Downscaling digital soil maps using electromagnetic induction and aerial imagery

Abstract Coarse-resolution soil maps at regional to national extents are often inappropriate for mapping intra-field variability. At the same time, sensor data, such as electromagnetic induction measurements and aerial imagery, can be highly useful for mapping soil properties that correlate with electrical conductivity or soil color. However, maps based on these data nearly always require calibration with local samples, as multiple factors can affect the sensor measurements. In this study, we present a downscaling method, which combines coarse-resolution, large extent soil maps with sensor data in order to improve predictions of soil properties. The method modifies values from coarse-resolution soil maps to predict soil properties at a location, using relationships between soil properties and sensor data from other locations. We test this method for predicting clay and soil organic matter contents at five agricultural fields located in Denmark. We test the method for one field at a time, using soil samples from the four other fields to predict soil properties. The maps produced with the method are generally more accurate than the coarse-resolution soil maps, especially for soil organic matter. The method generally overestimates prediction uncertainties, a disadvantage, which will require improvements. Overall, the method is a simple, promising tool for giving a quantitative estimate of soil properties, when no local soil samples are available.