82
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
fraction based on the scatterplot between the albedo and surface radiometric temperature. The analysis of models’ performances was realized by adopting two different data sets: the first one built up by seven airborne high-resolution images acquired
between June and October 2008, and the second data set acquired during the spring
and summer of 2005, including an ATM and CASI-2 high-resolution images as well
as an ASTER moderate resolution multispectral image. A preliminary validation of
the TSEB-modeled fluxes was performed using micrometeorological data acquired
over an olive field by means of an EC and scintillometer stations. The validation
highlights a good performance of the TSEB model, with average errors on ET d of
about 0.5 mm day –1 .
As a general conclusion, the analyses assess that the SEBAL model causes an overestimation of sensible heat fluxes H mainly due to an underestimation of the aerodynamic resistance to heat transport, which does not take into account the soil–canopy
interaction, as well as to difficulties to accurately define boundary conditions at low
spatial resolution. By focusing on the ET d evaluation, the effect can be compensated by
an overestimation of soil heat flux, especially for the high-spatial-resolution airborne
images. This compensation of fluxes in the single-source model, however, produces
similar ET d estimations to that retrieved by the more detailed two-source model.
The analysis of the three models’ differences (assuming the previously validated TSEB as reference) highlighted a good agreement between the average ET d
assessments on the whole scene. The comparison of MAD and RE statistical indices in five fields of the study area emphasized the greater discrepancies in areas
characterized by low vegetation coverage (vineyards) in case of high atmospheric
water demand. Additionally, in some cases, areas having high vegetation coverage
(citrus orchards) also showed significant differences. These areas, which generally
correspond to the boundary conditions in the self-calibration procedure adopted in
SEBAL and S-SEBI, seem to be the more sensitive to the arbitrary parameterization
of these models. Moreover, clearly, dates characterized by high water availability
show negligible or at least less significant ET d differences among models’ retrievals. Differences found on vineyards, characterized by strong heterogeneity and low
vegetation coverage, suggest the need for further improvement on the modeling of
energy flux partition for these kinds of crops.
Pixel-size dimension is crucial for surface energy balance applications over agricultural fields that are highly fragmented (of the order of hectares or less). From this
point of view, the case study showed that the spatial resolution of the ASTER thermal band could be considered as an upper limit to accurately identify the spatial distribution of ET d with TSEB. This is mainly due to the model capability to correctly
identify soil and canopy contributions to the surface energy budget. On the contrary,
the ASTER resolution is not appropriate to apply the SEBAL single-source model
over agricultural fragmentized landscape, since the hypothesis of homogeneous land
cover is not achieved and due to the complexity to define adequate boundary conditions, especially over small study areas. For all these reasons, the TSEB approach
seems the more “operational” method to be used with commonly available moderate
resolution remote sensing data, thus removing the dependency from an arbitrary
selection of boundary conditions, which is a strong limitation in a highly fragmented
landscape and in areas characterized by an elevated degree of water stress.
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
fraction based on the scatterplot between the albedo and surface radiometric temperature. The analysis of models’ performances was realized by adopting two different data sets: the first one built up by seven airborne high-resolution images acquired
between June and October 2008, and the second data set acquired during the spring
and summer of 2005, including an ATM and CASI-2 high-resolution images as well
as an ASTER moderate resolution multispectral image. A preliminary validation of
the TSEB-modeled fluxes was performed using micrometeorological data acquired
over an olive field by means of an EC and scintillometer stations. The validation
highlights a good performance of the TSEB model, with average errors on ET d of
about 0.5 mm day –1 .
As a general conclusion, the analyses assess that the SEBAL model causes an overestimation of sensible heat fluxes H mainly due to an underestimation of the aerodynamic resistance to heat transport, which does not take into account the soil–canopy
interaction, as well as to difficulties to accurately define boundary conditions at low
spatial resolution. By focusing on the ET d evaluation, the effect can be compensated by
an overestimation of soil heat flux, especially for the high-spatial-resolution airborne
images. This compensation of fluxes in the single-source model, however, produces
similar ET d estimations to that retrieved by the more detailed two-source model.
The analysis of the three models’ differences (assuming the previously validated TSEB as reference) highlighted a good agreement between the average ET d
assessments on the whole scene. The comparison of MAD and RE statistical indices in five fields of the study area emphasized the greater discrepancies in areas
characterized by low vegetation coverage (vineyards) in case of high atmospheric
water demand. Additionally, in some cases, areas having high vegetation coverage
(citrus orchards) also showed significant differences. These areas, which generally
correspond to the boundary conditions in the self-calibration procedure adopted in
SEBAL and S-SEBI, seem to be the more sensitive to the arbitrary parameterization
of these models. Moreover, clearly, dates characterized by high water availability
show negligible or at least less significant ET d differences among models’ retrievals. Differences found on vineyards, characterized by strong heterogeneity and low
vegetation coverage, suggest the need for further improvement on the modeling of
energy flux partition for these kinds of crops.
Pixel-size dimension is crucial for surface energy balance applications over agricultural fields that are highly fragmented (of the order of hectares or less). From this
point of view, the case study showed that the spatial resolution of the ASTER thermal band could be considered as an upper limit to accurately identify the spatial distribution of ET d with TSEB. This is mainly due to the model capability to correctly
identify soil and canopy contributions to the surface energy budget. On the contrary,
the ASTER resolution is not appropriate to apply the SEBAL single-source model
over agricultural fragmentized landscape, since the hypothesis of homogeneous land
cover is not achieved and due to the complexity to define adequate boundary conditions, especially over small study areas. For all these reasons, the TSEB approach
seems the more “operational” method to be used with commonly available moderate
resolution remote sensing data, thus removing the dependency from an arbitrary
selection of boundary conditions, which is a strong limitation in a highly fragmented
landscape and in areas characterized by an elevated degree of water stress.