Published: Jan 1986
| ||Format||Pages||Price|| |
|PDF Version (188K)||14||$25||  ADD TO CART|
|Complete Source PDF (2.7M)||14||$55||  ADD TO CART|
This paper identifies the critical problems of soil sampling for pollution monitoring and explains the solutions provided by new advances in spatial statistics such as Kriging. The U.S. EPA at the Environmental Monitoring Systems Laboratory, Las Vegas, is applying new types of spatial statistics to the chronic problems of pollution plume monitoring and contouring. Firstly, the most common soil sampling problem is the failure to recognize and utilize autocorrelation in space and/or time of the variable in question. Such correlation variables require a systematic rather than a random sampling design, and a spatial analysis by contouring rather than a t-test. This problem can be solved by the use of Kriging. Secondly, chemical analyses of pollution samples often give many readings below detection limit with a few relatively high readings indicating a non-normal, truncated, positively skewed distribution. Estimates or tests assuming normal theory are imprecise for such data, but nonparametric spatial statistics are more appropriate. Thirdly, data may be from different monitoring techniques or different chemical analyses, causing heteroprecision. Such data require an analysis such as “soft-data” Kriging that accounts for the precision of each datum in the analysis. Finally, if remediation decisions are to be based on the data, the probability of misjudging “clean areas to be dirty” or “dirty areas to be clean” must be calculable. Probability Kriging gives such probabilities. This paper discusses these new forms of spatial analysis as they apply to the recurring problems of pollution monitoring.
spatial statistics, Kriging, soil sampling, sampling design, monitoring system design, geostatistics
Mathematical Statistician, Environmental Monitoring Systems Laboratory-Las Vegas, Las Vegas, Nevada
Paper ID: STP18431S