In the future research of this area, more detailed studies such as the spatial
and temporal dependence analysis should be conducted. Studies are also needed to
consider soil moisture information from other satellite missions over a wider range
of catchment conditions with different hydrological models in order to find generalisation patterns of the error distribution models (this is especially important for
ungauged catchments).
6 The Need for New Hydrological Soil Moisture Product
Development
Although there have been significant investments by various organisations such
as the European Space Agency (ESA), NASA and United States Department of
Agriculture (USDA) in a wide range of soil moisture observational programs
(e.g. satellite missions such as ASCAT, SMOS and SMAP) and ground-based
networks such as Soil Climate Analysis Network, US Surface Climate Observing
Reference Networks and COSMOS, they are not sufficiently used in hydrology
mainly because they are calibrated by in situ soil moisture measurements or
airborne retrievals which have significant spatial mismatch (both horizontally and
vertically) to catchment scales and are therefore less applicable to hydrological
modelling [103].
Similar to many satellites and soil moisture estimation algorithms, SMOS uses
the L-band Microwave Emission of the Biosphere (LMEB) model for the data
retrieval purpose [104]. LMEB is applied to estimate L-band brightness temperatures
(T b s) for a set of physical parameters, soil composition, and moisture content and
vegetation opacity [74]. In order to estimate soil moisture, the simulated T b s are
compared with those measured by SMOS using an iterative process to minimise
the difference between them. This approach then requires in situ observation data
for soil moisture evaluation [87, 97]. However most areas do not have in situ sensors
because they are expensive to set up and impractical to maintain; and they are too
sparse for catchment-scale studies [6–10]. Another problem of using this type of
method is that by decoupling the effects of soil properties and vegetation cover
can significantly reduce its soil moisture accuracy and hence its useful application
[105, 106].
In order to retrieve accurate soil wetness information that can be directly used
in a hydrological model and avoid aforementioned shortcomings, a need for a
data-driven model is desirable, which can effectively link the inputs to the desired
output and is not computationally intensive. Works carried out by [7, 21, 99, 107]
are good foundations for future hydrological soil moisture product development.
For example, in [99], three artificial intelligence techniques along with the generalised linear model are used to improve the spatial resolution of the SMOS-derived soil
moisture. The land surface temperature data retrieved from MODIS satellite is used
for the data downscaling, and SMD data calculated from a hydrological model called
Satellite Remote Sensing of Soil Moisture for Hydrological Applications. . .
273
and temporal dependence analysis should be conducted. Studies are also needed to
consider soil moisture information from other satellite missions over a wider range
of catchment conditions with different hydrological models in order to find generalisation patterns of the error distribution models (this is especially important for
ungauged catchments).
6 The Need for New Hydrological Soil Moisture Product
Development
Although there have been significant investments by various organisations such
as the European Space Agency (ESA), NASA and United States Department of
Agriculture (USDA) in a wide range of soil moisture observational programs
(e.g. satellite missions such as ASCAT, SMOS and SMAP) and ground-based
networks such as Soil Climate Analysis Network, US Surface Climate Observing
Reference Networks and COSMOS, they are not sufficiently used in hydrology
mainly because they are calibrated by in situ soil moisture measurements or
airborne retrievals which have significant spatial mismatch (both horizontally and
vertically) to catchment scales and are therefore less applicable to hydrological
modelling [103].
Similar to many satellites and soil moisture estimation algorithms, SMOS uses
the L-band Microwave Emission of the Biosphere (LMEB) model for the data
retrieval purpose [104]. LMEB is applied to estimate L-band brightness temperatures
(T b s) for a set of physical parameters, soil composition, and moisture content and
vegetation opacity [74]. In order to estimate soil moisture, the simulated T b s are
compared with those measured by SMOS using an iterative process to minimise
the difference between them. This approach then requires in situ observation data
for soil moisture evaluation [87, 97]. However most areas do not have in situ sensors
because they are expensive to set up and impractical to maintain; and they are too
sparse for catchment-scale studies [6–10]. Another problem of using this type of
method is that by decoupling the effects of soil properties and vegetation cover
can significantly reduce its soil moisture accuracy and hence its useful application
[105, 106].
In order to retrieve accurate soil wetness information that can be directly used
in a hydrological model and avoid aforementioned shortcomings, a need for a
data-driven model is desirable, which can effectively link the inputs to the desired
output and is not computationally intensive. Works carried out by [7, 21, 99, 107]
are good foundations for future hydrological soil moisture product development.
For example, in [99], three artificial intelligence techniques along with the generalised linear model are used to improve the spatial resolution of the SMOS-derived soil
moisture. The land surface temperature data retrieved from MODIS satellite is used
for the data downscaling, and SMD data calculated from a hydrological model called
Satellite Remote Sensing of Soil Moisture for Hydrological Applications. . .
273