Scholarlink

<p>The systematic biases in the modelled precipitation datasets hinder their applications in various hydrological response studies and lead to ambiguous results. These biases are more pronounced in the complex Himalayan topography, mandating robust bias correction methodologies for the terrain. Previous studies often overlooked correcting the extreme part of the precipitation distribution and comparing bias correction methods. This paper presents a comparative analysis of the six conventional bias correction techniques and a novel ensemble method, EQMX-RF, to correct the satellite and reanalysis precipitation products from 2012 to 2023 for the Budhi Gandaki River Basin (BGRB), Nepal, using daily IMERG V07 Final run (GPM) and ERA5-Land (ERA5-L) precipitation datasets and 6 rain-gauge precipitation data by incorporating 10 statistical parameters both quantitative and categorical, quantile-quantile plots, upper-tail, and seasonal scale analysis. Results demonstrate the good performance of EQMX-RF compared to the conventional bias correction methods in the majority of the evaluation metrics. Using the EQMX-RF method, both GPM and ERA5-Land achieved good skill, with Kling-Gupta efficiency (KGE) values around 0.7, correlation coefficients close to 0.8, very low bias (~ 1–2%), and error magnitudes of about 3.2&#xa0;mm/day in RMSE and 1.5&#xa0;mm/day in MAE. Importantly, the method also showed substantial improvement in representing the extreme part of precipitation, reducing errors by more than 40% relative to conventional approaches. Thus, this comparative assessment study, grounded in region-based hydrometeorology, is operationally indispensable and proposes a novel technique that remains consistently reliable across varying seasonal conditions.</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

Improving satellite and reanalysis precipitation estimates in a Himalayan River Basin: a comparative study of bias correction methods with focus on extremes and ensemble method performance

  • 期刊论文

  • 发表日期：

    2025年10月29日

查看全文
  • Harshita Tiwari,
  • Rahul Dev Garg

摘要

The systematic biases in the modelled precipitation datasets hinder their applications in various hydrological response studies and lead to ambiguous results. These biases are more pronounced in the complex Himalayan topography, mandating robust bias correction methodologies for the terrain. Previous studies often overlooked correcting the extreme part of the precipitation distribution and comparing bias correction methods. This paper presents a comparative analysis of the six conventional bias correction techniques and a novel ensemble method, EQMX-RF, to correct the satellite and reanalysis precipitation products from 2012 to 2023 for the Budhi Gandaki River Basin (BGRB), Nepal, using daily IMERG V07 Final run (GPM) and ERA5-Land (ERA5-L) precipitation datasets and 6 rain-gauge precipitation data by incorporating 10 statistical parameters both quantitative and categorical, quantile-quantile plots, upper-tail, and seasonal scale analysis. Results demonstrate the good performance of EQMX-RF compared to the conventional bias correction methods in the majority of the evaluation metrics. Using the EQMX-RF method, both GPM and ERA5-Land achieved good skill, with Kling-Gupta efficiency (KGE) values around 0.7, correlation coefficients close to 0.8, very low bias (~ 1–2%), and error magnitudes of about 3.2 mm/day in RMSE and 1.5 mm/day in MAE. Importantly, the method also showed substantial improvement in representing the extreme part of precipitation, reducing errors by more than 40% relative to conventional approaches. Thus, this comparative assessment study, grounded in region-based hydrometeorology, is operationally indispensable and proposes a novel technique that remains consistently reliable across varying seasonal conditions.