Estimation of drought trends and comparison between SPI and SPEI with prediction using machine learning models in Rangpur, Bangladesh
| Autor: | Mst. Labony Akter, Md. Naimur Rahman, Syed Anowerul Azim, Md. Rakib Hasan Rony, Md. Salman Sohel, Hazem Ghassan Abdo |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2023 |
| Předmět: | |
| Zdroj: | Geology, Ecology, and Landscapes, Pp 1-15 (2023) |
| Druh dokumentu: | article |
| ISSN: | 24749508 2474-9508 |
| DOI: | 10.1080/24749508.2023.2254003 |
| Popis: | ABSTRACTThis study investigates drought trends, SPI-SPEI comparisons, and predictions in Rangpur, Bangladesh, from 1979 to 2020. We employed Modified Mann-Kendall for trend analysis, SPI and SPEI for drought assessment, and Pearson Correlation Coefficient and Simple Linear Regression for evaluating SPI and SPEI relationships. Additionally, we utilized ANN, SVM, and RF for prediction. The study revealed notable negative trends in seasonal and annual drought, with the highest z statistics observed for SPI 06 (-2.75), SPI 09 (-4.50), SPI 12 (5.60), SPI 24 (-8.40), SPEI 06 (-5.13), SPEI 09 (-6.82), SPEI 12 (-8.04), and SPEI 24 (-11.20). Strong correlations were identified across all SPI and SPEI indices, with coefficients peaking at 97%, 98%, 98%, and 97% for 06, 09, 12, and 24-month periods, respectively. The comparative assessment favored SPEI over SPI, highlighting its superiority and accuracy. The ANN prediction model showed significant results for short-term and seasonal drought forecasts, projecting SPEI 03 and SPEI 06 increases of 0.02 and 0.24, respectively. However, long-term drought estimation exhibited insignificant performance across all predictive models. This emphasizes the need for developing essential predictive tools for future drought variability. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|