This paper investigates the problem of aggregation of land surface pro
perties in a large area comparable in size with a model grid box of a
GCM. The study is based on 3D numerical results obtained for atmospher
ic situations encountered during HAPEX-MOBILHY 1986-that is. moderate
plant water stress at the beginning of summer. Estimating effective su
rface properties describing the spatial distribution of the vegetation
and the soil texture within the large area under study is proposed as
a first guess for accounting for spatial variability. Despite the non
linear dependence of surface fluxes on both vegetation and soil water
content, it is found that the effective surface fluxes computed from e
ffective parameters with a 1D column model match the areal-averaged fl
uxes estimated from 3D mesoscale model results with a relative error l
ess than 10%. On the other hand, fluxes computed with prescribed surfa
ce properties associated with the dominant land use of the large domai
n depart significantly from the averaged fluxes. For the cases examine
d. the effects of nonlinearity are found to be smaller for the vegetat
ion behavior than for the soil water transfers. The parameter aggregat
ion method has been tested successfully for a long time period within
the context of a 1D GCM grid cell representing the HAPEX-MOBILHY 1986
instrumented area. Given precipitation and solar radiation fluxes. pre
dictions of soil water content and total evaporation for 25 days compa
re well with aggregated observations within the large area.