In order to model the distribution of areas where groundwater resources are susceptible to nitrate contamination , the data-driven Weights of Evidence (WofE) and Weighted-Logistic Regression (WLR) methods were used. Using this couple of techniques, different tests were performed considering various combinations of predictor factors, deriving posterior probability maps (probability that a unit area contains a training point). In order to establish the best model, success rate curve and prediction rate curve were calculated. At the end simple statistical techniques were then used to individuate the best model between two tests that showed very similar values of prediction rate. This last test could be useful to determine the distribution of the probabilities to find wells with different values of nitrate concentration. Comparison between simulation shows the best performance of Weights of Evidence.
GIS and data-driven models for producing vulnerability maps. A case study: nitrate contamination in Milan District groundwater(Italy) / S. Poli, M. Masetti, S. Sterlacchini - In: Quantitative geology from multiple sources / [a cura di] E. Pirard, A. Dassargues, H. B. Havenith. - [s.l] : IAMG, 2006. (( convegno International IAMG Conference 2006 tenutosi a Liegi nel 2006.
GIS and data-driven models for producing vulnerability maps. A case study: nitrate contamination in Milan District groundwater(Italy)
M. Masetti;
2006
Abstract
In order to model the distribution of areas where groundwater resources are susceptible to nitrate contamination , the data-driven Weights of Evidence (WofE) and Weighted-Logistic Regression (WLR) methods were used. Using this couple of techniques, different tests were performed considering various combinations of predictor factors, deriving posterior probability maps (probability that a unit area contains a training point). In order to establish the best model, success rate curve and prediction rate curve were calculated. At the end simple statistical techniques were then used to individuate the best model between two tests that showed very similar values of prediction rate. This last test could be useful to determine the distribution of the probabilities to find wells with different values of nitrate concentration. Comparison between simulation shows the best performance of Weights of Evidence.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.