Omuna et al. (2026) Integrated GIS, remote sensing, hydrological modelling and machine learning assessment of climate-induced Water availability in Kumi District, eastern Uganda
Identification
- Journal: The Egyptian Journal of Remote Sensing and Space Science
- Year: 2026
- Date: 2026-11-20
- Authors: Daniel Omuna, Alum Esther Ugo, Manal Awad Khiry, Obaroh Israel Olusegun, Afam Uzorka
- DOI: 10.1016/j.ejrs.2026.11.002
Research Groups
- Department of Biological & Environmental Sciences, School of Natural & Applied Sciences, Kampala International University
- Department of Physical Sciences, School of Natural & Applied Sciences, Kampala International University
Short Summary
This study assessed climate-induced water availability in Kumi District, eastern Uganda, from 1994 to 2024 using an integrated approach, revealing increasing climatic water deficits and variable runoff despite stable rainfall, leading to spatially heterogeneous water stress and hydrological extremes.
Objective
- How has climate variability affected effective water availability in Kumi District from 1994 to 2024, and how do those effects vary spatially between drought-sensitive and flood-sensitive parts of the district?
Study Configuration
- Spatial Scale: Kumi District, eastern Uganda, with specific analysis for Nyero and Ongino sub-counties.
- Temporal Scale: 1994 to 2024.
Methodology and Data
- Models used: Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI), Standardized Runoff Index (SRI), Mann–Kendall test (for trend analysis), Soil and Water Assessment Tool (SWAT 2012, revision 681, with SUFI-2 algorithm), Random Forest (for flood-susceptibility modelling), Spearman's rank correlation.
- Data sources:
- Satellite: Landsat imagery (Google Earth Engine, 30 m), SRTM DEM (30 m), Copernicus CGLS-LC100 land cover (2019, 100 m), Sentinel-1 GRD SAR (for flood event mapping).
- Observation/Reanalysis: ERA5-Land monthly reanalysis (Copernicus C3S Climate Data Store, ~0.1° grid for precipitation, 2-meter air temperature, potential evaporation, surface runoff, subsurface runoff), NASA POWER daily meteorological forcing, OpenStreetMap Uganda/Geofabrik (roads), HydroRIVERS v1.0 (HydroSHEDS/WWF, rivers), Uganda MWE/DWRM (observed streamflow from "Kumi Watershed Outlet," station KUMI-SYN-01), SoilGrids 2.0 (ISRIC, 250 m for soil properties).
- Other: Kumi District disaster/field reports (flood inventory), Purposive stakeholder interviews (20 stakeholders).
Main Results
- The Standardized Precipitation Index (SPI) showed no significant monotonic trends (p > 0.05) across 3-, 6-, and 12-month scales.
- The Standardized Precipitation Evapotranspiration Index (SPEI) declined significantly (p < 0.001) across all 3-, 6-, and 12-month timescales, indicating increasing climatic water deficits.
- The Standardized Runoff Index (SRI) showed substantial variability, with a significant positive trend only at the 12-month scale (Z = 3.606, p = 0.003).
- Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) exhibited a strong negative Spearman's rank correlation (ρ ≈ −0.96, p < 0.001), indicating that increasing vegetation greenness did not consistently represent greater surface-water/moisture signals.
- The SWAT model demonstrated strong streamflow simulation performance with R² = 0.923, NSE = 0.897, RMSE = 9.79 m³/s, and PBIAS = -9.2% for the full period (2000-2024).
- The Random Forest model achieved high flood-prediction accuracy (AUC = 0.949 on the test set), with elevation, combined NDVI–NDWI, and distance to rivers being the most important predictors.
- Spatial analysis identified Nyero as more drought-sensitive (56.82% very low flood susceptibility) and Ongino as more flood-prone (over 72% medium-to-very-high flood susceptibility).
- Overall, increasing evaporative demand, variable runoff, declining surface-water signals, and uneven hydrological extremes explain water stress better than rainfall decline alone.
Contributions
- Integrated multiple geospatial, hydrological, and machine learning techniques to provide a comprehensive assessment of climate-induced water availability in a data-limited semi-humid tropical environment.
- Demonstrated that water stress in the study area is primarily driven by increasing evaporative demand and variable hydrological responses, rather than solely by changes in precipitation.
- Provided a spatially differentiated understanding of hydrological vulnerability, identifying distinct drought-sensitive (Nyero) and flood-prone (Ongino) sub-regions within the same district.
- Highlighted the "green-drought" phenomenon, where increasing vegetation greenness does not correspond to improved surface moisture conditions.
- Offered an integrated framework for water-availability assessment that considers the interplay of climatic water balance, runoff behavior, and land-surface indicators.
Funding
Not explicitly stated in the provided paper text.
Citation
@article{Omuna2026Integrated,
author = {Omuna, Daniel and Ugo, Alum Esther and Khiry, Manal Awad and Olusegun, Obaroh Israel and Uzorka, Afam},
title = {Integrated GIS, remote sensing, hydrological modelling and machine learning assessment of climate-induced Water availability in Kumi District, eastern Uganda},
journal = {The Egyptian Journal of Remote Sensing and Space Science},
year = {2026},
doi = {10.1016/j.ejrs.2026.11.002},
url = {https://doi.org/10.1016/j.ejrs.2026.11.002}
}
Original Source: https://doi.org/10.1016/j.ejrs.2026.11.002