<p>Anthropogenic activities are altering the hydrological cycle, accelerating drought phenomena. The duration, intensity, and frequency of this climate hazard have become more severe, leading to prolonged drought conditions over the past decades. Global climate models (GCMs) with varying regional capabilities provide detailed insights about future trends in this phenomenon. Under various Shared Socioeconomic Pathways (SSPs), the development of a more robust framework is necessary for accurate future drought assessments through effective selection and aggregation of GCMs. This study provides a more robust framework by minimizing biases in GCM evaluation under diverse conditions, thereby improving the reliability of multi-model ensembles (MMEs). We applied the proposed framework to precipitation simulations from 22 GCMs from CMIP6, covering 28 grid points for the historical period (1950–2014) and the three Shared Socioeconomic Pathways (SSPs) (2015–2100). The Garson algorithm, double input symmetrical relevance and <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="704_2025_5443_Article_IEq1.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="33" /> </InlineMediaObject> <EquationSource Format="TEX">$last$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="italic">last</mi> </mrow> </math></EquationSource> </InlineEquation> relative importance metric (RIM) approaches are used in the evaluation phase to ensure spatial and methodological consistency, along with a comprehensive rating metric (CRM) for unified ranking at each grid point. Eleven of the most consistent GCMs across metrics and grid points, including CNRM-CM6-1, AWI-CM-1–1-MR, and BCC-CSM2-MR, have been selected based on their net strength scores. These top-performing GCMs are then aggregated using various geometric and regression-based approaches. The Kling-Gupta Efficiency with knowable moments (KGEkm) validates the superiority of the Lp-norm ensemble (LpNE) in capturing historical precipitation patterns, recommending it for future trend projections. For drought categorization under the novel index Regional Gaussian Drought Index (RGDI), K-Component Gaussian Mixture Models (K-CGMM) is found to be optimal with the minimal Bayesian Information Criterion (BIC) value across the SSPs and various time scales. The long-term drought trend under the steady-state probabilities of the Markov chain reveals severe dry periods, particularly under SSP5-8.5, by the late twenty-first century. These results highlight the dire need for precautionary and mitigation strategies to address the extreme effects of climate change in the study region.</p>

Enhancing drought projection reliability: a framework for optimal GCM selection, aggregation, and trend analysis under shared socioeconomic pathways

期刊论文
发表日期：

2025年3月28日

查看全文

Muhammad Shakeel,
Hussnain Abbas,
Zulfiqar Ali,
Muhammad Mohsin

摘要

Anthropogenic activities are altering the hydrological cycle, accelerating drought phenomena. The duration, intensity, and frequency of this climate hazard have become more severe, leading to prolonged drought conditions over the past decades. Global climate models (GCMs) with varying regional capabilities provide detailed insights about future trends in this phenomenon. Under various Shared Socioeconomic Pathways (SSPs), the development of a more robust framework is necessary for accurate future drought assessments through effective selection and aggregation of GCMs. This study provides a more robust framework by minimizing biases in GCM evaluation under diverse conditions, thereby improving the reliability of multi-model ensembles (MMEs). We applied the proposed framework to precipitation simulations from 22 GCMs from CMIP6, covering 28 grid points for the historical period (1950–2014) and the three Shared Socioeconomic Pathways (SSPs) (2015–2100). The Garson algorithm, double input symmetrical relevance and $last$ $last$ relative importance metric (RIM) approaches are used in the evaluation phase to ensure spatial and methodological consistency, along with a comprehensive rating metric (CRM) for unified ranking at each grid point. Eleven of the most consistent GCMs across metrics and grid points, including CNRM-CM6-1, AWI-CM-1–1-MR, and BCC-CSM2-MR, have been selected based on their net strength scores. These top-performing GCMs are then aggregated using various geometric and regression-based approaches. The Kling-Gupta Efficiency with knowable moments (KGEkm) validates the superiority of the Lp-norm ensemble (LpNE) in capturing historical precipitation patterns, recommending it for future trend projections. For drought categorization under the novel index Regional Gaussian Drought Index (RGDI), K-Component Gaussian Mixture Models (K-CGMM) is found to be optimal with the minimal Bayesian Information Criterion (BIC) value across the SSPs and various time scales. The long-term drought trend under the steady-state probabilities of the Markov chain reveals severe dry periods, particularly under SSP5-8.5, by the late twenty-first century. These results highlight the dire need for precautionary and mitigation strategies to address the extreme effects of climate change in the study region.