In the past several decades there has been a dramatic increase in the use of scientific, quantitative methods for informing landscape change and decision-making in the presence of deep uncertainty. This increase has occurred in both the public and private sectors. The predominant approach in such assessments has been characterized as a predict-then-act paradigm, which pairs models of rational decision-making with methods for treating uncertainty derived largely from the sciences and engineering (Raiffa 1968; Lempert et al. 2003). The preferred course of action in predict-then-act assessments is the one that performs ‘‘best’’ given some (typically small) set of assumptions about the likelihood of various futures and the landscape processes that will be sustained if these assumptions prove true. Such assessments are strongly tied to the validity of these assumptions.

A second paradigm is emerging that differs from predict-then-act in important ways. Rather than seeking strategies and policies that are optimal against some small set of scenarios for the future, this explore-then-test approach seeks near-term actions that are shown to perform well across a large ensemble of plausible future scenarios. These approaches offer the promise (but less so the proof) of policies and patterns that are sufficiently robust against future surprise that they can seize unexpected opportunities, adapt when things go wrong and provide new avenues in forging consensus regarding the facts and values that steer landscape change (Lempert et al. 2003; van Notten et al. 2005; Davis et al. 2007). Agent-based models are central tools in the explore-then-test paradigm.

Envision was created to conduct research about the nature and properties of coupled human and natural environmental systems in the context of climate change. The approach employed scenarios, data and evaluative models produced by past research (Harmon ref, Gregory et al ref, Bolte ref), and built on prior work in agent-based modeling (Ostrom 1998; Janssen and Jager 2000; Parker et al. 2003; Brown et al. 2005b; Grimm et al. 2005) and biocomplexity theory (O’Neill et al. 1986; Levin 1998; Jager et al. 2000; Holling 2001; Michener et al. 2001; Beisner et al. 2003).

Central to Envision, and conceived at the simplest level, are the three-way interactions of agents, who have decision making authority over parcels of land, the landscape which is changed as these decisions are made, and the policies that guide and constrain decisions (Bolte et al. 2006; Guzy et al. 2008). In Envision, agents are entities that make decisions about the management of particular portions of the landscape for which they have management authority, based on balancing a set of objectives reflecting their particular values, mandates and the policy sets in force on the parcels they manage. They do this within the scope of policy sets that are consistent with the assumptions and intentions of a chosen future scenario. These policies are operative on particular landscape elements over which they have decision-making control.

Fundamentally, agents are characterized by the values they express through their behaviors, behaviors that, in turn, alter land use/land cover. These values are correlated with demographic characteristics and, in part, guide the process agents use to select policies to implement; policies consistent with agents’ values are more likely to be selected. Policies in Envision provide a fundamental construct guiding and constraining agent decisionmaking. As used in this context, policies are decisions or plans of action for accomplishing a desired outcome (Lackey 2006). They make scenario intentions operational and in so doing must integrate the facts of a situation with the values that motivate people to manage lands they control in the ways they do. Policies capture rules, regulations, incentives and other strategies promulgated by public agencies in response to demands for ecological and social goods, as well as considerations used by private landowners/land managers to make land and water use decisions. They contain information about site attributes defining the spatial domain of application of the policy, whether the policy is mandatory or voluntary, goals the policy is intended to accomplish, and the duration the policy, once applied, will be active at a particular site. Defining and specifying a scenario-compliant set of policies is a key step in configuring Envision to model alternative futures.

As agents assess alternative land management options, they weigh the relative utility of potentially relevant policies to determine what policies, if any, they will select to apply at any point in time/space. Once applied, a policy outcome is triggered that modifies one or more site attributes, resulting in landscape change. Policies may also be constrained to operating only with selected agent classes (e.g., all home owners, farmers with streams flowing through their property, forest owners with anadromous fish in adjacent streams, etc.).

Envision represents a landscape as a set of polygon-based geographic information system maps and associated information containing spatially explicit depictions of landscape attributes and patterns. Taken as a modeling approach, Envision employs a spatially explicit multi-agent construct that models relationships of agent’s values and behaviors, policy intentions and landscape metrics of production, as the agents attempt to avoid scarcity (Bolte et al. 2006; Guzy et al. 2008).