Vet. Res. 38 (2007) 669-683
DOI: 10.1051/vetres:2007025

Use of high spatial resolution satellite imagery to characterize landscapes at risk for bluetongue

Hélène Guis^{a, b, c}, Annelise Tran^{b, c}, Stéphane de La Rocque^b, Thierry Baldet^b, Guillaume Gerbier^b, Bruno Barragué^b, Fabienne Biteau-Coroller^b, François Roger^b, Jean-François Viel^a and Frédéric Mauny<sup>a

a</sup>  Department of Public Health, Biostatistics and Epidemiology Unit, Faculty of Medicine, St. Jacques University Medical Center, 2 place St. Jacques, 25030 Besançon Cedex, France
^b  French Agricultural Research Center for International Development (CIRAD), Epidemiology and Ecology of Animal Diseases Unit, Baillarguet Campus, TA 30 B, 34398 Montpellier Cedex 5, France
^c  Territories, Environment, Remote Sensing and Spatial Information Joint Research Unit (UMR TETIS), Maison de la Télédétection, 500 rue J.-F. Breton, 34093 Montpellier Cedex 5, France

(Received 18 October 2006; accepted 27 March 2007 ; published online 23 June 2007)

Abstract - The recent and rapid spread in the Mediterranean Basin of bluetongue, a viral disease of ruminants transmitted by some species of Culicoides (biting midges), highlights the necessity of determining the conditions of its emergence. This study uses high spatial resolution satellite imagery and methods from landscape ecology science to identify environmental parameters related to bluetongue occurrence in Corsica, a French Mediterranean island where the disease occurred for the first time in 2000. A set of environmental variables recorded in the neighborhood of 80 sheep farms were related to case occurrence through a logistic regression model computed within three subsequent buffer distances of 0.5, 1 and 2 km. The results reveal the role of landscape metrics, particularly those characterizing land-use units such as prairies and woodlands, as well as farm type, latitude and sunshine to explain the presence of bluetongue. Internal and external validation both indicate that the best results are obtained with the 1 km buffer size model (area under Receiver Operating Characteristic curve = 0.9 for internal validation and 0.81 for external validation). The results show that high spatial resolution remote sensing (i.e. 10 m pixels) and landscape ecology approaches contribute to improving the understanding of bluetongue epidemiology.

Key words: epidemiology / bluetongue / remote sensing / geographic information systems / landscape ecology

Corresponding author: annelise.tran@cirad.fr

© INRA, EDP Sciences 2007