1 Spatial Modelling of the Terrestrial Environment - Georeferencial

Coupled Land Surface and Microwave Emission Models 71
comparatively small error, but potentially significant, particularly when there are already
other sources of errors in the retrieval algorithm.
Crow and Wood (2002) demonstrated that area-average soil moisture, when used in a
land surface model, can lead to potentially large errors in surface energy fluxes of up to 50 W
m −2 . One method of incorporating subgrid scale heterogeneity into land surface models is
to represent such heterogeneity using a probability density function (PDF) (e.g. Famiglietti
and Wood, 1995). However, if the model grid scale corresponds to the scale of the satellite
footprint for which soil moisture information is available, estimating subgrid statistics is not
a trivial task. Crow and Wood (2002) discuss a possible soil moisture downscaling procedure
based on an assumption of spatial scaling (i.e., a power-law relationship between statistical
moments and scale), and demonstrate that the downscaled soil moisture derived from coarseresolution
soil moisture imagery can improve prediction of grid-scale surface energy fluxes.
Recently, Kim and Barros (2002) developed a modified fractal interpolation technique to
downscale soil moisture estimates, based on an analysis of the effects of topography, soil
properties, and vegetation on the measured distribution of soil moisture. Reichle et al.
(2001) used a four-dimensional assimilation algorithm to show that soil moisture can be
satisfactorily estimated at scales finer than the resolution of the microwave brightness
temperature image. Their downscaling experiment suggests that brightness temperature
images with a resolution of tens of kilometres can yield soil moisture profile estimates on a
scale of a few kilometres, provided that micrometeorological, soil texture, and land cover
inputs are available at the finer scale.
The high-resolution active and low-resolution passive combination of the HYDROS
mission may provide an opportunity to develop a novel method of downscaling. Bindlish and
Barros (2002) have demonstrated the potential of this using an L-band satellite-based radar
and an L-band aircraft-based radiometer. They successfully demonstrated downscaling of
soil moisture estimates from 200 m to 40 m.
The SMOS mission may also provide enough information to allow downscaled estimates
of soil moisture. Burke et al. (2002a) explored this potential by using MICRO-SWEAT
in a year-long simulation to define the patch-specific soil moisture, optical depth, and the
synthetic, pixel-average microwave brightness temperatures for the range of angles that will
be provided by SMOS. The microwave emission component of MICRO-SWEAT was used
as the basis of an exploratory SMOS retrieval algorithm in which the RMSE between the
synthetic and modelled pixel-average microwave brightness temperatures was minimized
by optimizing the soil moisture and optical depth in different patches of vegetation. The
optimization was made using the Shuffled Complex Evolution optimization procedure
(Duan et al., 1993). An example five-patch pixel comprising equal areas of water, bare soil,
short grass, crop, and shrub was studied, assuming a uniform sandy loam soil (75% sand,
5% clay). Simulation of the growth of the vegetation through the year was also included.
It is assumed that the portion of the pixel occupied by each land-cover type was known
and that each vegetation type was homogeneous. Figure 4.5 shows the prescribed and
retrieved near-surface soil moisture and vegetation optical depth for the four land patches
(it was assumed that the contribution of water to the area-average brightness temperature
is known). The last plot in Figure 4.5 is the area weighted mean soil moisture (excluding
the water patch) for the pixel. The SCE-UA algorithm, which was run ten times, gave
each time, ten possible values of soil moisture, optical depth and ten possible values of
the errors in brightness temperature. The ten possible solutions are plotted as individual