Spatial correlation in reflected radiation from the ground and its implications for sampling and mapping by ground-based radiometry

Abstract The radiation reflected from the ground can rarely be measured over the whole of a region, for example, a field, by ground-based radiometry. It must usually be estimated by sampling and, to avoid bias samples, should be probabilistic. The radiation is spatially correlated and largely stochastic, and in these circumstances systematic sampling is the most efficient. Assuming the Intrinsic Hypothesis of Regionalized Variable Theory, the variation in radiance can be described quantitatively by the variogram. This can then be used to design sampling schemes both for estimating the mean radiation from isolated blocks of land and for mapping the radiation over a region to meet some specified tolerance expressed in terms of standard error. The paper describes the procedure for doing this and illustrates it with examples from species-poor grassland near San Francisco Bay in California, winter barley near Thetford, Norfolk, England, and species-rich grassland near Monyash, Derbyshire, England.