Using adaptive cluster sampling based on order statistics for kriging estimation of heavy metal in soils

• Main Authors: Rong-Fu Wu, 吳榮富
• Other Authors: Dar-Yuan Lee
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/66762249633985364881

碩士 === 國立臺灣大學 === 農業化學研究所 === 89 === The spatial interpolation of pollutants in contaminated sites is essential for determining hazardous areas needed for remediation. The spatial distribution of pollutants in contaminated sites can be estimated by using geostatistics methods. However, the accuracy of spatial estimation is highly dependent on the sampling design. If the location of high pollution, i.e. hot spots where are priority areas needed for remediation, are not sampled in the sampling process, the accuracy of spatial estimation of pollutants in contaminated site will be decreased and therefore the effectiveness of remediation will be reduced. In this study, one method of adaptive samplings: adaptive cluster sampling based on order statistics, which can locate hot spots, was combined with geostatistics for spatial interpolation of heavy metal distribution in heavy-metal contaminated site. The simulation approach was used to comparing the effect of adaptive cluster sampling based on order statistics and random sampling on the spatial interpolation of pollutants in contaminated site using geostatistics. The results can be used for evaluate the feasibility of using adaptive sampling combined with geostatistics for increasing the accuracy of estimation of spatial distribution of pollutants. In this study, soil Cu concentration data of 177 blocks (1 ha per block ) from Hsing-chu were used. Five hundred simulations were performed using samples from adaptive cluster sampling based on order statistics and random sampling respectively to estimate the Cu concentration of each block by block kriging. Since avoiding the under-estimates of pollutant is most important for delineating hazardous areas of contaminated soils for remediation, the error of under-estimates were used to evaluate to the performance of estimation by block kriging using the two sampling designs. The 500 simulation results show that there fewer numbers of blocks having under-estimates by using adaptive cluster sampling based on order statistics than random sampling. It indicates that adaptive cluster sampling based on order statistics can reduce under-estimates in estimation of spatial distribution of heavy metal when compared to using random sampling. It suggests that adaptive cluster sampling based on order statistics can be used to locate hot spots in a heavy-metal contaminated site, and therefore , can be combined with kriging estimation for obtaining the spatial distribution of heavy metals in soils without loss of spatial information of hot spots.