Srarfi et al. (2025) Investigation of soil salinity and environmental indicators by Google Earth Engine/Machine Learning in Northeast Tunisia under climate changes

Identification

• Journal: Environmental Earth Sciences
• Year: 2025
• Date: 2025-11-01
• Authors: Feyda Srarfi, Zeineb Haj Ammar, Mohamed Salah Hamdi, Noura Guesmi, Mayssa El Yazidi, Hesam Ahmady‐Birgani
• DOI: 10.1007/s12665-025-12693-4

Research Groups

• Faculty of Sciences of Tunis, Department of Geology, Tunis El Manar University, Tunis, Tunisia
• Center for Studies and Activities for Space, University of Padova, Padua, Italy
• Faculty of Natural Resources, Department of Range & Watershed Management, Urmia University, Urmia, Iran

Short Summary

This study mapped and monitored soil salinity in Northeast Tunisia over 24 years (2000–2023) using Google Earth Engine and machine learning, revealing strong spatio-temporal variability driven by climatic and anthropogenic factors, with remote sensing offering a reliable monitoring tool for sustainable land management.

Objective

• To determine the spatio-temporal evolution of soil salinity in the Mejerda low valley from 2000 to 2023 using Landsat and MODIS imagery.
• To verify the presence and evolution of soil salts by calculating soil salinity, vegetation, and iron indices, linking them to ferrous processes.
• To monitor water transition in irrigated areas of the Mejerda low valley using the Google Earth Engine platform, investigating changes in salinity distribution patterns and contributing factors.

Study Configuration

• Spatial Scale: El Habibia–Mansoura zone in the Mejerda low valley, Northeast Tunisia, covering approximately 4000 hectares.
• Temporal Scale: 24-year period (2000–2023), analyzed in 5-year intervals (2000–2005, 2005–2010, 2010–2015, 2015–2020, 2020–2023).

Methodology and Data

• Models used: Google Earth Engine (GEE) platform, Support Vector Machine (SVM) machine-learning algorithm with a linear kernel.
• Data sources:
  • Satellite imagery: Landsat 5 TM, 7 ETM+, and 8 OLI (for 2000, 2005, 2010, 2015, 2020, 2023), Moderate Resolution Imaging Spectroradiometer (MODIS MOD11A1) (2000–2023).
  • Topographic data: Shuttle Radar Topography Mission (SRTM) for slope and altitude.
  • Climatic data: MODIS-derived Land Surface Temperature (LST) and evapotranspiration (ET, Ei, Es), CHIRPS (Climate Hazards Group Infra-Red Precipitation with Station data) for precipitation patterns (from Carthage weather station, 42 km from study area).
  • Ground control measurements: 90 topsoil samples (0–10 cm depth) collected in June 2023, analyzed for pH and electrical conductivity (EC).
  • Derived indices: Salinity Indices (SI, SI2, SI3), Normalized Difference Vegetation Index (NDVI), Iron Index (IIn).

Main Results

• The Support Vector Machine (SVM) model achieved robust predictive performance for soil salinity with an overall coefficient of determination (R²) of 0.82 and a Root Mean Squared Error (RMSE) of 1.34.
• Soil salinity exhibited strong spatio-temporal variability across the study area from 2000 to 2023.
• The most pronounced peak in salinity occurred between 2010 and 2015, with 46.72% of the area classified as moderately saline and 1.92% as highly saline, correlating with periods of high evapotranspiration and low precipitation.
• Extremely saline soils, though limited in extent, showed a gradual but consistent increase from 0.03% in 2000–2005 to 0.12% in 2020–2023.
• Iron nutrient availability sharply declined after 2013, showing a negative correlation with salinity, while alkaline pH (> 7) and high Na⁺ and Cl⁻ contents further constrained soil fertility.
• Salinity dynamics are driven by both climatic factors (drought, high evapotranspiration, low precipitation) and anthropogenic factors (poor irrigation water quality, dam construction reducing water supply, and inefficient drainage systems).
• Irrigation water quality from Wadi Mejerda, Wadi Chafrou, and groundwater sources showed increasing salinity values from 1966 to 2023, with surface water exceeding 3.5 dS/m and groundwater ranging from 3.2 to 12 dS/m.
• Water transition analysis revealed significant changes in surface water conditions, with less than 10% permanent water and 58.4% newly added seasonal water between 2005 and 2017, indicating vulnerability to climate warming.

Contributions

• Provided the first comprehensive spatio-temporal mapping and monitoring of soil salinity in the El Habibia–Mansoura zone of Northeast Tunisia over a 24-year period (2000–2023) using Google Earth Engine and machine learning.
• Investigated the relationship between soil salinity, ferrous mineral content (Iron Index), vegetation health (NDVI), and climatic/anthropogenic factors, offering a multi-indicator approach.
• Demonstrated the effectiveness of integrating remote sensing data with machine learning algorithms for reliable, large-scale, and long-term soil salinity assessment in arid and semi-arid regions.
• Offered critical information and decision-support tools for agricultural stakeholders to develop region-specific management strategies, improve land productivity, and enhance food security in Tunisia and globally.
• Utilized ground truth measurements to rigorously validate and select the most correlated salinity indices for accurate long-term mapping, improving upon previous studies that lacked detailed estimation of soil salinity across different time scales.

Funding

No external funding was obtained for the achievement of this study.

Citation

@article{Srarfi2025Investigation,
  author = {Srarfi, Feyda and Ammar, Zeineb Haj and Hamdi, Mohamed Salah and Guesmi, Noura and Yazidi, Mayssa El and Ahmady‐Birgani, Hesam},
  title = {Investigation of soil salinity and environmental indicators by Google Earth Engine/Machine Learning in Northeast Tunisia under climate changes},
  journal = {Environmental Earth Sciences},
  year = {2025},
  doi = {10.1007/s12665-025-12693-4},
  url = {https://doi.org/10.1007/s12665-025-12693-4}
}