Comparison of Ordinary Kriging and Inverse Weighting ‎Distance Methods in Estimating Shallow Groundwater ‎Contamination by Seawater in Tripoli, Libya

Ahmed Ibrahim Khamaj(1) and Abdul Rahman Ahmed Alriyani*(2)

(1). Soil and Water Department, University of Tripoli, Libya.

(2). Advanced Laboratory of Chemical Analysis, Authority of Natural Science Research and Technology, Tripoli, Libya.  

(*Corresponding Author e-mail: abdllrahman817@gmail.com ).

Received: 01/07/2020                               Accepted: 14/08/2020

Abstract

The Ordinary Kriging (OK) and the weighted inverse distance (IDW) methods were used to predict the spatial distribution of seawater contamination extent the shallow aquifer within Tripoli region, Libya. 42 water samples were analyzed to determine the Simpson ratio which was selected as an indicator of groundwater contamination by seawater. Ordinary Kriging method was aimed to be applied on the original and the logarithmic transformed data. The spatial structure was estimated using the spherical, exponential, and Gaussian models. The cross-validation technique was used to assess the interpolation process resulting from the use of the OK and IDW methods with an exponent value ranging from (1 to 5). The performance of the variogram models was evaluated using the determination coefficient (R2) and the sum squares of the residuals (RSS). The square root of mean square error (RMSE) and the relative mean error (MRE) were used to assess the accuracy of the results obtained from the cross-validation technique. The results showed that the performance of the variogram models before the logarithmic transformation was more accurate, than those obtained after the logarithmic transformation. The best performer is the Gaussian model before the logarithmic transformation with 0.463 and 4.30, for (R2) and (RSS), respectively. The cross-validation technique also showed that the method of (OK) by using the Gaussian model before the logarithmic transformation and the method of IDW with exponent of 1 gave the most accurate results, with the preference for the (OK) of calculating with values of 0.535, 0.246, 0.249 for RMSE, MRE, R2, respectively. The map obtained from applying the (OK) method using the Gaussian model showed that, although 76.55% of the area of the study area classified its groundwater as being slightly contaminated by seawater, there is an increase in the risk of groundwater pollution in the southeastern and northwestern borders the region.

Keywords: Groundwater, Seawater intrusion, Simpson ratio, Geostatistics, Cross validation.

Full paper in Arabic: PDF