Soil Survey: Prediction of Bulk Density Using k ‐Nearest Neighbor Approach

In soil survey, there is a need to estimate bulk density (at −33 kPa matric potential, D b,33 ) for the range of soils that occur in the United States. The objectives of this study were to develop a bulk density model using the k ‐nearest neighbor ( k ‐NN) approach and validate this model against an independent dataset. The nearest‐neighbor approach passively stores the development (or reference) dataset until the time of application, and then the dataset is searched for the 10 ( k ) most similar soils to that of the target soil, based on selected attributes (e.g., organic C, total sand, silt, and clay). Reference datasets were developed from the National Cooperative Soil Survey characterization database in Lincoln, NE. The model uses a decision tree to determine which reference dataset and attributes are to be used to predict D b,33. Validation results with an independent dataset showed a small, but significant bias or systematic error (ME = −0.0215 g cm −3 ). However, the overall prediction error (or RMSE) of 0.108 g cm −3 is smaller than most reported literature values. The k ‐NN model for estimating bulk density is suggested to provide an efficient and reasonably accurate tool for use in soil survey for a wide range of soils.