Forest biomass estimation from polarimetric SAR interferometry [Elektronische Ressource] / Tobias Mette

Administrator

Informations

• Publié par: technische_universitat_munchen
• Publié le: 01 janvier 2007
• Nombre de lectures: 30
• Langue: Deutsch
• Poids de l'ouvrage: 11 Mo

Extrait

Wissenschaftszentrum Weihenstephan
für Ernährung, Landnutzung und Umwelt
der Technischen Universität München



Forest Biomass Estimation from
Polarimetric SAR Interferometry




Dissertation von Tobias Mette

zur Erlangung des akademischen Grades eines Doktors rer.nat.


















durchgeführt am Institut für Hochfrequenztechnik und Radarsysteme
des Deutschen Zentrums für Luft- und Raumfahrt e.V.
vorgelegt am Lehrstuhl für Waldwachstumskunde der TU München Lehrstuhl für Waldwachstumskunde



Forest Biomass Estimation from
Polarimetric SAR Interferometry




Tobias Mette







Vollständiger Abdruck der von der Fakultät Wissenschaftszentrum Weihenstephan
für Ernährung, Landnutzung und Umwelt der Technischen Universität München
zur Erlangung des akademischen Grades des Doktors der Naturwissenschaften
(Dr. rer.nat.) genehmigten Dissertation.


Vorsitzender:
Univ.-Prof. Dr. Th. Knoke
Prüfer der Dissertation:
1. Univ.-Prof. Dr. H. Pretzsch (schriftliche Beurteilung)
1. Univ.-Prof. Dr. R. Mosandl (mündliche Beurteilung)
2. Univ.-Prof. Dr. A. Moreira (Universität Karlsruhe TH)
3. Univ.-Prof. Dr. R. Matyssek


Die Dissertation wurde am 11. Juli 2006 bei der Technischen Universität München
eingereicht und durch die Fakultät Wissenschaftszentrum Weihenstephan für
Ernährung, Landnutzung und Umwelt am 08. März 2007 angenommen. Abstract
[english] Polarimetric SAR interferometry (Pol-InSAR) is a radar remote sensing tech-
nique that allows to extract forest heights by means of model-based inversions. Forest
biomass is closely related to forest height, and can be derived from it with allometric re-
lations.
This work investigates the combination of the two methods to estimate forest biomass
from Pol-InSAR. It develops a concept for the use of height-biomass allometry, and out-
lines the Pol-InSAR height inversion. The methodology is validated against a set of for-
est inventory data and Pol-InSAR data at L-band of the test site Traunstein. The results
allow to draw conclusions on the potential of Pol-InSAR forest biomass missions.

[deutsch] Polarimetrische SAR Interferometrie (Pol-InSAR) ist eine Radar-
Fernerkundungs-Technik, die es ermöglicht, Waldhöhe mittels modell-basierter Inver-
tierung zu extrahieren. Waldbiomasse steht in engem Bezug zur Waldhöhe und lässt
sich mittels allometrischer Gleichungen daraus ableiten.
Diese Arbeit untersucht die Verknüpfung beider Methoden, mit dem Ziel, Biomasse aus
Pol-InSAR Daten zu schätzen. Ein Konzept zur Einbindung der Höhe-Biomasse Allo-
metrie wird entwickelt und die Höhen-Invertierung aus Pol-InSAR Daten dargelegt. Die
Methodologie wird anhand von Forst-Inventurdaten und L-Band ‚single baseline’ Pol-
InSAR Daten für eine Waldfläche in Traunstein (Süddeutschland) validiert. Die Ergeb-
nisse ermöglichen eine Einschätzung des Potentials von Pol-InSAR Waldbiomasse-
Bestimmungen. Preface i
Preface
I acknowledge and thank the following people who have contributed to this work:
Prof. Dr. Hans Pretzsch and Prof. Dr. Alberto Moreira for advice and encouragement,
Dr. Irena Hajnsek for continuous promotion, and Dr. Konstantinos Papathanassiou for
the entire supervision.
My colleagues in science for support, scientific interaction, corrections and proof read-
ing, only to name a few: Martin Hellmann, Ralf Horn, Angelo Liseno, Fifame Kou-
dogbo, Florian Kugler, Matteo Nannini, Maxim Neumann, Rolf Scheiber, Rafael
Schneider, Steffen Thölert, Reiner Zimmermann.
My family and friends for backing me up.

Munich, July 2006 Summary iii
Summary
The title, "Forest Biomass Estimation from Polarimetric SAR Interferometry", assigns
the presented work to the field of remote sensing based forest biomass estimations.
Biomass is the most integrative structural forest parameter and along with species com-
position the key parameter in forestry and ecology. Lately, it has been intensively dis-
cussed in terms of the global carbon cycle since ~50% of the dry plant matter consists of
carbon. However, current global estimates are still afflicted with great uncertainties.
The methodology proposed in this work is a two-step approach: First, forest height is
inverted from polarimetric interferometric (Pol-InSAR) radar data by using a coherent
forest scattering model. Then forest biomass is derived from the extracted forest heights
by means of allometric height-biomass relations. Both steps, the height extraction and
the height-biomass conversion, are based on physical and biological relationships, and
do not need to be parameterized from empirical regressions.
The challenge of this work was to link the two methodologies, Pol-InSAR height ex-
traction and forest height biomass allometry, and to investigate the potential of combin-
ing them in a forest biomass estimation from Pol-InSAR.
The first part of the work approaches the task from the forest side. The main concept of
how to use height-biomass allometry with remote sensing height data was addressed –
keeping in mind a possible global application. The idea was to establish an allometric
height-biomass reference function in the form of the allometric equation. This reference
function can be adjusted to different forest types by introducing an allometric level,
which serves as height-related density measure. However, before deriving the reference
allometry, it was necessary to agree on a rational definition of forest height, forest bio-
mass and consider a sensible forest type for the derivation of the reference allometry. In
this respect, upper canopy height and stem biomass were chosen as reference measures;
standard forestry yield tables were used to obtain the reference function.
Being a widespread, and economically important tree species, Norway spruce as de-
scribed by the yield tables of Assmann and Franz (1963, average yield level) was se-
lected for the definition of the reference function. Compared against the yield tables of
11 different common species the reference allometry of Norway spruce lies, for a given
height, at the upper biomass limit. Most other species reach allometric levels between
0.7 and 1.0 (70%-100%) with the exception of birch and poplar that are characterised by
significantly lower allometric levels. A tendency of a decreasing allometric level from
climax to pioneer character of the species was noted, but also regional differences seem
to play a role. On the contrary, the effect of site conditions – reflected by the yield
classes – was comparably low. For strong thinning intensities it was necessary to reduce
the biomass estimation by 10%-20%.
In the second part, the required radar remote sensing elements are introduced, discussed
and analyzed. Based on an interferometric system model the individual contributions of
the interferometric coherence are addressed. The most important contribution is the vol-
ume decorrelation which depends on the vertical scatterer distribution and allows to in-
vestigate the vertical structure of forests by means of interferometry. iv Summary
Interpreting the vertical forest dimension as a two-layer medium, in terms of the Ran-
dom-Volume-over-Ground (RVoG) model, the volume decorrelation is obtained as a
function of forest height, extinction, volume and ground scattering. Polarimetric SAR
interferometry allows to resolve these parameters and invert forest heights. Key of the
inversion is the polarimetric sensitivity to different scattering contributions – namely the
one of the ground. Since the RVoG inversion is based purely on volume decorrelation,
any system-related or temporal effects in the coherence must be compensated.
Finally, in the third and last part, the methodology is validated against available ground
measurements for the test site Traunstein. The test site is dominated by spruce stands,
-1which may reach heights up to 40 m and locally exceed biomass levels of 450 t ha . 20
homogeneous stands between 1 ha and 23 ha were selected for the validation. The radar
data were acquired with the E-SAR L-band sensor of DLR. The height inversion was
implemented on a fully polarimetric data set with a single interferometric baseline.
The concept of reference allometry and allometric level is validated first. Even if the
forest structure at the test site was not as homogeneous as in the yield tables, the power-
function character of the height-biomass allometry was clearly observed. Despite the
exceptionally well growing conditions of the site, it was necessary to adjust the allomet-
ric level to 0.87 (87 %) when regarding all inventory plots, and 0.92 (92 %) when con-
sidering only the inventory plots of the relatively homogeneous validation stands. The
low allometric level could partly be explained from the increased thinning in the regen-
eration stadiums, and the presence of leftover emergents in the youth stadium. The high