SPI-Based Hybrid Hidden Markov–GA, ARIMA–GA, and ARIMA–GA–ANN Models for Meteorological Drought Forecasting

Abstract

Drought is a severe environmental disaster that results in significant social and economic damage. As such, efficient mitigation plans must rely on precise modeling and forecasting of the phenomenon. This study was designed to enhance drought forecasting through developing and evaluating the applicability of three hybrid models—the hidden Markov model–genetic algorithm (HMM–GA), the auto-regressive integrated moving average–genetic algorithm (ARIMA–GA), and a novel auto-regressive integrated moving average–genetic algorithm–ANN (ARIMA–GA–ANN)—to forecast the standard precipitation index (SPI) in the Bisha Valley, Saudi Arabia. The accuracy of the models was investigated and compared with that of classical HMM and ARIMA based on a performance evaluation and visual inspection. Furthermore, the multi-class Receiver Operating Characteristic-based Area under the Curve (ROC–AUC) was applied to evaluate the ability of the hybrid model to forecast drought events. We used data from 1968 to 2008 to train the models and data from 2009 to 2019 for validation. The performance evaluation results confirmed that the hybrid models provided superior results in forecasting the SPI one month in advance. Furthermore, the results demonstrated that the GA-induced improvement in the HMM forecasts was matched by an approximate 16.40% and 23.46% decrease in the RMSE in the training and testing results, respectively, compared to the classical HMM model. Consequently, the RMSE values of the ARIMA–GA model were reduced by an average of 10.06% and 9.36% for the training and testing processes, respectively. Finally, the ARIMA–GA–ANN, which combined the strengths of the linear stochastic model ARIMA and a non-linear ANN, achieved a greater reduction values in RMSE by an average of 32.82% and 27.47% in comparison with ARIMA in the training and testing phases, respectively. The ROC–AUC results confirmed the capability of the developed models to distinguish between events and non-events with reasonable accuracy, implying the appropriateness of these models as a tool for drought mitigation and warning systems.

Evidence for Intensification in Meteorological Drought since the 1950s and Recent Dryness–Wetness Forecasting in China

“…Forecasting future dry-wet patterns in a region is essential for supporting drought risk assessments and sustainable strategies of water resource management [79,80]. There is a crucial necessity to perform an accurate forecast of drought occurrence especially for a longer timescale [81].…”

“…ARIMAs are normally divided into two categories of seasonal and nonseasonal [113]. A typical nonseasonal ARIMA model is characterized by three parameters (p,d,q), where d represents the order of the differences in time series and p and q represent the orders of the autoregressive and moving averages, respectively [79]. The expression for the nonseasonal ARIMA(p,d,q) model is given as follows:…”

“…The AIC, MAE, MSE, RMSE, R 2 , and NSE were used to evaluate the goodness-of-fit of the developed forecasting models. These indicators have been widely used to evaluate the performance of different prediction models (e.g., [46,47,74,75,[78][79][80]82,96,97]). To illustrate the uncertainty of forecasting results in space, the geographical distribution of the performance metrics was plotted, as shown in Figure 9, in which the blue (red) refers to the model error (accuracy) indicator.…”

“…Drought forecasting can effectively reduce the risk of drought [96] and has a key role in risk management and drought readiness and alleviation [112]. Given this, efficient mitigation plans must rely on precise modeling and forecasting of the phenomenon [79]. A key challenge is developing accurate drought forecast model and understanding a model's capability to examine different drought characteristics [97].…”

“…Although linear models such as ARIMA give better forecasting accuracy with stationary time series data, nonlinear methods such as the ANN are more appropriate for nonstationary datasets [77]. The studies in the literature mainly focus on linear or nonlinear modelings individually [46,47,74,76,78,[80][81][82]97,113,134,135] or a combination of them [75,77,79,96,111,112,136]. The combination of linear and nonlinear models exploits the advantages of each individual model type (i.e., linear and nonlinear) and complexity levels (i.e., single or hybrid) in time series forecasting [75].…”