113
These results indicate that still a relatively high level of uncertainty is present in the
simulation of surface processes by today's ESEMs. Identifying the causes for the range
found in inter-model response is difficult, because of the large number of model parameters
and the differences in model formulations and set up. The effects that the use of different
ESEMs has on climate models is also difficult to quantify. Changes of a few to several tens
of W 1m 2 in surface fluxes are sufficient to affect the surface climatology of atmospheric
models, with temperature responses of a few to several degrees and precipitation responses
up to a few tens of percent (Dickinson 1992, Seller 1992). Overall, however, the simulation
of large scale structures of the general circulation in the mid and upper troposphere and
above is not very sensitive to the use of different surface schemes (Dickinson, pers. comm.).
One of the main areas of model uncertainty resides in the partitioning of precipitation
into evaporation and runoff, and in the partitioning of energy into latent and sensible heat.
This partitioning depends on surface morphology and characteristics which are spatially
variable on scales much smaller than those resolved by current climate models. The
issue of the representation of surface heterogeneity is thus central to improving surface
modeling. This issue is examined in the next section.
4. Description of surface heterogeneity effects in ESEMs
In the previous section we have seen that state-of-the-art ESEMs can reproduce quite
accurately the observed point-values of surface fluxes of energy and water vapor when
driven by observed meteorological forcings. This is partly because of the presence of
multiple parameters which can be tuned to fit specific datasets. We have also seen that
for the same climatic forcing a wide range of surface response is produced by different
ESEMs. Even if an ESEM was capable of perfectly reproducing the surface energy and
water budget at a given location, however, coupling to an AM would still introduce a
possibly large level of uncertainty because of the fine, sub-grid scale structure of the
surface characteristics.
Present three-dimensional global climate models are typically run at resolutions of a
few hundred km. The recent development of regional climate models, which cover only
limited area domains, can allow to reach resolution of a few tens of km (e.g. Giorgi and
Mearns 1991). However, the variability of surface characteristics such as vegetation type,
soil properties and terrain morphology is almost fractal in nature. This is illustrated, for
example, by high resolution maps of surface vegetation types as inferred from satellite
data (Loveland et al. 1991).
Usually, AMs assign at a given grid point either the dominant surface type or gridbox averaged surface characteristics. Because of the non-linear nature of many of the
processes we have described in the previous two sections, and because of the strongly
These results indicate that still a relatively high level of uncertainty is present in the
simulation of surface processes by today's ESEMs. Identifying the causes for the range
found in inter-model response is difficult, because of the large number of model parameters
and the differences in model formulations and set up. The effects that the use of different
ESEMs has on climate models is also difficult to quantify. Changes of a few to several tens
of W 1m 2 in surface fluxes are sufficient to affect the surface climatology of atmospheric
models, with temperature responses of a few to several degrees and precipitation responses
up to a few tens of percent (Dickinson 1992, Seller 1992). Overall, however, the simulation
of large scale structures of the general circulation in the mid and upper troposphere and
above is not very sensitive to the use of different surface schemes (Dickinson, pers. comm.).
One of the main areas of model uncertainty resides in the partitioning of precipitation
into evaporation and runoff, and in the partitioning of energy into latent and sensible heat.
This partitioning depends on surface morphology and characteristics which are spatially
variable on scales much smaller than those resolved by current climate models. The
issue of the representation of surface heterogeneity is thus central to improving surface
modeling. This issue is examined in the next section.
4. Description of surface heterogeneity effects in ESEMs
In the previous section we have seen that state-of-the-art ESEMs can reproduce quite
accurately the observed point-values of surface fluxes of energy and water vapor when
driven by observed meteorological forcings. This is partly because of the presence of
multiple parameters which can be tuned to fit specific datasets. We have also seen that
for the same climatic forcing a wide range of surface response is produced by different
ESEMs. Even if an ESEM was capable of perfectly reproducing the surface energy and
water budget at a given location, however, coupling to an AM would still introduce a
possibly large level of uncertainty because of the fine, sub-grid scale structure of the
surface characteristics.
Present three-dimensional global climate models are typically run at resolutions of a
few hundred km. The recent development of regional climate models, which cover only
limited area domains, can allow to reach resolution of a few tens of km (e.g. Giorgi and
Mearns 1991). However, the variability of surface characteristics such as vegetation type,
soil properties and terrain morphology is almost fractal in nature. This is illustrated, for
example, by high resolution maps of surface vegetation types as inferred from satellite
data (Loveland et al. 1991).
Usually, AMs assign at a given grid point either the dominant surface type or gridbox averaged surface characteristics. Because of the non-linear nature of many of the
processes we have described in the previous two sections, and because of the strongly