Nonlinear Estimation

Abstract

Linear estimation through ordinary kriging provides local averages of a regionalized variable, and calculated estimation variance gives some idea of the expected precision of the estimates. However, some applications may call for knowledge of local spatial distributions. Placing confidence intervals around local estimates; stating the probability that a regionalized variable lies below, between, or above some values; computing mean values of a variable above a cutoff; drawing a map of estimated local medians rather than estimated local averages; and handling extreme values in markedly nonnormal data all require consideration of local distribution.