作者: Jorge Kazuo Yamamoto
关键词:
摘要: This paper presents an interpolation variance as alternative to the measure of reliability ordinary kriging estimates. Contrary traditional variance, is data-values dependent, variogram and a local accuracy. Natural phenomena are not homogeneous; therefore, variability expressed through data values must be recognized for correct assessment uncertainty. The simply weighted average squared differences between retained estimate. Ordinary or simple variances expected variances; these homoscedastic estimation cannot properly dispersion. More precisely, estimate conditional when weights interpreted probabilities associated n neighboring data. interpretation valid if, only all positive constrained such. Extensive tests illustrate that useful variance.