Uncertainties of nonlinearly estimated parameters from incubations of soil organic matter

Research output: Contribution to journalArticleResearchpeer review

Authors

  • Jürgen Böttcher
View graph of relations

Details

Original languageEnglish
Pages (from-to)293-302
Number of pages10
JournalJournal of Plant Nutrition and Soil Science
Volume167
Issue number3
Publication statusPublished - Jun 2004

Abstract

Parameters of first-order mineralization models are commonly estimated from incubation experiments by using nonlinear estimation techniques. The impact of incubation length and especially more or less small measurement errors on estimated mineralization parameters is largely unknown. The objective of this paper is to analyze the influence of errors in mineralization measurements on the nonlinearly estimated parameters, and to find out how the certainty of parameters can be enhanced. CO2 evolution from organic material (forest floor, L layer) of an acid sandy podzol under pine (Pinus silvestris L.) is taken as an example. For the nonlinear parameter estimation, Marquardt's method is used, and to evaluate uncertainties of estimated model parameters, Monte-Carlo simulations are performed. The results demonstrate that really small measurement errors in the order of random and unavoidable experimental errors may cause serious uncertainties (about 50 %) in the estimated parameters. Parameter uncertainties are the same for cumulative or rate measurements and models. The evaluations based on MonteCarlo simulations show that measurement accuracy and incubation length are key factors for the certainty of estimated parameters. Any decrease of measurement errors results in a disproportionate enhancement of parameter certainty. An increase of incubation length up to about 80% of the time needed to reach 90% of the maximum cumulative mineralization, t90, is optimal to minimize parameter uncertainty. This optimal incubation length is commonly not known a priori because t90 depends on the a priori unknown mineralization kinetics. A practical way for its determination is an accompanying estimation of the parameters while the incubation is going on, and the experiment may be terminated when incubation length has reached about 80% of the estimate of t90.

Keywords

    Incubation experiments, Mineralization, Nonlinear parameter estimation, Optimal incubation length, Uncertainties

ASJC Scopus subject areas

Cite this

Uncertainties of nonlinearly estimated parameters from incubations of soil organic matter. / Böttcher, Jürgen.
In: Journal of Plant Nutrition and Soil Science, Vol. 167, No. 3, 06.2004, p. 293-302.

Research output: Contribution to journalArticleResearchpeer review

Download
@article{8fd1352a5b114c1aa6e788fa5a7e392f,
title = "Uncertainties of nonlinearly estimated parameters from incubations of soil organic matter",
abstract = "Parameters of first-order mineralization models are commonly estimated from incubation experiments by using nonlinear estimation techniques. The impact of incubation length and especially more or less small measurement errors on estimated mineralization parameters is largely unknown. The objective of this paper is to analyze the influence of errors in mineralization measurements on the nonlinearly estimated parameters, and to find out how the certainty of parameters can be enhanced. CO2 evolution from organic material (forest floor, L layer) of an acid sandy podzol under pine (Pinus silvestris L.) is taken as an example. For the nonlinear parameter estimation, Marquardt's method is used, and to evaluate uncertainties of estimated model parameters, Monte-Carlo simulations are performed. The results demonstrate that really small measurement errors in the order of random and unavoidable experimental errors may cause serious uncertainties (about 50 %) in the estimated parameters. Parameter uncertainties are the same for cumulative or rate measurements and models. The evaluations based on MonteCarlo simulations show that measurement accuracy and incubation length are key factors for the certainty of estimated parameters. Any decrease of measurement errors results in a disproportionate enhancement of parameter certainty. An increase of incubation length up to about 80% of the time needed to reach 90% of the maximum cumulative mineralization, t90, is optimal to minimize parameter uncertainty. This optimal incubation length is commonly not known a priori because t90 depends on the a priori unknown mineralization kinetics. A practical way for its determination is an accompanying estimation of the parameters while the incubation is going on, and the experiment may be terminated when incubation length has reached about 80% of the estimate of t90.",
keywords = "Incubation experiments, Mineralization, Nonlinear parameter estimation, Optimal incubation length, Uncertainties",
author = "J{\"u}rgen B{\"o}ttcher",
note = "Copyright: Copyright 2008 Elsevier B.V., All rights reserved.",
year = "2004",
month = jun,
doi = "10.1002/jpln.200321323",
language = "English",
volume = "167",
pages = "293--302",
journal = "Journal of Plant Nutrition and Soil Science",
issn = "1436-8730",
publisher = "Wiley-VCH Verlag",
number = "3",

}

Download

TY - JOUR

T1 - Uncertainties of nonlinearly estimated parameters from incubations of soil organic matter

AU - Böttcher, Jürgen

N1 - Copyright: Copyright 2008 Elsevier B.V., All rights reserved.

PY - 2004/6

Y1 - 2004/6

N2 - Parameters of first-order mineralization models are commonly estimated from incubation experiments by using nonlinear estimation techniques. The impact of incubation length and especially more or less small measurement errors on estimated mineralization parameters is largely unknown. The objective of this paper is to analyze the influence of errors in mineralization measurements on the nonlinearly estimated parameters, and to find out how the certainty of parameters can be enhanced. CO2 evolution from organic material (forest floor, L layer) of an acid sandy podzol under pine (Pinus silvestris L.) is taken as an example. For the nonlinear parameter estimation, Marquardt's method is used, and to evaluate uncertainties of estimated model parameters, Monte-Carlo simulations are performed. The results demonstrate that really small measurement errors in the order of random and unavoidable experimental errors may cause serious uncertainties (about 50 %) in the estimated parameters. Parameter uncertainties are the same for cumulative or rate measurements and models. The evaluations based on MonteCarlo simulations show that measurement accuracy and incubation length are key factors for the certainty of estimated parameters. Any decrease of measurement errors results in a disproportionate enhancement of parameter certainty. An increase of incubation length up to about 80% of the time needed to reach 90% of the maximum cumulative mineralization, t90, is optimal to minimize parameter uncertainty. This optimal incubation length is commonly not known a priori because t90 depends on the a priori unknown mineralization kinetics. A practical way for its determination is an accompanying estimation of the parameters while the incubation is going on, and the experiment may be terminated when incubation length has reached about 80% of the estimate of t90.

AB - Parameters of first-order mineralization models are commonly estimated from incubation experiments by using nonlinear estimation techniques. The impact of incubation length and especially more or less small measurement errors on estimated mineralization parameters is largely unknown. The objective of this paper is to analyze the influence of errors in mineralization measurements on the nonlinearly estimated parameters, and to find out how the certainty of parameters can be enhanced. CO2 evolution from organic material (forest floor, L layer) of an acid sandy podzol under pine (Pinus silvestris L.) is taken as an example. For the nonlinear parameter estimation, Marquardt's method is used, and to evaluate uncertainties of estimated model parameters, Monte-Carlo simulations are performed. The results demonstrate that really small measurement errors in the order of random and unavoidable experimental errors may cause serious uncertainties (about 50 %) in the estimated parameters. Parameter uncertainties are the same for cumulative or rate measurements and models. The evaluations based on MonteCarlo simulations show that measurement accuracy and incubation length are key factors for the certainty of estimated parameters. Any decrease of measurement errors results in a disproportionate enhancement of parameter certainty. An increase of incubation length up to about 80% of the time needed to reach 90% of the maximum cumulative mineralization, t90, is optimal to minimize parameter uncertainty. This optimal incubation length is commonly not known a priori because t90 depends on the a priori unknown mineralization kinetics. A practical way for its determination is an accompanying estimation of the parameters while the incubation is going on, and the experiment may be terminated when incubation length has reached about 80% of the estimate of t90.

KW - Incubation experiments

KW - Mineralization

KW - Nonlinear parameter estimation

KW - Optimal incubation length

KW - Uncertainties

UR - http://www.scopus.com/inward/record.url?scp=3042819789&partnerID=8YFLogxK

U2 - 10.1002/jpln.200321323

DO - 10.1002/jpln.200321323

M3 - Article

AN - SCOPUS:3042819789

VL - 167

SP - 293

EP - 302

JO - Journal of Plant Nutrition and Soil Science

JF - Journal of Plant Nutrition and Soil Science

SN - 1436-8730

IS - 3

ER -