Landslide risk assessment considering socionatural factors: methodology and application to Cubatão municipality, São Paulo, Brazil

Abstract

The manuscript presents a methodology to integrate spatial information of susceptibility, vulnerability and rainfall thresholds to produce a dynamic landslide risk map. The inputs were combined in two matrices: combining susceptibility and vulnerability classes, constituting the socionatural criteria (SN); SN classes and rainfall thresholds (T) were coupled to determine the risk (R). The method was applied to the municipality of Cubatão (142,281 km2), state of São Paulo (Brazil), where there is an extensive landslide history, high rainfall rates, and communities living on hillsides. The susceptibility model was prepared using the Random Forest algorithm. Social vulnerability was based on socioeconomic and demographic indicators. Rainfall thresholds were generated by three approaches: intensity-duration (ID), rainfall event-duration (ED), and antecedent accumulated rainfall (A). Thus, each product was reclassified and entered into both 5 × 5 size matrices. The methodology results in an estimate of location and trigger rainfall thresholds of landslide events. In addition, the model offers three main advantages: easy adaptation and calibration as new data emerges; identification of deficiencies in public policies for certain groups of people with critical SN, allowing interventions to reduce vulnerability; a dynamic map that allows a real-time automation process in the case of weather forecasts, facilitating the concentration of efforts in specific areas. In conclusion, the method is a useful risk mitigation tool, through the development of the landslide early warning system and associated public policies, with potential for replication elsewhere.