© 2006 Heron Publishing—Victoria, Canada
Parameter sensitivity and uncertainty of the forest carbon flux model FORUG: a Monte Carlo analysis
Hans Verbeeck (1, 3, 4), Roeland Samson (1), Frederik Verdonck (2) and Raoul Lemeur (1)
1. Laboratory of Plant Ecology, Ghent University, Coupure Links 653, 9000 Ghent, Belgium / 2. Department of Applied Mathematics, Biometrics and Process Control, Ghent University, Coupure Links 653, 9000 Ghent, Belgium / 3. Research Group of Plant and Vegetation Ecology, Department of Biology, University of Antwerp, Universiteitsplein 1, 2610 Antwerp,
Belgium / 4. Corresponding author (Hans.Verbeeck@ua.ac.be) / Received April 13, 2005; accepted September 14, 2005; published online March 1, 2006
Summary
The Monte Carlo technique can be used to propagate input variable uncertainty and parameter uncertainty through a model to
determine output uncertainty. However, to carry out Monte Carlo simulations, the uncertainty distributions or the probability
density functions (PDFs) of the model parameters and input variables must be known. This remains one of the bottlenecks in
current uncertainty research in forest carbon flux modeling. Because forest carbon flux models involve many parameters, we
questioned whether it is necessary to take into account all parameters in the uncertainty analysis. A sensitivity analysis
can determine the parameters contributing most to the overall model output uncertainty. This paper illustrates the usefulness
of the Monte Carlo simulation technique for ranking parameters for sensitivity and uncertainty in process-based forest flux
models.
The uncertainty of the output (net ecosystem exchange, NEE) of the FORUG model was estimated for the Hesse beech forest (1997).
Based on the arbitrary uncertainty of ten key parameters, a standard deviation of 0.88 Mg C ha–1 year–1 NEE was found which is equal to 24% of the mean value of NEE. Sensitivity analysis showed that the overall output uncertainty
of the FORUG model can largely be determined by accounting for the uncertainty of only a few key parameters. The results led
to the identification of the key FORUG parameters and to the recommendation for a process-based description of the soil respiration
process in the FORUG model.
Keywords:
least square linearization, photosynthesis, probability density function, soil respiration.
