Abstract:
Soil fertility mapping is essential for optimizing agricultural practices and management. This
study was conducted in Mana district, southwestern Ethiopia. It aimed at evaluating and
comparing three GIS-based spatial interpolation methods (Inverse Distance Weighting: IDW;
Ordinary Kriging: OK; and Ordinary Cokriging: OCK) for estimating selected soil chemical
properties, and producing a set of accurate maps of selected soil chemical properties. The study
included 84 geo-referenced soil samples collected in April 2021 at 0 - 30 cm depth across the
entire district, using the systematic sampling technique with 2.5 km × 2.5 km grid. The soil
samples were analyzed for selected chemical properties. Descriptive statistics were first applied
to the data to evaluate and validate the normal distribution required for geostatistical analysis.
The performance of each interpolation method was assessed using cross-validation. The
descriptive statistical analyses revealed that besides the topographic aspect which were highly
variable, available phosphorus (AP) and available potassium (AK) were the most variable soil
properties (CV > 35%); while pH, soil organic carbon (SOC) and total nitrogen (TN) contents
were moderately variable (CV varying from 16.28% to 30.53%). Only pH and SOC were
normally distributed among all the variables. When comparing the resulting values of the
efficiency criteria of cross-validation (RMSE, MRE and RI) for each interpolation method, the
OCK technique was best performed for all the five soil chemical properties with lower values
of RMSE and MRE, and the best RI. However, for the TN, OK showed the same performance
as OCK. Interpolated maps were generated based on OCK for each soil property and indicated
their distribution throughout the study area of Mana.
