Singha et al. (2025) Machine learning-based mapping of fog water harvesting potential in Pithoragarh, Uttarakhand: Evaluating climate scenarios and geospatial influences

Identification

• Journal: Physics and Chemistry of the Earth Parts A/B/C
• Year: 2025
• Date: 2025-10-17
• Authors: Chiranjit Singha, Kishore Chandra Swain, Biswajeet Pradhan, Armin Moghimi, Babak Ranjgar, Shahid Gulzar, HemangM Shah
• DOI: 10.1016/j.pce.2025.104138

Research Groups

• Department of Agricultural Engineering, Institute of Agriculture, Visva-Bharati (A Central University), Sriniketan, Birbhum, West Bengal, India
• Centre for Advanced Modelling and Geospatial Information Systems (CAMGIS), School of Civil and Environmental Engineering, Faculty of Engineering and IT, University of Technology Sydney, Sydney, NSW, Australia
• Department of Photogrammetry and Remote Sensing, Geomatics Engineering Faculty, K. N. Toosi University of Technology, Tehran, Iran
• Department of Energy, Politecnico di Milano, Milan, Italy
• Department of Geoinformatics, International Institute of Geospatial Science and Technology (IIGST), Kolkata, India

Short Summary

This study maps current and future fog water harvesting (FWH) potential in Pithoragarh, Uttarakhand, using five machine learning models and 23 geo-environmental variables under CMIP6 climate scenarios, identifying significant areas with very high potential (up to 44.8% currently, and approximately 22% in future scenarios) to address water scarcity. The research provides a foundation for mitigating water scarcity and contributing to water security in the eastern Himalayas, aligning with Sustainable Development Goal 6 (SDG 6).

Objective

• To map current and future fog water harvesting (FWH) potential in the Pithoragarh district, Uttarakhand, by employing advanced machine learning models with geo-environmental variables and evaluating climate change scenarios (SSP2-4.5 and SSP5-8.5) for 2025-2055.

Study Configuration

• Spatial Scale: Pithoragarh district, Uttarakhand, India (80°–81°E, 29°4′–30°3′N), covering an area of approximately 489.26 km². Data were resampled to a spatial resolution of 30 m × 30 m.
• Temporal Scale:
  • Current FWH mapping: Data from various periods, including TerraClimate (1960–2023), NOAA/GFS0P25 (2015–2023), Landsat 8 (2021–2022), and Sentinel-2B (2021–2022). Ground-based fog sampling was conducted during peak fog periods in December and January of 2021 and 2022, with additional surveys in October and November 2022.
  • Future FWH mapping: Projections for the period 2025–2055 using CMIP6 climate models.

Methodology and Data

• Models used:
  • Machine Learning (ML) Models: Gradient Boosting Machine (GBM), AdaBoost (ADB), Model Averaged Neural Network (avNNet), Naive Bayes (NB), Shrinkage Discriminant Analysis (SDA).
  • Feature Selection/Importance: Boruta algorithm, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Gravitational Optimization (GO), Dragonfly Optimization (DFO), Harris Hawk Optimization (HHO), Grey Wolf Optimization (GWO).
  • Statistical Analysis: Ordinary Least Squares (OLS) Regression, Pearson correlation, Variance Inflation Factor (VIF), Mann-Kendall trend test.
  • Climate Models (CMIP6 Ensemble): EC-Earth3, NorESM2-LM, MIROC6.
  • Geospatial Tools: R (version 4.4.0), Python (version 3.10), ArcGIS (version 10.7), Quantum GIS v2.18.16, SAGA GIS.
• Data sources:
  • Satellite/Remote Sensing: TRMM, NOAH FLDAS, SRTM DEM (USGS/SRTMGL1_003), ESA World Cover (Sentinel-1 and Sentinel-2), Landsat 8 (L8), Sentinel-2B, GRACE satellite data (Centre for Space Research (CSR) and Jet Propulsion Laboratory (JPL) products).
  • Observation/Reanalysis: TerraClimate datasets (integrating WorldClim averages with CRU Ts4.0 and JRA55 data), Global Wind Atlas (at 50 m height), NOAA/GFS0P25, OpenStreetMap (OSM), HydroSHEDS, Google Earth.
  • Ground-based: Fog sampling at 200 georeferenced locations (100 foggy, 100 non-foggy) with on-site measurements of air temperature, humidity, and visibility.

Main Results

• Current FWH Potential: Machine learning models predicted high fog water potentiality classes in 34.16 % to 44.79 % of the study area. The ADB and GBM models demonstrated the highest predictive performance (AUC = 0.999), with ADB achieving an accuracy of 0.983, Kappa of 0.967, and F1 score of 0.983.
• Key Influencing Factors: Elevation (mean importance: 35.31), wind exposition index (14.92), wind speed (13.24), mean power density (10.78), and maximum temperature (7.70) were identified as the most influential factors for fog occurrence.
• Future FWH Potential (2025-2055):
  • Under the SSP2-4.5 scenario, approximately 22.81 % of the study area consistently exhibits very high potential for FWH, with avNNet predicting the largest "Very High" class coverage (49.24 %).
  • Under the SSP5-8.5 scenario, approximately 21.52 % of the study area consistently exhibits very high potential for FWH, with avNNet (51.27 %) and GBM (51.32 %) predicting the largest "Very High" class coverage.
  • Future FWH hotspots are consistently located in the north-central mountainous regions of the study area.
• Water Stress: GRACE satellite data (2002–2023) showed a significant decline in equivalent water thickness (EWT), decreasing from 0.421 m to −1.214 m, with annual losses ranging from −0.057 m to −0.068 m, indicating increasing water stress in the region.

Contributions

• Provides the first comprehensive machine learning-based assessment of fog water harvesting potential in the Pithoragarh district, Uttarakhand, integrating 23 geo-environmental variables.
• Evaluates future FWH potential under CMIP6 climate change scenarios (SSP2-4.5 and SSP5-8.5) for the period 2025-2055 using a multi-model ensemble approach.
• Identifies dominant topographic and atmospheric factors influencing FWH suitability through advanced statistical and machine learning-based feature selection techniques.
• Demonstrates the superior performance of ensemble machine learning models (AdaBoost and Gradient Boosting Machine) for high-precision FWH mapping in complex terrains.
• Lays a foundation for addressing environmental concerns related to FWH, representing a significant step towards mitigating water scarcity and contributing to water security in the eastern Himalayas, in line with Sustainable Development Goal 6 (SDG 6).

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Citation

@article{Singha2025Machine,
  author = {Singha, Chiranjit and Swain, Kishore Chandra and Pradhan, Biswajeet and Moghimi, Armin and Ranjgar, Babak and Gulzar, Shahid and Shah, HemangM},
  title = {Machine learning-based mapping of fog water harvesting potential in Pithoragarh, Uttarakhand: Evaluating climate scenarios and geospatial influences},
  journal = {Physics and Chemistry of the Earth Parts A/B/C},
  year = {2025},
  doi = {10.1016/j.pce.2025.104138},
  url = {https://doi.org/10.1016/j.pce.2025.104138}
}