Impact of compositional and configurational data loss on downscaling accuracy

Datasets collected at widely varying spatial scales are often merged to address questions related to global environmental change. Integrating data requires aggregation, which reduces data quality and introduces statistical biases collectively known as the Modifiable Areal Unit Problem (MAUP). These biases result from different forms of compositional and configurational data loss that occur during aggregation, but little is known about the relationship between data loss and MAUP biases for downscaling. This study uses the well-established process of landscape and surface metric scaling to examine how uncertainties related to the composition and configuration of land cover patterns propagate across scales when data are aggregated and ultimately impact downscaling results. Results suggest a link between compositional data loss and downscaling accuracy, particularly in the patch-based landscape paradigm. Further work is needed to determine if relationships exist between compositional and configurational data loss measures and downscaling error in the surface paradigm.