Database-related accuracy and uncertainty of pedotransfer functions

Marcel Schaap, Feike J. Leij

Research output: Contribution to journalArticle

309 Citations (Scopus)

Abstract

Pedotransfer functions (PTFs) are becoming a more common way to predict soil hydraulic properties from soil texture, bulk density, and organic matter content. Thus far, the calibration and validation of PTFs has been hampered by a lack of suitable databases. In this paper we employed three databases (RAWLS, AHUJA, and UNSODA) to evaluate the accuracy and uncertainty of neural network-based PTFs. Sand, slit, and clay percentages and bulk density were used as input for the PTFs, which subsequently provided retention parameters and saturated hydraulic conductivity, K(s) as output. Calibration and validation of PTFs were carried out on independent samples from the same database through combination with the bootstrap method. This method also yielded the possibility of calculating uncertainty estimates of predicted hydraulic parameters. Calibration and validation results showed that water retention could be predicted with a root mean square residual (RMSR) between 0.06 and 0.10 cm3 cm-3; the RMSR of log(K(s)) was between 0.4 and 0.7 log (cm day-1). Cross-validation was used to test how well PTFs that were calibrated for one database could predict the hydraulic properties of the other two databases. The results showed that systematically different predictions were made when the RMSR values increased to between 0.08 and 0.13 cm3 cm-3 for water retention and to between 0.6 and 0.9 log(cm day-1) for log(K(s)). The uncertainty in predicted K was one-half to one order of magnitude, whereas predicted water retention points had an uncertainty of about 0.04 to 0.10 cm3 cm-3. Uncertainties became somewhat smaller if the PTFs were calibrated on all available data. We conclude that the performance of PTFs may depend strongly on the data that were used for calibration and evaluation.

Original languageEnglish (US)
Pages (from-to)765-779
Number of pages15
JournalSoil Science
Volume163
Issue number10
StatePublished - Oct 1998
Externally publishedYes

Fingerprint

pedotransfer function
pedotransfer functions
uncertainty
calibration
water retention
hydraulic property
bulk density
fluid mechanics
bootstrapping
soil hydraulic properties
water
saturated hydraulic conductivity
soil texture
neural networks
hydraulic conductivity
soil organic matter
clay
sand
hydraulics
organic matter

Keywords

  • Hydraulic conductivity
  • Neural networks
  • Prediction
  • Uncertainty estimates
  • Water retention

ASJC Scopus subject areas

  • Soil Science
  • Earth-Surface Processes

Cite this

Database-related accuracy and uncertainty of pedotransfer functions. / Schaap, Marcel; Leij, Feike J.

In: Soil Science, Vol. 163, No. 10, 10.1998, p. 765-779.

Research output: Contribution to journalArticle

Schaap, Marcel ; Leij, Feike J. / Database-related accuracy and uncertainty of pedotransfer functions. In: Soil Science. 1998 ; Vol. 163, No. 10. pp. 765-779.
@article{c17c56952d0146f1b35b851620428172,
title = "Database-related accuracy and uncertainty of pedotransfer functions",
abstract = "Pedotransfer functions (PTFs) are becoming a more common way to predict soil hydraulic properties from soil texture, bulk density, and organic matter content. Thus far, the calibration and validation of PTFs has been hampered by a lack of suitable databases. In this paper we employed three databases (RAWLS, AHUJA, and UNSODA) to evaluate the accuracy and uncertainty of neural network-based PTFs. Sand, slit, and clay percentages and bulk density were used as input for the PTFs, which subsequently provided retention parameters and saturated hydraulic conductivity, K(s) as output. Calibration and validation of PTFs were carried out on independent samples from the same database through combination with the bootstrap method. This method also yielded the possibility of calculating uncertainty estimates of predicted hydraulic parameters. Calibration and validation results showed that water retention could be predicted with a root mean square residual (RMSR) between 0.06 and 0.10 cm3 cm-3; the RMSR of log(K(s)) was between 0.4 and 0.7 log (cm day-1). Cross-validation was used to test how well PTFs that were calibrated for one database could predict the hydraulic properties of the other two databases. The results showed that systematically different predictions were made when the RMSR values increased to between 0.08 and 0.13 cm3 cm-3 for water retention and to between 0.6 and 0.9 log(cm day-1) for log(K(s)). The uncertainty in predicted K was one-half to one order of magnitude, whereas predicted water retention points had an uncertainty of about 0.04 to 0.10 cm3 cm-3. Uncertainties became somewhat smaller if the PTFs were calibrated on all available data. We conclude that the performance of PTFs may depend strongly on the data that were used for calibration and evaluation.",
keywords = "Hydraulic conductivity, Neural networks, Prediction, Uncertainty estimates, Water retention",
author = "Marcel Schaap and Leij, {Feike J.}",
year = "1998",
month = "10",
language = "English (US)",
volume = "163",
pages = "765--779",
journal = "Soil Science",
issn = "0038-075X",
publisher = "Lippincott Williams and Wilkins",
number = "10",

}

TY - JOUR

T1 - Database-related accuracy and uncertainty of pedotransfer functions

AU - Schaap, Marcel

AU - Leij, Feike J.

PY - 1998/10

Y1 - 1998/10

N2 - Pedotransfer functions (PTFs) are becoming a more common way to predict soil hydraulic properties from soil texture, bulk density, and organic matter content. Thus far, the calibration and validation of PTFs has been hampered by a lack of suitable databases. In this paper we employed three databases (RAWLS, AHUJA, and UNSODA) to evaluate the accuracy and uncertainty of neural network-based PTFs. Sand, slit, and clay percentages and bulk density were used as input for the PTFs, which subsequently provided retention parameters and saturated hydraulic conductivity, K(s) as output. Calibration and validation of PTFs were carried out on independent samples from the same database through combination with the bootstrap method. This method also yielded the possibility of calculating uncertainty estimates of predicted hydraulic parameters. Calibration and validation results showed that water retention could be predicted with a root mean square residual (RMSR) between 0.06 and 0.10 cm3 cm-3; the RMSR of log(K(s)) was between 0.4 and 0.7 log (cm day-1). Cross-validation was used to test how well PTFs that were calibrated for one database could predict the hydraulic properties of the other two databases. The results showed that systematically different predictions were made when the RMSR values increased to between 0.08 and 0.13 cm3 cm-3 for water retention and to between 0.6 and 0.9 log(cm day-1) for log(K(s)). The uncertainty in predicted K was one-half to one order of magnitude, whereas predicted water retention points had an uncertainty of about 0.04 to 0.10 cm3 cm-3. Uncertainties became somewhat smaller if the PTFs were calibrated on all available data. We conclude that the performance of PTFs may depend strongly on the data that were used for calibration and evaluation.

AB - Pedotransfer functions (PTFs) are becoming a more common way to predict soil hydraulic properties from soil texture, bulk density, and organic matter content. Thus far, the calibration and validation of PTFs has been hampered by a lack of suitable databases. In this paper we employed three databases (RAWLS, AHUJA, and UNSODA) to evaluate the accuracy and uncertainty of neural network-based PTFs. Sand, slit, and clay percentages and bulk density were used as input for the PTFs, which subsequently provided retention parameters and saturated hydraulic conductivity, K(s) as output. Calibration and validation of PTFs were carried out on independent samples from the same database through combination with the bootstrap method. This method also yielded the possibility of calculating uncertainty estimates of predicted hydraulic parameters. Calibration and validation results showed that water retention could be predicted with a root mean square residual (RMSR) between 0.06 and 0.10 cm3 cm-3; the RMSR of log(K(s)) was between 0.4 and 0.7 log (cm day-1). Cross-validation was used to test how well PTFs that were calibrated for one database could predict the hydraulic properties of the other two databases. The results showed that systematically different predictions were made when the RMSR values increased to between 0.08 and 0.13 cm3 cm-3 for water retention and to between 0.6 and 0.9 log(cm day-1) for log(K(s)). The uncertainty in predicted K was one-half to one order of magnitude, whereas predicted water retention points had an uncertainty of about 0.04 to 0.10 cm3 cm-3. Uncertainties became somewhat smaller if the PTFs were calibrated on all available data. We conclude that the performance of PTFs may depend strongly on the data that were used for calibration and evaluation.

KW - Hydraulic conductivity

KW - Neural networks

KW - Prediction

KW - Uncertainty estimates

KW - Water retention

UR - http://www.scopus.com/inward/record.url?scp=0031783722&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031783722&partnerID=8YFLogxK

M3 - Article

VL - 163

SP - 765

EP - 779

JO - Soil Science

JF - Soil Science

SN - 0038-075X

IS - 10

ER -