In a comparative modeling study of landslide hazards in the Feresmay area of Central Tigray, Ethiopia, researchers evaluated slope instability across approximately 10.75 square kilometers and found that a logistic regression model achieved greater predictive accuracy than a frequency ratio approach. The quantitative investigation validated spatial predictions against documented slope failures, confirming that statistical modeling can effectively identify terrain prone to hazardous mass movements in the region.
Slope failures represent a persistent hazard in the hilly and mountainous terrains of the Ethiopian highlands, occurring most frequently during the summer season. In these upland landscapes, slope failures historically lead to loss of life, widespread destruction of agricultural lands, and severe damage to residential settlements. The study focused on the Feresmay locality, an area situated in central Tigray between Bizet and Adwa that is accessible via the primary asphalt road connecting Mekelle to Adwa.
The investigation, conducted by researchers Gebremedhin Berhane, Letebrhan Hagos, and Yowhans Birhanu and published in Discover Applied Sciences, was designed to assess landslide occurrences, identify failure mechanisms and causative factors, generate a landslide susceptibility map, and propose remedial measures. To achieve these objectives, the authors synthesized previous investigations and secondary records with systematic field data collection, establishing an inventory of 202 individual landslide events within the 10.75-square-kilometer study boundary.
To map hazard susceptibility across the landscape, the authors applied two separate quantitative modeling techniques: the frequency ratio method and logistic regression. The frequency ratio model derived a composite landslide susceptibility index by calculating and summing the individual frequency ratio values across each contributing factor. In contrast, the logistic regression framework determined spatial susceptibility by evaluating mathematical relationships among predictor variables assigned to each factor.
Model reliability was assessed through the area under the curve method using an independent validation dataset comprising 30 percent of the inventoried landslide occurrences. Both approaches showed good agreement with observed ground failures, but the logistic regression model demonstrated superior overall performance, attaining an area under the curve metric of 0.838 compared to 0.800 for the frequency ratio technique.
The two models produced differing estimates regarding the spatial extent of the most hazardous terrain. The frequency ratio model designated approximately 10.64 percent of the total area as highly susceptible to landslides, whereas the logistic regression model classified roughly 13 percent into the high susceptibility zone. Documented impacts in the evaluated area included the destruction of dwellings and farmland, alongside the displacement of local residents.
The analysis identified lithology, land use and land cover, and distance to streams as the primary causative factors controlling slope instability in Feresmay. Citing the superior predictive capability of the logistic regression approach, the authors concluded that high and very high susceptibility zones warrant detailed geophysical and geotechnical investigations to clarify local material properties and support effective hazard mitigation strategies.
The investigation established its empirical framework by combining previous records and secondary information with direct field data collection to interpret slope instability across the 10.75-square-kilometer study area. By documenting 202 distinct landslide events within Feresmay, the field mapping provided the observational baseline necessary to examine failure mechanisms and quantify terrain susceptibility. This inventoried distribution reflected ongoing slope vulnerability in the hilly upland terrain, where seasonal mass movements continue to displace local populations and cause severe destruction to residential structures and agricultural land.
Methodologically, the authors differentiated between two distinct analytical formulations to produce the landslide susceptibility maps. The frequency ratio model derived an integrated landslide susceptibility index by calculating and summing the individual frequency ratio values associated with each causative factor. In parallel, the logistic regression model evaluated spatial susceptibility based on predictor variables associated with each factor. Both modeling techniques incorporated key spatial controls, identifying lithology, land use and land cover, and distance to streams as the primary causative factors governing slope failure across the evaluated terrain.
Model evaluation demonstrated clear differences in predictive precision when tested against an independent validation sample comprising 30 percent of the inventoried landslide occurrences. While both models showed good overall agreement between predicted susceptibility classes and observed landslides, the logistic regression framework achieved higher predictive accuracy, registering an area under the curve value of 0.838 compared to 0.800 for the frequency ratio approach. The spatial models also produced divergent estimates for the extent of critical hazard zones, with the frequency ratio model classifying approximately 10.64 percent of the area as highly susceptible and the logistic regression model designating approximately 13 percent of the total area as high susceptibility zones.
To address the hazards identified through the susceptibility maps, the researchers recommended targeted remedial planning focused on areas categorized under high and very high susceptibility. The authors emphasized that detailed surveys utilizing geophysical methods and geotechnical approaches are necessary for investigating these high and very high susceptibility zones. According to the study, conducting these specialized geotechnical and geophysical surveys will provide a better understanding of the geological materials in the area and aid in developing effective mitigation strategies across central Tigray.
Subject of Research: Quantitative modeling and mapping of landslide susceptibility in Central Tigray
Article Title: Comparative quantitative modeling of landslide susceptibility in Feresmay, Central Tigray
Article References: Berhane, G., Hagos, L., & Birhanu, Y. (2026). Comparative quantitative modeling of landslide susceptibility in Feresmay, Central Tigray. Discover Applied Sciences. https://doi.org/10.1007/s42452-026-09632-8
Image Credits: AI Generated
DOI: 10.1007/s42452-026-09632-8
Keywords: landslide susceptibility, logistic regression, frequency ratio, Feresmay, Central Tigray, Ethiopian highlands, area under the curve, landslide inventory, lithology, land use land cover, distance to stream, geotechnical approaches, geophysical methods, slope failure, hazard mitigation, spatial susceptibility mapping, Adwa
News Source: Denise Maddox. (November 13, 2026). Logistic Regression Outperforms Frequency Ratio in Modeling Tigray Landslide Susceptibility. Scienmag.



