Premium
Estimating Pedotransfer Function Prediction Limits Using Fuzzy k ‐Means with Extragrades
Author(s) -
Tranter G.,
Minasny B.,
McBratney A. B.
Publication year - 2010
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/sssaj2009.0106
Subject(s) - pedotransfer function , outlier , calibration , mathematics , statistics , cumulative distribution function , range (aeronautics) , permanent wilting point , residual , probability density function , computer science , soil science , algorithm , environmental science , engineering , hydraulic conductivity , field capacity , soil water , aerospace engineering
We have developed a method for estimating the pedotransfer function uncertainty using fuzzy k ‐means with extragrades classification to, first, identify specific areas of covariate space that share similar residuals or distribution of residuals and, second, identify those observations that exist outside the convex hull of the calibration data. Those observations outside the convex hull are considered outliers of the calibration data and subsequently have their uncertainty penalized by a simple multiplier. Uncertainties are presented as upper and lower prediction limits. In addition, the method can be extended to estimate the cumulative density function for the prediction residual. The method was applied to a pedotransfer function for the prediction of gravimetric moisture content at permanent wilting points (‐1500 kPa). The method conformed closely to the desired confidence level; however, some instability was shown at confidence levels >80%. The ability to identify conditions “dissimilar” to those used for model calibration makes the method suitable for a diverse range of prediction problems.