Utilizing ensemble models for enhanced meteorological drought monitoring and analysis

Effective drought forecasting is crucial for mitigating the impacts of water scarcity and ensuring effective management of natural resources. This study employs three ensemble models based on Ensemble Streamflow Prediction (ESP) to monitor monthly drought predictions. The monthly Standardized Precipitation Temperature Index (SPTI) is initially used to identify droughts. Subsequently, the ESP model requires precipitation, temperature, and SPTI predictions. These forecasts employ a hybrid of the Autoregressive Integrated Moving Average and Artificial Neural Network (ARIMA-ANN) models. The Equal Ensemble Drought Prediction (EEDP), Weighted Ensemble Drought Prediction (WEDP), and Conditional Ensemble Drought Prediction (CEDP) models are established and extensively substantiated for drought prediction after the ESP model is premeditated. The findings show that the EEDP model consistently underperformed compared to the WEDP and CEDP models across multiple meteorological stations and evaluation metrics. For example, at Bahawalpur in April, the CEDP model achieved a lower Mean Absolute Error (MAE) of 0.7110, Root Mean Square Error (RMSE) of 0.8006, and Absolute Bias (AB) of 0.0156, while EEDP and WEDP recorded higher RMSE values of 1.0489 and equal AB values of 0.0626. In Muzaffargarh (April), the RMSE for CEDP was 0.5340, compared to 0.6797 for both EEDP and WEDP. Notably, at Narowal in May, CEDP significantly outperformed with an RMSE of 0.4612, whereas EEDP and WEDP both showed RMSE values of 0.8877. Across most stations, CEDP also reported markedly lower AB values, often below 0.02, while EEDP and WEDP generally exceeded 0.05. These results demonstrate that incorporating climate information, especially through the conditional logic of CEDP, substantially enhances forecasting performance. Overall, CEDP outperforms both EEDP and WEDP, confirming its superiority in achieving more accurate and reliable drought predictions.