Neural Error Regression Diagnosis (NERD): A tool for model bias identification and prognostic data assimilation

Gab Abramowitz, Hoshin Vijai Gupta, Andy Pitman, Yingping Wang, Ray Leuning, Helen Cleugh, Kuo Lin Hsu

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

Data assimilation in the field of predictive land surface modeling is generally limited to using observational data to estimate optimal model states or restrict model parameter ranges. To date, very little work has attempted to systematically define and quantify error resulting from a model's inherent inability to simulate the natural system. This paper introduces a data assimilation technique that moves toward this goal by accounting for those deficiencies in the model itself that lead to systematic errors in model output. This is done using a supervised artificial neural network to "learn" and simulate systematic trends in the model output error. These simulations in turn are used to correct the model's output each time step. The technique is applied in two case studies, using fluxes of latent heat flux at one site and net ecosystem exchange (NEE) of carbon dioxide at another. Root-mean-square error (rmse) in latent heat flux per time step was reduced from 27.5 to 18.6 W m-2 (32%) and monthly from 9.91 to 3.08 W m-2 (68%). For NEE, rmse per time step was reduced from 3.71 to 2.70 μmol m-2 s-1 (27%) and annually from 2.24 to 0.11 μmol m-2 s-1 (95%). In both cases the correction provided significantly greater gains than single criteria parameter estimation on the same flux.

Original languageEnglish (US)
Pages (from-to)160-177
Number of pages18
JournalJournal of Hydrometeorology
Volume7
Issue number1
DOIs
StatePublished - Feb 2006

Fingerprint

data assimilation
net ecosystem exchange
latent heat flux
prognostic data
artificial neural network
land surface
carbon dioxide
modeling
simulation

ASJC Scopus subject areas

  • Atmospheric Science

Cite this

Neural Error Regression Diagnosis (NERD) : A tool for model bias identification and prognostic data assimilation. / Abramowitz, Gab; Gupta, Hoshin Vijai; Pitman, Andy; Wang, Yingping; Leuning, Ray; Cleugh, Helen; Hsu, Kuo Lin.

In: Journal of Hydrometeorology, Vol. 7, No. 1, 02.2006, p. 160-177.

Research output: Contribution to journalArticle

Abramowitz, Gab ; Gupta, Hoshin Vijai ; Pitman, Andy ; Wang, Yingping ; Leuning, Ray ; Cleugh, Helen ; Hsu, Kuo Lin. / Neural Error Regression Diagnosis (NERD) : A tool for model bias identification and prognostic data assimilation. In: Journal of Hydrometeorology. 2006 ; Vol. 7, No. 1. pp. 160-177.
@article{ae2ea7357f914f77ba5ea302bc4168cf,
title = "Neural Error Regression Diagnosis (NERD): A tool for model bias identification and prognostic data assimilation",
abstract = "Data assimilation in the field of predictive land surface modeling is generally limited to using observational data to estimate optimal model states or restrict model parameter ranges. To date, very little work has attempted to systematically define and quantify error resulting from a model's inherent inability to simulate the natural system. This paper introduces a data assimilation technique that moves toward this goal by accounting for those deficiencies in the model itself that lead to systematic errors in model output. This is done using a supervised artificial neural network to {"}learn{"} and simulate systematic trends in the model output error. These simulations in turn are used to correct the model's output each time step. The technique is applied in two case studies, using fluxes of latent heat flux at one site and net ecosystem exchange (NEE) of carbon dioxide at another. Root-mean-square error (rmse) in latent heat flux per time step was reduced from 27.5 to 18.6 W m-2 (32{\%}) and monthly from 9.91 to 3.08 W m-2 (68{\%}). For NEE, rmse per time step was reduced from 3.71 to 2.70 μmol m-2 s-1 (27{\%}) and annually from 2.24 to 0.11 μmol m-2 s-1 (95{\%}). In both cases the correction provided significantly greater gains than single criteria parameter estimation on the same flux.",
author = "Gab Abramowitz and Gupta, {Hoshin Vijai} and Andy Pitman and Yingping Wang and Ray Leuning and Helen Cleugh and Hsu, {Kuo Lin}",
year = "2006",
month = "2",
doi = "10.1175/JHM479.1",
language = "English (US)",
volume = "7",
pages = "160--177",
journal = "Journal of Hydrometeorology",
issn = "1525-755X",
publisher = "American Meteorological Society",
number = "1",

}

TY - JOUR

T1 - Neural Error Regression Diagnosis (NERD)

T2 - A tool for model bias identification and prognostic data assimilation

AU - Abramowitz, Gab

AU - Gupta, Hoshin Vijai

AU - Pitman, Andy

AU - Wang, Yingping

AU - Leuning, Ray

AU - Cleugh, Helen

AU - Hsu, Kuo Lin

PY - 2006/2

Y1 - 2006/2

N2 - Data assimilation in the field of predictive land surface modeling is generally limited to using observational data to estimate optimal model states or restrict model parameter ranges. To date, very little work has attempted to systematically define and quantify error resulting from a model's inherent inability to simulate the natural system. This paper introduces a data assimilation technique that moves toward this goal by accounting for those deficiencies in the model itself that lead to systematic errors in model output. This is done using a supervised artificial neural network to "learn" and simulate systematic trends in the model output error. These simulations in turn are used to correct the model's output each time step. The technique is applied in two case studies, using fluxes of latent heat flux at one site and net ecosystem exchange (NEE) of carbon dioxide at another. Root-mean-square error (rmse) in latent heat flux per time step was reduced from 27.5 to 18.6 W m-2 (32%) and monthly from 9.91 to 3.08 W m-2 (68%). For NEE, rmse per time step was reduced from 3.71 to 2.70 μmol m-2 s-1 (27%) and annually from 2.24 to 0.11 μmol m-2 s-1 (95%). In both cases the correction provided significantly greater gains than single criteria parameter estimation on the same flux.

AB - Data assimilation in the field of predictive land surface modeling is generally limited to using observational data to estimate optimal model states or restrict model parameter ranges. To date, very little work has attempted to systematically define and quantify error resulting from a model's inherent inability to simulate the natural system. This paper introduces a data assimilation technique that moves toward this goal by accounting for those deficiencies in the model itself that lead to systematic errors in model output. This is done using a supervised artificial neural network to "learn" and simulate systematic trends in the model output error. These simulations in turn are used to correct the model's output each time step. The technique is applied in two case studies, using fluxes of latent heat flux at one site and net ecosystem exchange (NEE) of carbon dioxide at another. Root-mean-square error (rmse) in latent heat flux per time step was reduced from 27.5 to 18.6 W m-2 (32%) and monthly from 9.91 to 3.08 W m-2 (68%). For NEE, rmse per time step was reduced from 3.71 to 2.70 μmol m-2 s-1 (27%) and annually from 2.24 to 0.11 μmol m-2 s-1 (95%). In both cases the correction provided significantly greater gains than single criteria parameter estimation on the same flux.

UR - http://www.scopus.com/inward/record.url?scp=33645535077&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33645535077&partnerID=8YFLogxK

U2 - 10.1175/JHM479.1

DO - 10.1175/JHM479.1

M3 - Article

AN - SCOPUS:33645535077

VL - 7

SP - 160

EP - 177

JO - Journal of Hydrometeorology

JF - Journal of Hydrometeorology

SN - 1525-755X

IS - 1

ER -