Scholarlink

<p>Rainfall is recognized as one of the most important components of the water cycle, and their forecasting plays a vital role in catchment management, including drought and flood warnings way before their occurrence. The current study proposes novel rainfall forecasting approaches by coupling ground-observed data obtained from the Indian Meteorological Department (IMD) with Coupled Model Intercomparison Project Phase 6 (CMIP6) simulation data and three data-driven modeling techniques, namely multivariate linear regression (MVLR), artificial neural network (ANN), and support vector machine (SVM). To illustrate the methodologies proposed in this study, the CMIP6 and IMD data on&#xa0;pressure at mean sea level, wind speed, minimum and maximum temperatures, average temperature, and monthly precipitation derived for Chhattisgarh State in India have been utilized. Three scenarios are explored depending on the combination of data types (i.e., CMIP6, IMD, or both) utilized as input variables in data-driven models. The performance of various models in forecasting rainfall is assessed using standard statistical measures. The results obtained from this study suggest that out of all scenarios, the ANN model that incorporated both ground-observed IMD and CMIP6 simulation data as input performed the best compared to the other scenarios in all three techniques with the lowest values of AARE (18.1) and RMSE (0.066&#xa0;mm) and highest values of R (0.912), and all threshold values during testing. The SVM technique was also observed to outperform ANN and MVLR in forecasting rainfall for all scenarios explored in this study. The study presented here suggests the use of ground-observed data coupled with CMIP simulation outputs in forecasting rainfall to improve the accuracy of the forecasting models which can be useful in further strengthening of early warning systems and related water resources management.</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

Data-driven techniques in rainfall forecasting using CMIP6 simulation outputs and ground-observed data

  • 期刊论文

  • 发表日期：

    2025年6月17日

查看全文
  • Varun Dubey,
  • Vikas Kumar Vidyarthi

摘要

Rainfall is recognized as one of the most important components of the water cycle, and their forecasting plays a vital role in catchment management, including drought and flood warnings way before their occurrence. The current study proposes novel rainfall forecasting approaches by coupling ground-observed data obtained from the Indian Meteorological Department (IMD) with Coupled Model Intercomparison Project Phase 6 (CMIP6) simulation data and three data-driven modeling techniques, namely multivariate linear regression (MVLR), artificial neural network (ANN), and support vector machine (SVM). To illustrate the methodologies proposed in this study, the CMIP6 and IMD data on pressure at mean sea level, wind speed, minimum and maximum temperatures, average temperature, and monthly precipitation derived for Chhattisgarh State in India have been utilized. Three scenarios are explored depending on the combination of data types (i.e., CMIP6, IMD, or both) utilized as input variables in data-driven models. The performance of various models in forecasting rainfall is assessed using standard statistical measures. The results obtained from this study suggest that out of all scenarios, the ANN model that incorporated both ground-observed IMD and CMIP6 simulation data as input performed the best compared to the other scenarios in all three techniques with the lowest values of AARE (18.1) and RMSE (0.066 mm) and highest values of R (0.912), and all threshold values during testing. The SVM technique was also observed to outperform ANN and MVLR in forecasting rainfall for all scenarios explored in this study. The study presented here suggests the use of ground-observed data coupled with CMIP simulation outputs in forecasting rainfall to improve the accuracy of the forecasting models which can be useful in further strengthening of early warning systems and related water resources management.