Scholarlink

<p>The sparsity of ground gauges poses a significant challenge for evaluating and merging satellite-based and reanalysis-based precipitation datasets in lake regions. While the standard triple collocation (TC) method offers a solution without access to ground-based observations, it fails to address rain/no-rain classification and its suitability for assessing and merging lake precipitation has not been explored. This study combines categorical triple collocation (CTC) with standard TC to create an integrated framework (CTC-TC) tailored to evaluate and merge global gridded precipitation products (GPPs). We assess the efficacy of CTC-TC using six GPPs (ERA5-Land, SM2RAIN-ASCAT, IMERG-Early, IMERG-Late, GSMaP-MVK, and PERSIANN-CCS) across the five largest freshwater lakes in China. CTC-TC effectively captures the spatial patterns of metrics for all GPPs, and precisely estimates the correlation coefficient and root mean square error for satellite-based datasets apart from SM2RAIN-ASCAT, but overestimates the classification accuracy indicator V for all GPPs. Regarding multi-source fusion, CTC-TC leverages the strengths of individual products of triplets, resulting in significant improvements in the critical success index (CSI) by over 11.9% and the modified Kling-Gupta efficiency (KGE’) by more than 13.3%. Compared to baseline models, including standard TC, simple model averaging, one outlier removal, and Bayesian model averaging, CTC-TC achieves gains in CSI and KGE’ of no less than 24.7% and 3.6%, respectively. In conclusion, the CTC-TC framework offers a thorough evaluation and efficient fusion of GPPs, addressing both categorical and continuous accuracy in data-scarce regions such as lakes.</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

Integrating categorical and standard triple collocation to improve precipitation fusion over the five largest freshwater lakes in China

  • 期刊论文

  • 发表日期：

    2025年11月11日

查看全文
  • Lingjie Li,
  • Guoqiang Tang,
  • Yintang Wang,
  • Rui Gao,
  • Yong Liu,
  • Wenpeng Zhao,
  • Cheng Chen

摘要

The sparsity of ground gauges poses a significant challenge for evaluating and merging satellite-based and reanalysis-based precipitation datasets in lake regions. While the standard triple collocation (TC) method offers a solution without access to ground-based observations, it fails to address rain/no-rain classification and its suitability for assessing and merging lake precipitation has not been explored. This study combines categorical triple collocation (CTC) with standard TC to create an integrated framework (CTC-TC) tailored to evaluate and merge global gridded precipitation products (GPPs). We assess the efficacy of CTC-TC using six GPPs (ERA5-Land, SM2RAIN-ASCAT, IMERG-Early, IMERG-Late, GSMaP-MVK, and PERSIANN-CCS) across the five largest freshwater lakes in China. CTC-TC effectively captures the spatial patterns of metrics for all GPPs, and precisely estimates the correlation coefficient and root mean square error for satellite-based datasets apart from SM2RAIN-ASCAT, but overestimates the classification accuracy indicator V for all GPPs. Regarding multi-source fusion, CTC-TC leverages the strengths of individual products of triplets, resulting in significant improvements in the critical success index (CSI) by over 11.9% and the modified Kling-Gupta efficiency (KGE’) by more than 13.3%. Compared to baseline models, including standard TC, simple model averaging, one outlier removal, and Bayesian model averaging, CTC-TC achieves gains in CSI and KGE’ of no less than 24.7% and 3.6%, respectively. In conclusion, the CTC-TC framework offers a thorough evaluation and efficient fusion of GPPs, addressing both categorical and continuous accuracy in data-scarce regions such as lakes.