Coupled land surface and radiative transfer models for the analysis of passive microwave satellite observations

Soil moisture is one of the key variables controlling the water and
energy exchanges between Earth’s surface and the atmosphere.
Therefore, remote sensing based soil moisture information has
potential applications in many disciplines. Besides numerical
weather forecasting and climate research these include agriculture
and hydrologic applications like flood and drought forecasting. The
first satellite specifically designed to deliver operational soil
moisture products, SMOS (Soil Moisture and Ocean Salinity), was
launched 2009 by the European Space Agency (ESA). SMOS is a passive
microwave radiometer working in the L-band of the microwave domain,
corresponding to a frequency of roughly 1.4 GHz and relies on a new
concept. The microwave radiation emitted by the Earth’s surface is
measured as brightness temperatures in several look angles. A
radiative transfer model is used in an inversion algorithm to
retrieve soil moisture and vegetation optical depth, a measure for
the vegetation attenuation of the soil’s microwave emission. For
the application of passive microwave remote sensing products a
proper validation and uncertainty assessment is essential. As these
sensors have typical spatial resolutions in the order of 40 – 50
km, a validation that relies solely on ground measurements is
costly and labour intensive. Here, environmental modelling can make
a valuable contribution. Therefore the present thesis concentrates
on the question which contribution coupled land surface and
radiative transfer models can make to the validation and analysis
of passive microwave remote sensing products. The objective is to
study whether it is possible to explain known problems in the SMOS
soil moisture products and to identify potential approaches to
improve the data quality. The land surface model PROMET (PRocesses
Of Mass and Energy Transfer) and the radiative transfer model L-MEB
(L-band microwave emission of the Biosphere) are coupled to
simulate land surface states, e.g. temperatures and soil moisture,
and the resulting microwave emission. L-MEB is also used in the
SMOS soil moisture processor to retrieve soil moisture and
vegetation optical depth simultaneously from the measured microwave
emission. The study area of this work is the Upper Danube
Catchment, located mostly in Southern Germany. Since model
validation is essential if model data are to be used as reference,
both models are validated on different spatial scales with
measurements. The uncertainties of the models are quantified. The
root mean squared error between modelled and measured soil moisture
at several measuring stations on the point scale is 0.065 m3/m3. On
the SMOS scale it is 0.039 m3/m3. The correlation coefficient on
the point scale is 0.84. As it is essential for the soil moisture
retrieval from passive microwave data that the radiative transfer
modelling works under local conditions, the coupled models are used
to assess the radiative transfer modelling with L-MEB on the local
and SMOS scales in the Upper Danube Catchment. In doing so, the
emission characteristics of rape are described for the first time
and the soil moisture retrieval abilities of L-MEB are assessed
with a newly developed LMEB parameterization. The results show that
the radiative transfer modelling works well under most conditions
in the study area. The root mean squared error between modelled and
airborne measured brightness temperatures on the SMOS scale is less
than 6 – 9 K for the different look angles. The coupled models are
used to analyse SMOS brightness temperatures and vegetation optical
depth data in the Upper Danube Catchment in Southern Germany. Since
the SMOS soil moisture products are degraded in Southern Germany
and in different other parts of the world these analyses are used
to narrow down possible reasons for this. The thorough analysis of
SMOS brightness temperatures for the year 2011 reveals that the
quality of the measurements is degraded like in the SMOS soil
moisture product. This points towards radio frequency interference
problems (RFI), that are known, but have not yet been studied
thoroughly. This is consistent with the characteristics of the
problems observed in the SMOS soil moisture products. In addition
to that it is observed that the brightness temperatures in the
lower look angles are less reliable. This finding could be used to
improve the brightness temperature filtering before the soil
moisture retrieval. An analysis of SMOS optical depth data in 2011
reveals that this parameter does not contain valuable information
about vegetation. Instead, an unexpected correlation with SMOS soil
moisture is found. This points towards problems with the SMOS soil
moisture retrieval, possibly under the influence of RFI. The
present thesis demonstrates that coupled land surface and radiative
transfer models can make a valuable contribution to the validation
and analysis of passive microwave remote sensing products. The
unique approach of this work incorporates modelling with a high
spatial and temporal resolution on different scales. This makes
detailed process studies on the local scale as well as analyses of
satellite data on the SMOS scale possible. This could be exploited
for the validation of future satellite missions, e.g. SMAP (Soil
Moisture Active and Passive) which is currently being prepared by
NASA (National Aeronautics and Space Administration). Since RFI
seems to have a considerable influence on the SMOS data due to the
gained insights and the quality of the SMOS products is very good
in other parts of the world, the RFI containment and mitigation
efforts carried out since the launch of SMOS should be continued.