Development and Extrapolation of a General Light Use Efficiency Model for the Gross Primary Production [Elektronische Ressource] / Judith Horn. Betreuer: Karsten Schulz

Judith Horn

Informations

• Publié par: ludwig-maximilians-universitat_munchen
• Publié le: 01 janvier 2011
• Nombre de lectures: 35
• Langue: English
• Poids de l'ouvrage: 16 Mo

Extrait

Development and Extrapolation of a
General Light Use Efficiency Model
for the Gross Primary Production
Dissertation der Fakulät für Geowissenschaften der
Ludwig–Maximilians–Universität München
zur Erlangung des akademischen Grades
"doctor rerum naturalium" (Dr. rer. nat.)
vorgelegt von
Judith E. Horn
aus Offenburg
München, den 14.2.2011Erstgutachter: Prof. Dr. KarstenSchulz
Zweitgutachter: Prof. Dr. AndreasHuth
TagdermündlichenPrüfung: 13.7.2011"That’sthewholeproblemwithscience. You’ve
gotabunchofempiriciststryingtodescribe
thingsofunimaginablewonder."
Bill Watterson in "Calvin And Hobbes"Acknowledgments
This PhD study was funded by the Deutsche Forschungsgemeinschaft (DFG, Grant
#SCHU1271/4 1&2). ItwasmadepossiblebyaccesstotheFLUXNETdatabase;the
data are provided by the data centers "CarboEuropeIP Ecosystem Component Database"
supported by the European Commission, as well as the"AmeriFlux" data archive at the
"CarbonDioxideInformationAnalysisCenter"(CDIAC)ofDOE’s(U.S.Departmentof
Energy) Oak Ridge National Laboratory(ORNL).
IamverygratefultoVanessaStauch. In hercheerful andwarmmanner sheintroduced
meto Leipzigandthe workinggroup. She made it easy for me to become familiar with
thedata and required methods andthephilosophybehindtheproject.
I would like to thank Prof. Andreas Huth for spontaneously accepting the second review
ofthisthesis.
IwishtoexpressmywarmestgratitudetomyadvisorProf. KarstenSchulzforsupporting
andencouragingmeoverthepastyears. HegavenewimpulseswhenIwasstuckand
sharedhisthoughts,experiencesandinspiringideaswithme. Andhetaughtme: Science
canbefun!
My thanks also go to my brothers Pirmin, Dominik and Simon for being the reliable
companionstheyare;IamverygladIhavethesethreebrothersandIknowIcancount
on them when it matters. I always enjoy the refreshing and inspiring talks with them on
the meaningful senselessness and the important things in life. Thanks to Virginia for her
valuable assistance in the field ofgraphicdesign.
I am indebted to my parents and thank them sincerely for their unconditional love,
support and faith in me. It is themIwanttodedicatethisworkto.
And last but not least, I thank you with all my heart, Peter, for all the love you give.
Thankyouforyourunlimitedsupport,foryourstrongconfidenceinme,foryourinfinite
optimisms, for sharing your worries, happiness and dreams with me. Thank you for
beingthere.Summary
The globalcarboncycleissubstantiallyinfluencedbytheterrestrialbiosphere. Against
the background of accelerating global change and in particular of the rising CO -concen-
2
tration in the atmosphere and associated climate change, the scientific community is
highlyinterestedinanalyzingandunderstandingthedynamicsoftheglobalcarboncycle.
Thehighlycomplexprocessesdeterminingthegrossprimaryproduction–thecarbon
uptakeandassimilationbyphotosyntheticallyactiveplantsatecosystemlevel–areof
particularrelevanceinthiscontextandformthefocusofthisthesis. Onlyacomprehensive
observation network and thorough analyses of those interdependencies enable scientists
toachievetheobjectiveofdevelopingfuturescenariosand,finally,establishingadaptation
andmitigation strategies to combattheconsequencesofclimatechange.
To meet this challenge, measurement networks such as FLUXNET have been initiated.
Aroundtheworld,FLUXNETparticipantscollectandprovidedataontheexchangeof
energy and matter between vegetation stands and the atmosphere with the so-called
eddy-covariance (EC) technique. These EC measurements support the identification
and understanding of ecosystem processes and are indispensable for purposes of devel-
oping,calibratingandvalidatingsimulationmodels. However,ifthesesoil-vegetation-
atmosphere-transfer(SVAT)modelsaretobeappliedonlargerscales,spatiallycontinuous
inputdataare needed. Remote sensingasonlysourceoflarge-scaleinformationonthe
land surface can serve this purpose. MODIS, a satellite sensor specifically designed
for this task, has proven to be a key sensor for observing ecosystem states and track-
ingchanges. Modeling,ECmeasurementsandremotesensingcomplementeachother
synergistically.
ExistingSVATmodelshaveshowntosufferfromlimitations: Amismatchbetweenthe
smallscaleandhighcomplexityoftheprocessdescriptionsimplementedinthemodel
andthelargermodelapplicationscalewithscarcedatasourceshasbeendetected. This
mismatchcanrenderparametercalibrationsdifficultandpredictionsuncertain. Other
models specifically established for large-scale applications are, on the contrary, typically
formulatedwithparsimoniousmodelstructures. Theyareparametrizedwithfixedor
coarselygroupedvaluesaswellaspredefinedrelationshipsbetweenecosystemprocesses
andenvironmentaldriversthatoftendonotholdwhencomparedwithECmeasurement
data.
Theoverarchingobjectiveofthisthesisistoexploitthepowerfulcombinationofmodeling,
EC measurements and remote sensing in order to develop a simple but robust model
for the gross primary production of vegetation stands. Furthermore, an extrapolationviii Summary
scheme issoughtwithwhichthemodelparameterscalibratedatFLUXNETsitescanbe
regionalized. Toachievethisaim,thewell-establishedconceptoflightuseefficiencyis
chosenasmodelingframework. Thebasicequationissuccessivelyrefineduntilthemodel
complexity commensurates with the observations made at daily time steps. To serve this
purpose, non-linear data analysis tools are applied to exploit the information content
of EC measurement data. The functional forms describing the relationships between
the photosynthetic gross CO flux and environmental drivers are directly identified
2
from the measurement data without making strong a priori assumptions. In this way,
physical process understanding is coupled with the information on ecosystem processes
as reflected in the measurement data. To apply the MODIS data products in a best
possiblewayandavoidusagepitfalls,athoroughanalysisofacoreMODISproductused
throughout this thesis is carried out.
Thederivedmodeldemandsnotmorethanthreeenvironmentalvariables,namelytheab-
sorbed photosynthetically active radiation, temperature and a water availability measure.
Despite its simplicity, the model captures a great proportion of the day-to-day variations
of the gross primary production as measured at the study sites. The unique model struc-
ture accounts for variable influences of temperature and droughts on the photosynthetic
gross CO flux at different sites. This inherent model characteristic renders the model
2
widely applicable and enables its usage for differing vegetation types and environmental
conditions ranging from boreal needleleaf forests to semi-arid grasslands. The optimized
set of model parameters is well defined and the model uncertainty due to the parameter
calibration is generally found to below.
To allowformodel applicationsat largerscales, thecalibratedmodel parametersare re-
latedtoclimaticandbiophysicalsitecharacteristicsbymeansofsupportvectorregression,
apowerfulmachinelearningtechnique. Anovelframeworkissetupwhichautomatically
andobjectivelyselectstheexplanatoryfeaturesforeachmodelparameteroutofalarge
set of site characteristics. In a cross-validation, the time series of the photosynthetic
grossCO fluxmodeledwiththe extrapolatedparameterscorrelatesverywellwith the
2
measured dynamics. Likewise, the cumulative sums of the gross primary production as
stringentperformancecriterioncomparesatisfyinglywiththemeasuredsums. Overall,
the performed cross-validation proved the proposed scheme to be highly suitable for the
extrapolationofmodelparametersandthusallowsthemodelapplicationatlargerscales.
The modeling and extrapolation framework presented in this thesis contributes substan-
tiallytotheeffortsofthescientificcommunitytopredictthegrossprimaryproduction
underfutureenvironmentalconditions. Thedata-drivenapproachfollowedinthisthesis
likewise appears to be appropriate for modeling other ecosystem fluxes such as the evap-
otranspiration. This crucial component of the water cycle, which is intrinsically linked to
thecarbon cycle, is also of major importanceinachangingenvironment.Contents
Acknowledgments v
Summary vii
Table of Contents xi
List of Figures xiv
List of Tables xv
List of Symbols andAbbreviations xix
Chapter 1 Introduction 1
1.1. Thematic Background andMotivation . . . . . . . . . . . . . . . . . . . . . 1
1.1.1. The Terrestrial BiosphereandtheCarbonCycle . . . . . . . . . . . 1
1.1.2. MicrometeorologicalMeasurements . . . . . . . . . . . . . . . . . . 2
1.1.3. Remote Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.4. Modeling of GrossPrimaryProduction . . . . . . . . . . . . . . . . 4
1.1.5. Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2. Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1. Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.2. Investigating UsageOptionsofMODISLAI/FPARData . . . . . . 9
1.2.3. Deriving a Light UseEfficiencyModel . . . . . . . . . . . . . . . . . 9
1.2.4. Finding an ExtrapolationScheme . . . . . . . . . . . . . . . . . . . . 10
Chapter 2 Data and Data ProcessingMethods 13
2.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2. Micrometeorological Data . . .