Scholarlink

<p>Satellite precipitation products (SPPs) are crucial for applications such as water resource management and flood forecasting. A key challenge remains the accurate evaluation and subsequent correction of SPP biases. This study develops a machine-learning-driven hierarchical framework for both evaluating and correcting SPPs, demonstrated through a case study in Guangxi, China. First, we account for the uncertainty in rain gauge observations by formulating them as interval-valued data. A novel distance-based evaluation index is proposed, and a corresponding bias correction rule is established based on the resulting performance hierarchy. The precipitation observation is modeled as a function of the satellite precipitation at a target grid point, its <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\mathbf{K}\)</EquationSource> </InlineEquation>-nearest neighbors, and the geographic attributes of the ground station. This functional relationship is captured by a specially designed neural network. Second, using daytime IMERG data over Guangxi, we illustrate the proposed evaluation index and correction model. By implementing a hierarchical training strategy for the neural network, the bias in SPPs is significantly reduced, particularly for heavy precipitation events exceeding 50&#xa0;mm. Finally, comparative results and discussions are presented, highlighting the advantages of the proposed model over existing approaches.</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

Evaluating and enhancing the performance of satellite precipitation products by considering uncertainty in rain gauge observations

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • Tai Wei,
  • Xian-Ci Zhong,
  • Yang Gao

摘要

Satellite precipitation products (SPPs) are crucial for applications such as water resource management and flood forecasting. A key challenge remains the accurate evaluation and subsequent correction of SPP biases. This study develops a machine-learning-driven hierarchical framework for both evaluating and correcting SPPs, demonstrated through a case study in Guangxi, China. First, we account for the uncertainty in rain gauge observations by formulating them as interval-valued data. A novel distance-based evaluation index is proposed, and a corresponding bias correction rule is established based on the resulting performance hierarchy. The precipitation observation is modeled as a function of the satellite precipitation at a target grid point, its \(\mathbf{K}\) -nearest neighbors, and the geographic attributes of the ground station. This functional relationship is captured by a specially designed neural network. Second, using daytime IMERG data over Guangxi, we illustrate the proposed evaluation index and correction model. By implementing a hierarchical training strategy for the neural network, the bias in SPPs is significantly reduced, particularly for heavy precipitation events exceeding 50 mm. Finally, comparative results and discussions are presented, highlighting the advantages of the proposed model over existing approaches.