Sensitivity of the Error in Multivariate Statistical Interpolation to Parameter Values

Abstract

The sensitivity of multivariate optimum interpolation to variations in values of its parameters is investigated, including missing observation values. The influence of observation error misspecification and parameters in the spatial correlation function are also considered. The calculations are carried out on three different observation patterns: fairly uniform, partly uniform?partly sparse, and sparse. The decay rate of the correlation function is an important parameter to estimate properly, and estimates of height and wind errors should be consistent.