Inverse method using boosted regression tree and k-nearest neighbor to quantify effects of point and non-point source nitrate pollution in groundwater

摘要: Abstract Nitrate pollution of groundwater has increased dramatically worldwide due to increase population and agricultural productivity. The resulting nitrate concentration in is usually a combination various types point non-point pollutant sources. It often difficult distinguish between these sources since formed large complex catchments with natural processes anthropogenic influence that contribute certain downstream concentration. For such conditions, this paper uses methodology can be used inversely determine type location main source. builds on two state-of-the-art data mining techniques, boosted regression tree (BRT) k-nearest neighbor (KNN). These techniques are produce vulnerability map. mitigate effects subjective judgement determining importance different mechanisms for transport. investigated hydrogeological, hydrological, anthropogenic, topography, soil conditioning factors. Thus, the proposed separate pollution. To calculate maps, 40 mg/L (suggested by WHO 20% risk margin) was selected as general threshold identifying polluted areas resulted 96 wells. Non-polluted locations were from well less than 15 mg/L (96 non-polluted). models trained 70% non-polluted site data. remaining data, 30% sites, validate simulation results. Results showed BRT produced outputs higher performance KNN algorithm. final ranking results based model hydraulic conductivity, river density, soil, slope percent, net recharge, distance villages, order, relative other