Premium
Regression Models for Estimating Soil Properties by Landscape Position
Author(s) -
Brubaker S. C.,
Jones A. J.,
Frank K.,
Lewis D. T.
Publication year - 1994
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj1994.03615995005800060026x
Subject(s) - linear regression , soil science , regression analysis , surface runoff , confidence interval , environmental science , hydrology (agriculture) , sampling (signal processing) , soil water , mathematics , statistics , geology , ecology , geotechnical engineering , filter (signal processing) , computer science , computer vision , biology
Slope geometry and the associated variation in soil properties influence runoff, drainage, soil temperature, the extent of soil erosion and deposition, and crop yields. With the current emphasis on prescription farming, approaches are needed to more effectively match inputs to production system needs while accounting for variation in soil and water resources within a field. The objective of the study was to develop simplified regression models to predict soil properties on different landscape positions from observed values on the nearly level upper interfluve. Soil samples were taken from the upper and lower interfluve, shoulder, upper and lower linear, and footslope at each of four sites in eastern Nebraska. Predictive equations were developed for 20 soil properties using multiple linear regression. Independent variables included were observed values of the property being modeled from the upper interfluve, sampling depth, and an irrigation code. Of the 100 models developed, only eight included significant contributions from all three independent variables. Models for pH, organic matter, electrical conductivity, exchangeable K, base saturation percentage, and available P and K consistently had R 2 values greater than 0.50. The upper interfluve contributed significantly to the prediction of each of these properties except electrical conductivity. A comparison between average observed and predicted values for each soil property at each sampling depth revealed that the observed values generally fell within a 95% confidence interval about the predicted values. The confidence interval half‐width was generally <10% of the mean for the observed values. Further evaluation with independent data sets could be used to help strengthen and refine such generalized or geographically based models.