Ali et al. (2026) Regional drought assessment using multi-site probabilistically integrated precipitation by Bayesian network

Identification

• Journal: Natural Hazards
• Year: 2026
• Date: 2026-02-25
• Authors: Farman Ali, J. Han, Dong Su, Zulfiqar Ali, Yang Zhou, Yuefei Huang, Munir Ahmad, Shafeeq Ur Rahman
• DOI: 10.1007/s11069-026-07987-0

Research Groups

• Water Science and Environmental Engineering Research Centre, College of Chemistry and Environmental Engineering, Shenzhen University, Shenzhen, China
• College of Civil and Transportation Engineering, Shenzhen University, Shenzhen, China
• Key Laboratory of Coastal Urban Resilient Infrastructures (Shenzhen University), Ministry of Education, Shenzhen, China
• College of Statistical Sciences, University of the Punjab, Lahore, Pakistan
• Laboratory of Ecological Protection and High Quality Development in the Upper Yellow River, School of Civil Engineering and Water Resources, Qinghai University, Xining, China

Short Summary

This study proposes a Regional Standardized Precipitation Drought Index (RSPDI) for regional drought assessment by integrating precipitation dynamics from multiple stations using Bayesian Network (BN) theory. Application to two regions in Pakistan showed strong agreement between RSPDI and SPI, demonstrating enhanced spatial coherence and robust probabilistic consistency for regional drought monitoring.

Objective

• To develop a novel Regional Standardized Precipitation Drought Index (RSPDI) that quantifies spatiotemporal precipitation dynamics from multiple meteorological stations for regional drought analysis, addressing limitations of traditional single-station indices.

Study Configuration

• Spatial Scale: Two distinct hydro-climatic regions in Pakistan (mountainous north and Punjab plains), encompassing eleven meteorological stations.
• Temporal Scale: Long-term monthly accumulated precipitation data spanning from 1970 to 2017 (48 years).

Methodology and Data

• Models used: Bayesian Network (BN) model for probabilistic integration and representative station selection, K-Component Gaussian Mixture Model (K-CGMM) for precipitation standardization, Regional Standardized Precipitation Drought Index (RSPDI), and Standardized Precipitation Index (SPI) for comparison. Markov Chain Monte Carlo (MCMC) simulations (200,000 iterations) were used for BN model precision.
• Data sources: Monthly accumulated precipitation data collected from the Pakistan Meteorological Department.

Main Results

• The proposed RSPDI demonstrated strong agreement with the traditional SPI, showing correlations ranging from 0.319 to 0.756 for 1-month timescales and 0.181 to 0.824 for 3-month timescales across the studied regions.
• The Astor station in Region 1 exhibited the highest Average Joint Dependency Probability (AJDP) of 0.622, confirming its strong regional representativeness within the Bayesian Network framework.
• The K-Component Gaussian Mixture Model (K-CGMM) provided an excellent fit to the precipitation data for both SPI and RSPDI, validated by Q-Q plots and consistent Bayesian Information Criterion (BIC) values across individual and regional datasets.
• MCMC simulations for the BN model consistently demonstrated convergence and stability, ensuring the robustness of the probabilistic station selection.

Contributions

• Introduces a novel Regional Standardized Precipitation Drought Index (RSPDI) that probabilistically integrates multi-site precipitation data using a Bayesian Network (BN) to enhance spatial coherence and consistency in regional drought assessment.
• Addresses the limitations of traditional station-based drought indices by accounting for spatiotemporal dependencies and the challenges posed by uneven meteorological station distribution.
• Employs a K-Component Gaussian Mixture Model (K-CGMM) for flexible standardization of aggregated precipitation, which more accurately captures the complex, often multimodal, distributional characteristics of regional precipitation.
• Provides a statistically grounded and transferable framework for regional drought monitoring and early-warning systems, offering practical utility for water resource management and policymakers in heterogeneous climatic zones.

Funding

• National Natural Science Foundation of China (No. 52479019)
• Major Basic Research Development Program of the Science and Technology, Qinghai Province (2025-HZ-805)
• Shenzhen Talent Research Start-up Fund

Citation

@article{Ali2026Regional,
  author = {Ali, Farman and Han, J. and Su, Dong and Ali, Zulfiqar and Zhou, Yang and Huang, Yuefei and Ahmad, Munir and Rahman, Shafeeq Ur},
  title = {Regional drought assessment using multi-site probabilistically integrated precipitation by Bayesian network},
  journal = {Natural Hazards},
  year = {2026},
  doi = {10.1007/s11069-026-07987-0},
  url = {https://doi.org/10.1007/s11069-026-07987-0}
}