Potential of adaptive neuro-fuzzy inference system for evaluation of drought indices

Drought as a natural hazard is characterized using quantitative measures named drought indices. Thus, accurate drought monitoring requires approaches for assessment of drought indices. This work investigates precision of an adaptive neuro-fuzzy computing technique (ANFIS) for drought index estimation through the obtained ANFIS-index. The input data was collected from six meteorological stations in Serbia during the period 1980–2010. Based on selected data, the drought indices such as the water surplus variability index (WSVI) and standardized precipitation index (SPI) for 12 month time scale were calculated. To approve the proposed approach, the ANFIS-index is statistically and graphically compared with SPI and WSVI values. The root-mean-square error ranged between 0.11 and 0.24. The ANFIS-index was highly correlated with SPI and WSVI. The results also show that ANFIS can be efficient applied for reliable drought indices estimation.