Integrating Geospatial Information and Artificial Intelligence for Irrigation Planning in Kwali Area Council, Federal Capital Territory, Nigeria

Agriculture in Nigeria is increasingly threatened by climate variability, irregular rainfall, inefficient water management and land degradation, which collectively affect food security and rural livelihoods. This study evaluated irrigation suitability within the Wako Irrigation Site, Kwali Area Council, Federal Capital Territory (FCT), Nigeria, using an integrated GeoAI framework combining geospatial technology and artificial intelligence for precision irrigation planning. Environmental variables comprising slope, Soil Moisture Index (SMI), drainage density and vegetation condition represented by the Normalised Difference Vegetation Index (NDVI) were analysed and integrated using a Multilayer Perceptron Artificial Neural Network (MLP-ANN). Topographic analysis indicated that nearly level terrain dominated the study area, covering 8.72 km² (58.23%), while gentle and moderate slopes accounted for 4.68 km² (31.22%) and 1.58 km² (10.55%), respectively. Soil-moisture analysis showed that moist areas constituted the dominant class, covering 6.52 km² (43.52%), followed by wet areas at 4.96 km² (33.13%) and dry areas at 3.50 km² (23.35%). Low drainage density was the dominant drainage class, occupying 6.62 km² (44.19%), while moderate and very low drainage densities covered 4.85 km² (32.38%) and 3.51 km² (23.43%), respectively. NDVI values ranged from 0.031 to 0.473, indicating vegetation conditions ranging from no vegetation to low vegetation, with comparatively higher values observed around Dagara. The MLP-ANN model integrated the environmental variables to delineate irrigation-suitability zones. Highly irrigable areas constituted the dominant class, covering 9.91 km² (66.20%) of the 14.98 km² study area. Moderately irrigable areas occupied 3.68 km² (24.50%), while low irrigable areas accounted for 1.39 km² (9.27%). Model validation using 100 field-reference observations produced an overall accuracy of 79.00%, sensitivity of 80.00%, specificity of 76.00% and ROC–AUC of 0.780, indicating acceptable and reasonably balanced predictive performance. The findings demonstrate that GeoAI-based irrigation-suitability modelling can support evidence-based irrigation planning, improve water-use efficiency and promote climate-smart agricultural development in Nigeria.