Aquatic Monitoring: Data Analysis and Network Design Using Regionalized Variable Theory

摘要

Standard statistical techniques for the testing of hypotheses, both parametric, and nonparametric, are based on a few key assumptions such as independence, randomness, and, often, equal variance of the observations. In the case of spatially distributed variables such as descriptors of water quality and aquatic biology, these assumptions are usually not met. The recently developed theory of regionalized variables, together with a related estimation technique, kriging, provide a methodology for the statistical analysis of spatially and/or temporally distributed variables which do not meet the standard assumptions. In this report, regionalized variable theory is reviewed, and a technique is developed to assess possible differences in the level of a spatial process in two regions (the so-called two sample problem) when the observations are highly correlated, heteroscedastic, and a trend is present. The effectiveness of the proposed method is assessed using synthetically generated data; the results indicate the kriging method to be substantially more powerful than conventional analysis of variance. An algorithm is also developed to assist in the placement of sample stations.