson 2016,
Pennington et al. 2017), and are expected to
change in the coming century (Palmquist et al.
2016a, b, Schlaepfer et al. 2017). Several lines of
evidence suggest dryland ecosystems are already
being exposed to increasingly extreme climate
conditions, including increases in the frequency
of extreme events (Trenberth et al. 2014, Wu
et al. 2015), large-scale drought (Seager et al.
2007, Cook et al. 2015), and drought-related mor-
tality (Fensham et al. 2015, Meddens et al. 2015).
In addition, many dryland ecosystems have
experienced substantial changes in disturbance
regimes, most notably increases in the size, fre-
quency, and severity of fire (Westerling et al.
2006, Balch et al. 2013) and heavy grazing by
livestock (Bai et al. 2012). When these distur-
bances interact with climatic change, the impacts
can exacerbate. Climate change-induced drought
and warming will likely continue to increase fire
frequency in the future at least in areas where
there is enough fuel to carry fire (McKenzie and
Littell 2017). Grazing impacts are also likely to
vary with soil moisture, as drought-adapted spe-
cies are often more resistant to grazing (Adler
et al. 2004, 2005), especially in ecosystems that
have a long evolutionary history of grazing
(Milchunas et al. 1988).
Simulation modeling, particularly individual-
based modeling, is an important tool to under-
stand how plant communities will respond to
changing environmental conditions and distur-
bance regimes. In particular, it is useful for explor-
ing questions over long time periods and under
future conditions for a range of environmental set-
tings, which are not feasible using experiments
(Smith and Huston 1990). The strengths of
individual-based models are that they simu-
late intraspecific and interspecific competition
between plant individuals and individual plant
responses (establishment, growth, and mortality)
to fluctuating environmental conditions, from
which population- and community-level proper-
ties emerge (Huston et al. 1988, DeAngelis and
Grimm 2014, Grimm et al. 2017). Thus, they can
be utilized to understand hierarchical (individual,
population, community, ecosystem) responses to
environmental variability and trajectories between
dynamic equilibrium (Huston et al. 1988). Because
individual-based simulation modeling explicitly
represents the key demographic stages of plants,
it can provide a more process-based representa-
tion of plant responses to climate change and dis-
turbance than bioclimatic envelope modeling,
albeit at a smaller scale.
Gap dynamics models are a subset of individ-
ual-based models, which have been used primar-
ily as a tool for understanding the dynamics in
forest stands after individual trees die, where
light is the limiting resource plants compete for
(Botkin et al. 1972, Shugart 1984, Seidl et al. 2012,
Morin et al. 2018). This approach has been
applied in non-forested ecosystems, but less fre-
quently (shortgrass steppe, Coffin and Lauenroth
1990, Coffin et al. 1993, big sagebrush ecosys-
tems, Bradford and Lauenroth 2006, desert grass-
lands, Peters 2000, 2002). In non-forested
drylands, gap dynamics models are similar to
gap dynamics models implemented in forests,
except that the primary resource for which plants
compete is soil water, which is the most frequent
limiting resource in drylands (Sala et al. 1992,
Loik et al. 2004, Gremer et al. 2015).
Here, we present STEPWAT2, an individual-
based, gap dynamics plant simulation model, as
a tool for exploring the impacts of climate, cli-
mate change, and disturbance regimes on
❖www.esajournals.org 2August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

dryland plant communities under current and
future conditions. STEPWAT2 can be applied to
any dryland ecosystem where soil water controls
plant growth and plant community structure.
Specifically, this paper has three goals: (1) to
describe the core modules and functions within
STEPWAT2 (model description), (2) to use field
data from big sagebrush plant communities to
validate STEPWAT2 model output (model vali-
dation), and (3) to highlight the usefulness of
STEPWAT2 as a framework for examining the
joint impacts of climate change and disturbance
regimes on future dryland plant communities
(model application). To pursue goal 2 and to
highlight the types of research questions that can
be addressed and results that can be produced
by STEPWAT2 (goal 3), we focus on 15 sites that
span the spatial extent of big sagebrush plant
communities in the western United States. For
these sites, we ask: How will climate change and
disturbances (fire and grazing) interact to influ-
ence plant functional type biomass and composi-
tion? We focus on big sagebrush ecosystems
because they are one of the most widespread
dryland ecosystems in North America and pro-
vide critical habitat for ~350 species of conserva-
tion concern (Davies et al. 2011), while providing
important ecosystem services (e.g., recreation,
energy resources; Knick et al. 2003, Rowland
et al. 2006).
STEPWAT2 MODEL DESCRIPTION
Overview of STEPWAT2 and model revisions
STEPWAT2 is a simulation framework that
combines a stochastic, individual-based plant
simulation model (based on STEPPE, Coffin and
Lauenroth 1990, Coffin et al. 1993, Bradford and
Lauenroth 2006) and a deterministic, process-
based soil water balance model (SOILWAT2, Sch-
laepfer et al. 2012a). STEPWAT2 simulates plant
demographics on patches that correspond to the
belowground resource space utilized by a full-
sized individual of the dominant species (Coffin
and Lauenroth 1990). Resources (soil water avail-
able for transpiration) are generated each year in
SOILWAT2, passed to STEPPE, and end-of-year
plant aboveground biomass is then passed back
to SOILWAT2 to update the vegetation parame-
ters before the next year’s SOILWAT2 run
(Fig. 1).
STEPPE, on which STEPWAT2 depends, was
originally conceptualized for the shortgrass
steppe (Coffin and Lauenroth 1990, Coffinetal.
1993), a semiarid grassland ecosystem in the cen-
tral United States, but has since been modified for
big sagebrush ecosystems (Bradford and Lauen-
roth 2006). The model was easily translatable to
big sagebrush ecosystems as soil water is the key
limiting resource in both ecosystems. STEPPE
simulates establishment, growth, and death of
SOILWAT2
Resource
availability
Soil
properties growth
Daily
weather
STEPPE
mortality
Establishment
Plant available
soil water
STEPWAT2
Biomass
Root depth
End-of-year biomass
Fig. 1. STEPWAT2 conceptual model diagram. STEPWAT2 is a coupled simulation model framework that links
SOILWAT2 and STEPPE. Each year, plant available soil water for transpiration is generated by SOILWAT2 and is
passed to STEPPE and serves as resource availability for plant growth. After establishment, growth, and mortal-
ity in STEPPE, the end-of-year biomass for multiple plant functional types is calculated and passed to SOILWAT2
to serve as input for the next year. Variables that are passed between SOILWAT2 and STEPPE are shown in bold.
❖www.esajournals.org 3August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

individual plants, whose life-history traits (phe-
nology, root distribution, growth rate, maximum
size, life span) are dictated by the species and
functional type to which the individual belongs
(Bradford and Lauenroth 2006). Individuals, spe-
cies, and functional types are all tracked within
STEPPE. Annual output from STEPPE includes
the number of individuals established for each
species (density), the total aboveground biomass
for each species and functional type (g/m
2
,here-
after “biomass”), and the number of individuals
that died according to their age.
SOILWAT2 is a daily time step, process-based
soil water model that simulates water balance
fluxes and pools for multiple soil layers (Sch-
laepfer et al. 2012a, Bradford et al. 2014). Output
from SOILWAT2 includes daily water fluxes and
pools. Estimates of transpiration, derived from
SOILWAT2 output, are used within STEPPE to
represent abiotic influences on resource alloca-
tion, plant growth, and mortality. Resource avail-
ability is determined by the ability of each
functional type to extract water through time
and with depth in the soil. This representation of
resource utilization with depth and across each
month for multiple plant functional types makes
STEPWAT2 particularly useful for understanding
how climate-driven changes in soil moisture will
impact demographic and community-level
responses in drylands. Resource availability from
SOILWAT2 and demand for resources in STEPPE
are captured by the parameter PR: the ratio of
resources required to resources available. PR
greater than one indicates resource limitation
(resources required exceed resources available),
while PR of one indicates there are sufficient
resources.
STEPWAT2 is run via the Monte Carlo method
to simulate vegetation dynamics in a single cell
at each site (non-gridded mode) or for a grid of
cells that are linked by dispersal (gridded mode)
to model spatially explicit vegetation dynamics
across a site (based on ideas presented by Coffin
et al. 1993). In either case, the size of each cell is
based on the belowground resource space of a
full-sized individual of the dominant species and
can be adjusted based on the dryland ecosystem
and dominant species in question. Both single
cell and gridded simulations use the same core
modules. However, when run in gridded mode,
seed dispersal among cells is enabled. The size
and arrangement of the gridded landscape are
specified by providing the number of cells within
the landscape and how those cells are configured
in 2D space. Each cell within the grid can have
its own soil properties, disturbance regime, and
species composition, but climate is held constant
for all cells within a site (Coffin et al. 1993).
Because soil characteristics and species composi-
tion can vary, resource availability for plant
establishment, growth, and mortality can differ
among cells in a gridded landscape. This can
result in divergence of species composition
among cells in the grid in response to spatial
heterogeneity. In addition, gridded mode allows
for examination of how disturbances that vary in
their spatial extent (e.g., wildfire, an oil or gas
well-pad, road network) and spatial patterns
(e.g., configuration, patch size, and number) may
impact dryland plant communities in the face of
climatic change. Two features in gridded mode
allow for vegetation to become established
before simulations begin. The first is a module in
which seed dispersal for the specified species is
allowed for a set number of years before the first
year of simulation. The second option, called
spin-up, runs the full model for a number of
years until steady-state vegetation is reached,
after which the experimental simulations begin.
In contrast, simulations in the non-gridded mode
begin with no established plant community and
assume unlimited seed rain of every species.
We have substantially revised STEPWAT2
since its original presentation in Coffin and
Lauenroth (1990) and Bradford and Lauenroth
(2006), which include changes to SOILWAT2 and
STEPPE. Within STEPPE, we added functionality
to simulate fire and grazing in both non-gridded
and gridded mode (see Disturbances), along with
addition of spin-up and seed dispersal options in
the gridded mode to aid in establishment of veg-
etation prior to simulation experiments. Annual
plant functional types are now fully represented
by individual plants within the model: Establish-
ment of annuals now occurs at the same time as
perennials and annual individuals are tracked
and grow throughout the year (previously indi-
viduals of annual species were not tracked and
had no capacity for growth). The most substan-
tial changes to SOILWAT2 include addition of
the following: a snow module, hydraulic redistri-
bution, plant functional types, an option to limit
❖www.esajournals.org 4August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

percolation and allow for saturated soils and
ponding of water, and an overhaul of soil water
that is available for transpiration (see Appendix S1
for more detail on SOILWAT2).
Below we briefly describe the main modules of
STEPWAT2 in the order of their execution: plant
establishment, generation of soil water resources
from SOILWAT2, partitioning of soil water
resources, growth, mortality, and implementation
of disturbances. A more detailed description of
each module is in Appendix S2. STEPWAT2 is writ-
teninC++ and is publicly available on Zenodo:
http://doi.org/10.5281/zenodo.1306924 (Palmquist
et al. 2018).
Plant establishment
The first process in STEPWAT2 is plant estab-
lishment for perennial and annual species. Estab-
lishment is stochastic, but also is a function of the
number of individuals that can establish each
year and the probability of establishment (P
estab
)
for each species (Table 1). P
estab
reflects the aver-
age probability of suitable environmental condi-
tions for germination and survival and the
species-specific requirements for regeneration. If
the establishment conditions for perennial spe-
cies are met, the number of individuals that
establish in that year is determined based on the
maximum number of seedlings that can establish
per year (maxindivs; Table 1; Appendix S2). For
species of annual functional types, the number of
individuals that establish is based on the number
of viable seeds in the seedbank (see Appendix S2
for further description of the annual seedbank)
and P
estab
: More seedlings are added as the
viable seedbank increases and as P
estab
increases.
Seedlings are added at the size of a 1-yr-old indi-
vidual for perennials and at the size of a recent
germinate for annuals. The biomass of added
individuals is specified for each species (minbio,
Table 1).
Each time establishment, growth, or mortality
occurs, the relative sizes of individuals, species,
and functional types are updated. The relative
size of each individual (relsize
ndv
) is a function of
its size relative to a full-sized individual of that
species (range =0–1). The species relative size
(relsize
sp
) is the summation of the relsize
ndv
and
reflects both the number and size of established
individuals belonging to that species. The mean
relative size of the functional type (relsize
ft
) is the
summed relsize
sp
divided by the number of
species in the functional type.
SOILWAT2: generation of soil water resources for
plants to use
After establishment, resource availability for
the year is generated by SOILWAT2. Key
processes within SOILWAT2 are influenced by
three types of inputs: weather and climate, vege-
tation parameters, and physical properties for
each soil layer (Schlaepfer et al. 2012a). Here, we
Table 1. STEPWAT2 model parameters for species.
Species FT age rP
estab
maxindivs minbio maxbio clonal vegindv slow
Artemisia tridentata 1 150†0.02 0.30|| 2 2# 400.0 N NA 4
Cryptantha sp. 2 1 0.947 0.03 5 0.035 4.7 N NA NA
Phlox hoodii 3 10 0.426# 0.15 10 0.035# 5.0 N NA 2
Bromus tectorum 4 1 0.947# varied 20 0.02‡5.0 N NA NA
Pseudoroegnaria spicata 5 35 0.474# 0.15 10 0.5# 15.096 Y 3 2
Sporobolus curtissii 629§0.474 0.05 3 0.605# 6†† Y32
Chrysothamnus viscidiflorus 717¶0.474‡0.15 2 0.986‡70.726‡NNA2
Opuntia polyacantha 8 30 0.289‡0.05‡1 2.25‡15‡Y3NA
Notes: FT, STEPPE functional types listed in Table 2; age, maximum age (years); r, intrinsic rate of growth; P
estab
, probability
of establishment; maxindivs, the maximum number of individuals for each species that can establish in each year; minbio,
biomass (g/m
2
) of a 1-yr-old individual for perennials and of a recent germinate for annuals; maxbio, the biomass at maturity;
clonal, whether the species can reproduce clonally; vegindv, the maximum number or ramets added during clonal growth; and
slow, the number of continuous slow-growth years.
†Ferguson (1964).
‡Coffin and Lauenroth (1990).
§Lauenroth and Adler (2008).
¶P. B. Adler, unpublished data.
# Bradford and Lauenroth (2006).
|| Schlaepfer et al. (2014).
†† Lucero et al. (2008).
❖www.esajournals.org 5August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

describe the differences in how weather and veg-
etation inputs are calculated when SOILWAT2 is
running within the STEPWAT2 framework.
Additional detail on how these inputs influence
water balance fluxes and pools is provided in
Appendix S1.
Weather data include daily precipitation and
daily minimum and maximum temperature,
while climate data consist of monthly mean rela-
tive humidity, wind speed, and cloud cover. His-
torical weather data or future weather data
extracted from global circulation models (GCMs)
and downscaled to the relevant temporal and
spatial scales can be used as input to SOILWAT2.
Because the establishment of a steady-state plant
community within STEPWAT2 often takes 50–
100 yr and because daily weather data are most
often not available for 100 yr or more, additional
weather data must be generated to run STEP-
WAT2. We use a first-order Markov weather gen-
erator within SOILWAT2, which is a modified
WGEN weather generator (Richardson and
Wright 1984), to produce daily weather data for
the required number of years. The Markov pro-
cess uses 30-yr daily precipitation averages and
wet- and dry-day probabilities and weekly tem-
perature averages and covariances calculated
from historical or predicted future weather to
generate daily weather data with statistical char-
acteristics that are equivalent to the input data
(Appendix S3).
There are four plant functional types in SOIL-
WAT2 (trees, shrubs, grasses, and forbs), whose
monthly parameters (estimates of total above-
ground biomass, litter, living aboveground bio-
mass, and root depth profile) influence a number
of ecohydrological processes (Appendix S1).
Each year before SOILWAT2 runs within the
STEPWAT2 framework, vegetation parameters
are updated based on the biomass (g/m
2
) of each
established functional type in STEPPE. STEPPE
functional type biomass is tracked on an annual
basis and must be converted to monthly biomass,
along with monthly values of litter (g/m
2
) and
percent live (Appendix S4: Table S1). In addition,
STEPPE functional type biomass must be con-
verted to biomass for each of the SOILWAT2
functional types, as multiple STEPPE functional
types can make up a single SOILWAT2 functional
type. Appendix S2 provides additional detail on
the conversion of annual STEPPE functional type
biomass to monthly SOILWAT functional type
biomass.
In addition to updating biomass and litter each
year, the relative proportion of active roots for
each STEPPE functional type in each soil layer is
updated before SOILWAT2 runs (Appendix S2).
Thereafter, transpiration coefficients are updated
for each established functional type in SOIL-
WAT2 based on the relative proportion of active
roots for each STEPPE functional type in each
soil layer. Finally, the relative fractional composi-
tion of each established SOILWAT2 functional
type is updated.
Available SOILWAT2 output within the STEP-
WAT2 framework include the following: precipi-
tation (cm), snowpack (SWE), air temperature
(°C), soil temperature (°C), potential evapotran-
spiration (cm), annual evapotranspiration (cm),
interception by vegetation and litter (cm), bare-
soil evaporation (cm), evaporation of intercepted
water (cm), surface water (cm), transpiration
(cm), infiltration (cm), percolation (cm), ground-
water recharge (cm), hydraulic redistribution
(cm), available soil water (cm), soil water content
(cm), soil water potential (MPa), volumetric water
content (m
3
/m
3
), and the number of wet days
(>1.5 MPa). These variables are available on
daily, weekly, monthly, and annual time frames.
The most important SOILWAT2 variable within
the STEPWAT2 framework is monthly transpira-
tion (cm) from each soil layer, which determines
resource availability for plant growth. Monthly
transpiration is calculated by summing daily tran-
spiration for each month for each soil layer for
each SOILWAT2 functional type. Since transpira-
tion is strongly coupled to leaf area index (LAI),
when plant biomass and LAI are low, so is tran-
spiration. For example, in a year after a distur-
banceevent,suchasafire, STEPPE passes low
biomass to SOILWAT2, which simulates low tran-
spiration for that year. However, some plant spe-
cies and functional types can recover quickly after
disturbance and produce a substantial amount of
new biomass within a year in response to low bio-
mass and high available soil moisture. To approxi-
mate this within-year new growth, we calculate
the amount of water that could be available to be
transpired and augment resource availability for
plant functional types in below-average transpira-
tion years. For each year, we compare the annual
transpiration/annual precipitation ratio to a
❖www.esajournals.org 6August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

moving average of the ratio of annual transpira-
tion/annual precipitation across simulation years.
If the current year’s ratio is within one standard
deviation of the moving average, then no adjust-
ments are made and resource availability is based
solely on actual transpired water. Otherwise,
additional soil water is added to actual transpira-
tion to augment resource availability (hereafter
“added transpired water,”Appendix S2). These
additional soil water resources would otherwise
not be utilized when plant biomass is low (e.g.,
disturbance years) if resource availability were
based solely on actual transpiration.
Resource partitioning
Resource partitioning to functional types in
STEPPE is determined by their relative proportion
of active roots in each soil layer in each month
and the matching of those temporal and spatial
root distributions to transpiration from SOIL-
WAT2 in each month and in each soil layer. Thus,
resource partitioning is sensitive to each functional
type’s phenology and root distribution. If added
transpired water is available, it is partitioned to
each STEPPE functional type based on the propor-
tion of actual transpired water each functional
type received and then is added to actual tran-
spired water (Appendix S2). After resource parti-
tioning to functional types, resources are then
divided among the individuals of each species in
each functional type. Individuals are sorted by
size and resources are allocated beginning with
the largest individuals and partitioned until none
remain (Appendix S2). In some years, extra
resources remain after partitioning to individuals.
These extra resources are divided proportionally
to functional types that can utilize other types’
resources according to each functional types’con-
tribution to total biomass (types with larger bio-
mass receive a greater share of extra resources).
Thereafter, extra resources are partitioned to indi-
viduals based on size, where larger individuals
receive extra resources first, along with a greater
share of extra resources.
Plant growth
There are two components of growth within
STEPWAT2: annual increase in size and annual bio-
mass increment. Annual increase in size is the
increase in the size of the perennial parts of the
plant, and biomass increment is the increase in
biomass that will senesce at the end of the growing
season. The increase in size of each plant individual
is a function of its optimum growth rate, resource
availability and resource competition with other
plants, and mean annual temperature (MAT). The
optimum growth rate (dR/dt)iscalculatedas:
dR=dt ¼rð1:0relsizendvðtÞÞgmod (MAT) (1)
where ris the intrinsic rate of growth,relsize
ndv
(t)
is the relative size of an individual at time t,and
gmod is the growth year modifier based on tem-
perature and resource availability in the current
year (Appendix S2). The intrinsic rates of growth
were calculated from years to reach full size
under optimal growing conditions (Coffinand
Lauenroth 1990, Bradford and Lauenroth 2006,
Table 1). Temperature effects on growth rates via
gmod give C
3
functional types an advantage
under cool conditions and C
4
functional types an
advantage under warm conditions (Appendix S2).
For succulents, increase in size does not occur in
wet years, due to the negative impact of wet
conditions on this functional type’s growth and
persistence (see Plant mortality).
Interspecific competition is represented in three
ways. First, functional types (and ultimately indi-
viduals of different species) receive different
amounts of resources each year based on the
matching of each functional type’s active root dis-
tribution to transpiration (see Resource partition-
ing). Second, species-specificrvalues simulate
differences in competitive ability between species
such that some species grow more rapidly in
response to available resources. Third, optimum
growth rates for each individual are modified by
gmod based on PR
ndv
(the ratio of resources
required/resources available for each individual)
such that higher PR
ndv
results in lower gmod val-
ues (see Appendix S2). Gmod values are also
modified based on whether that individual
belongs to a species that is C
3
or C
4
. Intraspecific
competition is represented by resource allocation
by size, in which larger individuals receive
resources first and a larger share of the resources.
Since larger individuals receive resources first,
the PR
ndv
for these individuals is more likely to
be near 1, resulting in higher gmod values.
Clonal species are handled differently from
unitary species in that clonal species can either
increase in size (add ramets) or decrease in size
❖www.esajournals.org 7August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

(lose ramets) at each time step. Increase in size is
handled identically to unitary species. However,
clonal species can recover following loss of ram-
ets by a disturbance. Each species is identified as
having or lacking the ability to recover from such
losses. For clonal species, we specify a probabil-
ity of regrowth after slow-growth mortality or
age-independent mortality and a probability of
recovery after mortality due to resource limita-
tion (Coffin and Lauenroth 1990).
Increase in biomass (that will senesce at the
end of the year) occurs when there are extra
resources available after the resource needs of
each individual in a functional type have been
met (Appendix S2).
Plant mortality
Mortality occurs through multiple mecha-
nisms: resource limitation, age-independent mor-
tality, slow-growth mortality, and succulent
mortality in wet years. Annual functional types
are not subject to these sources of mortality and
instead die at the end of each year.
The first type of mortality occurs when there
are not enough resources available in a given year
to support the current biomass of each individual
in a functional type (e.g., resource limitation). This
algorithm is controlled by PR
ft
, the ratio of
resources required to resources available for each
functional type, and the number of continuous
years low resources can be tolerated before low-
resource mortality occurs (stretch, Table 2;
Appendix S2). Low-resource mortality occurs
when PR
ft
>1. Individuals in each functional type
are sorted according to their relsize
ndv,
and then,
mortality occurs beginning with the smallest indi-
vidual. As resource limitation increases (e.g.,
higher PR
ft
values), the number of individuals
experiencing mortality increases (Appendix S2).
Clonal functional types are subject to an alter-
native source of mortality when resources are
limited (PR
ft
>1) and when the number of years
where PR
ft
>1 exceeds stretch. In cases where an
entire clonal individual is not eliminated from
resource limitation, its size can be reduced by
partial mortality. Smaller clonal individuals are
most susceptible to reduction (Appendix S2).
After low-resource mortality, killing occurs
for functional types that are susceptible to age-
independent mortality (mort, Table 2). Age-
independent mortality simulates the probability
that only a small percentage of individuals within
a cohort will reach their maximum age (age;
Table 1, Shugart 1984, Coffin and Lauenroth
1990) and represents a type II survivorship curve
(here, P
age-indep
=1%; Pearl 1928). However, in
combination with all other sources of mortality,
most species simulated within STEPWAT2 follow
type III survivorship curves. To implement age-
independent mortality, the age of each individual
is divided by the age for that species (ageratio).
The age, ageratio, and proportion of the cohort
surviving to maturity are used to calculate the
probability of mortality by year n(Appendix S2).
Thereafter, the slow-growth mortality routine
is implemented. Individuals that are growing
slowly may have a higher risk of mortality due
to greater vulnerability to stressors than individ-
uals with average or above average growth rates
(Coffin and Lauenroth 1990). A slow-growth rate
is 5% of the maximum growth rate (90% of the r)
and can result in mortality if the number of con-
tinuous slow-growth years (slow) exceeds the
number that is allowed (Table 2; Appendix S2).
An additional source of mortality in STEP-
WAT2 is succulent mortality in wet years. Several
lines of evidence suggest succulents are vulnera-
ble in wet years (Cook 1942, Lauenroth et al.
2009). First, the cactus moth and other herbivores
have greater larval survivorship in wet years,
which results in greater mortality of succulents
due to insect damage (Cook 1942). Second, when
water resources are abundant, grass, forb and/or
shrub biomass increases. This can lead to
decreases in succulents (Lauenroth et al. 2009),
due to shading and greater humidity surrounding
Table 2. STEPWAT2 model parameters for STEPPE
functional types.
Functional type FT Space Stretch
Big sagebrush 1 0.32 8
Annual C3 forbs 2 0.05 NA
Perennial C3 forbs 3 0.18 2
Annual C3 grasses 4 0.05 NA
Perennial C3 grasses 5 0.22 2
Perennial C4 grasses 6 0.04 2
Non-sagebrush shrubs 7 0.09 2
Succulents 8 0.05 8
Note: Space is the proportion of total resources each func-
tional type receives and stretch is the number of continuous
years low resources can be tolerated before low-resource
mortality occurs.
❖www.esajournals.org 8August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

cladodes leading to mortality from fungal patho-
gens. To implement wet-year mortality, a proba-
bility of succulent mortality is determined for
each year as a function of precipitation during the
growing season (April–September; see Eq. 16 in
Coffin and Lauenroth 1990). If it is a wet year,
partial mortality can occur, which increases with
growing-season precipitation (see Eq. 10 in Coffin
and Lauenroth 1990). At the end of the year, mor-
tality of annuals occurs, along with senescence of
biomass produced due to extra resources.
Disturbances
There are several types of disturbances within
STEPWAT2, including plant mortality in a partic-
ular year, extirpation of plants (and the seed-
bank) in a year after which they do not return,
prescribed fire, wildfire, and grazing. Prescribed
fire, wildfire, and grazing can be turned on or
off. Prescribed fire is implemented for each func-
tional type by specifying: the fire-return interval
(FRI, the fixed number of years between fires),
the proportion of biomass removed by fire
(which combines fire severity and a species’sen-
sitivity to fire), and the proportion of biomass
recovered after fire via resprouting (0–1). In con-
trast, wildfire occurs probabilistically, but bio-
mass removed and recovered after fire is
implemented identically to prescribed fire. Both
prescribed fire and wildfire can result in plant
mortality if all plant biomass is removed by fire.
If grazing is implemented, grazing frequency
and intensity must be provided for each func-
tional type, which represents the proportion of
biomass removed by grazers at the frequency that
is specified. The impacts of different grazing ani-
mals (e.g., cattle vs. native herbivores) and forag-
ing preference for particular plant functional
types (e.g., perennial grasses) can be manipulated
by adjusting the proportion of biomass removed
for each plant functional type. Grazing can
directly kill plants if all biomass is removed by
grazers or indirectly through low-resource mor-
tality, as larger plant individuals receive resources
before smaller (grazed) individuals.
FIELD DATA FOR MODEL VALIDATION
We validated our model with vegetation data
collected in 2013 and 2014 from 15 big sagebrush
plant communities (Fig. 2; Pennington et al.
2017). These 15 sites were distributed across the
spatial extent of big sagebrush ecosystems in the
western United States and were selected based on
several criteria. First, sites were chosen to capture
areas in five of the seven Greater Sage-grouse
Management Zones (MZs, Manier et al. 2013)
with high relative density of greater sage-grouse
(Centrocercus urophasianus) breeding populations
(Appendix S5; Fig. 10 in Doherty et al. 2016):
Great Plains (MZ I), Wyoming Basins (MZ II),
Southern Great Basin (MZ III), Snake River Plain
(MZ IV), and Northern Great Basin (MZ V). Sec-
ond, all sites had few signs of disturbance from
oil and gas development, invasive species (e.g.,
cheatgrass, Bromus tectorum), fire, and heavy graz-
ing (Pennington et al. 2017). Finally, sites had
level topography (range of slope, 0–1.3°).
Within each site, we established three 100 m
2
plots. Within each 100-m
2
plot, we randomly sam-
pled 30 20 950 cm quadrats and recorded per-
cent canopy cover for each plant species, using
cover classes (1 =0.1–5%, 2 =5–15%, 3 =15–
25%, 4 =25–40%, 5 =40–60%, 6 =60–100%;
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
Field sites
Fig. 2. Map of the 15 big sagebrush field sites in the
western United States on a background of sagebrush
ecosystem occurrence (green shading) from Schlaepfer
et al. (2012a). Sagebrush ecosystem occurrence was
summarized for 10 910 km
2
pixels from 30 930 m
2
GAP grid cells. Solid black lines delineate Sage-grouse
Management Zones (MZ).
❖www.esajournals.org 9August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

Daubenmire 1959). We assigned each taxon found
in each of the 15 field sites to a functional type:
annual C
3
forbs, perennial C
3
forbs, annual C
3
grasses, perennial C
3
grasses, perennial C
4
grasses, big sagebrush, non-sagebrush shrubs,
and succulents (Appendix S6: Table S1). To calcu-
late absolute cover for functional types in each of
the 90 quadrats in each site, we summed cover
class midpoints for each functional type for each
quadrat. We then averaged the quadrat-level
cover values to determine a site level mean cover
value for each functional type. To calculate func-
tional type percent relative cover at the site level,
we divided the mean functional type cover by the
total cover and multiplied by 100.
To characterize big sagebrush stand structure,
we counted the number of big sagebrush individu-
als in each 100-m
2
plot and assigned each individ-
ual a canopy volume size class, which captures
both average canopy diameter and height: class
1=diameter 0–5cm,2=5–15 cm, 3 =15–30 cm,
4=30–50 cm, 5 =50–75 cm, 6 =75–100 cm, 7 =
100–150 cm, and 8 =>150 cm. For each big sage-
brush individual, we calculated canopy volume
(m
3
)usingthemidpointofthecanopydiameter
class that was assigned to that individual. There-
after, we calculated mean canopy volume (m
3
)for
each site. For 10% of the big sagebrush individuals
within each plot, we also measured the length,
weight, and height of the canopy in cm. We then
calculated big sagebrush biomass (g/m
2
)foreach
of the 15 sites using length, width, and height
measurements of individual shrubs, the allometric
equation provided in Cleary et al. (2008), and big
sagebrush density (individuals/m
2
). We collected
soil samples from three locations in each plot at
depths of 0–10 cm, 10–20 cm, and 20–30 cm, for a
total of nine samples per plot and determined soil
texture using a method modified from Bouyoucos
(1962).
MODEL PARAMETERS FOR BIG SAGEBRUSH
PLANT COMMUNITIES
We simulated plant community dynamics
using the non-gridded mode for the 15 sites on
patches the size of a full-sized big sagebrush indi-
vidual (1 m
2
), which was determined from esti-
mates of average rooting characteristics of
A. tridentata (~1m
2
; Sturges 1977) and average
density of individuals/m
2
from our 15 sites
(mean =1.5 individuals/m
2
,range=0.6–3.5).
Eight plant functional types were included in our
simulations: annual C
3
forbs, perennial C
3
forbs,
annual C
3
grasses, perennial C
3
grasses, perennial
C
4
grasses, big sagebrush, non-sagebrush shrubs,
and succulents. C
4
forbs are uncommon in big
sagebrush ecosystems, and only one C
4
forb
genus (Amaranthus; Sage 2004) occurred in one of
our sites. We removed this taxon and represented
all forbs as C
3
. For simplicity, we represented each
functional type by a single plant species (Table 1),
which are both common and widely distributed
in big sagebrush ecosystems in our data sets (Mar-
tyn et al. 2016, Pennington et al. 2017). For each
of the 15 sites, we turned off functional types that
were not detected in the field.
We specified several key parameters for each
functional type including space, stretch, rooting
depth distribution, and phenological activity
(Table 2; Appendix S7: Tables S1, S2). Space is an
estimate of the resource space each functional
type can access and is represented as the propor-
tion of total resources each functional type
receives and is a function of the temporal match-
ing of available soil water with depth and active
root distribution with depth (Coffin and Lauen-
roth 1990, Bradford and Lauenroth 2006). We
modified the space parameters for C
3
grasses, C
4
grasses, and shrubs across our 15 field sites to
represent differences in the relative abundance
and distribution of these functional types in the
western United States (Paruelo and Lauenroth
1996, Epstein et al. 2002). First, we extracted the
specificC
3
grass, C
4
grass, and shrub potential
relative abundance values from Epstein et al.
(2002) for each field site. We rescaled those val-
ues so they summed to 1, summed the previous
space parameters for those functional types, and
then multiplied the summed value by the
rescaled relative abundance values from Epstein
et al. (2002). For shrubs, we partitioned the new
space parameter proportionally between big
sagebrush and non-sagebrush shrubs based on
the previous space parameters. For C
3
grasses,
we partitioned the new space parameter
proportionally between perennial C
3
grasses and
annual C
3
grasses based on the previous space
parameters.
For functional types that are drought-tolerant
(big sagebrush, succulents), we increased the
stretch parameters (the number of years a
❖www.esajournals.org 10 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

functional type can survive low-resource condi-
tions) relative to other functional types (Table 2).
Rooting depth distributions were based on
parameters identified in the literature (App-
endix S7: Table S1). Phenological activity was
based on published literature whenever possible,
but otherwise was estimated from life-history
strategies for each functional type (Appendix S7:
Table S2). Big sagebrush is phenologically active
throughout the year, but the bulk of activity is
concentrated in March–August (Lauenroth et al.
2014). Phenological activity for C
3
forbs and C
3
grasses begins earlier in the year than for C
4
grasses, as C
3
species begin growth earlier to
exploit moist conditions in spring and early sum-
mer (Paruelo and Lauenroth 1996). Annual C
3
species have most of their phenological activity
concentrated in May and June and complete
their life cycle by mid-summer before their peren-
nial counterparts (Campbell and Harris 1977,
Appendix S7: Table S2). The one exception is
annual C
3
grasses, represented by the winter
annual Bromus tectorum, which germinate in fall
(Jones et al. 2015), and are phenologically active
from October to June. All functional types were
subject to age-independent and slow-growth
mortality except annuals. For clonal species, we
specified a 90% probability of regrowth after
slow-growth mortality or age-independent mor-
tality and a 75% probability of regrowth after
low-resource mortality.
Species-specific parameters were obtained
from the literature for age, r,P
estab
, minbio, and
maxbio (see Table 1). In cases where we could
not find specific parameters, we estimated values
for each species based on functional type infor-
mation provided in Coffin and Lauenroth (1990)
and Bradford and Lauenroth (2006). The pres-
ence and abundance of cheatgrass (Bromus tecto-
rum) varies considerably across the range of big
sagebrush ecosystems, with greater abundance
in the Great Basin, Columbia River Plateau, Col-
orado Plateau, and parts of the Snake River Plain
(Brummer et al. 2016). To accurately represent
differences in the presence and abundance of
cheatgrass across our 15 sites, we allowed the
P
estab
for cheatgrass to vary. We assigned one of
three p
estab
values to sites (0.015, 0.15, 0.25) based
on whether sites occur in a region with high
cheatgrass abundance, as documented by Brum-
mer et al. (2016). All sites in the Great Plains (MZ
I) and Wyoming Basins (MZ II) were assigned
P
estab
values of 0.015, sites in the Southern Great
Basin (MZ III) were assigned values of 0.15, and
sites in the Snake River Plain (MZ IV) and North-
ern Great Basin (MZ V) were assigned values of
0.25.
We specified maxindivs, the maximum num-
ber of individuals for each species that can estab-
lish in each year, to reflect average density for
each species in 1 m
2
(Table 1). We set vegindv,
the number of seedling-sized units added when
clonal growth occurs, to three for all clonal spe-
cies: Pseudoroegneria spicata,Sporobolus curtissii,
and Opuntia polyacantha. Finally, slow was set to
four for Artemisia tridentata and two for all other
perennial species.
STEPWAT2 SIMULATIONS
We ran STEPWAT2 for our 15 sites for 300 yr
and 100 iterations under current climatic condi-
tions (1980–2010) and future climatic conditions
derived from 13 GCMs (Appendix S8: Table S1)
for RCP8.5 for end of century (2070–2100). We
simulated conditions for 300 yr because it
typically takes 50–100 yr for the vegetation to
reach steady-state conditions in non-gridded
mode. We chose the 13 GCMs to capture models
that perform well in the western United States
(Rupp et al. 2013) and span the families of
GCMS (Knutti et al. 2013), according to the rec-
ommendations of Baker et al. (2017). Current and
future monthly gridded climate data for each
GCM were extracted from the Downscaled
CMIP3 and CMIP5 Climate and Hydrology Pro-
jects archive at http://gdo-dcp.ucllnl.org/downsca
led_cmip_projections/ (data accessed on February
3, 2014; Maurer et al. 2007). Hybrid-delta down-
scaling was implemented to obtain future daily
forcing (Hamlet et al. 2010, Tohver et al. 2014).
We used rSFSTEP2, an R program that runs
STEPWAT2 for multiple sites, climate scenarios,
and disturbance regimes (Appendix S9), to create
300 yr of weather data for each site. Specifically,
we used rSFSTEP2 to create site-specific versions
of the required files for the Markov weather gen-
erator (Appendix S9). We created these files
using 30 yr of weather data for each site for cur-
rent conditions and all GCMs and then utilized
the Markov weather generator in tandem with
the site-specificfiles to run STEPWAT2. To
❖www.esajournals.org 11 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

estimate parameters for soil properties for each
site, we calculated mean soil texture across all
nine soil samples collected in the field in each site
and specified those values as inputs.
Using rSFSTEP2, we ran simulations for all
combinations of three FRIs (no fire, 50-yr, and 10-
yr return intervals) and three grazing intensities
(light, moderate, and heavy) under all of the cli-
matic conditions described above. We utilize
these fire and grazing treatments as potential sce-
narios that represent the range of fire and grazing
regimes that big sagebrush ecosystems are cur-
rently experiencing. Big sagebrush is fire-intoler-
ant and does not re-sprout after fire (Schultz
2006); thus, there was no recovery of big sage-
brush biomass the year after a fire. We also
assumed no recovery for non-sagebrush shrubs
and succulents. We implemented a 40% recovery
for perennial C
3
forbs, and a 70% recovery for
perennial C
3
and C
4
grasses in the year after fire,
reflecting evidence that perennial grasses are typ-
ically not killed by fire and can respond rapidly
post-fire (Ellsworth and Kauffman 2017). Grazing
occurred every year. Under light grazing, we
assumed only grasses and forbs were grazed, as
those functional types are preferred by native
ungulates and cattle (Davies et al. 2014). As graz-
ing intensity increased, more biomass was
removed for each functional type. For grasses
and forbs, the proportion of biomass removed for
light, moderate, and heavy grazing treatments
was 0.24, 0.41, and 0.58, respectively. For big
sagebrush and non-sagebrush shrubs, the propor-
tion of biomass removed was 0, 0.02, and 0.10 for
light, moderate, and heavy grazing intensities,
respectively. We assumed succulents were never
grazed. Light, moderate, and heavy grazing
intensities were based on estimated consumption
of aboveground net primary production data
under light, moderate, and heavy grazing in
shrublands around the globe, including big sage-
brush ecosystems (see Fig. 1 in Milchunas and
Lauenroth 1993).
MODEL VALIDATION
We compared model output from STEPWAT2
under current climate conditions, light grazing,
and no fire to field data from the 15 sites for the
dominant functional types: big sagebrush, peren-
nial C
3
grasses, and perennial C
4
grasses, along
with total herbaceous and total shrub abun-
dance. We utilized simulated values under light
grazing and no fire because the field sites we
sampled exhibited no signs of fire or heavy graz-
ing. The values presented here are means across
the 100 iterations and the last 100 yr. We also
compared simulated density values for big sage-
brush and forb and perennial grass species to
density values recorded in big sagebrush sites in
Wyoming during 2017. Finally, we compared
simulated biomass to published values from big
sagebrush plant communities.
Shrub functional types
For big sagebrush, we compared mean canopy
volume (m
3
) for each site to mean simulated bio-
mass (g/m
2
), with the expectation that sites with
larger mean canopy volume would also have
larger mean biomass. We also compared mean
big sagebrush biomass (g/m
2
) calculated from
field measurements to mean simulated bio-
mass (g/m
2
). In addition, we examined differ-
ences in total shrub relative abundance between
simulated data and field data.
Big sagebrush field-measured canopy volume
was positively correlated with simulated big sage-
brush biomass (r=0.47): Sites with the largest
canopy volume also had the largest simulated
biomass (Fig. 3A). In addition, field-derived esti-
mates of big sagebrush biomass were positively
correlated (r=0.50) with simulated big sagebrush
biomass (Fig. 3B). Our mean simulated biomass
values for big sagebrush (113–332 g/m
2
)arewithin
the range reported in other big sagebrush field
studies (Rickard 1985, 100–148 g/m
2
; Vora 1988,
9.8–1534 g/m
2
; Perfors et al. 2003, 86–365 g/m
2
).
Mean big sagebrush density from the 15 field sites
ranged from 0.6 to 3.5 stems/m
2
(V. Pennington,
unpublished data). Our simulated density for big
sagebrush closely overlap those values (simulated
range across all sites =1.6–4.1 stems/m
2
).
For most sites, total shrub relative biomass
from simulations was greater than total shrub
relative cover from field data (Fig. 4B). However,
simulated relative biomass for most sites was
within one standard deviation (SD) of relative
cover documented in the field (Fig. 4B).
Herbaceous functional types
We compared the mean simulated relative bio-
mass of perennial C
3
and C
4
grasses to the mean
❖www.esajournals.org 12 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

relative cover of perennial C
3
and C
4
grasses
from field data. We also examined differences in
the total herbaceous relative abundance from
simulations and from field data.
Perennial C
3
grass simulated mean relative
biomass was greater than perennial C
3
grass
mean relative cover recorded in field sites, except
for the three sites in the Great Plains (MZ I,
Fig. 5A). Despite this, seven sites had simulated
mean relative biomass that was within 10% of
mean relative cover values and all of the simu-
lated values were within one SD of relative cover
recorded in the field (Fig. 5A). In our simula-
tions, perennial C
3
grasses were represented by
Pseudoroegneria spicata (bluebunch wheatgrass).
Mean simulated densities for P. spicata varied
from 9 to 17 individuals/m
2
across the 15 field
sites. These are very similar to density values we
have for P. spicata from 2017 in three different big
sagebrush field sites in Wyoming: 1–17.5 indi-
viduals/m
2
(L. Smith, unpublished data).
For perennial C
4
grasses, our simulated mean
relative biomass was nearly identical to relative
cover for two of the sites in the Great Plains
(Fig. 5B). For the remaining two sites with peren-
nial C
4
grasses, one site in the Great Plains had
slightly greater simulated relative biomass (5.4%)
than relative cover, while the second site in the
Southern Great Basin (MZ III) had slightly lower
simulated relative biomass (10.7%) than relative
cover. However, in both cases, simulated relative
biomass was within one SD of relative cover val-
ues from the field (Fig. 5B).
In our simulations, Phlox hoodii, a common and
ubiquitous species in big sagebrush plant com-
munities (Pennington et al. 2017), represented
the perennial C
3
forb functional type. Mean sim-
ulated density for P. hoodii across our sites was
3.9 individuals/m
2
, which fell within the range of
densities (1–22 individuals/m
2
) that we recorded
in 2017 across 19 big sagebrush sites in Wyoming
(L. Smith, unpublished data). In addition, our
mean simulated biomass of 7.2 g/m
2
for peren-
nial forbs is similar to values reported for peren-
nial forbs in other big sagebrush studies (Rickard
1985, 3.2 g/m
2
; Bates et al. 2006, 3–13 g/m
2
).
Total herbaceous simulated relative biomass
was in most cases slightly lower than total herba-
ceous relative cover from our field data (Fig. 4A).
At most, these differences in mean relative abun-
dance were 27.8%, and 9 of 15 sites have differ-
ences of 10% or less. In all cases, the values we
100 150 200 250 300 350
0.1 0.2 0.3 0.4 0.5 0.6
Simulated biomass (g/m2)
Field-measured canopy volume (m3)
r = 0.47
A)
MZ I
MZ II
MZ III
MZ IV
MZ V
100 150 200 250 300 350
100 200 300 400 500 600
Simulated biomass (g/m2)
Field-derived biomass (g/m2)
r = 0.5
B)
MZ I
MZ II
MZ III
MZ IV
MZ V
Fig. 3. Comparisons of mean big sagebrush canopy volume (m
3
) from field data to simulated mean big sage-
brush biomass (g/m
2
) (A), and mean big sagebrush biomass (g/m
2
) estimated from allometric equations and field
measurements to simulated mean big sagebrush biomass (g/m
2
) (B). Simulated biomass values represent means
over 100 iterations and the last 100 yr of the simulations for current conditions, light grazing, and no fire. Sites
are colored according to their Sage-grouse Management Zone (MZ).
❖www.esajournals.org 13 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

simulated for relative biomass were within one SD
of relative cover measured in the field (Fig. 4A).
MODEL APPLICATION—IMPACTS OF CLIMATE
CHANGE AND DISTURBANCE ON BIG
SAGEBRUSH PLANT COMMUNITIES
We summarized simulated biomass (g/m
2
)for
each site for current and future conditions (across
13 GCMs for RCP8.5 by 2070–2100) and for dif-
ferent fire-return intervals and grazing intensi-
ties. We focus on presenting results for the two
dominant functional types in big sagebrush plant
communities: big sagebrush and perennial C
3
grasses. However, we calculated absolute
changes in biomass from current conditions to
2070–2100 for all functional types. Results for
each site are presented as the median biomass
across simulations forced by the 13 GCMs and as
the mean biomass across the last 100 yr of the
simulations, unless otherwise noted. All values
represent the mean across 100 iterations.
Under the no fire and light grazing treatment,
most sites were expected to have increases in big
sagebrush biomass by 2100 (Fig. 6A), with the
largest increases in our sites in the Northeast
(MZs I and II: Great Plains and Wyoming Basins)
and one high elevation (2115 m) site in the
Southwest (MZ III: Southern Great Basin,
Fig. 7B). In contrast, the two sites farthest west
Total herb relative abundance
020 40 60 80
Blak Malt Winn Cowd Jeff Pine Rock Loa Wells Yomb Gras Jack Laid MCD Adel
A)
MZ I MZ II MZ III MZ IV MZ V
Field relative cover Sim relative biomass
Total shrub relative abundance
Site
0 20406080
100
Blak Malt Winn Cowd Jeff Pine Rock Loa Wells Yomb Gras Jack Laid MCD Adel
B)
MZ I MZ II MZ III MZ IV MZ V
Fig. 4. (A) Total herbaceous percent relative cover from field sites (light blue) vs. total herbaceous percent rela-
tive biomass from simulations (green) for each site. (B) Total shrub percent relative cover from field sites (light
blue) vs. total shrub percent relative biomass from simulations (green) for each site. Simulated biomass values
represent means over 100 iterations and the last 100 yr of the simulations for current conditions, light grazing,
and no fire. Error bars for field data represent 1 SD of relative cover across 90 Daubenmire quadrats in each site.
Error bars for simulated biomass represent 1 SD of relative biomass across 300 simulation years. Sites are ordered
according to their Sage-grouse Management Zone (MZ).
❖www.esajournals.org 14 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

will likely have decreases in big sagebrush bio-
mass in the future (14.4 and 1.4 g/m
2
, respec-
tively; Figs. 6A, 7B). For perennial C
3
grasses,
both increases (N =5) and decreases of biomass
(N =10) were projected (Figs. 6A, 7E). The lar-
gest predicted increases in perennial C
3
grass
biomass (5–10 g/m
2
) were for the three sites in
Wyoming (Fig. 7E). Biomass for almost all of the
remaining functional types will likely decrease or
not change by 2100, with the exception of peren-
nial C
4
grasses, for which increases in biomass
were projected in all four sites where perennial
C
4
grasses currently occur (Fig. 6B).
Grazing interacted with climate to influence
functional type biomass by 2100. Under heavy
grazing, increases in big sagebrush biomass
were smaller for almost all sites, relative to
light grazing (light grazing mean absolute
change across all sites =15.2 g/m
2
, heavy graz-
ing mean absolute change across all sites =
12.7 g/m
2
; Fig. 7B, C). The impact of heavy
grazing on perennial C
3
grass biomass was con-
siderable: 14 sites were expected to decrease in
biomass under heavy grazing, relative to 10
sites under light grazing. Furthermore, there
were large declines in perennial C
3
grass bio-
mass under heavy grazing (heavy grazing
mean absolute change across all sites =9.8 g/
m
2
, light grazing mean absolute change across
all sites =0.1 g/m
2
; Fig. 7E, F).
C3 grass relative abundance
010
20 30 40 50 60
Blak Malt Winn Cowd Jeff Pine Rock Loa Wells Yomb Gras Jack Laid MCD Adel
A)
MZ I MZ II MZ III MZ IV MZ V
Field relative cover Sim relative biomass
C4 relative abundance
Site
010 20 30 40 50
Blak Malt Winn Cowd Jeff Pine Rock Loa Wells Yomb Gras Jack Laid MCD Adel
B)
MZ I MZ II MZ III MZ IV MZ V
Fig. 5. (A) Perennial C3 grass percent relative cover from field sites (light blue) vs. perennial C3 grass percent
relative biomass from simulations (green) for each site. (B) Perennial C4 grass percent relative cover from field
sites (light blue) vs. perennial C4 grass percent relative simulated biomass from simulations (green) for each site.
Simulated biomass values represent means over 100 iterations and the last 100 yr of the simulations for current
conditions, light grazing, and no fire. Error bars for field data represent 1 SD of relative cover across 90 Dauben-
mire quadrats in each site. Error bars for simulated biomass represent 1 SD of relative biomass across 300 simula-
tion years. Sites are ordered according to their Sage-grouse Management Zone (MZ).
❖www.esajournals.org 15 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

Fire substantially reduced the mean future bio-
mass of big sagebrush across all sites: 38.3 g/m
2
under light grazing and FRI of 10 yr vs. 266.8
g/m
2
under light grazing and no fire (Fig. 8A, B).
Frequent fire had less of an impact on future
perennial C
3
grass biomass: 68 g/m
2
for FRI of
10 yr and light grazing versus 81 g/m
2
for no fire
and light grazing (Fig. 8C, D). Furthermore,
perennial C
3
grass biomass recovered more
rapidly after each fire event than big sagebrush
(Fig. 8B, D).
DISCUSSION
Climate change and altered disturbance
regimes are expected to impact drylands in the
future (Palmquist et al. 2016a, b, Schlaepfer et al.
2017) and alter competitive hierarchies. Thus,
quantifying the direct and indirect effects of cli-
mate change will be critical for understanding
future plant community responses. STEPWAT2 is
a useful tool for understanding demographic
and community-level outcomes as emergent phe-
nomena of individual plant responses to chang-
ing climate and disturbance regimes, as it
simulates soil water availability in multiple soil
layers and competitive interactions in response
to fluctuating soil water resources. Furthermore,
STEPWAT2 can elucidate how climate change,
fire, and grazing will interact to influence plant
community structure for multiple future time
periods throughout the 21st century in any dry-
land ecosystem. The STEPWAT2 modeling
framework has the capacity to identify specific
disturbance regimes that will push dryland
ecosystems across thresholds and to inform con-
servation planning and land management deci-
sions in the face of environmental change.
Model validation
Our model output for big sagebrush biomass
was positively correlated with field-derived big
sagebrush canopy volume and biomass, indicat-
ing that STEPWAT2 is adequately representing
variability in big sagebrush abundance across
sites. Furthermore, simulated big sagebrush
density was similar to field-derived density,
despite the fact that density varies considerably
with time since disturbance, grazing history,
and land-use history (Miller and Eddleman
Sagebrush C3PG
–10
0
10
20
30
40
Functional group
Change in Biomass (g/m2)
A)
C3AF C3PF C3AG C4PG Shrub Succ
–4
–2
0
2
Functional group
Change in Biomass (g/m2)
B)
Fig. 6. Change in functional type biomass from current conditions (1980–2010) to end of century (2070–2100)
with no fire and light grazing. Biomass values represent means over 100 iterations and the last 100 yr of the sim-
ulations for the established functional types at each site. Functional types include big sagebrush, perennial C
3
grasses (C3PG), annual C
3
forbs (C3AF), perennial C
3
forbs (C3PF), annual C
3
grasses (C3AG), perennial C
4
grasses (C4PG), shrubs, and succulents (succ). Values shown are the median (black horizontal line), first and
third quartiles (ends of the box), range (ends of the whiskers), and outliers (circles) across all sites for the median
biomass across simulations forced by 13 GCMs for RCP8.5.
❖www.esajournals.org 16 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

2001), factors we did not explicitly control for or
simulate. For total shrub abundance, simulated
mean relative biomass was almost always
greater than mean relative cover estimated in
the field, despite being within one standard
deviation of relative cover. We did not expect a
one-to-one relationship between total shrub rela-
tive biomass and total shrub relative cover, as
there is greater biomass per unit cover for
shrubs than for other functional types (e.g.,
forbs, grasses). Instead, we expected greater rel-
ative biomass than relative cover and that is
what we documented.
Simulated mean relative biomass and mean
density values for perennial C
3
and C
4
grasses
were similar and in all cases within one standard
deviation of mean relative cover and mean den-
sity documented in the field. In addition, mean
density for Phlox hoodii, which represented
perennial C
3
forbs, was within the range of den-
sity values recorded for Phlox hoodii in a different
set of big sagebrush sites in Wyoming (L. Smith,
unpublished data). However, total herbaceous
mean simulated relative biomass was often less
than total herbaceous mean relative cover from
field sites. This difference occurs because there is
less biomass per unit cover for grasses and forbs,
relative to shrubs.
A challenge in validating STEPWAT2 was
relating the most frequently used metric of
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
120 –180
180 – 240
240 – 320
320 – 380
Biomass
A) Current Sagebrush Biomass, light grazing
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
Biomass
B) Change in Sagebrush Biomass, light grazing
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
Biomass
C) Change in Sagebrush Biomass, heavy grazing
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
63 – 73
73 – 83
83 – 93
93 – 103
Biomass
D) Current C3Pgrass Biomass, light grazing
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
–15 to –10
–10 to 0
0 – 5
5 – 10
Biomass
–15 to 0
0 – 10
10 – 20
20 – 30
30 – 45
E) Change in C3Pgrass Biomass, light grazing
MZ 1
MZ 2
MZ 7
MZ 3
MZ 4
MZ 5
MZ 6
–15 to –10
–10 to 0
0 – 5
5 – 10
Biomass
–15 to 0
0 – 10
10 – 20
20 – 30
30 – 45
F)Change in C3Pgrass Biomass, heavy grazing
Fig. 7. Maps of big sagebrush and perennial C
3
grass (C3Pgrass) biomass for the 15 field sites. Panels show
current (1980–2010) biomass (A, D), change in biomass from current to end of century (2070–2100) with no fire
and light grazing (B, E), and change in biomass from current to end of century with no fire and heavy grazing
(C, F). Biomass values represent means over 100 iterations and the last 100 years of the simulations for the med-
ian biomass across simulations forced by 13 GCMs for RCP8.5. Maps are color-coded according to current bio-
mass or change in future biomass: Warm colors indicate decreases in biomass, while cool colors indicate
increases in biomass in the future.
❖www.esajournals.org 17 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

abundance in plant community field studies
(percent cover) to the metric of abundance avail-
able from STEPWAT2 (biomass). Whenever pos-
sible, we included additional biomass estimates
from other big sagebrush studies to validate
simulated biomass from STEPWAT2. There is a
clear need for studies that collect both biomass
and percent cover data in the field to build
relationships between these two metrics of
abundance.
While our modeling approach lacks some of
the detailed ecophysiological complexity of
plant growth implemented in other simulation
models (e.g., photosynthesis, nutrient cycling),
STEPWAT2 is designed to be sensitive to water
availability, the key resource for plants in dry-
land ecosystems, by utilizing a process-based
ecosystem water balance model to determine
soil water availability coupled with classic
logistic growth equations. STEPWAT2 follows a
long tradition of gap dynamic models that are
conceptualized and implemented at the individ-
ual patch or stand level (Botkin et al. 1972, Shu-
gart 1984), which are utilized extensively for
forests and other ecosystems (Morin et al.
2018). The advantages of gap dynamics models
Fig. 8. End-of-century big sagebrush and perennial C
3
grass (C3Pgrass) biomass (g/m
2
) across the 300-yr simu-
lation with light grazing and no fire (A, C) and light grazing and fire every 10 yr (B, D). The colored lines are the
response of the 15 sites, while the black line is the average response for all sites over the final 100 yr. Biomass
values represent means over 100 iterations for the median biomass across simulations forced by 13 GCMs for
RCP8.5.
❖www.esajournals.org 18 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

are that they simulate both interspecificand
intraspecific competition for limiting resources
at the scales at which plants compete for them
(e.g., fine spatial scales or patch scales) and rep-
resent individual plant responses to disturbance
and fluctuating environmental conditions.
Thus, while representing plant growth simplis-
tically, gap dynamics models like STEPWAT2
have some advantages over other approaches
that explicitly represent detailed ecophysiologi-
cal aspects of plant growth, albeit at larger spa-
tial scales.
Model application
Simulations for our 15 field sites under future
climatic conditions and different disturbance
regimes highlight the utility of this model in
understanding how climate change, fire, and
grazing will interact to influence plant functional
type biomass and composition. Under no fire
and light grazing, for most of our sites, big sage-
brush biomass is expected to increase in the
future, except for the two sites farthest west,
which will have decreases in big sagebrush bio-
mass due to declines in available soil water in the
future. These results are consistent with other
studies that have documented drying leading to
declines in big sagebrush in warmer sites in the
Desert Southwest or along the western edge of
the big sagebrush distribution (Neilson et al.
2005, Schlaepfer et al. 2012b, Renwick et al.
2018). Most sites were projected to have
decreased biomass of C
3
grasses and forbs, while
perennial C
4
grasses were expected to increase
by 2100. These results suggest that plant func-
tional types in big sagebrush plant communities
will respond differently to climatic change, with
potential increases in perennial C
4
grasses due to
warming and decreases in C
3
herbaceous species.
Thus, STEPWAT2 can be utilized to explore spe-
cies-specific, functional type-specific, and com-
munity-level responses to climate change and
identify sites or geographic regions that would
continue to support the dominant functional
types in dryland plant communities in the
future.
Grazing intensity and fire-return interval influ-
enced changes in functional type biomass by
2100: Heavy grazing significantly reduced peren-
nial C
3
grass biomass, while frequent fire caused
a large reduction in big sagebrush biomass. Our
simulation results indicate that fire-return inter-
val (FRI) has large implications for functional
type biomass under climatic change, especially
for fire-intolerant species, such as big sagebrush.
These results demonstrate that frequent fire can
shift these shrub-dominated communities to
grass-dominated states. STEPWAT2 can be uti-
lized to understand how specific management
strategies (reduced grazing intensity or pre-
scribed fire) will impact dryland plant communi-
ties and guide management action now and in
the future.
ACKNOWLEDGMENTS
Funding for this work came from the U.S. Geologi-
cal Survey, North Central Climate Science Center
(Grant G12AC 20504), the U.S. Fish and Wildlife Ser-
vice (F13AC00865 and Interagency Agreement
#4500054042 to USGS), Yale University, and the
University of Wyoming. We would like to thank sev-
eral ecologists and computer programmers who have
contributed to the development of STEPWAT2: Debra
Peters, Ruihua Liu, Donovan Miller, Ryan Murphy,
Eric Murphy, Ashish Tiwari, Brenden Bernal, Karan
Sodhi, and Caitlin Andrews. We would also like to
acknowledge Victoria Pennington, Lukas Lindquist,
Lexi Smith, Ashley Wildeman, Jonathan Paklaian,
Margarita Reza, Vincent Irizarry, and Kim Hill, who
collected field data that was used in the Model Valida-
tion section. Any use of trade, product, or firm names
is for descriptive purposes only and does not imply
endorsement by the U.S. Government.
LITERATURE CITED
Adler, P. B., D. G. Milchunas, W. K. Lauenroth, O. E.
Sala, and I. C. Burke. 2004. Functional traits of
graminoids in semi-arid steppes: a test of graz-
ing histories. Journal of Applied Ecology 41:653–
663.
Adler, P. B., D. G. Milchunas, O. E. Sala, I. C. Burke,
and W. K. Lauenroth. 2005. Plant traits and ecosys-
tem grazing effects: comparison of U.S. sagebrush
steppe and Patagonian steppe. Ecological Applica-
tions 15:774–792.
Bai, Y., J. Wu, C. M. Clark, Q. Pan, L. Zhang, S. Chen,
Q. Wang, and X. Han. 2012. Grazing alters ecosys-
tem functioning and C:N: P stoichiometry of grass-
lands along a regional precipitation gradient.
Journal of Applied Ecology 49:1204–1215.
Baker, D. J., A. J. Hartley, J. W. Pearce-Higgins, R. G.
Jones, and S. G. Willis. 2017. Neglected issues in
❖www.esajournals.org 19 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

using weather and climate information in ecology
and biogeography. Diversity and Distributions
23:329–340.
Balch, J. K., B. A. Bradley, C. M. D’Antonio, and J.
G
omez-Dans. 2013. Introduced annual grass
increases regional fire activity across the arid west-
ern USA (1980–2009). Global Change Biology
19:173–183.
Bates, J. D., T. Svejcar, R. F. Miller, and R. A. Angell.
2006. The effects of precipitation timing on sage-
brush steppe vegetation. Journal of Arid Environ-
ments 64:670–697.
Botkin, D. B., J. F. Janak, and J. R. Wallis. 1972. Some
ecological consequences of a computer model of
forest growth. Journal of Ecology 60:849–872.
Bouyoucos, G. J. 1962. Hydrometer method improved
for making particle size analyses of soils 1. Agron-
omy Journal 54:464–465.
Bradford, J. B., and W. K. Lauenroth. 2006. Controls
over invasion of Bromus tectorum: the importance
of climate, soil, disturbance and seed availability.
Journal of Vegetation Science 17:693–704.
Bradford, J., D. Schlaepfer, and W. Lauenroth. 2014.
Ecohydrology of adjacent sagebrush and lodgepole
pine ecosystems: the consequences of climate
change and disturbance. Ecosystems 17:590–605.
Brummer, T. J., K. T. Taylor, J. Rotella, B. D. Maxwell,
L. J. Rew, and M. Lavin. 2016. Drivers of Bromus
tectorum abundance in the western North
American sagebrush steppe. Ecosystems 19:986–
1000.
Butterfield, B. J., and S. M. Munson. 2016. Temperature
is better than precipitation as a predictor of plant
community assembly across a dryland region.
Journal of Vegetation Science 27:938–947.
Campbell, G. S., and G. A. Harris. 1977. Water
relations and water use patterns for Artemisia tri-
dentata Nutt. in wet and dry years. Ecology 58:652–
659.
Chen, I.-C., J. K. Hill, R. Ohlem€
uller, D. B. Roy, and C.
D. Thomas. 2011. Rapid range shifts of species
associated with high levels of climate warming.
Science 333:1024.
Cleary, M. B., E. Pendall, and B. E. Ewers. 2008. Testing
sagebrush allometric relationships across three fire
chronosequences in Wyoming, USA. Journal of
Arid Environments 72:285–301.
Coffin, D. P., and W. K. Lauenroth. 1990. A gap
dynamics simulation model of succession in a
semiarid grassland. Ecological Modelling 49:229–
266.
Coffin, D. P., W. K. Lauenroth, and I. C. Burke. 1993.
Spatial dynamics in recovery of shortgrass steppe
ecosystems. Lectures on Mathematics in the Life
Sciences 23:75–108.
Cook, C. W. 1942. Insects and weather as they influ-
ence growth of cactus on the central Great Plains.
Ecology 23:209–214.
Cook, B. I., T. R. Ault, and J. E. Smerdon. 2015.
Unprecedented 21st century drought risk in the
American Southwest and Central Plains. Science
Advances 1:e1400082.
Daubenmire, R. 1959. A canopy-coverage method of
vegetational analysis. Northwest Science 33:43–64.
Davies, K. W., C. S. Boyd, J. L. Beck, J. D. Bates, T. J.
Svejcar, and M. A. Gregg. 2011. Saving the sage-
brush sea: an ecosystem conservation plan for big
sagebrush plant communities. Biological Conserva-
tion 144:2573–2584.
Davies, K. W., M. Vavra, B. W. Schultz, and N. R. Rim-
bey. 2014. Implications of longer term grazing rest
in the sagebrush steppe. Journal of Rangeland
Applications 1:14–34.
DeAngelis, D. L., and V. Grimm. 2014. Individual-
based models in ecology after four decades.
F1000Prime Reports 6:39.
Doherty, K. E., J. S. Evans, P. S. Coates, L. M. Juliusson,
and B. C. Fedy. 2016. Importance of regional varia-
tion in conservation planning: a rangewide exam-
ple of the Greater Sage-Grouse. Ecosphere 7:
e01462.
Ellsworth, L. M., and J. B. Kauffman. 2017. Plant com-
munity response to prescribed fire varies by pre-
fire condition and season of burn in mountain big
sagebrush ecosystems. Journal of Arid Environ-
ments 144:74–80.
Epstein, H. E., R. A. Gill, J. M. Paruelo, W. K. Lauen-
roth, G. J. Jia, and I. C. Burke. 2002. The relative
abundance of three plant functional types in tem-
perate grasslands and shrublands of North and
South America: effects of projected climate change.
Journal of Biogeography 29:875–888.
Fensham, R. J., J. Fraser, H. J. MacDermott, and J. Firn.
2015. Dominant tree species are at risk from exag-
gerated drought under climate change. Global
Change Biology 21:3777–3785.
Ferguson, C. W. 1964. Annual rings in big sagebrush:
Artemisia tridentata. University of Arizona Press,
Tucson, Arizona, USA.
Gornish, E. S., and C. M. Prather. 2014. Foliar
functional traits that predict plant biomass
response to warming. Journal of Vegetation Science
25:919–927.
Gremer, J. R., J. B. Bradford, S. M. Munson, and M. C.
Duniway. 2015. Desert grassland responses to cli-
mate and soil moisture suggest divergent vulnera-
bilities across the southwestern United States.
Global Change Biology 21:4049–4062.
Grimm, V., D. Ayll
on, and S. F. Railsback. 2017. Next-
generation individual-based models integrate
❖www.esajournals.org 20 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

biodiversity and ecosystems: Yes we can, and yes
we must. Ecosystems 20:229–236.
Hamlet, A. F., E. P. Salath
e, and P. Carrasco. 2010. Sta-
tistical downscaling techniques for global climate
model simulations of temperature and precipita-
tion with application to water resources planning
studies. Chapter 4 in Final Report for the Columbia
Basin Climate Change Scenarios Project, Climate
Impacts Group, Center for Science in the Earth Sys-
tem, Joint Institute for the Study of the Atmosphere
and Ocean, University of Washington, Seattle,
USA.
Huston, M., D. DeAngelis, and W. Post. 1988. New
computer models unify ecological theory. BioS-
cience 38:682–691.
Jones, R. O., J. C. Chambers, D. I. Board, D. W. John-
son, and R. R. Blank. 2015. The role of resource lim-
itation in restoration of sagebrush ecosystems
dominated by cheatgrass (Bromus tectorum). Eco-
sphere 6:1–21.
Knick, S. T., D. S. Dobkin, J. T. Rotenberry, M. A.
Schroeder, W. M. Vander Haegen, and C. van
Riper. 2003. Teetering on the edge or too late? Con-
servation and research issues for avifauna of sage-
brush habitats. Condor 105:611–634.
Knutti, R., D. Masson, and A. Gettelman. 2013. Cli-
mate model genealogy: generation CMIP5 and
how we got there. Geophysical Research Letters
40:1194–1199.
Lauenroth, W. K., and P. B. Adler. 2008. Demography
of perennial grassland plants: survival, life expec-
tancy and life span. Journal of Ecology 96:1023–
1032.
Lauenroth, W. K., R. L. Dougherty, and J. S. Singh.
2009. Precipitation event size controls on long-term
abundance of Opuntia polyacantha (plains prickly-
pear) in Great Plains grasslands. Great Plains
Research 19:55–64.
Lauenroth, W. K., D. R. Schlaepfer, and J. B. Bradford.
2014. Ecohydrology of dry regions: storage versus
pulse soil water dynamics. Ecosystems 17:1469–
1479.
Lin, D., J. Xia, and S. Wan. 2010. Climate warming and
biomass accumulation of terrestrial plants: a meta-
analysis. New Phytologist 188:187–198.
Loik, M. E., D. D. Breshears, W. K. Lauenroth, and J.
Belnap. 2004. A multi-scale perspective of water
pulses in dryland ecosystems: climatology and eco-
hydrology of the western USA. Oecologia 141:269–
281.
Lucero, M. E., J. R. Barrow, P. Osuna, I. Reyes, and S.
E. Duke. 2008. Enhancing native grass productivity
by cocultivating with endophyte-laden calli. Ran-
geland Ecology & Management 61:124–130.
Manier, D. J., et al. 2013. Summary of science, activi-
ties, programs, and policies that influence the ran-
gewide conservation of greater sage-grouse
(Centrocercus urophasianus). US Geological Survey,
Reston, Virginia, USA.
Martyn, T. E., J. B. Bradford, D. R. Schlaepfer, I. C.
Burke, and W. K. Lauenroth. 2016. Seed bank and
big sagebrush plant community composition in a
range margin for big sagebrush. Ecosphere 7:
e01453.
Maurer, E. P., L. Brekke, T. Pruitt, and P. B. Duffy. 2007.
Fine-resolution climate projections enhance regio-
nal climate change impact studies. Eos, Transac-
tions American Geophysical Union 88:504–504.
McKenzie, D., and J. S. Littell. 2017. Climate change
and the eco-hydrology of fire: Will area burned
increase in a warming western USA? Ecological
Applications 27:26–36.
Meddens, A. J. H., J. A. Hicke, A. K. Macalady, P. C.
Buotte, T. R. Cowles, and C. D. Allen. 2015. Pat-
terns and causes of observed pi~
non pine mortality
in the southwestern United States. New Phytolo-
gist 206:91–97.
Milchunas, D. G., and W. K. Lauenroth. 1993. Quanti-
tative effects of grazing on vegetation and soils
over a global range of environments. Ecological
Monographs 63:328–366.
Milchunas, D. G., O. E. Sala, and W. K. Lauenroth.
1988. A generalized model of the effects of grazing
by large herbivores on grassland community struc-
ture. American Naturalist 132:87–106.
Miller, R. F., and L. L. Eddleman. 2001. Spatial and
temporal changes of sage grouse habitat in the
sagebrush biome. Technical Bulletin 151, Oregon
State University Agricultural Experiment Station,
Corvallis, Oregon, USA.
Morin, X., L. Fahse, H. Jactel, M. Scherer-Lorenzen, R.
Garc
ıa-Vald
es, and H. Bugmann. 2018. Long-term
response of forest productivity to climate change is
mostly driven by change in tree species composi-
tion. Scientific Reports 8:5627.
Neilson, R. P., J. M. Lenihan, D. Bachelet, and R. J. Dra-
pek. 2005. Climate change implications for sage-
brush ecosystems. Transactions of the 70yh North
American Wildlife and Natural Resources Confer-
ence 70:145–159.
Palmquist, K. A., D. R. Schlaepfer, J. B. Bradford,
and W. K. Lauenroth. 2016a. Mid-latitude shrub
steppe plant communities: climate change conse-
quences for soil water resources. Ecology 97:
2342–2354.
Palmquist, K. A., D. R. Schlaepfer, J. B. Bradford, and
W. K. Lauenroth. 2016b. Spatial and ecological
variation in dryland ecohydrological responses to
❖www.esajournals.org 21 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

climate change: implications for management.
Ecosphere 7:e01590.
Palmquist, K., D. Schlaepfer, T. Martyn, J. Bradford, and
W. Lauenroth. 2018. DrylandEcology/STEPWAT2:
STEPWAT2 Model Description (Palmquist et al.
2018 Ecosphere) (Version v1.2.0). Zenodo. http://
doi.org/10.5281/zenodo.1306924
Paruelo, J. M., and W. K. Lauenroth. 1996. Relative
abundance of plant functional types in grasslands
and shrublands of North America. Ecological
Applications 6:1212–1224.
Pearl, R. 1928. The rate of living. A. E. Kopf, New York,
New York, USA.
Pennington, V. E., K. A. Palmquist, J. B. Bradford, and
W. K. Lauenroth. 2017. Climate and soil texture
influence patterns of forb species richness and
composition in big sagebrush plant communities
across their spatial extent in the western U.S. Plant
Ecology 218:957–970.
Perfors, T., J. Harte, and S. E. Alter. 2003. Enhanced
growth of sagebrush (Artemisia tridentata)in
response to manipulated ecosystem warming. Glo-
bal Change Biology 9:736–742.
Peters, D. P. C. 2000. Climatic variation and simulated
patterns in seedling establishment of two dominant
grasses at a semi-arid-arid grassland ecotone. Jour-
nal of Vegetation Science 11:493–504.
Peters, D. P. 2002. Plant species dominance at a grass-
land–shrubland ecotone: an individual-based gap
dynamics model of herbaceous and woody species.
Ecological Modelling 152:5–32.
Renwick, K. M., C. Caroline, A. R. Kleinhesselink, D.
Schlaepfer, B. A. Bradley, C. L. Aldridge, B. Poulter,
and P. B. Adler. 2018. Multi-model comparison
highlights consistency in predicted effect of warm-
ing on a semi-arid shrub. Global Change Biology
24:424–438.
Richardson, C. A., and D. A. Wright. 1984. A model
for generating daily weather variables. Pages 1–86.
United States Department of Agriculture, Agricul-
ture Research Service, Washington, D.C., USA.
Rickard, W. H. 1985. Biomass and shoot production in
an undisturbed sagebrush-bunchgrass community.
Northwest Science 59:126–133.
Rowland, M. M., M. J. Wisdom, L. H. Suring, and C.
W. Meinke. 2006. Greater sage-grouse as an
umbrella species for sagebrush-associated verte-
brates. Biological Conservation 129:323–335.
Rupp, D. E., J. T. Abatzoglou, K. C. Hegewisch, and P.
W. Mote. 2013. Evaluation of CMIP5 20th century
climate simulations for the Pacific Northwest USA.
Journal of Geophysical Research: Atmospheres
118:2013JD020085.
Sage, R. F. 2004. The evolution of C4 photosynthesis.
New Phytologist 161:341–370.
Sala, O. E., W. K. Lauenroth, and W. J. Parton. 1992.
Long-term soil water dynamics in the shortgrass
steppe. Ecology 73:1175–1181.
Schlaepfer, D. R., W. K. Lauenroth, and J. B. Bradford.
2012a. Ecohydrological niche of sagebrush ecosys-
tems. Ecohydrology 5:453–466.
Schlaepfer, D. R., W. K. Lauenroth, and J. B. Bradford.
2012b. Effects of ecohydrological variables on current
and future ranges, local suitability patterns, and model
accuracy in big sagebrush. Ecography 35:374–384.
Schlaepfer, D. R., W. K. Lauenroth, and J. B. Bradford.
2014. Modeling regeneration responses of big sage-
brush (Artemisia tridentata) to abiotic conditions.
Ecological Modelling 286:66–77.
Schlaepfer, D. R., et al. 2017. Climate change reduces
extent of temperate drylands and intensifies drought
in deep soils. Nature Communications 8:14196.
Schultz, L. M. 2006. Artemisia. Pages 503–534 in Flora
of North America Editorial Committee, editor.
Flora of North America North of Mexico. Oxford
University Press, Oxford, UK.
Seager, R., et al. 2007. Model projections of an immi-
nent transition to a more arid climate in southwest-
ern North America. Science 316:1181–1184.
Seidl, R., W. Rammer, R. M. Scheller, and T. A. Spies.
2012. An individual-based process model to simu-
late landscape-scale forest ecosystem dynamics.
Ecological Modelling 231:87–100.
Seidl, R., M.-J. Schelhaas, W. Rammer, and P. J. Verk-
erk. 2014. Increasing forest disturbances in Europe
and their impact on carbon storage. Nature Cli-
mate Change 4:806–810.
Shugart, H. H. 1984. A theory of forest dynamics: the
ecological implications of forest succession models.
Springer-Verlag, New York, New York, USA.
Smith, T., and M. Huston. 1990. A theory of the spatial
and temporal dynamics of plant communities.
Pages 49–69 in G. Grabherr, L. Mucina, M. B. Dale,
and C. J. F. Ter Braak, editors. Progress in theoreti-
cal vegetation science. Springer, Dordrecht, The
Netherlands.
Sturges, D. L. 1977. Soil water withdrawal and root
characteristics of big sagebrush. American Midland
Naturalist 98:257–274.
Thom, D., W. Rammer, and R. Seidl. 2017. Distur-
bances catalyze the adaptation of forest ecosystems
to changing climate conditions. Global Change
Biology 23:269–282.
Tohver, I. M., A. F. Hamlet, and S.-Y. Lee. 2014.
Impacts of 21st-century climate change on hydro-
logic extremes in the Pacific Northwest Region of
North America. JAWRA Journal of the American
Water Resources Association 50:1461–1476.
Trenberth, K. E., A. Dai, G. van der Schrier, P. D. Jones,
J. Barichivich, K. R. Briffa, and J. Sheffield. 2014.
❖www.esajournals.org 22 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.

Global warming and changes in drought. Nature
Climate Change 4:17–22.
Tylianakis, J. M., R. K. Didham, J. Bascompte, and D.
A. Wardle. 2008. Global change and species inter-
actions in terrestrial ecosystems. Ecology Letters
11:1351–1363.
Vora, R. S. 1988. Predicting biomass of five shrub spe-
cies in northeastern California. Journal of Range
Management 41:63–65.
Westerling, A. L., H. G. Hidalgo, D. R. Cayan, and T.
W. Swetnam. 2006. Warming and earlier spring
increase western U.S. forest wildfire activity.
Science 313:940–943.
Wu, W., C. Lan, M. Lo, J. T. Reager, and J. S. Famiglietti.
2015. Increases in the annual range of soil water
storage at northern middle and high latitudes
under global warming. Geophysical Research
Letters 42:3903–3910.
SUPPORTING INFORMATION
Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.
2394/full
❖www.esajournals.org 23 August 2018 ❖Volume 9(8) ❖Article e02394
EMERGING TECHNOLOGIES PALMQUIST ET AL.