LightResponseCurveFitter_fitLRC: LightResponseCurveFitter fitLRC

View source: R/LRC_base.R

LightResponseCurveFitter_fitLRCR Documentation

LightResponseCurveFitter fitLRC

Description

Optimize rectangular hyperbolic light response curve in one window

Usage

LightResponseCurveFitter_fitLRC(dsDay, E0, 
    sdE0, RRefNight, controlGLPart = partGLControl(), 
    lastGoodParameters = rep(NA_real_, 7L))

Arguments

dsDay

data.frame with columns NEE, Rg, Temp_C, VPD, and no NAs in NEE

E0

temperature sensitivity of respiration

sdE0

standard deviation of E_0.n

RRefNight

basal respiration estimated from night time data

controlGLPart

further default parameters (see partGLControl)

lastGoodParameters

numeric vector returned by last reasonable fit

Details

Optimization is performed for three initial parameter sets that differ by beta0 (* 1.3, * 0.8). From those three, the optimization result is selected that yielded the lowest misfit. Starting values are: k = 0, beta = interpercentileRange(0.03, 0.97) of respiration, alpha = 0.1, R_ref from nightTime estimate. E0 is fixed to the night-time estimate, but varies for estimating parameter uncertainty.

If controlGLPart$nBootUncertainty == 0L then the covariance matrix of the parameters is estimated by the Hessian of the LRC curve at optimum. Then, the additional uncertainty and covariance with uncertainty E0 is neglected.

If controlGLPart.l$nBootUncertainty > 0L then the covariance matrix of the parameters is estimated by a bootstrap of the data. In each draw, E0 is drawn from N ~ (E_0, sdE_0).

If there are no estimates for more than 20% of the bootstrapped samples The an NA-result with convergence code 1001L is returned.

Value

a list, If none of the optimizations from different starting conditions converged, the parameters are NA.

thetaOpt

numeric vector of optimized parameters including the fixed ones and E0

iOpt

index of parameters that have been optimized, here including E0, which has been optimized prior to this function.

thetaInitialGuess

the initial guess from data

covParms

numeric matrix of the covariance matrix of parameters, including E0

convergence

integer code specifying convergence problems: \ 0: good convergence \ , 1-1000: see optim \ , 1001: too few bootstraps converged \ , 1002: fitted parameters were outside reasonable bounds \ , 1003: too few valid records in window \ , 1004: near zero covariance in bootstrap indicating bad fit \ , 1005: covariance from curvature of fit yielded negative variances indicating bad fit \ , 1006: prediction of highest PAR in window was far from saturation indicating insufficient data to constrain LRC \ , 1010: no temperature-respiration relationship found \ , 1011: too few valid records in window (from different location: partGLFitLRCOneWindow) \

Author(s)

TW, MM Department for Biogeochemical Integration at MPI-BGC, Jena, Germany <REddyProc-help@bgc-jena.mpg.de> [cph], Thomas Wutzler <twutz@bgc-jena.mpg.de> [aut, cre], Markus Reichstein <mreichstein@bgc-jena.mpg.de> [aut], Antje Maria Moffat <antje.moffat@bgc.mpg.de> [aut, trl], Olaf Menzer <omenzer@bgc-jena.mpg.de> [ctb], Mirco Migliavacca <mmiglia@bgc-jena.mpg.de> [aut], Kerstin Sickel <ksickel@bgc-jena.mpg.de> [ctb, trl], Ladislav <U+0160>igut <sigut.l@czechglobe.cz> [ctb]

See Also

partGLFitLRCWindows

LightResponseCurveFitter_optimLRCBounds


bgctw/REddyProc documentation built on March 26, 2024, 11:35 p.m.