ABSTRACT
Natural disasters like floods, earthquakes, droughts, cyclones, and heat/cold waves typically wreak havoc to the environment and human society. They are classified into three categories: hydrometeorological, geological, and biological. Hydro-meteorological disasters (HMD) are caused by changes in temperature, humidity, air density, pressure, wind speed, and direction. Recently, various extreme climatic observations generate vast amounts of data, i.e., big data from multiple sources that can be used to forecast the extreme climatic events. To overcome the issues faced by conventional numerical weather prediction (NWP), many researchers are turning to deep learning (DL) algorithms to develop an accurate and effective extreme climatic event forecasting system. DL can offer a flexible way of processing ample data and provides scalable solutions for data-driven applications. This chapter discusses the drawbacks of NWP and highlights the robustness of the deep learning-based weather prediction (DLWP) in prognosing HMD. Also, a DL-based framework for the prediction of HMD is proposed that includes spatiotemporal models. The final section of this chapter provides a summary of the various metrics used to assess the effectiveness of the models. In conclusion, the study elaborates the big data handling efficiency of DLWP, which is at above par with the NWP method.
