Effects of cell sizes on resistance surfaces in GIS-based cost distance modeling for landscape analyses

摘要

GIS-based cost distance modeling has been increasingly applied to evaluate habitat quality of existing landscapes and to test "what-if" scenarios for habitat restoration planning. The validity of the raster-based modeling tool is inevitably affected by the cell size of the modeling input. Such effects may be assessed at two levels: resistance and accumulated cost surfaces. This study assesses the resistance-level effects by scaling-up a base landscape at 25 m resolution with six aggregation approaches: the nearest-neighbor assignment, the bilinear interpolation, the cubic convolution, the BlockMajority, the BlockMean and the AggregateMean in ArcGIS~(TM). The effects were measured by the difference between the aggregated and base resistance surfaces using the proposed measures: the mean absolute error, the mean squared error and the autocorrelation coefficient. The results showed that increasing aggregation sizes would generally increase the difference between the generated and the base resistance surfaces for all aggregations. Furthermore, the BlockMajority performed best in terms of the first two measures, whereas the BlockMean and AggregateMean performed best if judged by the third measure.