Sampling strategies for estimating accuracy and area in a long-term land-cover monitoring program

Accuracy assessment of land-cover change products creates challenging problems of sampling design and analysis particularly because change is typically rare, and in a detailed land-cover monitoring program specific types of change may be exceedingly rare. Stratified sampling is often implemented to increase the sample size from rare classes, but how to define strata and allocate the sample to these strata has not been investigated when a time series of land-cover products must be assessed. An evaluation of different options for implementing stratified sampling for a multiple date accuracy assessment will be conducted and results used to provide recommendations on the design and analysis of such studies.