Mazhar et al. (2025) Enhancing aridity index assessment in Pakistan's dryland ecosystems: A machine learning approach integrating remote sensing and seasonal lag effects

Identification

• Journal: Physics and Chemistry of the Earth Parts A/B/C
• Year: 2025
• Date: 2025-10-04
• Authors: Nausheen Mazhar, Asad K. Ghalib, Issam Malki, Noreena, Sana Arshad
• DOI: 10.1016/j.pce.2025.104135

Research Groups

• Lahore College for Women University, Lahore, Pakistan
• Liverpool Hope Business School, Liverpool Hope University, UK
• School of Finance and Accounting, Westminster Business School, University of Westminster, UK
• Department of Geography & Geoinformatics, The Islamia University of Bahawalpur, Bahawalpur, Pakistan

Short Summary

This study evaluated the aridity index (AI) and Standardized Precipitation Index (SPI-3) in Pakistan's dryland ecosystems from 1990 to 2023 using machine learning and remote sensing, revealing that Gradient Boosting Regression with a three-month lag accurately predicted AI and highlighted the significance of seasonal effects and biophysical indicators for regional water management.

Objective

• To evaluate the aridity index (AI) and Standardized Precipitation Index at a three-month scale (SPI-3) across three arid stations of Pakistan from 1990 to 2023, integrating remote sensing and seasonal lag effects using machine learning for enhanced prediction.

Study Configuration

• Spatial Scale: Three arid stations in Pakistan.
• Temporal Scale: 1990 to 2023 (34 years); analysis of 3-month seasonal lag effects.

Methodology and Data

• Models used:
  • Machine Learning: Gradient Boosting Regression, Random Forest (and two other unspecified well-optimized models).
  • Statistical: Mann-Kendall trend analysis, Sen’s slope analysis, Cross-correlation.
  • Time Series: Seasonally decomposed time series.
• Data sources:
  • Remote sensing indices (seven used as covariates, including NDVI, NDWI, Dry Barren Soil Index (DBSI)).
  • Standardized Precipitation Index at a three-month scale (SPI-3).
  • Mean temperature.

Main Results

• Mann-Kendall and Sen’s slope analysis showed a significant (p < 0.001) increasing trend of AI and SPI-3 values, indicating comparatively lower aridity in recent years.
• NDVI also showed a consistent increasing trend with Sen’s slope ranging from 0.0002 to 0.003.
• Cross-correlation revealed a seasonal effect of biophysical indicators on AI, with positive correlations of r = 0.4 with NDVI and r = 0.6 with NDWI at lag 0, suggesting a late lag effect.
• Machine learning prediction of AI with a three-month lag demonstrated:
  • Gradient Boosting Regression outperformed other models with a coefficient of determination (R²) = 0.806 and a root mean square error (RMSE) = 0.076.
  • Random Forest followed with R² = 0.732 and RMSE = 0.089.
• Dry Barren Soil Index (DBSI), NDWI, and SPI-3 were identified as features with high importance in the best-performing model.

Contributions

• Highlights the significance of temporal and seasonal relationships between aridity and biophysical indicators in dryland ecosystems.
• Provides an enhanced approach for aridity index assessment by integrating machine learning, remote sensing, and seasonal lag effects.
• Offers insights to inform region-specific land and water resource management policies aimed at mitigating hydroclimatic extremes.

Funding

• Not specified in the provided text.

Citation

@article{Mazhar2025Enhancing,
  author = {Mazhar, Nausheen and Ghalib, Asad K. and Malki, Issam and Noreena and Arshad, Sana},
  title = {Enhancing aridity index assessment in Pakistan's dryland ecosystems: A machine learning approach integrating remote sensing and seasonal lag effects},
  journal = {Physics and Chemistry of the Earth Parts A/B/C},
  year = {2025},
  doi = {10.1016/j.pce.2025.104135},
  url = {https://doi.org/10.1016/j.pce.2025.104135}
}