ORIGINAL PAPER Exporting a Google Earth TM aided earth-flow susceptibility model: a test in central Sicily D. Costanzo • C. Cappadonia • C. Conoscenti • E. Rotigliano Received: 12 February 2010 / Accepted: 3 May 2011 / Published online: 8 June 2011 Ó Springer Science+Business Media B.V. 2011 Abstract In the framework of a regional landslide susceptibility study in southern Sicily, a test has been carried out in the Tumarrano river basin (about 80 km 2 ) aimed at char- acterizing its landslide susceptibility conditions by exporting a ‘‘source model’’, defined and trained inside a limited (about 20 km 2 ) representative sector (the ‘‘source area’’). Also, the possibility of exploiting Google Earth TM software and photo-images databank to pro- duce the landslide archives has been checked. The susceptibility model was defined, according to a multivariate geostatistic approach based on the conditional analysis, using unique condition units (UCUs), which were obtained by combining four selected con- trolling factors: outcropping lithology, steepness, plan curvature and topographic wetness index. The prediction skill of the exported model, trained with 206 landslides, is compared with the one estimated for the whole studied area, by using a complete landslide archive (703 landslides), to see to what extent the largest time/money costs needed are accounted for. The investigated area stretches in the fore-deep sector of southern Sicily, where clayey rocks, mainly referring to the Numidian Flysch and the Terravecchia Formations, largely crop out. The results of the study confirm both the exploitability of Google Earth TM to produce landslide archive and possibility to adopt in assessing the landslide susceptibility for large basin, a strategy based on the exportation of models trained in limited repre- sentative sectors. Keywords Landslide susceptibility Á Exportation of models Á Google Earth TM Á Validation 1 Introduction A susceptibility map of an area has to provide an image in which the probability of new landslides is estimated and would enable us to predict where a landslide is more likely to occur (Guzzetti et al. 2006). The most widely adopted methods for landslide susceptibility assessment on a basin or regional scale are those which exploit functional relationships D. Costanzo (&) Á C. Cappadonia Á C. Conoscenti Á E. Rotigliano Dipartimento di Scienze della Terra e del Mare, Universita ` degli Studi di Palermo, Palermo, Italy e-mail: costanzodario@gmail.com 123 Nat Hazards (2012) 61:103–114 DOI 10.1007/s11069-011-9870-0