Scholarlink

<p>This study assessed the metrological drought patterns in Pakistan using the Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI) during 1960–2023. Choosing data from 44 meteorological stations and five timescales (1, 3, 6, 9, and 12-month), this study analyzed the spatial and temporal trends using the Mann–Kendall test and Sen’s slope estimator. Time series analysis revealed that the impact of temperature on drought severity becomes more pronounced in recent decades, as evidenced by the divergence between SPI and SPEI. Further, the study applied Random Forest model to evaluate the predictive importance of each index. Results showed that SPI-12 and SPEI-12 consistently emerge as the most informative indicators across diverse climatic zones, while shorter timescales contribute less to model accuracy. The average R² values for the testing dataset show a consistent trend for SPI and SPEI. For SPI, R² value decreases from 0.86 at SPI-1 to 0.61 at SPI-12. For SPEI, R² also declines from 0.87 at SPEI-1 to 0.66 at SPEI-12. This is further evidenced by the increasing average Root Mean Square Error values, which rise from 0.38 to 0.78 for SPI and from 0.37 to 0.75 for SPEI over the same timescales. Spatial patterns of feature importance vary by region, with SPEI gaining significance in areas more sensitive to evapotranspiration. These findings offer valuable insight into drought monitoring and highlight the relevance of long-term indices for improved forecasting and climate adaptation strategies in Pakistan.</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

Predicting long-term meteorological drought using random forest and multi-scale drought indices

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Riaz Ul Hissan,
  • Nusrat Parveen

摘要

This study assessed the metrological drought patterns in Pakistan using the Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI) during 1960–2023. Choosing data from 44 meteorological stations and five timescales (1, 3, 6, 9, and 12-month), this study analyzed the spatial and temporal trends using the Mann–Kendall test and Sen’s slope estimator. Time series analysis revealed that the impact of temperature on drought severity becomes more pronounced in recent decades, as evidenced by the divergence between SPI and SPEI. Further, the study applied Random Forest model to evaluate the predictive importance of each index. Results showed that SPI-12 and SPEI-12 consistently emerge as the most informative indicators across diverse climatic zones, while shorter timescales contribute less to model accuracy. The average R² values for the testing dataset show a consistent trend for SPI and SPEI. For SPI, R² value decreases from 0.86 at SPI-1 to 0.61 at SPI-12. For SPEI, R² also declines from 0.87 at SPEI-1 to 0.66 at SPEI-12. This is further evidenced by the increasing average Root Mean Square Error values, which rise from 0.38 to 0.78 for SPI and from 0.37 to 0.75 for SPEI over the same timescales. Spatial patterns of feature importance vary by region, with SPEI gaining significance in areas more sensitive to evapotranspiration. These findings offer valuable insight into drought monitoring and highlight the relevance of long-term indices for improved forecasting and climate adaptation strategies in Pakistan.