DYNAMIC NAIVE BAYES CLASSIFIER FOR HYDROLOGICAL DROUGHT RISK ASSESSMENT

Hydrological drought requires specific and effective tools for quantification and estimation considering its dependence on both climatic and catchment characteristics. In this study, the Dynamic Naive Bayes Classifier (DNBC) was developed and employed for the assessment of onset and end of hydrological drought. We consider five classes that represent different severities of hydrological drought. The results showed that the probabilities of occurrence of different classes of hydrological drought based on the DNBC were quite suitable and can be employed to estimate the onset of each class and transition to other classes for hydrological droughts. For performance evaluation of classification results, a confusion matrix was made to calculate prediction accuracy and its results were also found appropriate. In comparison with SRI, the accuracies of estimating five classes: class 1, class 2, class 3, class 4, and class 5 by DNBC varied from 50% to 63%, 46% to 70%, 59% to 67%, 45% to 67%, and 33% to 50%, respectively in all the watersheds, respectively. The overall results indicate that the DNBC is an effective tool in predicting the onset and end of hydrological drought events and can be employed for monitoring, improving preparedness and resilience to cope with the risk of this natural disaster.