COMBINATION OF DIFFERENTIAL INTERFEROMETRY AND SUB-PIXEL IMAGE CORRELATION IN MEASUREMENT OF THE 2005 KASHMIR EARTHQUAKE DISPLACEMENT FIELD Y. Yan 1,2 , V. Pinel 2 , E. Trouv´ e 1 , E. Pathier 2 , S. Galichet 1 , G. Mauris 1 , A. Bisserier 1 1 : Laboratoire d’Informatique, Syst` emes, Traitement de l’Information et de la Connaissance Universit´ e de Savoie - Polytech Annecy-Chamb´ ery - BP 80439 - F-74944 - Annecy-le-Vieux Cedex - FRANCE Tel: +33 450 096 533 - Fax: +33 450 096 559 - Email: {yajing.yan,emmanuel.trouve,sylvie.galichet,gilles.mauris,amory.bisserier}@univ-savoie.fr 2 : Laboratoire de G´ eophysique Interne et Tectonophysique, CNRS, IRD Universit´ e de Savoie - Campus Scientifique - 73376 Le Bourget du Lac Cedex - FRANCE Tel: +33 479 758 651 - Fax: +33 479 759 406 - virginie.pinel@univ-savoie.fr Universit´ e de Joseph Fourier - BP53 - 38041- Grenoble - FRANCE Tel: +33 476 635 209 - Fax: +33 476 635 252 - erwan.pathier@obs.ujf-grenoble.fr ABSTRACT In this paper, we apply both sub-pixel image corre- lation and differential interferometry techniques on a series of ENVISAT images (ascending + descending) in order to map the surface 3D displacement field due to the Kashmir earthquake of October 8, 2005. A particular chain for phase unwrapping based on multi-scales local frequencies is developed with aim to get around of the coherence loss problem. In order to retrieve the surface 3D displacement field, the dis- placement measurements from both sub-pixel image correlation and differential interferometry are com- bined. The measurement uncertainty is represented by the conventional approach and a fuzzy approach respectively. Moreover, the evolution of the mea- surement uncertainty while adding redundant infor- mation of sub-pixel image correlation and/or differ- ential interferometry is analyzed and compared for both approaches. Key words: sub-pixel image correlation, D-InSAR, combination, fuzzy representation. 1. INTRODUCTION The deformation of the Kashmir earthquake of Octo- ber 8, 2005, has been analyzed by using sub-pixel cor- relation of SAR images [1] as well as optical images [2]. However, the application of differential interfer- ometry (D-InSAR) to the displacement measurement due to this earthquake has not been successful be- cause of the coherence loss. In this paper, we apply both sub-pixel image correlation and D-InSAR on a series of ENVISAT images from October 2004 to June 2006 (Table 1) in order to map the deformation due to this earthquake. The sub-pixel image correlation can provide displace- ment measurement in LOS and azimuth directions with an accuracy of decimeters, while the D-InSAR gives a measure of displacement in LOS direction with an accuracy of centimeters. The displacement maps from both techniques are considered as projec- tions of the full 3D surface displacement field in LOS and/or azimuth directions specified by the satellite position and orbit at the acquisition time. Therefore, one can construct the full 3D displacement field from these projection measurements using the projection vectors, where three or more such projections are available. In general, the 3 components (E, N, Up) of the surface displacement are estimated by a con- ventional weighted least square method [3] [4] [5] [1]. Sub-pixel image correlation has been proven robust to measure large amplitude displacements induced by a magmatic intrusion or a seismic event in the near field, close to the dyke or fault [1]. Whereas D- InSAR can provide displacement measurements in the far-field, at largest distance from the dyke or fault, with a better precision. As a consequence, the combination of sub-pixel image correlation and D- InSAR allows to take advantage of both techniques and has been applied successfully to deduce the full 3D surface displacement [6] [7]. However, the strate- gies of combination and the evolution of measure- ment uncertainty have not been well analyzed yet. In this paper, we aim to combine all of the measure- ments from both sub-pixel image correlation and D- InSAR to deduce the full 3D displacement field by in-