The Terrestrial Vegetation Model (TVM) of the Integrated Model to Asse
ss the Greenhouse Effect (IMAGE 2.0) determines potential land cover f
or natural ecosystems and potential productivity for agrosystems. A mo
del analysis of this TVM using Monte Carlo simulations is described fo
r the eighteen Costa Rican (0.5 degrees x 0.5 degrees latitude-longitu
de) IMAGE 2.0 grid cells. It is demonstrated that the TVM is not overl
y sensitive to climate input variability. Much observed output variabi
lity is related to model parameters. These model parameter related eff
ects can strongly determine the intra/inter grid variability for Costa
Rican potential yields. The model analysis indicates that more model
refinement is required to make any effort to improve data quality effe
ctive. It is therefore proposed to remove as many model criteria as po
ssible or to replace them by more gradual criteria based on crop ecoph
ysiological characteristics. A first tentative comparison of maize, ri
ce and pulses yield potentials and their distribution in Costa Rica du
ring 1973 and 1984, suggests that the assumption that the two an relat
ed, as assumed in the Land Cover Model (LCM) in IMAGE 2.0, has some va
lidity.