Interpolation methods for soil properties are often dependent on independent variables for which measurements are available for the entire surface being interpolated. In modeling soil organic carbon (SOC) for example, management related factors for which data do not exist for the entire area become nuisance factors for they may exhibit a strong influence in the estimation of the spatial continuity function. This study demonstrates an approach for controlling such factors. Soils were sampled from agricultural landscapes of western Kenya at nine sites, each site measuring about 1 km in radius. A Y-shaped sampling design was established at each site and three farms were selected on each arm of the Y and one at the center, making ten farms per site. Each farm was divided into fields according to the land use history and type of enterprises. Soil samples were collected from each field for analysis of SOC, pH and particle size distribution. Mixed effects autoregression modeling was applied to the data to estimate the effect of landscape covariates and to control for this effect in the estimation of the spatial continuity function. Size of the plot (p<0.0001) and application of fertilizers (p<0.08) were the landscape factors that had significant influence of the spatial structure of SOC landscape scale. A variogram model was developed after removing the effect of these factors and used to interpolate SOC. A correlation coefficient of 0.55 and mean squared error (MSE) of 9.3 g/kg were obtained from the predictions. The Prediction Accuracy (PE) was 55 percent. These results demonstrate a protocol for controlling for such nuisance factors.