Spatial variability and risk assessment of groundwater pollution in El-Outaya region, Algeria

Abstract

Groundwater contaminations’ spatial distribution and risk assessment are among the most important issues for human consumption, water resources managers and decision makers. For this purpose, stochastic and deterministic methods are employed in this paper based on the empirical Bayesian Kriging (EBK), ordinary kriging (OK), and radial basis function (RBF) with multi-quadratic function, which help to predict the spatial distribution and risk assessments. These methods are based on the contamination data from 68 wells in El-Outaya plain (Biskra, southeastern Algeria. Nitrate-nitrogen (NO3–N) is the most important contamination in this region. Among the methods indicator Kriging (IK) is used to elaborate probability maps based on three thresholds, namely, [NO3–N] > 3 mg/l, [NO3–N] > 8.9 mg/l and [NO3–N] > 11.3 mg/l. The probability maps are then transformed to categories using binary maps of two of exceeding thresholds levels as (P ≥ 0.5 and P ≥ 0.9). The determination coefficient (R2), the root mean square error (RMSE) and the average standard error (ASE) of the obtained raster maps are used to compare the different interpolation methods and to select the best one. The results show that all the used interpolation techniques have accurately predicted the spatial distribution of [NO3–N], which provides opportunity to predict the contamination values even at unsampled locations in the region. The raster maps indicate the same overall spatial distribution of groundwater quality categories and their measurement errors indicate the superiority of RBF over EBK and OK. RBF is an accurate interpolation technique to identify groundwater pollution in El-Outaya plain. The transformation of probability maps to categorical maps based on IK is important for decision makers to identify regions with high nitrate-N contamination probability and high risk to human health.