Forecasting compound drought-heatwaves using Burg entropy spectral analysis with multi-frequency resolutions

Abstract

Due to the significant impacts of compound drought and heatwave (CDHW) events, which are projected to increase in frequency under global warming, establishing a proactive management plan for CDHW events is highlighted in this study. However, studies on forecasting CDHWs have been limited, with most done on a monthly scale. Since heatwaves evolve at a short time scale of less than a week, monitoring and forecasting at sub-monthly scales can provide important information. Thus, this study developed a standardized compound drought-heatwave index (SCDHI) at a 7-day time scale (SCDHI-7D) for South Korea by coupling the standardized antecedent precipitation and evapotranspiration index (SAPEI) and the standardized temperature index (STI) using the copula method. The developed SCDHI-7D effectively monitored CDHW events that occurred in South Korea. To forecast SCDHI-7D, Burg entropy spectral analysis (BESA) with maximal overlap discrete wavelet transform (MODWT), referred to as BESA, was developed. BESA showed good forecasting accuracy, with a median Kling-Gupta efficiency (KGE) value greater than 0.79 up to a 24-day lead time. Besides, BESA forecasted CDHW events effectively, with a hit rate greater than 75 % and a false positive less than 11.52 % up to a 24-day lead time. BESA also maintained good forecasting accuracy for extended lead times longer than 3 weeks. Under the same training conditions, BESA outperformed the Long short-term memory (LSTM), Support Vector Regression (SVR), and XGBoost models that were considered benchmark models. This study could be useful for providing accurate early warning information to establish a proactive management plan for CDHW events.