Banca de DEFESA: MAYARA PEIXOTO COSTA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : MAYARA PEIXOTO COSTA
DATE: 31/08/2026
TIME: 14:00
LOCAL: VIDEOCONFERÊNCIA
TITLE:

Mapping of Natural Landslide Susceptibility in the Municipality of Angra dos Reis, Rio de Janeiro

KEY WORDS:

Natural susceptibility map, mass movement, landslide, Random Forest, iForest, GIS.

PAGES: 70
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUBÁREA: Ciência do Solo
SUMMARY:

Landslides (LS) are multicausal natural hazards whose occurrence is mainly associated with the interaction among geomorphological, hydrological, and pedological factors. In this context, the mapping of terrain attributes is fundamental for characterizing locations with greater natural susceptibility to the occurrence of these events. The present study aimed to evaluate which vector geometry (VG) provides the best representation of LS conditioning factors in order to develop a natural susceptibility map (NSM) for the municipality of Angra dos Reis, located in the state of Rio de Janeiro (SRJ), Brazil. The NSM was developed through the extraction of topographic and hydrological attributes from a digital elevation model (DEM) with a spatial resolution of 10 m, generated from contour lines with a 5 m contour interval provided by the municipal government.The topographic attributes of elevation, slope, aspect, plan curvature (Cplan), and profile curvature (Cprofile) were used, in addition to the hydrological attributes of flow direction (FD), flow accumulation (FA), flow width (FW), specific catchment area (SCA), watershed area (WA), and a topographic wetness index (TWI). To identify the most relevant attributes for event characterization, the Isolation Forest (iForest) algorithm was used, which separated the events into anomalous and normal events and calculated importance scores for this distinction. These scores were subjected to permutation to analyze the importance of each variable in the model generated by iForest. The VGs used in the different extraction methods (EMs) of these attributes were the point of highest elevation within the landslide scar (SP), a 50 m buffer generated from this point, and the event scar itself. The sets of pixels corresponding to the three EMs were used to develop a natural susceptibility map (NSM) using the Random Forest algorithm. The performance of the EMs was compared using the confusion matrix and the metrics of Overall Accuracy (OA), Kappa Index, Producer's Accuracy (PA), and User's Accuracy (UA). The results of this study demonstrated that hydrological variables and the buffer VG better represent areas with greater susceptibility to the occurrence of LS.

COMMITTEE MEMBERS:
Externo à Instituição - DIEGO NUNES BRANDÃO - CEFET/RJ
Externo à Instituição - GUSTAVO DE MATTOS VASQUES - EMBRAPA
Presidente - 1220296 - MARCOS BACIS CEDDIA