Scholarlink

<p>A better understanding of regional water resources may help improve management of these resources. Snow is a resource that cannot be accurately measured. Research in this field has been increasing in recent years with the advancements in machine learning methods and satellite imagery. This research measured Snow Water Equivalent (SWE) using eight machine learning approaches; Furthermore, the input data used in this study are monthly. Ten meteorological parameters related to snow, and also two parameters for albedo and snow cover fraction obtained from satellite imagery from 1981 to 2022, were provided as inputs to the models. The best input combination was determined through the gamma test, and the impact percentage of each input on the result was evaluated using sensitivity analysis. The results showed that the XGBoost (XGB) method with a Nash-Sutcliffe Efficiency (NSE) coefficient of 0.99 is the best, and Ridge regression with an NSE coefficient value of 0.6486 has the worst result. The dataset was divided into 80% for training and 20% for testing. The trend of snow water equivalent (SWE) during this period was evaluated utilizing the Mann-Kendall test and Sen’s slope analysis. The findings indicate a decreasing trend in SWE. The utilization of current machine learning methods helps a more precise estimation of the water volume in the region, providing improved decision-making for the management of regional water resources.</p> Graphical Abstract <p></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

Uncertainty Analysis of Machine Learning Methods To Estimate Snow Water Equivalent Using Meteorological and Remote Sensing Data

  • 期刊论文

  • 发表日期：

    2025年4月5日

查看全文
  • Mohammad Reza Goodarzi,
  • Ali Barzkar,
  • Maryam Sabaghzadeh,
  • Meysam ghanbari,
  • Nasrin Fathollahzadeh Attar

摘要

A better understanding of regional water resources may help improve management of these resources. Snow is a resource that cannot be accurately measured. Research in this field has been increasing in recent years with the advancements in machine learning methods and satellite imagery. This research measured Snow Water Equivalent (SWE) using eight machine learning approaches; Furthermore, the input data used in this study are monthly. Ten meteorological parameters related to snow, and also two parameters for albedo and snow cover fraction obtained from satellite imagery from 1981 to 2022, were provided as inputs to the models. The best input combination was determined through the gamma test, and the impact percentage of each input on the result was evaluated using sensitivity analysis. The results showed that the XGBoost (XGB) method with a Nash-Sutcliffe Efficiency (NSE) coefficient of 0.99 is the best, and Ridge regression with an NSE coefficient value of 0.6486 has the worst result. The dataset was divided into 80% for training and 20% for testing. The trend of snow water equivalent (SWE) during this period was evaluated utilizing the Mann-Kendall test and Sen’s slope analysis. The findings indicate a decreasing trend in SWE. The utilization of current machine learning methods helps a more precise estimation of the water volume in the region, providing improved decision-making for the management of regional water resources.

Graphical Abstract