Meteorological Drought Forecasting Based on Climate Signals Using Artificial Neural Network – A Case Study in Khanhhoa Province Vietnam

Abstract In Khanhhoa Province (Vietnam) long-lasting droughts often occur, causing negative consequences for this region, so accurate drought forecasting is of paramount importance. Normally, drought index forecasting model uses previously lagged observations of the index itself and rainfall as input variables. Recently, climate signals are being also used as potential predictors. In this study, we use 3-month, 6-month, and 12-month of Standardized Precipitation Evapotranspiration Index (SPEI), with a calculation time during the period from 1977 to 2014. This paper aims at examining the lagged climate signals to predict SPEI at Khanhhoa province, using artificial neural network. Climate signals indices from Indian Ocean and Pacific Ocean surrounding study area were analysed to select five predictors for the model. These were combined with local variables (lagged SPEI and rainfall) and used as input variables in 16 different models for different forecast horizons. The results show that adding climate signals can achieve better prediction. Climate signals can be also used solely as predictors without using local variables – in this case they explain the variation SPEI (longer horizons, e.g.12-month) reaching 61 – 80%. The developed model can benefit developing long-term policies for reservoir and irrigation regulation and plant alternation schemes in the context of drought hazard.