By Pierre Jongerius
In order to get a better insight into the prediction of the spatial distribution of soil organic matter (SOM) using different types of data in semi-arid regions a field survey has been conducted at the West Bijou creek in Colorado. The objective is to find out which combination of soil and landscape properties predict the spatial distribution of SOM best in the erosion affected areas at the West Bijou creek. In total 108 sample points consisting out of calibration, validation within the study area and validation points outside of the study area were sampled on different soil and landscape properties. Also multispectral images and hyperspectral reflectance data for every sample was collected. Prediction are based on stepwise multiple linear regression and location specific prediction methods. This study shows that predicting the SOM at the semi-arid region of the West Bijou creek using stratification has shown to be a good way to indicate what the spatial distribution of SOM is. For prediction of SOM outside of the initial study area applying an average SOM to every stratum results in the best prediction. The combination of landscape properties used for this stratification are altitude and whether an area is within or outside a gully. The best combination for predicting SOM within the study area consists out of 5 hyperspectral bands.
Keywords: West Bijou creek; High Plains; Colorado; Digital soil mapping; Soil organic matter; global carbon cycle; Stepwise multiple linear regression; stratification