Premium
Accuracy and Reliability of Pedotransfer Functions as Affected by Grouping Soils
Author(s) -
Pachepsky Ya. A.,
Rawls W. J.
Publication year - 1999
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/sssaj1999.6361748x
Subject(s) - pedotransfer function , soil science , soil water , water content , environmental science , reliability (semiconductor) , soil test , bulk density , hydraulic conductivity , mathematics , statistics , geotechnical engineering , geology , power (physics) , physics , quantum mechanics
Pedotransfer functions (PTFs; i.e., dependencies of soil water retention and soil hydraulic conductivity on basic soil parameters available from soil surveys) are widely used to predict soil functioning in agricultural and environmental systems. The reliability of PTFs needs to be assessed by examining the correspondence between measured and estimated data for data set(s) other than the one used to develop a PTF. Our objective was to see whether grouping according to taxonomic unit, soil moisture regime, soil temperature regime, and soil textural class would improve both PTF accuracy and reliability. We estimated soil water contents at matrix potentials of −33 kPa and −1500 kPa for the 447 soil samples from the Oklahoma National Resource Conservation Service database. Dry bulk density, the ratio of cation‐exchange capacity (CEC) to clay content, and contents of clay, sand, coarse fragments, and organic matter were used as predictors. The Group Method of Data Handling (GMDH) was used to develop PTFs. To assess accuracy and reliability of the PTFs, we used cross‐validation; i.e., repeated random splitting of the data set into subsets for development and validation. The PTF accuracy and reliability was quantified by the root mean square error in the development and validation data set, respectively. Grouping improved the accuracy of PTFs in most cases. None of the grouping criteria proved to be clearly superior. Although PTFs developed from the groups were more accurate than the PTFs developed from the whole database, they were not more reliable. Improving PTF reliability may be an issue distinctly different from improving PTF accuracy.