Here, we support future studies of 137Cs-derived net soil redistribution to apply their often limited resources
across scales of variation (field, catchment, region etc.) without compromising the quality of the estimates
at any scale. We describe a hybrid, design-based and model-based, stratified random sampling
design with composites to estimate the sampling variance and a cost model for fieldwork and laboratory
measurements. Geostatistical mapping of net (1954e2012) soil redistribution as a case study on the
Chinese Loess Plateau is compared with estimates for several other sampling designs popular in the
literature. We demonstrate the cost-effectiveness of the hybrid design for spatial estimation of net soil
redistribution. To demonstrate the limitations of current sampling approaches to cut across scales of
variation, we extrapolate our estimate of net soil redistribution across the region, show that for the same
resources, estimates from many fields could have been provided and would elucidate the cause of differences
within and between regional estimates. We recommend that future studies evaluate carefully
the sampling design to consider the opportunity to investigate 137Cs-derived net soil redistribution
across scales of variation.
