Flash drought prediction using deep learning
Bibliographic Data
| ID | 15548610 |
|---|---|
| Authors | Preminda Jacob (Carnegie Mellon University, corresponding author), Pratik Elias Jacob (0009-0006-7065-9857), Nurendra Choudhary (0000-0002-4471-8968, Virginia Tech), Abhirup Dikshit (0000-0003-2876-4080, UNSW Sydney), Jason P Evans (0000-0003-1776-3429, UNSW Sydney), Biswajeet Pradhan (0000-0001-9863-2054, University of Technology Sydney), Alfredo Huete (0000-0003-2809-2376, University of Technology Sydney) |
| Year | 2025 |
| Volume | 20 |
| Issue | 7 |
| Pages | 074006-074006 |
| Publication date | 2025-05-21 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Environmental Research Letters (JOURNAL) |
| Journal identifiers | ISSN: 1748-9326 • E-ISSN: 1748-9326 |
| Publisher | IOP Publishing (PUBLISHER • GB) |
| DOI | 10.1088/1748-9326/addb65 |
| OpenAlex | W4410578926 |
| Language | EN |
| References cited | 22 |
Flash droughts are rapid, short-term drought events that develop within weeks, driven by factors such as low rainfall, high temperatures, and strong winds, which deplete soil moisture and stress vegetation. These events have profound agricultural, economic, and ecological impacts, yet the use of machine learning to predict flash droughts remains underexplored, hindered by challenges like imbalanced datasets and limited data. This study addresses these issues by applying Convolutional neural networks (CNNs) to predict flash droughts in Eastern Australia, a region prone to such events. We identified flash droughts from 2001 to 2022, training the model with data from 2001–2015, validating it on 2016–2017 data, and testing it on 2018–2022 data. The model’s performance was evaluated across drought duration, spatial distribution, and seasonal variability. Achieving a balanced accuracy of 80% and an Area under the curve of 93%, the CNN demonstrated strong predictive capability. However, it tended to overestimate the spatial extent of droughts, indicating areas for future improvement. These results highlight the potential of deep learning in flash drought prediction, offering valuable insights for early warning systems and drought management strategies
Flash (photography · Machine learning · Computer Science · Energy Load and Power Forecasting · Flood Risk Assessment and Management · Hydrology and Drought Analysis · Artificial Intelligence
| Citation velocity | historical |
|---|---|
| Highly cited | No |