Hydro-climatological forecasting: A view from the spectral domain

Abstract:

As of late July 2020, Australia has seen the most devastating bushfire and the world has been altered radically due to the coronavirus pandemic. Urgent actions should be taken to tackle poverty and inequality, health, education, education, biodiversity, and climate. As hydro-climatologists, we aim to resolve the challenge in hydro-climatological forecasting, focusing on the drought risk assessment across different temporal scales. Wavelet-based methods for Hydro-climatological forecasting have been adopted in many recent studies due to their capability of time-frequency localization. The wavelet decompositions of a time series can represent periodicities, temporal short- and long-range dependence, and non-stationarities existed in the time series. As a result, we have developed a novel wavelet-based methodology for transforming predictor variables so as to force greater consistency in spectral attributes with the response of interest. The proposed spectral transformation technique is assessed with both synthetic and real-world examples (Jiang, Sharma, & Johnson, 2020). In the study, a commonly adopted drought index (Standardized Precipitation Index, SPI) was predicted with greater accuracy using the transformed predictor variables. First, this approach provides values in applications particularly when model simulations are used for future impact assessment. However, it can be extended to cases where only concurrent information is available if another wavelet transform alternative is adopted. While our application focuses on the Australian mainland, the method is generic and can be adopted anywhere. Meanwhile, we are looking for big problems to implement our methodology.