Meteorological drought forecasting in the Mekong delta using artificial neural networks (ANN)

Abstract

    The objective of this study is to forecast meteorological droughts in the Mekong Delta using Artificial Neural Networks (ANN). Standardized Precipitation Index (SPI) was first calculated to determine 3 and 6 - month droughts for current period of 1980-2013; and SDF (Severity - Duration - Frequency) curve was then established. ANN for meteorological droughts (SPI 3 và 6) was established, calibrated, validated and then used to forecast SPI3và SPI6, leading time t+1 and t+3. Results of SPI calculation for period of 1980-2013 are found to change in terms of spatial and frequency. From results of drought index maps and SDF curve, solutions will be proposed to adapt to meteorological drought under different severity level and frequency. The results of calibration and validation at 3 stations (Can Tho, Bac Lieu and Chau Doc) for SPI show a relatively good agreement between simulated and observed SPI (calculated from rainfall data) and the accuracy has reduced when the forecasting lead time increased.