Scholarlink

<p>Drought is a complex natural hazard, and its prolonged events have severe socio-economic and environmental impacts worldwide. The study aims to introduce a comprehensive framework for efficient and accurate drought assessment based on the mutual information of spatiotemporal precipitation patterns. We propose the Regionally Integrated Standardized Drought Index (RISDI), which integrates advanced statistical methods to deliver accurate cluster analysis and cluster-specific drought assessments. This hybrid approach is based on three phases, each of which significantly contributes to the overall procedure. Overall, the proposed framework integrates Model-Based Clustering (MBC) with the Bayesian Information Criterion (BIC), Principal Component Analysis (PCA), and the K-Component Gaussian Mixture Distribution (K-CGMD) to estimate the Regional Integrated Standardized Drought Index (RISDI) for each cluster. To validate our proposed framework, we use monthly-averaged precipitation data for all Districts of Punjab, Pakistan, for the period 1981 to 2021. A comparative assessment is conducted to assess the consistency of the proposed index (i.e., RISDI) across the respective clusters. Overall, nine regional clusters were identified as optimal in the model-based clustering (MBC) approach, corresponding to a Bayesian Information Criterion (BIC) value of − 181,796. After the aggregation of precipitation data at each cluster, it has been observed that Clusters 1 to 4 exhibit lower mean, median, and lower-quartile precipitation with higher kurtosis, indicating generally dry conditions punctuated by occasional heavy rainfall, whereas Clusters 5 to 9 show higher central and upper quantiles, reflecting wetter and more consistent rainfall regimes. The weighted mean aggregation preserves these contrasts by scaling each location’s contribution according to its relative importance. During the calculation of RISDI for each cluster, the low Bayesian Information Criterion (BIC) values, QQ plots, and histograms obtained in the standardization procedure indicate that the 12-component Gaussian mixture distribution provides the best fit and accurately captures precipitation variability and density patterns across all nine clusters. For each cluster, the proposed index (RISDI) has high correlations, comparable standard deviations, and low-centered RMSE values across all SPIs of districts within the respective cluster. The findings indicate that RISDI serves as a well-tuned tool that reduces the complexity of precipitation large datasets while remaining sensitive to local changes. It efficiently assesses multi-regional drought severity and its pattern. It is concluded that the proposed aggregated approach, instead of relying on a single key location for the entire cluster, improves the accuracy and reliability of the drought index for future drought prediction. The proposed framework can be considered a valuable approach for accurate drought monitoring and helps produce efficient drought mitigation policies.</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

A Novel Integrated Standardized Index for Drought Assessment of Homogeneous Regions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月30日

查看全文
  • RegionsAsad Ellahi,
  • Hamza Amin,
  • Ijaz Hussain,
  • Jianyi Lin,
  • Hanen Louati,
  • Mhassen. E. E. Dalam,
  • Mohammed M. A. Almazah

摘要

Drought is a complex natural hazard, and its prolonged events have severe socio-economic and environmental impacts worldwide. The study aims to introduce a comprehensive framework for efficient and accurate drought assessment based on the mutual information of spatiotemporal precipitation patterns. We propose the Regionally Integrated Standardized Drought Index (RISDI), which integrates advanced statistical methods to deliver accurate cluster analysis and cluster-specific drought assessments. This hybrid approach is based on three phases, each of which significantly contributes to the overall procedure. Overall, the proposed framework integrates Model-Based Clustering (MBC) with the Bayesian Information Criterion (BIC), Principal Component Analysis (PCA), and the K-Component Gaussian Mixture Distribution (K-CGMD) to estimate the Regional Integrated Standardized Drought Index (RISDI) for each cluster. To validate our proposed framework, we use monthly-averaged precipitation data for all Districts of Punjab, Pakistan, for the period 1981 to 2021. A comparative assessment is conducted to assess the consistency of the proposed index (i.e., RISDI) across the respective clusters. Overall, nine regional clusters were identified as optimal in the model-based clustering (MBC) approach, corresponding to a Bayesian Information Criterion (BIC) value of − 181,796. After the aggregation of precipitation data at each cluster, it has been observed that Clusters 1 to 4 exhibit lower mean, median, and lower-quartile precipitation with higher kurtosis, indicating generally dry conditions punctuated by occasional heavy rainfall, whereas Clusters 5 to 9 show higher central and upper quantiles, reflecting wetter and more consistent rainfall regimes. The weighted mean aggregation preserves these contrasts by scaling each location’s contribution according to its relative importance. During the calculation of RISDI for each cluster, the low Bayesian Information Criterion (BIC) values, QQ plots, and histograms obtained in the standardization procedure indicate that the 12-component Gaussian mixture distribution provides the best fit and accurately captures precipitation variability and density patterns across all nine clusters. For each cluster, the proposed index (RISDI) has high correlations, comparable standard deviations, and low-centered RMSE values across all SPIs of districts within the respective cluster. The findings indicate that RISDI serves as a well-tuned tool that reduces the complexity of precipitation large datasets while remaining sensitive to local changes. It efficiently assesses multi-regional drought severity and its pattern. It is concluded that the proposed aggregated approach, instead of relying on a single key location for the entire cluster, improves the accuracy and reliability of the drought index for future drought prediction. The proposed framework can be considered a valuable approach for accurate drought monitoring and helps produce efficient drought mitigation policies.