Using model analysis to design monitoring programs for landscape management and impact assessment

While ecologists have long recognized the key role of monitoring programs in natural-resource management, we have only recently come to appreciate the logistical difficulties of designing powerful yet efficient schemes for monitoring large, heterogeneous landscapes. Such designs are especially challenging if the signal to be monitored is uncertain, such as in the case of ecosystem response to climate change. I illustrate an approach in which a simulation model is used to design a monitoring scheme that focuses on application-specific sensitivities or uncertainties. Formal model analysis defines these sensitivities in the model's parameter space. These parametric domains are then mapped into geographic space by regressing model sensitivity on terrain variables in a geographic information system. Specific sites for monitoring are then selected by sampling with a two-stage stratified-cluster design from these parametrically sensitive or uncertain locations. As an example, I use a forest simulation model to design a monitoring scheme as part of a climate-change research program in the southern Sierra Nevada of California (USA). I analyze the model to summarize its sensitivity to variation in temperature and precipitation, and then add a consideration of uncertainty due to the influence of topographic convergence on soil moisture - an influence not simulated by the model. Sensitive and uncertain sites are further constrained by logistical concerns about ease of access, resulting in a target sampling domain that represents less than 2% of the study area.

Duke Scholars

Author Dean L. Urban Environmental Sciences and Policy