Environmental and geostatistical modelling of soil properties toward precision agriculture

Understanding the spatial distribution of soil properties is critical for achieving precision agriculture. The study aims to model soil property heterogeneity in the context of food sustainability using remote sensing (RS) and geostatistical techniques at Federal University of Agriculture, Abeokuta, Nigeria. We combined RS metrics like Number patches (NP), Largest-path (LP), and effective MESH alongside Normalized difference vegetation (NDVI), and Enhanced vegetation (EVI) indices from 2014 and 2024, with a particular focus on built-up, vegetation, farmlands, and wetlands in the area. We collected and analyzed 70 geocoded composite soil sample (0 to 30 cm) for their physical, chemical, and biological conditions, interpolated by kriging and added to the exponential, spherical and gaussian to model the soil properties. NP, LP, and MESH showed substantial discontinuity and landscape fragmentation, especially in the built-up areas. At the same time, NDVI, and EVI highlight a significant decrease in vegetation cover, respectively. The modelling of soil properties based on cross-validation showed that soil properties in the studied area ranged between strong (< 0.25) and weak (0.25 to 0.75) spatial autocorrelations. The findings could aid in mitigating anthropogenic climate shocks on soil properties and thus ensuring landscape sustainability and precision agriculture.