Title: Using expert knowledge to map the level of risk of shallow landslides in Brazil

Abstract

Shallow landslides are common in Brazil’s urban areas. Geomorphology and land use are contributing factors, and rainfall is the triggering one. In these urban areas, anthropogenic activities that increase the level of landslide risk are common, such as cutting and filling or discharging wastewater onto the slopes. The Brazilian Government has developed a methodology to map the risk level in landslide-prone areas. The methodology is based on field observation and divides the risk into four main categories: low, moderate, high, and very high. Technicians in the field decide the sector’s landslide risk level based on their professional and personal experiences, but without mathematical calculations or without using specific weights for the contributing factors. This study proposes a method for automatically computing the risk level by involving experts in deriving each classifier weight, thereby reducing the bias in selecting the final risk level. The weight calculation involved 23 experts on urban landslides, it was computed using the Analytical Hierarchical Process, and standard deviation was used to define the risk level threshold. We validated the study using a prior risk mapping of Sao Paulo city. Finally, an application (app) that can be used on a tablet, computer, or smartphone was created to facilitate data collection during fieldwork and to automatically compute the risk level. Risk areas in Brazil are frequently changing as new residents move to the area or changes in the buildings or terrain are made. In addition, mapping the risk areas is expensive and time-demanding for municipalities. Therefore, an application that gathers the data easily and automatically computes the risk level can help municipalities rapidly update their risk sectors, allowing them to use an update risk mapping during the rainy season and be less dependent on rarely available financial resources to hire a risk mapping service.