%0 Journal Article %T Mitigating bias in inversion of InSAR data resulting from radar viewing geometries %+ Laboratoire Magmas et Volcans (LMV) %A Dumont, Quentin %A Cayol, Valérie %A Froger, Jean-Luc %< avec comité de lecture %Z Challenge 4 %@ 0956-540X %J Geophysical Journal International %I Oxford University Press (OUP) %V 227 %N 1 %P 483-495 %8 2021-10-08 %D 2021 %R 10.1093/gji/ggab229 %K Radar interferometry %K Indian Ocean %K Interferometry %K Inverse theory %K Numerical modelling %K Remote sensing of volcanoes %Z Sciences of the Universe [physics]/Earth Sciences/VolcanologyJournal articles %X InSAR data acquired from ascending and descending orbits are often characterized by different magnitudes of the observed line-of-sight displacements, which may potentially bias inverse models. Using synthetic numerical models of dyke intrusions, we show that biased solutions are obtained when carrying out ‘conventional’ inversions where only observation and modelling errors are taken into consideration. To mitigate the impact of the relative magnitudes of the data, we propose two methods: a covariance weighting inversion and a wrapped data inversion. These methods are compared to a conventional inversion using synthetic data generated by models of dykes of known geometry. We find that the covariance weighting method allows to retrieve an initial source geometry better than the other methods. These methods are then applied to the July 2017 eruption of Piton de la Fournaise. Using a covariance weighting inversion, the difference in fit between data sets decreases from 50% to 20 % and the newly estimated source is in better agreement with the geological context. %G English %2 https://uca.hal.science/hal-03356622/document %2 https://uca.hal.science/hal-03356622/file/Dumontetal_GJI2021_selfCompiled-1.pdf %L hal-03356622 %U https://uca.hal.science/hal-03356622 %~ IRD %~ INSU %~ PRES_CLERMONT %~ CNRS %~ LMV %~ ACL-SF %~ OPGC %~ CAP20-25 %~ TEST-HALCNRS %~ ANR %~ TEST3-HALCNRS