Advances in soil salinity diagnosis for mangrove swamp rice production in Guinea Bissau, West Africa

Rice is one of the most important crops in many West African countries and has a direct impact on food security. Mangrove swamp cultivation is the most productive rice system in this area but is highly vulnerable to changes in rainfall patterns due to soil salinity. Diagnosing and identifying areas of high salinity concentration are essential strategies for adapting to climate change and mitigating its impacts. The aim of this study is to provide a methodological approach to identify the causes of soil salinity and map the spatial distribution of hypersaline areas, focusing on three case studies in Guinea Bissau. At three study sites in the north, center, and south of the country, 382 soil samples were collected under initial conditions before rice cultivation. Indices derived from spectral bands and soil texture raster of the Planet Scope project were used to calibrate the three machine learning based models: Random Forest (RF), Support Vector Machine, and Convolutional Neural Networks. Chemical analysis of the soil revealed that Mg²⁺ and Na⁺ were the extractable cations with the highest concentration in all three study sites. The RF showed the highest accuracy for salinity prediction (ρ = 0.90, R² = 0.80, MAE = 15.41 dS m⁻¹, RMSE = 25.49 dS m⁻¹, NRMSE = 51 %, BIAS = 0.18, PBIAS = 0.36 %, RPIQ = 2.25), with normalized difference salinity index (RNDSI, calculated with red edge). Silt raster, normalized salinity index (NDSI), and normalized difference water index (NDWI) were the main contributors in the predicted data for soil electrical conductivity of the saturation paste extract (EC_e, dS m⁻¹). This approach produced a reliable approximation during validation for the three study sites (ρ = 0.84 to 0.90, R² = 0.68 to 0.78, MAE = 11.74 dS m⁻¹ to 24.85 dS m⁻¹, RMSE = 17.26 dS m⁻¹ to 38.98 dS m⁻¹, NRMSE = 42 %–54 %, BIAS = −2.25 to 2.24, PBIAS = −5.49 %–7.01 %, RPIQ = 2.01 to 2.43), each exhibiting unique edaphoclimatic characteristics. This study highlights the critical importance of diagnosing hypersaline sites to improve agronomic management practices by introducing improved water management infrastructures, conserving mangrove forests, and promoting regional ecological resilience.