This paper evaluates errors and uncertainties in representing landscapes that arise from different data rasterization methods, spatial resolutions, and downscaled land-use change (LUC) scenarios. A vector LU dataset for Luxembourg (minimum mapping unit: 0.15 ha; year 2000) was used as the baseline reference map. This map was rasterized at three spatial resolutions using three cell class assignment methods. The landscape composition and configuration of these maps were compared. Four alternative scenarios of future LUC were also generated for the three resolutions using existing LUC scenarios and a statistical downscaling method creating 37 maps of LUC for the year 2050. These maps were compared in terms of composition and spatial configuration using simple metrics of landscape fragmentation and an analysis of variance (ANOVA). Differences in landscape composition and configuration between the three cell class assignment methods and the three spatial resolutions were found to be at least as large as the differences between the LUC scenarios. This occurred in spite of the large LUC projected by the scenarios. This demonstrates the importance of the rasterization method and the level of aggregation as a contribution to uncertainty when developing future LUC scenarios and in analysing landscape structure in ecological studies.
|Number of pages||18|
|Journal||International Journal of Geographical Information Science|
|Publication status||Published - 1 Sep 2008|