An integrated hidden Markov model with an adaptive exponential weighting scheme for forecasting a meteorological drought index

Abstract

Drought is a naturally occurring climate phenomenon that significantly affects human and environmental activity, and can be considered one of the most widespread and destructive natural disasters. This makes drought forecasting vital for drought risk reduction. In this study, a hidden Markov model (HMM) with an adaptive exponential weighting (AEW) scheme (HMM-AEW) was proposed to develop a new framework to forecast the standardized precipitation index (SPI) as a meteorological drought measurement considering historically-similar patterns of the SPI. The model was applied to a monthly SPI series for a sub-basin of Han River in South Korea that covered more than 30 years. The performance of the HMM-AEW was measured using two commonly-used statistical criteria. The results indicated that the HMM-AEW is able to forecast most of the key points (i.e., fluctuation and extreme points) of the SPI series, with root mean squared error values of 0.612 and 0.368 and adjusted R-square values of 0.671 and 0.670, respectively. Furthermore, in terms of drought category, the proposed model exhibited a satisfying forecasting performance with hit rates of 80% and 50% for fluctuation and extreme points, respectively. © New Zealand Hydrological Society (2017).