Statistical Models and Water Quality Characterization of Source Waters in Nepal

Available sample data of 270 water samples collected from different intake sources of Nepal is analyzed. Analysis showed that well waters were the hardest with more than 60% samples recording hardness above 180 mg/l whereas stream and spring waters showed only 20% and 10% in the same range, respectively. Total dissolved solid (TDS) contents in spring and stream waters were below 500 mg/l while wells waters showed 16% above 500 mg/l. More than 80% from all sources showed pH values within WHO limit. Electric conductivity (EC) was found very low in almost all samples from stream and spring waters. Only well waters showed about 55% with EC in between 500 and 1000 μmhos/cm. Spring and stream waters were corrosive compared to well waters. In well waters 60% were found stable and 15% scalable. Two nonlinear statistical models, a log-linear and a Box-Cox transformed model are found appropriate for estimating total hardness and TDS, respectively with predictor variables like temperature, pH, and electric conductivity. Models are useful for the purpose of prediction when measurements are unavailable for total hardness and TDS.