Causes of interannual variability in ecosystem-atmosphere CO₂ exchange in a northern Wisconsin forest using a Bayesian model calibration.
Authors:
Ricciuto, Daniel M. Butler, Martha P. Davis, Kenneth J. Cook, Bruce D. Bakwin, Peter S. Andrews, Arlyn Teclaw, Ronald M. USDA, FS
Source:
Agricultural and forest meteorology. 2008 Feb. 13, v. 148, no. 2 [Oxford]: Elsevier Science Ltd., p. 309-327.
NALT Subjects:
temperate forests carbon dioxide gas exchange temporal variation soil-plant-atmosphere interactions Bayesian theory simulation models calibration model validation carbon sequestration turbulent flow data analysis friction diurnal variation seasonal variation soil water content soil temperature soil respiration cell respiration Monte Carlo method Wisconsin
Other Subjects:
net ecosystem exchange
Issue Date:
13-Feb-2008
Abstract:
Variability in fluxes of CO₂ observed at the WLEF tall tower in northern Wisconsin was analyzed for the years 1997-2004. During this time, the WLEF region was a source of CO₂ to the atmosphere averaging 120gCm⁻² year⁻¹, with a range of interannual variability of 140gCm⁻² year⁻¹. Random uncertainty in annual sums of net ecosystem exchange (NEE) due to turbulent variability and gap-filling was estimated to be 15-20gCm⁻² year⁻¹. Although magnitudes of NEE sums were affected systematically by the choice of friction velocity (u*) threshold, this choice had little effect on interannual variability of annual sums. The WLEF region was, on average, a source of carbon from 1997 to 2004 regardless of the u* threshold applied. Interannually, daytime NEE sums varied more than nighttime NEE sums, and spring and summer NEE sums varied more than autumn and winter NEE sums. Interannual variations in seasonal sums of daytime, nighttime and total NEE were often strongly correlated with changes in soil moisture and soil temperature. Standard nonlinear gap-filling regression models of respiration and gross ecosystem productivity were extended to incorporate the effects of soil moisture and phenology and combined into a single model of NEE. The Markov Chain Monte Carlo (MCMC) data assimilation technique was performed using observed WLEF NEE to derive full probability density functions (PDFs) of time-invariant model parameters. Prior values had little effect on posterior parameter PDFs, but significant differences in parameter PDFs occurred depending on whether daytime NEE, nighttime NEE, or total NEE data were used. This simple model was moderately successful in producing statistically significant correlations with interannual variations in annual and growing season NEE sums, but was generally unsuccessful in spring and autumn. In all cases, the model underestimated the degree of variability in NEE sums.
