Scholarlink

<p>Precipitation prediction is crucial for managing floods and issuing timely warnings. As a vital part of the water cycle, precipitation provides valuable insights into regional climate patterns. In this research, Decision Tree, Random Forest, and XGBoost algorithms were employed to model monthly precipitation and monthly maximum 24-hour precipitation data for the city of Lamerd, located in Fars Province, Iran. The average temperature, absolute minimum temperature, absolute maximum temperature, average minimum temperature, average maximum temperature, average humidity, maximum humidity, minimum humidity, and evaporation were considered as input data. The data were studied from 1996 to 2024. 70% of the data was used for training and 30% for testing. The results indicated that all three models exhibited satisfactory performance. The XGBoost model outperformed the other two models in estimating monthly precipitation and monthly maximum 24-hour precipitation. The performance metrics of the XGB model for the test dataset are summarized below: For predicting Pmonth, the R² value is 0.9337, with a Mean Absolute Error (MAE) of 6.5032, a Root Mean Squared Error (RMSE) of 11.2456, and a Normalized RMSE (NRMSE) of 9043.4. In the case of predicting PMax24hr, the model achieves an R² of 0.9450, an MAE of 3.0358, an RMSE of 6.3950, and an NRMSE of 6.0559. Sensitivity analysis with the XGBoost model showed that the input parameters average humidity and evaporation had the most and least effect on monthly precipitation and monthly maximum 24-hour precipitation modeling, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of monthly precipitation and maximum 24 h precipitation using Random Forest, Decision Tree and XGBoost models

  • 期刊论文

  • 发表日期：

    2026年1月6日

查看全文
  • Mahdi Kashefi,
  • Hojat Karami,
  • Mehdi Niksefat,
  • Hamidreza Ghazvinian

摘要

Precipitation prediction is crucial for managing floods and issuing timely warnings. As a vital part of the water cycle, precipitation provides valuable insights into regional climate patterns. In this research, Decision Tree, Random Forest, and XGBoost algorithms were employed to model monthly precipitation and monthly maximum 24-hour precipitation data for the city of Lamerd, located in Fars Province, Iran. The average temperature, absolute minimum temperature, absolute maximum temperature, average minimum temperature, average maximum temperature, average humidity, maximum humidity, minimum humidity, and evaporation were considered as input data. The data were studied from 1996 to 2024. 70% of the data was used for training and 30% for testing. The results indicated that all three models exhibited satisfactory performance. The XGBoost model outperformed the other two models in estimating monthly precipitation and monthly maximum 24-hour precipitation. The performance metrics of the XGB model for the test dataset are summarized below: For predicting Pmonth, the R² value is 0.9337, with a Mean Absolute Error (MAE) of 6.5032, a Root Mean Squared Error (RMSE) of 11.2456, and a Normalized RMSE (NRMSE) of 9043.4. In the case of predicting PMax24hr, the model achieves an R² of 0.9450, an MAE of 3.0358, an RMSE of 6.3950, and an NRMSE of 6.0559. Sensitivity analysis with the XGBoost model showed that the input parameters average humidity and evaporation had the most and least effect on monthly precipitation and monthly maximum 24-hour precipitation modeling, respectively.