Erosion probability maps: Calibrating precision agriculture data with soil surveys using logistic regression

ABSTRACT: Soil surveys provide information about the location of eroded areas across landscapes, but not at a scale that may be necessary for land use planning, precision agriculture, and conservation management. The objective of this paper was to determine whether sitespecific information and logistic regression could be used to improve the spatial resolution of soil surveys. This study was conducted on fragipan soils developed from loess in a western Kentucky agricultural field. Information about the presence and severity of erosion was obtained from a highly detailed first-order soil survey and less detailed second-order county soil surveys. Digital terrain attributes (slope, length-slope factor, wetness), reflectance (visible, red-NIR, and NIR), soil electrical conductivity, and direct contact electrical conductivity were used as regressor variables. Binary variables were assigned a value of one if they were located in eroded map phases and if slope values were greater than or equal to two percent. For all other cases they were assigned values of 0. Stepwise multiple logistic regression was used to develop models that were used to map probability that substantial soil erosion had occurred in the past. The resulting probability maps were remarkably similar for both survey orders indicating that this approach was robust to soil map unit inclusions and classification errors. Erosion probability maps created using the second order soil survey matched in many cases with the boundaries of the first order survey. Our results demonstrated that precision agriculture technologies and logistic regression analysis could potentially be used to improve the value and utility of existing second order soil surveys. Soil and water conservation, management, and planning will be more effective and economical if these methods can be adapted for soils in other regions of the United States.