Coupled modelling of land surface microwave interactions using ENVISAT ASAR data [Elektronische Ressource] / vorgelegt von Alexander Löw

Löw, Charles Wesley Alexander

Informations

• Publié par: ludwig-maximilians-universitat_munchen
• Publié le: 01 janvier 2004
• Nombre de lectures: 36
• Langue: English
• Poids de l'ouvrage: 10 Mo

Extrait

Coupled modelling of land surface
microwave interactions using
ENVISAT ASAR data


Dissertation der Fakultät für Geowissenschaften
der Ludwig Maximilians Universität München





vorgelegt von:
Alexander Löw


Eingereicht: 25.08.2004




























1. Gutachter: Prof. Dr. Wolfram Mauser
2. Gutachter: Prof. Dr. Friedrich Wieneke

Tag der mündlichen Prüfung: 05.11.2004

II

Abstract
In the last decades microwave remote sensing has proven its capability to provide
valuable information about the land surface. New sensor generations as e.g.
ENVISAT ASAR are capable to provide frequent imagery with an high information
content. To make use of these multiple imaging capabilities, sophisticated
parameter inversion and assimilation strategies have to be applied. A profound
understanding of the microwave interactions at the land surface is therefore
essential.
The objective of the presented work is the analysis and quantitative description of
the backscattering processes of vegetated areas by means of microwave
backscattering models. The effect of changing imaging geometries is investigated
and models for the description of bare soil and vegetation backscattering are
developed. Spatially distributed model parameterisation is realized by synergistic
coupling of the microwave scattering models with a physically based land surface
process model. This enables the simulation of realistic SAR images, based on bio-
and geophysical parameters.
The adequate preprocessing of the datasets is crucial for quantitative image
analysis. A stringent preprocessing and sophisticated terrain geocoding and
correction procedure is therefore suggested. It corrects the geometric and
radiometric distortions of the image products and is taken as the basis for further
analysis steps.
A problem in recently available microwave backscattering models is the inadequate
parameterisation of the surface roughness. It is shown, that the use of classical
roughness descriptors, as the rms height and autocorrelation length, will lead to
ambiguous model parameterisations. A new two parameter bare soil backscattering
model is therefore recommended to overcome this drawback. It is derived from
theoretical electromagnetic model simulations. The new bare soil surface scattering
model allows for the accurate description of the bare soil backscattering coefficients.
A new surface roughness parameter is introduced in this context, capable to
describe the surface roughness components, affecting the backscattering
coefficient. It is shown, that this parameter can be directly related to the intrinsic
fractal properties of the surface.

III

Spatially distributed information about the surface roughness is needed to derive
land surface parameters from SAR imagery. An algorithm for the derivation of the
new roughness parameter is therefore suggested. It is shown, that it can be
derived directly from multitemporal SAR imagery.
Starting from that point, the bare soil backscattering model is used to assess the
vegetation influence on the signal. By comparison of the residuals between
measured backscattering coefficients and those predicted by the bare soil
backscattering model, the vegetation influence on the signal can be quantified.
Significant difference between cereals (wheat and triticale) and maize is observed in
this context.
It is shown, that the vegetation influence on the signal can be directly derived from
alternating polarisation data for cereal fields. It is dependant on plant biophysical
variables as vegetation biomass and water content.
The backscattering behaviour of a maize stand is significantly different from that of
other cereals, due to its completely different density and shape of the plants. A
dihedral corner reflection between the soil and the stalk is identified as the major
source of backscattering from the vegetation. A semiempirical maize backscattering
model is suggested to quantify the influences of the canopy over the vegetation
period.
Thus, the different scattering contributions of the soil and vegetation components
are successfully separated. The combination of the bare soil and vegetation
backscattering models allows for the accurate prediction of the backscattering
coefficient for a wide range of surface conditions and variable incidence angles.
To enable the spatially distributed simulation of the SAR backscattering coefficient,
an interface to a process oriented land surface model is established, which provides
the necessary input variables for the backscattering model. Using this synergistic,
coupled modelling approach, a realistic simulation of SAR images becomes possible
based on land surface model output variables. It is shown, that this coupled
modelling approach leads to promising and accurate estimates of the backscattering
coefficients. The remaining residuals between simulated and measured backscatter
values are analysed to identify the sources of uncertainty in the model. A detailed
field based analysis of the simulation results revealed that imprecise soil moisture
predictions by the land surface model are a major source of uncertainty, which can
be related to imprecise soil texture distribution and soil hydrological properties.

IV

The sensitivity of the backscattering coefficient to the soil moisture content of the
upper soil layer can be used to generate soil moisture maps from SAR imagery. An
algorithm for the inversion of soil moisture from the upper soil layer is suggested
and validated. It makes use of initial soil moisture values, provided by the land
surface process model. Soil moisture values are inverted by means of the coupled
land surface backscattering model. The retrieved soil moisture results have an RMSE
of 3.5 Vol %, which is comparable to the measurement accuracy of the reference
field data.
The developed models allow for the accurate prediction of the SAR backscattering
coefficient. The various soil and vegetation scattering contributions can be
separated. The direct interface to a physically based land surface process model
allows for the spatially distributed modelling of the backscattering coefficient and
the direct assimilation of remote sensing data into a land surface process model.
The developed models allow for the derivation of static and dynamic landsurface
parameters, as e.g. surface roughness, soil texture, soil moisture and biomass from
remote sensing data and their assimilation in process models. They are therefore
reliable tools, which can be used for sophisticated practice oriented problem
solutions in manifold manner in the earth and environmental sciences.


V

Zusammenfassung
Die Erkenntnisse der letzten Jahrzehnte haben gezeigt, dass sich aus Daten von
Mikrowellensensoren wertvolle Informationen über Eigenschaften und Prozesse der
Landoberfläche ableiten lassen. Neue Sensoren, wie beispielsweise der ENVISAT
ASAR, ermöglichen die häufige Abdeckung und Beobachtung eines Gebietes. Damit
werden sie für operationelle und insbesondere auch zeitkritische Anwendungen, wie
beispielsweise die Hochwasservorhersage interessant. Um dieses Potential nutzen
zu können ist es notwendig, die Effekte der daraus resultierenden unterschiedlichen
Aufnahmegeometrien zu kompensieren. Dazu sind problemorientierte,
anspruchsvolle Lösungsansätze notwendig. Grundlage hierfür sind Erkenntnisse über
die Rückstreumechanismen an der Landoberfläche unter verschiedenen
Aufnahmegeometrien.
Ein Schwerpunkt der vorliegenden Arbeit liegt in der Analyse und quantitativen
Beschreibung der Rückstreumechanismen von offene Böden, sowie
vegetationsbestandenen Flächen. Neue Ansätze zur theoretischen und
semiempirischen Beschreibung des Radarrückstreukoeffizienten werden hierzu
entwickelt. Unterschiedlichste Aufnahmegeometrien finden dabei Berücksichtigung.
Eine Grundvoraussetzung zur flächenhaften Modellierung der Radarrückstreuung ist
die flächige Bereitstellung der notwendigen Modelleingabeparameter. Dies wird
durch die Kopplung der Radarrückstreumodelle mit einem physikalisch basierten
Prozessmodell erreicht, welches die notwendigen bio- und geophysikalischen
Eingabeparameter flächig verteilt bereitstellen kann.
Unabdingbare Grundlage für die quantitative Auswertung der SAR Daten ist eine
adäquate und genaue geometrische und radiometrische Vorprozessierung der
Datensätze. Insbesondere den reliefbedingten geometrischen und radiometrischen
Einflüssen auf das Bildprodukt muss hierbei Rechnung getragen werden. Ein
entsprechendes, anspruchsvolles Korrekturverfahren zur Eliminierung der
reliefbedingten Lagefehler sowie radiometrischen Unterschiede wurde daher auf
Basis eines vorhandenen Verfahrens weiterentwickelt. Es ist die Grundlage für alle
weiteren quantitativen Auswertungen d