Scholarlink

<p>Accurate precipitation representation in climate models is crucial for reliable climate projections, which underpin impact assessments in various sectors. Despite advancements in climate modeling, quantifying and addressing precipitation biases remains a significant challenge. This study investigates systematic biases in precipitation simulated by Global Climate Models (GCMs) and Regional Climate Models (RCMs), with the goal of enhancing model reliability and reducing the need for post-processing adjustments like bias correction. While bias correction methods are commonly used in impact studies, they serve as post-hoc fixes that do not address underlying model deficiencies. To improve next-generation models, a comprehensive assessment of existing biases is necessary. We evaluated 55 high-resolution RCMs from EURO-CORDEX driven by CMIP5 GCM simulations, and 93 GCMs from CMIP3, CMIP5, and CMIP6 over southern Italy. The analysis focused on both mean and extreme precipitation indices to assess spatial and temporal patterns. Our results indicate a consistent dry bias in GCMs, largely attributable to their coarse spatial resolution, which limits the accurate representation of localized atmospheric processes and orographic influences. In contrast, most RCM simulations display a wet bias, likely resulting from their finer spatial resolution. This enhanced resolution enables improved representation of local topography and convective processes but can also lead to an overestimation of precipitation. Overall, GCMs exhibit higher total errors, exceeding 50% in both spatial and temporal analyses. RCMs, while prone to a wet bias, demonstrate improved spatial correlation with observed precipitation patterns. Among the GCMs, EC-Earth3-CC (CMIP6) showed the best performance for spatial patterns of annual total precipitation, with a total error of 63%. For annual maxima, HadGEM2-CC (CMIP5) was the top-performing GCM, with a total error of 83%. In terms of interannual temporal analysis, BCCR-BCM-PT3 (CMIP3) achieved the lowest error (80%) for annual total precipitation, while NESM3-NUIST (CMIP6) performed best for annual maxima, with a total error of 75%. Despite their advantages in spatial representation, RCMs continue to exhibit significant uncertainty in the simulation of extreme precipitation events, primarily due to limitations related to boundary conditions and model resolution. Using the Aras diagram as an objective evaluation method, we quantitatively assessed these biases and discussed the implications for model improvement, particularly with respect to the refinement of sub-grid parameterizations (e.g., for convection and cloud processes), with the long-term goal of reducing reliance on bias correction techniques and foster the development of more accurate climate models for robust impact assessments.</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

Dry GCMs vs. wet RCMs: precipitation biases over Southern Italy

  • 期刊论文

  • 发表日期：

    2025年6月5日

查看全文
  • Aras Izzaddin,
  • Andreas Langousis,
  • Marwah Yaseen,
  • Vito Iacobellis

摘要

Accurate precipitation representation in climate models is crucial for reliable climate projections, which underpin impact assessments in various sectors. Despite advancements in climate modeling, quantifying and addressing precipitation biases remains a significant challenge. This study investigates systematic biases in precipitation simulated by Global Climate Models (GCMs) and Regional Climate Models (RCMs), with the goal of enhancing model reliability and reducing the need for post-processing adjustments like bias correction. While bias correction methods are commonly used in impact studies, they serve as post-hoc fixes that do not address underlying model deficiencies. To improve next-generation models, a comprehensive assessment of existing biases is necessary. We evaluated 55 high-resolution RCMs from EURO-CORDEX driven by CMIP5 GCM simulations, and 93 GCMs from CMIP3, CMIP5, and CMIP6 over southern Italy. The analysis focused on both mean and extreme precipitation indices to assess spatial and temporal patterns. Our results indicate a consistent dry bias in GCMs, largely attributable to their coarse spatial resolution, which limits the accurate representation of localized atmospheric processes and orographic influences. In contrast, most RCM simulations display a wet bias, likely resulting from their finer spatial resolution. This enhanced resolution enables improved representation of local topography and convective processes but can also lead to an overestimation of precipitation. Overall, GCMs exhibit higher total errors, exceeding 50% in both spatial and temporal analyses. RCMs, while prone to a wet bias, demonstrate improved spatial correlation with observed precipitation patterns. Among the GCMs, EC-Earth3-CC (CMIP6) showed the best performance for spatial patterns of annual total precipitation, with a total error of 63%. For annual maxima, HadGEM2-CC (CMIP5) was the top-performing GCM, with a total error of 83%. In terms of interannual temporal analysis, BCCR-BCM-PT3 (CMIP3) achieved the lowest error (80%) for annual total precipitation, while NESM3-NUIST (CMIP6) performed best for annual maxima, with a total error of 75%. Despite their advantages in spatial representation, RCMs continue to exhibit significant uncertainty in the simulation of extreme precipitation events, primarily due to limitations related to boundary conditions and model resolution. Using the Aras diagram as an objective evaluation method, we quantitatively assessed these biases and discussed the implications for model improvement, particularly with respect to the refinement of sub-grid parameterizations (e.g., for convection and cloud processes), with the long-term goal of reducing reliance on bias correction techniques and foster the development of more accurate climate models for robust impact assessments.