Recognition of shallow landslides using object-based image analysis: preliminary results in the Gurutuba watershed, Brazil

Shallow landslide recognition and mapping are essential for susceptibility, vulnerability, and risk assessment. Efficient mapping approaches can be used to identify source, transport, and deposition areas and enable the creation of geomorphological, seasonal, and event-based inventories. In Brazil, landslides are a frequent natural hazard that occurs every year, mostly in the summer season (Dec-Mar). Despite their frequency, no guidelines or common procedures for landslide mapping exist. The application of new methods for recognition and mapping may improve the creation of standards for landslide inventory mapping and the enhancement of consistent landslide databases in Brazil. Manual, semi-automated, and automated methods are commonly used to identify hazards worldwide, with remote sensing and geomorphological information being the primary data for landslide mapping. Object-based image analysis (OBIA) is a well-established method for mapping natural hazards and geomorphological features. Although several studies have demonstrated the applicability of OBIA for landslide mapping in various environments, its application for landslide mapping in tropical environments in Brazil is still incipient. In 2014, the city of Itaóca (São Paulo state, Brazil) was affected by a high-magnitude mass movement event responsible for infrastructure damages and deaths. The aim of this study is to identify rainfall-induced shallow landslides using the OBIA method in the Gurutuba watershed (area of 4,5 km²), Itaóca. The recognition and mapping were performed using a RapidEye satellite image (5 m), dated 2014/01/30, in eCognition 10.0 (Trimble) software. The OBIA classification process considers spectral, spatial, contextual, and hierarchical information. Preliminary analysis indicated promising results and good applicability of the OBIA method. Many shallow landslide scars were identified, whereby large scars were more easily recognized by the method than small ones. The mapping accuracy was assessed by comparison to a reference dataset, which was created by visual interpretation of the RapidEye image. However, more research is needed to increase the transferability of the approach to allow the analysis of a more complex and larger area.