Scholarlink

<p>Meteorological data record is essential for developing various hydrological models. The accuracy of these models is affected by the continuously available time series data of various climatic variables. These time series climatic data sometimes contain missing values for various reasons such as faults in the measuring devices, loss of records and other anthropological activities. Various methods are proposed previously to fill these missing values in the climatic data. In this paper, total fifteen conventional approaches used previously for filling missing data have been assessed and two popular machine learning (ML) methods, commonly known as artificial neural network (ANN) and long short-term memory (LSTM), have been proposed for this purpose. The precipitation data of different rain gauge stations, spatial distances among stations and the actual locations in terms of latitude and longitude extracted from the rain gauge near Raipur rain gauge station which is assumed to have missing values are utilized to illustrate the methodology assessed and proposed in this study. The results show that the ANN technique performed better than all the fifteen conventional methods assessed, with the MAE, R and RMSE values as 1.82&#xa0;mm, 0.970 and 3.78&#xa0;mm, respectively, for 6&#xa0;months data, 2.62&#xa0;mm, 0.976 and 4.22&#xa0;mm, respectively, for 1&#xa0;month data, and 2.67&#xa0;mm, 0.990 and 3.73&#xa0;mm, respectively, for 15&#xa0;days' data. The efficiency of the model increased as the number of missing data decreased from 6&#xa0;months to 15&#xa0;days. The ANN model performance is found to be better than LSTM model also in filling missing precipitation for 6&#xa0;months and 1&#xa0;month and comparable for 15&#xa0;days data. The results are based on the filling of precipitation data for a particular area; however, it can be extended to fill the missing data for other climatic variables for different regions.</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

Assessment of performance of conventional and machine learning methods for estimating missing precipitation data

  • 期刊论文

  • 发表日期：

    2025年11月7日

查看全文
  • Akhilesh Poyam,
  • Vikas Kumar Vidyarthi,
  • Manikant Verma

摘要

Meteorological data record is essential for developing various hydrological models. The accuracy of these models is affected by the continuously available time series data of various climatic variables. These time series climatic data sometimes contain missing values for various reasons such as faults in the measuring devices, loss of records and other anthropological activities. Various methods are proposed previously to fill these missing values in the climatic data. In this paper, total fifteen conventional approaches used previously for filling missing data have been assessed and two popular machine learning (ML) methods, commonly known as artificial neural network (ANN) and long short-term memory (LSTM), have been proposed for this purpose. The precipitation data of different rain gauge stations, spatial distances among stations and the actual locations in terms of latitude and longitude extracted from the rain gauge near Raipur rain gauge station which is assumed to have missing values are utilized to illustrate the methodology assessed and proposed in this study. The results show that the ANN technique performed better than all the fifteen conventional methods assessed, with the MAE, R and RMSE values as 1.82 mm, 0.970 and 3.78 mm, respectively, for 6 months data, 2.62 mm, 0.976 and 4.22 mm, respectively, for 1 month data, and 2.67 mm, 0.990 and 3.73 mm, respectively, for 15 days' data. The efficiency of the model increased as the number of missing data decreased from 6 months to 15 days. The ANN model performance is found to be better than LSTM model also in filling missing precipitation for 6 months and 1 month and comparable for 15 days data. The results are based on the filling of precipitation data for a particular area; however, it can be extended to fill the missing data for other climatic variables for different regions.