Full text: Proceedings, XXth congress (Part 7)

ul 2004 
ymeniul 
ehnico- 
(editor- 
'à unei 
Revista 
underii 
1ediului 
sar XX, 
iintifica 
alizarea 
JFCOT, 
/9/6, p. 
15 
PREDICTION OF BIODIVERSITY - 
CORRELATION OF REMOTE SENSING DATA WITH LICHEN FIELD SAMPLES 
L. T. Waser^*, M. Kuechler", M. Schwarz", S. Stofer", Ch. Scheidegger *, E. Ivits", B. Koch" 
* WSL, Swiss Federal Research Institute, 8903 Birmensdorf, Switzerland - (waser, kuechler, schwarz, stofer, 
scheidegger)@wsl.ch 
? Dept. of Remote Sensing and LIS, Freiburg University, 70106 Freiburg, Germany — (eva.ivits, 
barbara.koch)@felis.uni-freiburg.de 
Commission TS SS 1 EOS for Sustainable Development 
KEY WORDS: Orthoimage, Modelling, Land Cover, Prediction, Correlation, High resolution, Segmentation, Environment 
ABSTRACT: 
The objective of the present study was to develop a model to predict lichen species richness for six test sites in the Swiss Pre-Alps 
following a gradient of land use intensity combining airborne remote sensing data and regression models. This study ties in with the 
European Union Project BioAssess which aimed at quantifying patterns in biodiversity and developing “Biodiversity Assessment 
Tools" that can be used to rapidly assess biodiversity. For this study lichen surveys were performed on a circular area of 1ha on 96 
sampling plots in the six test sites. Lichen relevés were carried out on three different substrates: trees, rocks and soil. 
In a first step, ecological meaningful variables derived from CIR orthoimages were calculated using both spatial and spectral 
information and additional lichen expert knowledge. In a second step, all variables were calculated for each sampling plot and 
correlated with the different lichen relevés. Multiple linear regression models were built containing all extracted variables and a 
stepwise variable selection was applied to optimize the final models. The predictive power of the models (r ranging from 0.79 for 
lichens on trees to 0.48 for lichens) can be regarded as good to satisfactory, respectively. Species richness for each pixel within the 
six test sites was then calculated. The present ecological modelling approach also reveals two main restrictions 1) this method only 
indicates the potential presence or absence of species and 2) the models may only be useful for calculating species richness in 
neighboring regions with similar landscape structures. 
KURZFASSUNG: 
Die vorliegende Studie hatte zum Ziel, flugzeuggestiitze Fernerkundungsdaten mit Feldaufnahmen von Flechten zu korrelieren, um 
die Anzahl Flechtenarten in sechs Untersuchungsgebieten in den schweizer Voralpen zu modellieren. Diese Studie kniipft an das 
EU-Projekt BioAsses an, welches zum Ziel hatte Biodiversitäts-Indikatoren zu entwickeln, mittels welcher rasch die Biodiversität 
eines Gebietes abgeschätzt werden kann. Basis für diese Studie bilden Feldaufnahmen von Flechten, welche in den sechs 
Untersuchungsgebieten an insgesamt 96 Aufnahmeorten durchgeführt wurden — jedes der Grósse von 1 Hektare, was einem Kreis 
mit 56 m Radius entspricht. Flechten wurden auf den drei Substraten Baum, Stein und Boden aufgenommen. In einem ersten Schritt 
wurden ókologisch-relevante Variablen aus CIR Orthobildern, hauptsächlich basierend auf räumlicher und spektraler Information 
abgeleitet und mittels zusätzlichem Expertenwissen berechnet. In einem zweiten Schritt wurden für alle 96 Aufnahmeorte die 
entsprechenden Variablen berechnet und anschliessend mit den Feldaufnahmen korreliert. Um optimale Modelie zu erhalten wurden 
multiple lineare Regressionsmodelle mit einer schrittweisen Variablen Selektion verwendet. Für das Baumflechten-Modell wurde 
ein r von 0.79, für Bodenflechten ein r von 0.48 erreicht, was als gut bis genügend eingestuft werden kann. Schliesslich wurde mit 
Hilfe dieser Modelle die potentielle Anzahl Arten für jedes Pixel in den sechs Untersuchungsgebieten berechnet. Die Studie zeigt 
ferner, dass der ókologische Modellierungsansatz auch seine Grenzen hat: 1l) mittels dieser Methode kann nämlich nur die 
potentielle Anzahl Flechten berechnet werden und 2) das Anwendungspotential der Modelle beschránkt sich wahrscheinlich auf 
benachbarte Regionen mit einer ähnlichen Landschaftsstruktur. 
1. INTRODUCTION almost impossible to have a complete biodiversity survey at 
regional scale of 1-100 square kilometers. Therefore methods 
The need for conserving biodiversity has become increasingly for extrapolations are needed that provide information that is 
imperative during the last decade as rates of habitat and species remotely similar to field samples and which would allow to 
destruction continue to rise (Noss and Cooperrider, 1994, ^ considerably reduce extensive field surveys. New techniques 
Nagendra 2001). At the same time inventorying biodiversity ^ and data sets now enable remote sensing, in conjunction with 
and monitoring efficacy of measures for its conservation have ecological models, to shed more light on some of the 
emerged as important scientific challenges of recent years fundamental questions regarding biodiversity (Cousins and Ihse 
(Jorgensen 1997, Nagendra and Gadgil 1999). For monitoring ^ 1998). Furthermore, remote sensing also may help calculating 
biodiversity on a general level, homogenous consistent land biodiversity hotspots to facilitate biodiversity field surveys, e.g. 
cover information is primarily required as it is obtained using to focus the sampling of biological data on these hotspots (Kerr 
remote sensing data (Townshend et al. 1991, Chuvievo 1999). and Ostrovsky 2003). 
According to Palmer (1995) and Wohlgemuth (1998) it is 
  
845 
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.