PARENT SESSION
Oral Session #74: Spatial Ecology: Models and Methods. Presiding: C. Loehle.
Friday, August 10, 2001. 8:00 AM to 11:30 AM. Hall of Ideas E.
Validating spatially-explicit, predictive land-cover models - how does the question shape the method?
Meretsky, Vicky1, Evans, Tom1, Parker, Dawn1, 1
ABSTRACT- Landcover change modeling is a powerful tool to inform researchers about biophysical and social drivers of landcover change. Validation options for such models depend on the complexity of model inputs and outputs, and more importantly on the nature of the questions the model was designed to answer. We develop a conceptual framework to match validation approaches to model questions, and review a range of landscape metrics, the ecological inquiries they may address, and application of these metrics to specific conservation questions in our southern Indiana study area. Reforestation is a primary landscape change occurring on private and public land in the area. Validation of overall landscape composition and summary patch statistics permits modellers to address questions in which patch size and shape are of primary concern (area-sensitive species). If models correctly predict inter-patch relationships then problems involving movement across the landscape - dispersal, migration to breeding areas - can be addressed, provided model output includes appropriate detail and is at the appropriate scale. Finally, if a model can reproduce correlations with input variables such as slope, aspect, ownership, or land-use history, then understanding of underlying forces producing landscape change is improved, and modellers can address landscape conservation concerns such as predicting change in specific community/cover types. In contrast, validation of point-specific cover predictions sets a higher standard than many models need to meet, without guaranteeing that important landscape patterns can be discerned from model output.
KEY WORDS: land cover, validation, modeling, spatial metrics