A comparison of two methodologies to assess shallow landslide risk

Landslides are natural processes that can be hazardous events when they threaten human lives and assets (Keller, 2011), and their incidence can be increased by improper land use, such as irregular urbanization and the construction of dams and roads without good engineering planning. Landslide disasters and catastrophes are not new, and the loss of human life to them is common in developing countries. Many people live in risk prone-areas in the Southeastern part of Brazil (IBGE, 2018). Major cities such as Sao Paulo and Belo Horizonte have 6% and 16% of their population living in at-risk areas, respectively (IBGE, 2018). Moreover, mapping these risk areas are critical for managing and mitigating the risk. Municipalities can use the mappings to monitor the risk areas and to allocate resources or prioritize to relocate residents living in high and very high risk levels sectors. The Brazilian government through its Ministry of Cities and the Institute for Technological Research (Instituto de Pesquisas Tecnológicas – IPT) developed a landslide risk mapping methodology that we are calling the Brazilian Government Methodology (BGM) to map landslides and river-bank erosion risk (Carvalho, Macedo, & Ogura, 2007). Within the BGM, two technicians do field inventory, observing local geo-environmental variables, and based on their previous experience and personal observation, they decide on the level of risk. This mapping is biased since it does not involve any objective mathematical computation and is influenced by each technician's personal experience. The purpose of this study is to discuss and compare the performance of two methodologies that quantifies the BGM and automatically classify the risk level of shallow landslide prone-areas. The first methodology is based on the Analytical Hierarchical Process (AHP) and it is heuristic/expert based. The second is statistically-based and uses Ordinal Logistic Regression (OLR). We compared how similar their risk level classifications are and how similar they are with the BGM. Additionally, we are using a user-friendly application (app) developed by the authors, that computes the level of risk and it can be used on a tablet, smartphone, or computer with internet access or with RStudio software installed. Both methodologies can use the app since the classifiers are both based on the BGM