Model prediction of soil drainage classes based on digital elevation model parameters and soil attributes from coarse resolution soil maps

Abstract

High-resolution soil drainage maps are important for crop production planning, forest management, and environmental assessment. Existing soil classification maps tend to only have information about the dominant soil drainage conditions and they are inadequate for precision forestry and agriculture planning purposes. The objective of this research was to develop an artificial neural network (ANN) model for producing soil drainage classification maps at high resolution. Soil profile data extracted from coarse resolution soil maps (1:1 000 000 scale) and topographic and hydrological variables derived from digital elevation model (DEM) data (1:35 000 scale) were considered as candidates for inputs. A high-resolution soil drainage map (1:10 000) of the Black Brook Watershed (BBW) in northwestern New Brunswick (NB), Canada, was used to train and validate the ANN model. Results indicated that the best ANN model included average soil drainage classes, average soil sand content, vertical slope position (VSP), sediment delivery ratio (SDR) and slope steepness as inputs. It was found that 52% of model-predicted drainage classes were exactly the same as field assessment observations and 94% of model-predicted drainage classes were within ±1 class. In comparison, only 12% of maps indicated drainage classes were the same as field assessment observations based on coarse resolution soil maps and only 55% of points were within ±1 class of field assessed drainage classes. Results indicated that the model could be used to produce high-resolution soil drainage maps at relatively low cost. Key words: Soil drainage, artificial neural network model, ANN model, high-resolution soil maps, DEM, hydrology model