Document Type : Original Article
Authors
1
Department of Civil Engineering, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran
2
Civil Engineering Department, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran.
Abstract
Drought is considered one of the most hazardous natural phenomena for any country. Therefore, monitoring and forecasting this phenomenon are of critical importance in today's world. To this end, the Standardized Precipitation Evapotranspiration Index (SPEI), one of the most applied drought indices, was employed for drought prediction. To calculate this index, a combination of ground-based and remote sensing data was utilized. Given the different weights of each of these data types in calculating the drought index, the most relevant parameters were first selected using feature selection methods such as the Filter method and the LASSO method and were considered as input parameters for the model. Furthermore, artificial intelligence was utilized to apply various machine learning algorithms, resulting in the development of several models. These algorithms included Bias-Corrected Random Forest, Random Forest, Support Vector Machine, and Multilayer Perceptron. To validate the results of each of these models, indices such as RMSE, R², MSE, and MAE were used. Based on the values of these indices, the Bias-Corrected Random Forest model with R² = 0.9858 and RMSE = 0.1190 for cluster 1 and R² = 0.9809 and RMSE = 0.1375 for cluster 2 was selected as the best-performing model. Finally, the results of the optimized model were used to create drought zoning maps to identify and classify areas prone to drought conditions. These maps can provide valuable information on the distribution and intensity of drought across different regions to stakeholders and experts.
Keywords
)
Send comment about this article