Spatial Interpolation for Identifying Soil Contamination Area
Spatial Interpolation for Identifying Soil Contamination Area Author: Hsiao-Hui Chen Po-Huang Chiang Hsiao-Lei Chen Ta-Chien Chan I-Fang Mao Dennis P. H. Hsieh August 9, 2006
Synopsis • • • • •
August 9, 2006
Background Goal Method Result Discussion
Background • • • • •
Soil pollution in An-Shun superfund area in Tainan city, Taiwan Major product: pentachlorophenol (PCP) pesticides Byproducts---1.dioxin, 2.mercury To evaluate and predict the range and severity in this area. The health effect of human beings
August 9, 2006
History of An-Shun superfund area Since the pentachlorophenol (PCP) plant of the China-Petroleum-Chemical company at Tainan city was closed in 1982, the base of the plant has been found contaminated with dioxins, mercury, and PCP. Ponds and streams near the base could also been contaminated with these chemicals and may lead to the contamination of fishes, shrimps, and oysters raised. August 9, 2006
Geographical environment
Annan district, Sicao neighbor, Luer neighbor, & Xiangong neighbor in Tainan City
August 9, 2006
Sampling points-1
August 9, 2006
Sampling points-2 • Whole sample size is including 265 samples. Because of avoiding the bias of sample points without value, we excluded samples with null value: 1. Dioxin: 249 samples 2. Mercury: 189 samples
August 9, 2006
Goal • We use GeoStatistics and spatial analysis to predict and evaluate the data of soil and sediment in An-Shun area from Taiwan’s EPA, these including dioxin and mercury; in order to understand the range, severity of soil pollution and fish farms pollution in this area. August 9, 2006
Method • Ordinary Kriging • Simple Kriging • Universal Kriging
August 9, 2006
Kriging ¾Prediction – Trend – Spatially-autocorrelated random error (fluctuation)
¾Z(s) = μ(s) + ε (s)
August 9, 2006
Ordinary Kriging
Ordinary Kriging assumes the data have a constant mean (no trend) and that the mean value is not known in advance. The data values (orange dots) are thought of as random errors that fluctuate around the unknown mean. The random errors are9,autocorrelated, meaning they tend to be above or below the mean in a way August 2006 similar to their neighbors.
Simple Kriging • Regardless of whether the trend is constant or varies, if its terms (all parameters and covariates) are completely known in advance, then you have the model for Simple Kriging.
August 9, 2006
Universal Kriging
Universal Kriging assumes there is a trend in the data, but the terms of the trend function are not known in advance. The data values (orange dots) are thought of as random errors that fluctuate around the unknown trend. The random errors are autocorrelated, meaning they tend to be above or below the trend in a way similar to August 9, 2006 their neighbors.
Variogram models Variogram models must be “positive definite” so that the covariance matrix based on it can be inverted (which occurs in the kriging process). Because of this, only certain models can be used. August 9, 2006
Source:Arthur J. Lembo, Jr. Cornell University
Result
August 9, 2006
Dioxin Sampling Sites
August 9, 2006
Summary of Semivariogram ComparisonDioxin Root Mean Square Error
Average Standard Error
Spherical_OK
66110
44620
Exponential_OK
67130
42470
Spherical_UK
66060
44470
Exponential_UK
67020
42830
Spherical_SK
67280
34290
Exponential_SK
67910
30140
August 9, 2006
Prediction and probability map-Dioxin
August 9, 2006
Mercury Sampling Sites
August 9, 2006
Summary of Semivariogram ComparisonMercury Root Mean Square Error
Average Standard Error
Spherical_OK
1069
487.9
Exponential_OK
1052
597.2
Spherical_UK
1070
306.8
Exponential_UK
1052
367.7
Spherical_SK
1068
512.6
Exponential_SK
1052
658.3
August 9, 2006
Prediction and probability map-Mercury
August 9, 2006
Present investigation from Taiwan EPA 1. 2. 3. 4. 5. 6.
Dormitory area---mercury’s concentration is over the Soil and Groundwater Pollution Remediation Act (SGPRA). Fish farms’ dike in southern of factory---dioxin’s concentration is over the SGPRA. Soil of the northeast of dormitory area- dioxin’s concentration is higher than background value. Zufagong river is the most polluted area with its sediment---dioxin’s concentration is over the SGPRA Sediment of Luermin river: high concentration dioxin is centralized in midstream、upstream. The most polluted area of neighbor fish farms with sediment: locations are centralized in factory’s westside、southwestside
August 9, 2006
Discussion • It’s a convenient method to identify the polluted area rapidly. • The major limitation is that we cannot always predict the range and severity exactly. If possible, it’s better to have sampling on-the-spot, in order to validate the present predicted model. August 9, 2006
August 9, 2006
Synopsis • • • • •
August 9, 2006
Background Goal Method Result Discussion
Background • • • • •
Soil pollution in An-Shun superfund area in Tainan city, Taiwan Major product: pentachlorophenol (PCP) pesticides Byproducts---1.dioxin, 2.mercury To evaluate and predict the range and severity in this area. The health effect of human beings
August 9, 2006
History of An-Shun superfund area Since the pentachlorophenol (PCP) plant of the China-Petroleum-Chemical company at Tainan city was closed in 1982, the base of the plant has been found contaminated with dioxins, mercury, and PCP. Ponds and streams near the base could also been contaminated with these chemicals and may lead to the contamination of fishes, shrimps, and oysters raised. August 9, 2006
Geographical environment
Annan district, Sicao neighbor, Luer neighbor, & Xiangong neighbor in Tainan City
August 9, 2006
Sampling points-1
August 9, 2006
Sampling points-2 • Whole sample size is including 265 samples. Because of avoiding the bias of sample points without value, we excluded samples with null value: 1. Dioxin: 249 samples 2. Mercury: 189 samples
August 9, 2006
Goal • We use GeoStatistics and spatial analysis to predict and evaluate the data of soil and sediment in An-Shun area from Taiwan’s EPA, these including dioxin and mercury; in order to understand the range, severity of soil pollution and fish farms pollution in this area. August 9, 2006
Method • Ordinary Kriging • Simple Kriging • Universal Kriging
August 9, 2006
Kriging ¾Prediction – Trend – Spatially-autocorrelated random error (fluctuation)
¾Z(s) = μ(s) + ε (s)
August 9, 2006
Ordinary Kriging
Ordinary Kriging assumes the data have a constant mean (no trend) and that the mean value is not known in advance. The data values (orange dots) are thought of as random errors that fluctuate around the unknown mean. The random errors are9,autocorrelated, meaning they tend to be above or below the mean in a way August 2006 similar to their neighbors.
Simple Kriging • Regardless of whether the trend is constant or varies, if its terms (all parameters and covariates) are completely known in advance, then you have the model for Simple Kriging.
August 9, 2006
Universal Kriging
Universal Kriging assumes there is a trend in the data, but the terms of the trend function are not known in advance. The data values (orange dots) are thought of as random errors that fluctuate around the unknown trend. The random errors are autocorrelated, meaning they tend to be above or below the trend in a way similar to August 9, 2006 their neighbors.
Variogram models Variogram models must be “positive definite” so that the covariance matrix based on it can be inverted (which occurs in the kriging process). Because of this, only certain models can be used. August 9, 2006
Source:Arthur J. Lembo, Jr. Cornell University
Result
August 9, 2006
Dioxin Sampling Sites
August 9, 2006
Summary of Semivariogram ComparisonDioxin Root Mean Square Error
Average Standard Error
Spherical_OK
66110
44620
Exponential_OK
67130
42470
Spherical_UK
66060
44470
Exponential_UK
67020
42830
Spherical_SK
67280
34290
Exponential_SK
67910
30140
August 9, 2006
Prediction and probability map-Dioxin
August 9, 2006
Mercury Sampling Sites
August 9, 2006
Summary of Semivariogram ComparisonMercury Root Mean Square Error
Average Standard Error
Spherical_OK
1069
487.9
Exponential_OK
1052
597.2
Spherical_UK
1070
306.8
Exponential_UK
1052
367.7
Spherical_SK
1068
512.6
Exponential_SK
1052
658.3
August 9, 2006
Prediction and probability map-Mercury
August 9, 2006
Present investigation from Taiwan EPA 1. 2. 3. 4. 5. 6.
Dormitory area---mercury’s concentration is over the Soil and Groundwater Pollution Remediation Act (SGPRA). Fish farms’ dike in southern of factory---dioxin’s concentration is over the SGPRA. Soil of the northeast of dormitory area- dioxin’s concentration is higher than background value. Zufagong river is the most polluted area with its sediment---dioxin’s concentration is over the SGPRA Sediment of Luermin river: high concentration dioxin is centralized in midstream、upstream. The most polluted area of neighbor fish farms with sediment: locations are centralized in factory’s westside、southwestside
August 9, 2006
Discussion • It’s a convenient method to identify the polluted area rapidly. • The major limitation is that we cannot always predict the range and severity exactly. If possible, it’s better to have sampling on-the-spot, in order to validate the present predicted model. August 9, 2006
August 9, 2006
