COMPREHENSIVE DROUGHT ANALYSIS: MACHINE LEARNING-BASED METEROLOGICAL DROUGHT FORECASTING AND PCA-DRIVEN AGRICULTURAL DROUGHT MONITORING.

Author

Bishal Poudel, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2025

Degree Name

Master of Science

Department

Civil Engineering

First Advisor

Kalra, Ajay

Abstract

Drought is a long-term natural disaster that affects many aspects of human life from health, water supply to ecosystems and agriculture. Drought’s occurrence and intensity has increased in recent years because of global warming, which urges for proper and accurate monitoring and forecasting of drought. For better understanding of droughts, this study uses two approaches. In first part, the study uses the historical temperature and precipitation data from period of 1960 to 2021 as input features for three different machine learning models – Artificial Neural Network (ANN), Support Vector Machine (SVM) and Random Forest (RF). The research focuses on calculating the Standardized Precipitation Index(SPI) and Standardized Precipitation Evapotranspiration Index(SPEI) for temporal drought assessment of the watershed. Statistical parameters like the coefficient of determination (R2), Mean Absolute Error (MAE), Root mean square error (RMSE) and Nash–Sutcliffe Efficiency (NSE) were determined to evaluate the model accuracy. Overall, the ANN and SVM models outperformed the RF model for forecasting long-term drought. The second approach utilizes Principal Component Analysis(PCA) for creating a Combined Drought Indictor (CDI-NM) for New Mexico State, using four agrometeorological variables- Vegetation Condition Index (VCI), temperature, Smoothed Normalized Difference Vegetation Index (SMN), and gridded rainfall data for the period of 2003-2019. The performance of the CDI-NM was evaluated with SPI-3, which results a strong correlation (R² > 0.8 and RMSE=0.03) between both indices for the entire period of analysis. Negative correlation between crop yield and CDI-NM suggests the applicability of CDI-NM in drought monitoring. By combining machine learning models and PCA-based analysis this research enhances both meteorological drought forecasting and agricultural drought monitoring.