Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Drought forecasts using satellite data based on deep learning over East Asia

Auteur : Park, Sumin

Université de soutenance : Ulsan National Institute of Science and Technology

Grade : Doctor of Philosophy (PhD) 2021

Résumé
This thesis/dissertation seeks to 1) forecast drought conditions effectively considering temporal patterns of drought indices and upcoming weather conditions through the deep learning approach, and 2) forecast drought by identifying the teleconnection effect based on the sea surface temperature through the deep learning approach. In this thesis/dissertation, there are four chapters. Chapter 1 summarizes the background of the research and overviews of the thesis research. In Chapter 2, drought-forecasting models on a short-term scale (8 days) were developed considering the temporal patterns of satellite-based drought indices and numerical model outputs through the synergistic use of convolutional long short term memory (ConvLSTM) and random forest (RF) approaches over a part of East Asia. Through the combination of temporal patterns and the upcoming weather conditions (numerical model outputs), the overall performances of drought-forecasting models (ConvLSTM and RF combined) produced competitive results. Furthermore, our short-term drought-forecasting model can be effective regardless of drought intensification or alleviation. The proposed drought-forecasting model can be operationally used, providing useful information on upcoming drought conditions with high resolution (0.05°). In Chapter 3, the Drought forecasting model on a mid-and long-term scale (one-three lead time) over East Asia was developed using temporal patterns of drought indices and teleconnection phenomena of SST through the CNN. Reanalysis based drought index, SPI, were selected with a mid- and long-timescale (one to three months), and satellite-based variable, precipitation and SST across the Pacific Ocean. As the lead time increased, the accuracy tended to fall, but it showed good results compared to CFS. When compared to a drought case, the SST of 8 months ago influenced on the results. Chapter 4 provides a brief summary of these studies

Présentation

Version intégrale