Bitte verwenden Sie diesen Link, wenn Sie dieses Dokument zitieren oder verlinken wollen: https://nbn-resolving.org/urn:nbn:de:gbv:9-000493-4

Geostatistical tools for better characterization of the groundwater quality - case studies for the coastal quaternary aquifers in the Nam Dinh Area/Vietnam

  • The primary objective of this study is to practically apply geostatistical tools that can help to improve an evaluation of groundwater quality for a particular area. The Nam Dinh area, an area of 70 x 70 km2, located in the Southern part of the Red River Delta, was selected as a source for different data sets to be used as case studies. A set of geostatistical tools has thus been applied to the different real data sets which were collected from the coastal Quaternary aquifers in the different campaigns. This gives us a yardstick by which the success of a specific approach can be measured. Throughout the thesis a series of the case studies are, in turn, represented in order to get insight into and an understanding of what various geostatistical tools can do and, more importantly, what their shortcomings are. There are nine different methods of data analyses use in this thesis, which include: (1) Major Ion Comparison, (2) Graphic Plots, (3) Exploratory Statistical Data Analysis, (4) Variogram Analysis, (5) Spatial Estimation Using Kriging, (6) Cluster Analysis, (7) Principle Component Analysis, (8) Multivariate Regionalization Analysis, and (9) Contamination Risk Mapping Using Indicator Kriging. First, major ion comparison and graphic analysis are performed in order to get a general hydrogeochemical view of the collected datasets, before stepping into a further geostatistical approach. By these analyses, various groundwater types are observed and a general hydrochemical trend is visualized using Stiff- and Piper diagrams as well as site maps. The major ion ratios in relation to TDS concentration are compared to investigate the origin of water. Second, a statistical exploratory data analysis is applied to describe the important features of the data by which the character of a specific hydrogeochemical variable might be recognized. Deviations from the Gaussian probability model are detected and appropriate transformations for a formal analysis in geostatistics are selected. Since the Theory of Regionalized Variables (Mathéron 1971; short term: Geostatistics) assumes Gaussian distribution the hydrochemical variables used here are checked for normality. These analyses show that although the data are facing some problems such as outliers and they are very positively skewed at the linear scale, this can effectively be minimized by transforming the data to log-scale. Third, both variogram analyses and Kriging techniques are used to spatially estimate a rectangular 36x36 estimation grid within an area of 70 x 70 km based on the sampled locations (85, 45 and 74 and 38 visited locations for the Pleistocene RS, the Pleistocene DS, the Holocene RS, the Holocene DS, respectively). These estimated values are then used to map the spatiotemporal variability of groundwater quality. In practice, estimation of unknown values and mapping of concentrations of a specific variable can, of course, easily be created by many available software programs. However, error variances are always present in any estimation due to a level of uncertainty, so the reliability of how these estimates could be yielded has also been evaluated in this case study. A critical assessment of all possible variations, tightly related to the seasonal change, directional influence, spatial distribution and prediction error is conducted and concluded. Fourth, Cluster Analysis (CA), Principle Component Analysis (PCA) and Multivariate Regionalization Analysis (RA) are applied to three main datasets of all Quaternary aquifers in the Nam Dinh area to discover the relationships among measured hydrochemical parameters by which we can detect and regionalize major factors which have an impact upon groundwater quality. These approaches are also to overcome the plethora of data that is usually a common problem for any one who has already tackled groundwater data. In this case study both clustering and R mode principal component analyses are thus performed based on the following parameters: The log-transformed concentrations of all measured major ions and of NO3-, NO2-, NH4+, PO42-, i.e. 11 variables from three different datasets of the main aquifers. By cluster analysis three classes of water types, ranging from freshwater to brackish-saltwater types, are typically grouped. Finally, Indicator Kriging (IK) is performed to evaluate the risks of arsenic contamination. The focus of this approach is to assess contamination risk expressed as probability of exceeding threshold- values. The region may thus be subdivided into “safe” and “unsafe” zones on the basis of probability maps which mark contaminated all places where the risk of arsenic contamination exceeds a given threshold for drinking water purpose. By this case study, it is shown that Indicator Kriging is a useful method which has some advantages for many contamination studies. Firstly, it is well known as the non-parametric technique which can be appreciably used when a dataset does not reach normal shape or nearly normal shape as in this situation. Secondly, the outlier problem that often exists in any analysis can be overcome when applying this method. Thirdly, it can be applied in practice to delimit a study area into “safe zone” or “unsafe zone” from which decision-making on the water supply can be decided for the remediation of a contaminated water source or selecting an appreciative source for exploitation. The combined use of spatial (Geostatistics) and multivariate statistical measures have proven to be of major assistance in questions of assessing groundwater quality especially in less sampled regions. A major advantage lies in the possibility of simultaneously creating spatial estimates as well as estimation confidence limits.
  • The primary objective of this study is to practically apply geostatistical tools that can help to improve an evaluation of groundwater quality for a particular area. The Nam Dinh area, an area of 70 x 70 km2, located in the Southern part of the Red River Delta, was selected as a source for different data sets to be used as case studies. A set of geostatistical tools has thus been applied to the different real data sets which were collected from the coastal Quaternary aquifers in the different campaigns. This gives us a yardstick by which the success of a specific approach can be measured. Throughout the thesis a series of the case studies are, in turn, represented in order to get insight into and an understanding of what various geostatistical tools can do and, more importantly, what their shortcomings are. There are nine different methods of data analyses use in this thesis, which include: (1) Major Ion Comparison, (2) Graphic Plots, (3) Exploratory Statistical Data Analysis, (4) Variogram Analysis, (5) Spatial Estimation Using Kriging, (6) Cluster Analysis, (7) Principle Component Analysis, (8) Multivariate Regionalization Analysis, and (9) Contamination Risk Mapping Using Indicator Kriging. First, major ion comparison and graphic analysis are performed in order to get a general hydrogeochemical view of the collected datasets, before stepping into a further geostatistical approach. By these analyses, various groundwater types are observed and a general hydrochemical trend is visualized using Stiff- and Piper diagrams as well as site maps. The major ion ratios in relation to TDS concentration are compared to investigate the origin of water. Second, a statistical exploratory data analysis is applied to describe the important features of the data by which the character of a specific hydrogeochemical variable might be recognized. Deviations from the Gaussian probability model are detected and appropriate transformations for a formal analysis in geostatistics are selected. Since the Theory of Regionalized Variables (Mathéron 1971; short term: Geostatistics) assumes Gaussian distribution the hydrochemical variables used here are checked for normality. These analyses show that although the data are facing some problems such as outliers and they are very positively skewed at the linear scale, this can effectively be minimized by transforming the data to log-scale. Third, both variogram analyses and Kriging techniques are used to spatially estimate a rectangular 36x36 estimation grid within an area of 70 x 70 km based on the sampled locations (85, 45 and 74 and 38 visited locations for the Pleistocene RS, the Pleistocene DS, the Holocene RS, the Holocene DS, respectively). These estimated values are then used to map the spatiotemporal variability of groundwater quality. In practice, estimation of unknown values and mapping of concentrations of a specific variable can, of course, easily be created by many available software programs. However, error variances are always present in any estimation due to a level of uncertainty, so the reliability of how these estimates could be yielded has also been evaluated in this case study. A critical assessment of all possible variations, tightly related to the seasonal change, directional influence, spatial distribution and prediction error is conducted and concluded. Fourth, Cluster Analysis (CA), Principle Component Analysis (PCA) and Multivariate Regionalization Analysis (RA) are applied to three main datasets of all Quaternary aquifers in the Nam Dinh area to discover the relationships among measured hydrochemical parameters by which we can detect and regionalize major factors which have an impact upon groundwater quality. These approaches are also to overcome the plethora of data that is usually a common problem for any one who has already tackled groundwater data. In this case study both clustering and R mode principal component analyses are thus performed based on the following parameters: The log-transformed concentrations of all measured major ions and of NO3-, NO2-, NH4+, PO42-, i.e. 11 variables from three different datasets of the main aquifers. By cluster analysis three classes of water types, ranging from freshwater to brackish-saltwater types, are typically grouped. Finally, Indicator Kriging (IK) is performed to evaluate the risks of arsenic contamination. The focus of this approach is to assess contamination risk expressed as probability of exceeding threshold- values. The region may thus be subdivided into “safe” and “unsafe” zones on the basis of probability maps which mark contaminated all places where the risk of arsenic contamination exceeds a given threshold for drinking water purpose. By this case study, it is shown that Indicator Kriging is a useful method which has some advantages for many contamination studies. Firstly, it is well known as the non-parametric technique which can be appreciably used when a dataset does not reach normal shape or nearly normal shape as in this situation. Secondly, the outlier problem that often exists in any analysis can be overcome when applying this method. Thirdly, it can be applied in practice to delimit a study area into “safe zone” or “unsafe zone” from which decision-making on the water supply can be decided for the remediation of a contaminated water source or selecting an appreciative source for exploitation. The combined use of spatial (Geostatistics) and multivariate statistical measures have proven to be of major assistance in questions of assessing groundwater quality especially in less sampled regions. A major advantage lies in the possibility of simultaneously creating spatial estimates as well as estimation confidence limits.

Download full text files

Export metadata

Additional Services

Share in Twitter Search Google Scholar
Metadaten
Author: Duc Nghia Hoang
URN:urn:nbn:de:gbv:9-000493-4
Title Additional (German):Geostatistische Methoden zur besseren Charakterisierung der Grundwasserqualität – Fallstudien an quartären Aquiferen in der Küstenregion Nam Dinh / Vietnam
Advisor:Prof. Dr. Maria-Theresia Schafmeister
Document Type:Doctoral Thesis
Language:English
Date of Publication (online):2008/09/25
Granting Institution:Ernst-Moritz-Arndt-Universität, Mathematisch-Naturwissenschaftliche Fakultät (bis 31.05.2018)
Date of final exam:2008/04/24
Release Date:2008/09/25
Tag:Indikator-Kriging, Kluster-Kriging, Nam Dinh Gebiet, Variogramm
Cluster Analysis, Geostatistics, Indicator Kriging, Kriging, Nam Dinh area, Principle Component Analysis, Variogram
GND Keyword:Geostatistik, Komponentenanalyse, Kriging
Faculties:Mathematisch-Naturwissenschaftliche Fakultät / Institut für Geographie und Geologie
DDC class:500 Naturwissenschaften und Mathematik / 550 Geowissenschaften, Geologie