Copyright © 2014 by the author(s). Published here under license by the Resilience Alliance.
Miller, B. W., and J. T. Morisette. 2014. Integrating research tools to support the management of social-ecological systems under
climate change. Ecology and Society 19(3): 41. http://dx.doi.org/10.5751/ES-06813-190341
Synthesis
Integrating research tools to support the management of social-ecological
systems under climate change
Brian W. Miller
1
and Jeffrey T. Morisette
1,2
ABSTRACT. Developing resource management strategies in the face of climate change is complicated by the considerable uncertainty
associated with projections of climate and its impacts and by the complex interactions between social and ecological variables. The
broad, interconnected nature of this challenge has resulted in calls for analytical frameworks that integrate research tools and can
support natural resource management decision making in the face of uncertainty and complex interactions. We respond to this call by
first reviewing three methods that have proven useful for climate change research, but whose application and development have been
largely isolated: species distribution modeling, scenario planning, and simulation modeling. Species distribution models provide data-
driven estimates of the future distributions of species of interest, but they face several limitations and their output alone is not sufficient
to guide complex decisions for how best to manage resources given social and economic considerations along with dynamic and uncertain
future conditions. Researchers and managers are increasingly exploring potential futures of social-ecological systems through scenario
planning, but this process often lacks quantitative response modeling and validation procedures. Simulation models are well placed to
provide added rigor to scenario planning because of their ability to reproduce complex system dynamics, but the scenarios and
management options explored in simulations are often not developed by stakeholders, and there is not a clear consensus on how to
include climate model outputs. We see these strengths and weaknesses as complementarities and offer an analytical framework for
integrating these three tools. We then describe the ways in which this framework can help shift climate change research from useful to
usable.
Key Words: agent-based modeling; complex-adaptive systems; natural resource management; scenario planning; simulations; species
distribution modeling; state-and-transition modeling
INTRODUCTION
Developing strategies for adapting to climate change not only
requires regionally and locally relevant climate projections, but
also information on the expected impacts of climate change on a
range of key social and ecological variables. Crafting management
plans is especially challenging because social-ecological systems
are complex and adaptive, exhibiting heterogeneity, nonlinearity,
interactions across scales, and sensitivity to initial conditions
(Malanson et al. 2006, Portugali 2006). These systems also
produce aggregate patterns, which emerge from simple
interactions between individual components (Holland 1995,
Manson 2001, Abel and Stepp 2003). Emergent properties,
feedbacks, unanticipated consequences, and thresholds all pose
challenges to measuring and predicting system behavior and its
responses to potential management actions.
The broad, interconnected nature of these challenges demands
that a variety of assessment and planning tools be brought to bear
on making decisions for how to manage social-ecological systems
under climate change. We followed the suggestion that research
should focus on developing analytical frameworks for assessing
climate change impacts and their interactions with other stressors
(Staudt et al. 2013). We first reviewed three tools, which have been
promoted as useful for guiding natural resource management
decisions under climate change: species distribution modeling,
scenario planning, and simulation modeling. Although each has
proven useful, their use and development has been fairly isolated
even though there are potentially strong synergies when
considered collectively. As a result, we propose a novel analytic
framework for integrating these tools and drawing on their
complementary strengths to inform natural resource
management under climate change. We then discuss how this
workflow contributes to the goal of producing ‘actionable’
climate science (ACCCNS 2013, Asrar et al. 2013) and to the
shifting climate information from ‘useful to usable’ by
emphasizing interaction with stakeholders, providing added value
to climate data, and offering opportunities for information
customization (Lemos et al. 2012).
SPECIES DISTRIBUTION MODELING
A fundamental aspect of ecology, conservation biology, and
resource management is understanding the environmental
conditions and geographic areas that are suitable for a given
species to inhabit. Species distribution modeling (SDM) is
commonly used to determine such areas. In these models,
environmental variables are used as the explanatory
(independent) variables to estimate the potential habitat
(dependent variable) for a given species (Guisan and
Zimmermann 2000, Guisan and Thuiller 2005). These models are
built on field data consisting of known presence and possible
absence locations of a given species as well as the values of
environmental or climatic covariates thought to define the species’
habitat suitability at these locations. These covariate data are
generally extracted from remotely sensed imagery, interpolated/
gridded historical climate data, or downscaled climate model
output.
It has been argued that climate is often the most basic determinant
of a species fundamental niche in that it limits the species’ range
at the broadest spatial scale (Araújo and Peterson 2012). Climate
1
Department of the Interior North Central Climate Science Center, Natural Resource Ecology Laboratory, Colorado State University,
2
U.S.
Geological Survey
Ecology and Society 19(3): 41
http://www.ecologyandsociety.org/vol19/iss3/art41/
change and SDMs can be connected through the exploration of
climate variables, or climate derivatives such as BIOCLIM
summaries (Nix 1986), as the explanatory variable. The modeling
thus provides a data-driven method to explore the relationship
between species occurrence locations and climate based on
contemporary, or previous (i.e., historic) information. If the
model diagnostics indicate a decent fit and strong predictive
capacity, then it is reasonable to apply future climate projections
to the relationships from the model to derive corresponding
projections of potential habitat. There are significant
assumptions and uncertainties in such coupled modeling and
projections analysis (Heikkinen et al. 2006, Thuiller 2007),
including the uncertainty in both the contemporary and future
predictor variables and the models’ general inability to fully
capture species interactions, migration, dispersal, time lags, or
adaptation mechanisms (Sinclair et al. 2010). Given the numerous
caveats for SDM, as well as their ability to incorporate and
associate significant climate and species location information, it
seems prudent to exploit these modeling techniques to the furthest
extent possible as a means to develop, as opposed to test,
hypotheses regarding how climate can influence current and
future species distributions.
So, for the practical application of SDMs to natural resource
management decisions under a changing climate, we are left with
powerful statistical data-driven models with considerable
uncertainties and a long list of caveats. There are numerous papers
(only a small sample are referenced above) that describe the
limitations and proper use of these models (for a summary, see
Araújo and Peterson 2012). At this point, it seems that some of
these limitations are inherent to the underlying concept of
statistically, data-driven SDMs. That is, even significant advances
in statistical, machine learning, or any other techniques will not
allow models based on current observations to predict the future
state of complex coupled social-ecological-climate conditions.
The appropriate application of these models is largely dependent
on placing SDMs within a clear conceptual framework, and
thereby identifying their purpose and assumptions (Araújo and
Peterson 2012). In fact, SDMs are likely to be more compelling
as “part of a methodological toolkit” that addresses the
interactions with other dynamics such as climate and land use
change (Franklin 2013:1220). In other words, “combining SDM
with other data and approaches, such as spatially explicit models
of population and community dynamics, may improve
forecasting and impact assessment of environmental change at
larger spatial scales” (Elith and Franklin 2013:704). Given this
conclusion, we propose the integration of SDM with scenario
planning and simulation modeling. We go on to provide
background information on these topics and propose an
analytical framework for such integration.
The proposed workflow builds on previous suggestions for using
multiple scenarios of climate change to deal with climate model
uncertainty when estimating future species distributions (Araújo
and New 2007, Wang et al. 2012, Talbert et al. 2013). Other studies
have used SDMs for conservation or restoration prioritization
under climate uncertainty (Kujala et al. 2013, Veloz et al. 2013)
and to assess alternatives for allocating conservation resources
under different emissions scenarios (Carvalho et al. 2011).
Although these approaches have made considerable progress
toward addressing the uncertainty associated with climate change
projections, SDM research typically does not develop future
scenarios in conjunction with a variety of stakeholders to account
for other dynamics, which might interact with climate changes.
In the following sections, we detail how coupling SDMs with the
participatory development of place-based, social-ecological-
climate scenarios, and simulation models could help overcome
some of the remaining challenges facing SDM.
SCENARIO PLANNING
Projected changes in local climates and environments are
associated with substantial uncertainty. Moreover, climate change
is playing out over large spatial and temporal scales and also
interacts with diverse social and biophysical factors, which are
not included in SDMs, making it largely uncontrollable in the
context of local or regional resource management. Scenario
planning is an effective tool for creating management plans under
such uncertainty and largely uncontrollable external drivers
(Peterson et al. 2003a).
Scenario planning is the process of developing a set of plausible
futures for a given system. In this instance, scenarios are not
necessarily equivalent to climate scenarios, which capture a set of
modeled future situations with different assumptions (i.e.,
emission scenarios that represent a range of deviations in
precipitation or temperature), but are structured accounts of
plausible futures, which are developed to better understand
current trajectories, drivers of change, and response options
(Peterson et al. 2003a). Each scenario is based on a set of
assumptions about important relationships and driving forces in
a given area or location; different scenarios reflect alternative
economic, environmental, social, and technological conditions
(Walker et al. 2003). As such, scenario planning can incorporate
a variety of data types and perspectives (Peterson et al. 2003a).
It is especially useful for incorporating factors that are difficult
to model or include in quantitative approaches. For instance,
values and agency, which are not readily represented in SDMs,
are at the center of scenario development (Kass et al. 2011).
Scenario planning has been applied to a variety of sectors,
including energy (Wack 1985), advertising (Schoemaker 1995),
governance (Kahane 1998), and natural resource management
(Weeks et al. 2011; for a recent review of scenario development
for environmental decision making, see Mahmoud et al. 2009). It
can also be applied across multiple locations and geographic
scales (Rotmans et al. 2000, Foran et al. 2013). Scenario planning
has proven to be a useful tool for creating strategies under
uncertainty, and it can have the additional benefits of engaging
communities in planning (Tompkins et al. 2008) and natural
resource management (Dowsley et al. 2013), and it can improve
organizational resilience and the adoption of innovations (Cobb
and Thompson 2012).
Rose and Star (2013) described the process by which resource
managers can develop and apply scenarios to the management of
their own systems. The process, adapted from the Global Business
Network, entails five phases and associated tasks:
1. Orientation - determine the central challenge, question,
goals, and team
2. Exploration - identify key forces, variables, trends, and
uncertainties related to the central question
Ecology and Society 19(3): 41
http://www.ecologyandsociety.org/vol19/iss3/art41/
3. Synthesis - generate three to five plausible, thought-
provoking, and divergent scenarios
4. Application - create actions and strategies based on
scenarios
5. Monitoring - identify indicators of change for monitoring
The scenario planning process has been used to guide natural-
resource management decisions that are faced with substantial
uncertainty. For instance, Peterson et al. (2003b) developed three
divergent social-ecological scenarios, which described potential
futures for the Northern Highlands Lake District of Wisconsin.
These scenarios captured key uncertainties in future development
and the use of ecological services in the area, and served as a
starting point for a participatory discussion of how to plan under
these uncertainties. The scenario-development process itself
brought together a diverse group of people and provided them
with a better understanding of ongoing changes and new tools
for environmental management (Peterson 2007).
Through this process, which generally consists of a series of
workshops, scenario planning helps participants overcome biases,
identify uncertainties, and distill a variety of information into a
manageable number of possible states (Schoemaker 1995). In this
way, it can aid in the identification of otherwise unexpected
futures, but it is bounded by our capacity to integrate information.
In other words, the development of scenarios can still be
influenced by participant biases (Schoemaker 1995).
SIMULATION MODELING
Although scenario planning is an effective tool for describing
possible system future states in the face of substantial uncertainty,
it is less well suited for predicting complex dynamics (e.g.,
thresholds, interactions, emergent patterns), the secondary effects
of natural resource management, or climate change mitigation
plans. Social-ecological systems are complex, adaptive, and are
characterized by diverse and independent components,
interaction between components, nonlinearities, and selection
processes, which all produce variation and novelty (Levin 1998).
As a result, it is unreasonable to assume that we are able to predict
overall system behavior based on understandings of individual
system components (Walker and Salt 2006). Moreover, system
thresholds can only be detected by crossing them, a relatively rare
and unexpected occurrence, which is potentially unpredictable,
extreme, and an irreversible form of experimentation on large,
complex systems (Carpenter 2003, cited in Carpenter et al. 2005).
However, the identification of thresholds is necessary for
preventing sudden undesirable changes.
Computer-based simulation models have been advocated as ways
of reproducing complex dynamics and implementing ‘what if’
scenarios (Axelrod 1997, Erlien et al. 2006) and are also promising
means of identifying thresholds. We reviewed two simulation
tools, which have been increasing in use and diversity, and that
we see as key to addressing the challenge of managing social-
ecological systems under climate change: agent-based models and
state-and-transition simulation models. In addition, we provide
suggestions for how to combine these simulation tools with one
another and with SDM and scenario planning; although
simulation models can be effectively used to explore management
and climate scenarios, they are not appropriate for developing the
scenarios themselves (Peterson et al. 2003a).
State-and-transition models
State-and-transition models originated as conceptual models of
rangeland vegetation, which represented groups of vegetation
communities and the shifts between them (Westoby et al. 1989).
This was proposed as an alternative to the dominant equilibrium-
based view of rangelands, wherein succession proceeded steadily
toward a climax community. In state-and-transition models,
states can either be defined broadly as “...climate/soil/vegetation
domains that encompass a large amount of variation in species
composition,” or more narrowly, “as serial stages or phases of
vegetation development” (Stringham et al. 2001:2). Generally,
states are suites of vegetation communities, which have distinct
functional groups and ecosystem processes, as well as associated
vegetation structure and management requirements (Bestelmeyer
et al. 2003). Transitions can result from natural events,
management interventions, or a combination of both (Stringham
et al. 2001), and can be reversible or unidirectional. Conceptual
state-and-transition models are typically represented using box
and arrow diagrams, in which boxes or nested boxes represent
vegetation states, and arrows represent the transitions within and
between them.
More recently, there has been a movement toward quantitative
computer-based state-and-transition models. These state-and-
transition simulation models (STSMs) are based upon conceptual
models of vegetation types and their shifts between different
possible states. In STSMs, transitions can be deterministic or
probabilistic. Deterministic transitions are those that take place
after a given period of time, such as aging and growth.
Probabilistic transitions occur with a given probability and
include processes such as fire, invasion, and restoration.
Spatially explicit STSMs can provide considerable benefit to
quantitative simulations. These are analogous to joint cellular
automata-Markov models (Parker et al. 2003), hybrid Markov-
cellular automaton models (Li and Reynolds 1997), and spatio-
temporal Markov chains (Balzter et al. 1998). As described by
Bestelmeyer et al. (2011), spatially explicit STSMs are useful
because spatial dynamics such as contagion, feedbacks between
patches, spatial patterns in historical legacies, and variation in
soils, topography, and climate can all affect the likelihood and
location of transitions. In terms of practical implications,
accounting for spatial patterns and processes can improve
inventory and monitoring by identifying early-warning indicators
and sites that are more susceptible to transitions (Bestelmeyer et
al. 2011).
Overall, STSMs have proven useful for comparing and evaluating
different resource management scenarios (Forbis et al. 2006,
Provencher et al. 2007, Frid and Wilmshurst 2009, Strand et al.
2009, Costanza et al. 2012). They can also incorporate climate
effects on vegetation (Keane et al. 2008, Strand et al. 2009,
Halofsky et al. 2013), but STSMs that include dynamic climate
inputs, which directly affect vegetation, have not yet been
developed. Finally, although STSMs can be used to engage
resource managers and experts (Forbis et al. 2006), there is still a
need to clarify and standardize STSM terminology, development,
and validation to improve participatory modeling (Knapp et al.
2011).
Ecology and Society 19(3): 41
http://www.ecologyandsociety.org/vol19/iss3/art41/
Agent-based models
Agent-based models (ABMs) are another type of computer-based
simulation that have seen increased application in resource
management research. Agent-based models are composed of
three interacting components: an environment, agents, and rules.
The environment can range from a simple grid to a detailed
landscape, which reflects real-world patterns and processes.
Agents are autonomous units, such as animals, livestock, or
people, which interact with one another and the environment
through a set of user-defined rules. One of the fundamental
benefits of ABMs is that they can incorporate individual-level
variation, which more accurately reflects heterogeneity in human-
environment interactions (DeAngelis and Mooij 2005). Agent-
based models can not only include detail and variation in
resource-use decision making, but can also incorporate
uncertainty in particular variables or processes (e.g., climate and
population projections).
Agent-based models have been promoted as tools for studying
properties of complex-adaptive systems, such as feedbacks and
emergence (Berry et al. 2002; see the special issue in PNAS).
Recent studies suggest that ABMs are also useful for assessing
resilience (Schlüter and Pahl-Wostl 2007, Schlüter et al. 2009,
Schouten et al. 2013). These models range from basic and
abstracted simulations, which can be used to explore the simple
rules that underpin complex patterns (Schelling 1971, Epstein and
Axtell 1996), to complicated and highly parameterized
simulations of particular places and populations (Walsh et al.
2013). Agent-based models have been applied in many fields and
are particularly useful for the exploration of policy and
management options (Boone et al. 2002, 2006, Walsh et al. 2006,
2013, Anselme et al. 2010, Miller et al. 2010).
Like STSMs, there have been some initial applications of ABMs
to climate change research. In terms of adaptation to climate
change, “adaptive capacity can be thought of as an emergent
property, one for which ABM is a suitable analytic tool” (Patt and
Siebenhüner 2005:317). Agent-based models have been used to
explore a range of topics, such as the relationships between climate
change and local institutions (Wang et al. 2013), adaptive decision
making (Janssen and de Vries 1998, Ziervogel et al. 2005,
Aurbacher et al. 2013), migration (Smith et al. 2008, Kniveton et
al. 2011, Hassani-Mahmooei and Parris 2012), and land use (Yan
et al. 2013, Zhang et al. 2013). However, there is not yet a clear
strategy for incorporating climate data into ABMs as a dynamic
input. Although agent-based modeling of climate change issues
is still in its infancy, it is a promising area of research, especially
because ABMs are useful for spatially investigating adaptive
capacity and possible system futures, including possible climate
futures (Patt and Siebenhüner 2005).
Simulation challenges and opportunities
Simulation models have been applied to a variety of systems and
questions, but they face important challenges related to model
specification and validation. First, modelers must balance model
abstraction and specification. Highly parameterized models can
be useful for evaluating policy recommendations for particular
locations and populations, whereas abstract models can yield
more generalized information about system dynamics that
underlie complex patterns. Stylized models seek to strike a
compromise in complicatedness and are useful for exploring
scenarios, refining conceptualizations of complex systems, and
uncovering data gaps (Miller et al. 2010).
Validation is another major challenge across computer-based
simulations (Cooley and Solano 2011). Highly specified models
can be validated by comparing model output to aggregate
statistics, and individual agent behavior to observed behaviors in
a comparable population (Bruch and Atwell 2013; for a summary
of empirical validation approaches, see Berk 2008). On the other
hand, abstract models can be compared to observed patterns
(Grimm et al. 2005) or to hypothesized empirical relationships,
which are based on observed mechanisms (Bruch and Atwell
2013). The time period over which simulation models are run is
also relevant to validation procedures. Simulation models
oftentimes initialize at present and move forward in time to
simulate future conditions and dynamics. It is more appropriate
to run the model over a historical time period, compare model
output to historical and current observations, and then project
the model forward into the future. However, model validation can
be limited by data availability. In the absence of data for
validation, an alternative strategy is to compare model outputs
to information from the literature (Evers et al. 2013) or to more
widely applied and tested models (Yan et al. 2013).
Despite the shared challenges of model specification and
validation, integrating ABMs and STSMs is a plausible and
potentially valuable prospect. State-and-transition simulation
models are adept at representing spatially explicit ecological
dynamics and incorporating thresholds, which are key aspects of
complex-adaptive systems. Agent-based models are especially
useful for representing agent decision making, albeit in a
simplified form, and accounting for variability and uncertainty.
The common use of cellular automata (i.e., a grid of cells that
each exists in a defined state, which can change based on transition
rules and each cell’s neighborhood conditions; Parker et al. 2003)
to represent dynamic environments in ABMs and STSMs suggests
that it is possible to combine them through a common modeling
platform. Some researchers have started moving in this direction;
Millington et al. created an agent-based model (2008) and a state-
and transition model (2009) with the goal of combining these
approaches. Both ABMs and STSMs are being applied to climate
change issues, but the incorporation of climate data requires
further refinement. And although ABMs and STSMs have been
proven useful for experimentation and scenario testing, both
simulation methods could be improved through more transparent
and participatory scenario development.
Linking scenarios with simulations faces several challenges. First,
simulations can either be used as a basis for comparison with
scenarios that are developed independently by stakeholders, or
simulations can be codeveloped through participatory modeling.
In both cases, using simulations to explore future scenarios often
yields findings that contrast with expectations; it can be difficult
to determine if these surprises are caused by unanticipated
complex system dynamics or are artifacts of model specification
that do not reflect real-world processes. Second, participatory
modeling approaches can be problematic because of limitations
associated with time, cost, bias, and validation (Matthews et al.
2007). However, participatory modeling is a growing area of
research, which can take many forms (Voinov and Bousquet
2010), and is especially promising given the advent of more
Ecology and Society 19(3): 41
http://www.ecologyandsociety.org/vol19/iss3/art41/
Table 1. Summary of three tools for assisting climate change planning.
Key Citations Strengths Weaknesses Relation to Other Methods
Species
Distribution
Modeling
Guisan and Zimmermann
2000, Guisan and Thuiller
2005, Araújo and
Peterson 2012, Elith and
Franklin 2013, Franklin
2013
Developing empirical
relationships between
biophysical variables and
species ranges; spatially
explicit; statistically rigorous;
data driven
Simplifying assumptions
regarding species interactions
and behaviors; uncertainty in
predictor variables; accounting
for social dynamics
Estimate projected distributions of
key species or resources (identified
through scenario planning) under
climate change; create input data for
simulation models
Scenario
Planning
Peterson et al. 2003,
Mahmoud et al. 2009,
Cobb and Thompson
2012, Rose and Star 2013
Engaging stakeholders;
planning under uncertainty
and uncontrollability;
challenging conventional
thinking; assessing multiple
aspects of social-ecological
systems
Scenario validation;
accounting for complex
dynamics (thresholds,
emergence, feedbacks);
generating quantified
estimates of outcome variables
Specify focal challenge and question
to be addressed through quantitative
methods; identify key social and
ecological variables; create descriptive
scenarios to inform simulation
development
Simulation
Modeling
Epstein and Axtell 1996,
Axelrod 1997, Berry et al.
2002, Bestelmeyer et al.
2003, 2011, Parker et al.
2003
Capturing complex system
dynamics; exploring “what
if” scenarios; identifying
data gaps
Developing scenarios;
identifying empirical
relationships between species
and biophysical conditions
Validate scenarios (check for internal
consistency); identify unexpected
outcomes; test possible management
strategies
accessible software platforms (e.g., NetLogo, ST-Sim). A number
of studies have already demonstrated the feasibility and utility of
integrating various forms of participatory scenario development
with simulation modeling (Gurung et al. 2006, Bousquet et al.
2007, Smajgl 2010, Flaxman and Vargas-Moreno 2011, Vargas-
Moreno and Flaxman 2012, Smajgl and Ward 2013).
INTEGRATION
Currently there is a fairly long, and sometimes disconnected,
chain between climate modeling output, analysis of ecological
responses to climate forcings, and management options related
to those ecological implications. Species distribution models are
useful for providing hypotheses of the future distributions of
species and resources of interest, but this information alone is not
sufficient to guide decisions for how best to manage resources
under dynamic and uncertain future conditions, in which these
conditions will depend on much more than the predictor layers
included in the SDM. Researchers and managers are increasingly
collaborating through scenario planning to explore potential
futures. Although scenario-planning exercises are often
conducted in conjunction with a range of stakeholders in such a
way as to encourage broad, creative thinking, they often lack
rigorous, data-driven ecological response modeling. Scenarios
currently reflect more of a thought exercise, albeit a very useful
and productive one, than a repeatable analytical approach
combining climate science, ecological response modeling, and
management options. As a result, scenarios are rarely thoroughly
validated or tested for unexpected outcomes. Simulation models
can add quantitative rigor to the exploration of scenarios by
accounting for complex system dynamics. However, to date,
simulated management options may or may not have been
specified by the managers themselves, and there is a lack of clear
empirical relationships between climate model output and
uncertainty, and related ecosystem responses.
We see these strengths and weaknesses as opportunities for
integrating the three tools, particularly in the context of climate
change planning (Table 1). Our suggestion that these three tools
are complementary is consistent with the idea that ‘scenarios need
ecologists’ and ‘ecologists need scenarios’ (Bennett et al. 2003).
Researchers have described the benefits of scenario planning to
particular modeling approaches and the general utility of
quantitative approaches (Westhoek et al. 2006, Rinaudo et al.
2013), but have not demonstrated how simulation models can
contribute to the scenario planning process, and ultimately, to
resource management. So although there have been calls for
combining scenario planning with quantitative methods, there is
not a clear consensus on how to do so (Kemp-Benedict 2004). We
propose a more specific analytical framework for integrating
SDM, scenario planning, and simulation modeling in a way that
is mutually beneficial and could better serve resource managers
dealing with climate change (Fig. 1). It is important to note that
this is not a management framework (e.g., adaptive management,
conservation planning) or a conceptual framework for
understanding broader concepts (e.g., social-ecological systems,
vulnerability). The ultimate outcomes of applying the proposed
analytical framework, or workflow, would be a set of coupled and
repeatable qualitative and quantitative scenarios of expected
social-ecological responses to climate change, and spatially
explicit implications for implementing adaptation actions.
We recognize that climate and social science are terms that
encompass large and diverse bodies of research. We have
discussed them very little to limit the scope of the literature review,
but we describe their important role in our framework and in
climate change planning below. In fact, all of the boxes in Figure
1 are simplifications of complex methodologies; nested within
each box is a set of data collection and processing efforts. For
example, SDM is a multistep process, which draws on a variety
of input data, involves a series of pre- and post-processing steps,
and often includes a set of modeling options and outputs, rather
than a single model and output (Morisette et al. 2013).
Climate science provides key inputs for multiple methods within
this framework (Fig. 1, arrows A, B, C). These data inputs include
historical, hindcasted, and projected gridded climate data, as well
as summaries of projected climate futures. Gridded climate data