Data analysis, quality indexing and prediction of water quality for the management of rawal watershed in Pakistan

In contrast to managing the water quality only at the command level (where water is being consumed), one should also give importance to the water quality in the areas where water is being produced i.e. the watersheds. The failure to do so deteriorates the water quality for down streams and poses serious challenges for the water managers in order to meet the water quality requirements on sustainable basis. In order to have an effective water management in command areas, it is essential to assess different aspects of water quality. Rawal watershed is a relatively small watershed area which is being affected by the anthropogenic activities e.g. urbanization, deforestation etc. In this paper, we present the last four years (2009-2012) trends of water quality related parameters along with month-wise as well as source-wise parametric satisfactory analysis against WHO quality standards. Moreover, we applied regression models to check the seasonal water quality trends. The quality indices were analyzed by the combination of supervised and unsupervised machine learning techniques. Different sources of fecal coliforms contamination were also identified. Lastly the possible reasons for high contamination were identified by studying the watershed land covers. Our research suggests that in order to find the quality index of water, Average Linkage (Within Groups) method of Hierarchical Clustering using Euclidean distance is an accurate unsupervised learning technique. Similarly, for classifications, Multi-Layer Perceptron (MLP) has been found to be more accurate supervised learning technique. Higher values of fecal coliforms were found in the months of March, June, July, and October. Some of the possible reasons are land-covers especially scrub forest and rain-fed agriculture areas, poultry farms, and population settled around the streams.