Pontius, R. Gil Jr. In Review. Value of information in land cover modeling. Ecological Modelling. Volume ?? (Issue ??):Pages ??.


Value of information in land cover modeling

R Gil Pontius Jr

Abstract

This paper defines quantitative measurements of agreement between two maps of land categories, for example a map of real land categories and a map of simulated land categories. I demonstrate the method to compute the value of perfect information of quantity (VPIQ) and value of perfect information of location (VPIL) in the context of land change simulation models. VPIQ and VPIL show the increase in percent correct that a model would attain if it had the specified information. These measurements can be generated at any point in the modeling process. For example, even before modeling begins, the characteristics of the landscape determine the most valuable type of information, quantity or location. At intermediate steps in the modeling process, the modeler can see which additional information would be most valuable in improving a model. Hence the modeler can decide how to improve the model and how to collect additional data. The paper shows how VPIQ, VPIL, and several other indices of agreement change as a function of the quantity of land change in a simulation. I illustrate the concepts with a land use change model of Costa Rica.

Keywords

kappa, land change model, GIS, value of perfect information

Figures

See the figures page for this paper.

 

Copyright 2000 Graduate School of Geography, Clark University