Multiscale Effects on Spatial Variability Metrics in Global Water Resources Data

Spatial scales and methods for dealing with scale have been widely discussed in the water resources literature. Different spatial processes operate at different scales so interpretations based on data from one scale may not apply to another. Understanding the behavior of phenomena at multiple-scales of data aggregation is thus imperative to accurate integrations of data and models at different geographic resolutions. This study tests theoretical concepts of scale by presenting empirical results of multiscale GIS and statistical analyses on gridded water-availability, water use and population data for the Danube Basin in Europe, with results corroborated by similar tests in the Ganges (South Asia) and Missouri (North America) Basins. Fine-resolution datasets were aggregated to coarser grid sizes and standard statistical measures of spatial variability were computed. Statistical analysis of spatial variability demonstrated two distinctly different cases for unscaled and scaled variables. Results show that variance (and standard deviation) in unscaled variables like freshwater supply, use and population increases at coarser scales—contrary to the common assumption of decreasing variability as grid-cell size increases. On the other hand, a decreasing trend in variability with scale is noted for variables scaled to area or population (like population density, water availability per capita etc.). Moreover, relationships between variability and scale show strong non-linear trends. No mention of these relationships has been found in the water resources or socio-economic literature for scale and variability. Regression analyses suggest that power functions are the most appropriate model to fit trends in increasing variability at multiple scales. These results can be applied to interpretations of water-stress and water scarcity data and their locations relative to water sources or topographic barriers.