Skip to main content
ARS Home » Research » Publications at this Location » Publication #72464
Title: PREDICTION OF SOIL HYDRAULIC PROPERTIES FROM BASIC SOIL DATA: EXPLORATION WITH NEURAL NETWORKS

Author
item SCHAAP, M. - U.C. RIVERSIDE
item LEIJ, F. - U.C. RIVERSIDE

Submitted to: Agronomy Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: 7/2/1996
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: The modeling of flow and transport processes in soils requires a knowledge of unsaturated hydraulic properties (water retention and hydraulic conductivity). Unfortunately, measuring these properties is often difficult and expensive. A considerable amount of attention has been focused on predicting hydraulic properties from soil characteristics such as the particle size distribution, bulk density and organic matter content using (non-) linear regression analysis. Neural networks are well-suited to accomplish the same task because of their ability of "learning" relationships between different types of data. We will predict unsaturated soil hydraulic properties with both published regression equations and neural networks analyses. Neural networks often provide more accurate results. Furthermore, neural network analyses can be used to asses which soil properties are important factors in determining the curve shapes of hydraulic functions. The hydraulic properties are from different independent data sets.