Kinneret watershed analysis tool: a cell-based decision tree model for watershed flow and pollutants predictions

This paper presents the methodology and application underlying the Kinneret Watershed Analysis Tool (KWAT), developed for flow and contaminant predictions for Lake Kinneret (the Sea of Galilee) watershed located in northern Israel. Lake Kinneret watershed is about 2730 km2 (2,070 in Israel, the rest in Lebanon), inhabited by about 200,000 people organized in 25 municipalities, and three cities (the Israeli part). The model aims to predict flow and contaminant transports within the watershed, down to its outlet-Lake Kinneret, the most important surface water resource in Israel. The model is comprised of two sections: quantity and quality. The objective of the quantity section is to tune the values of a vector of coefficients alpha that multiply the average rainfall time series intensity I(t) (the input) imposed on given sub-sets (i.e., cells) of the basin so as to calibrate their outlet flows Q(t); the quality section then uses these optimal flows Q(t) and the effective optimal rainfall intensities to adjust the values of a vector of coefficients beta so as to calibrate the sub-watersheds outlet concentrations C(t). The model uses decision trees coupled with a genetic algorithm for optimally tuning the KWAT coefficients for each of the watershed cells, which taken together comprise the flow and contamination amounts measured at the watershed outlet.