Abstract
The Simple Biosphere Model, version 2 (SiB2), was designed for use within atmospheric general circulation models as a soil–vegetation–atmosphere transfer scheme that includes CO2 flux prediction. A stand-alone version of SiB2 was used to simulate a grassland at Station 16 of the First ISLSCP Field Experiment (FIFE) located near Manhattan, Kansas, for a period of 142 days of the 1987 growing season. Modeled values of soil temperature and moisture were initialized, using field measurements from the soil profile, and thereafter updated solely by model calculations. The model was driven by half-hourly atmospheric observations and regular observations of canopy biophysics. This arrangement was intended to mimic model forcing in a GCM. Three model versions are compared: (i) a Control run using parameter values taken from look-up tables used for running the Colorado State University GCM; (ii) a Tuned run with many adjustments to optimize SiB2 to this ecosystem; and (iii) a Calibrated run, which calibrated the Control version soil to the local site and incorporated two important changes from the Tuned version. Modeled fluxes of latent heat, sensible heat, soil heat, net radiation, and net site CO2 were compared to over 800 half-hourly observations; modeled surface and deep soil temperatures compared to 6500 observations; and three layers of modeled soil water content compared to 15 measurements of the soil water profile. Statistical methods were used to analyze these results. In the absence of water stress all three versions accurately simulated photosynthesis and canopy conductance. However, during episodes of drought, only the Tuned and Calibrated versions accurately simulated physiological control of canopy fluxes. The largest errors were encountered in the simulation of soil respiration. These were traced to problems predicting water content and temperature in the soil profile. These results highlight the need for improved simulation of soil biophysics to obtain accurate estimates of net CO2 balance. The accuracy of the Tuned version was improved by changes that (i) allowed water extraction by roots from all soil layers, (ii) matched the soil texture specification to the site, and (iii) calibrated the expressions used for diffusion of water and heat within the soil profile.
* Deceased 17 July 1996.
Corresponding author address: Dr. Joseph A. Berry, Department of Plant Biology, Carnegie Institution of Washington, Stanford, CA 94305.
Email: gcolello@mendelbio.com