Geospatial-Based
Assessment of Land Susceptibility Mapping using the Bivariate Statistical Frequency
Ratio Model: A Case Study of Idukki District, Kerala, India
Chitikela N. Vara Laxmi and Kollipara Padma Kumari
Disaster Advances; Vol. 18(9); 11-21;
doi: https://doi.org/10.25303/189da011021; (2025)
Abstract
Landslides are one of the most hazardous threats worldwide, posing significant geo-environmental
challenges including loss of life, destruction of infrastructure, damage to properties,
degradation of agricultural lands and impacting human society. An attempt has been
made to prepare the Landslide Susceptibility Zonation (LSZ) map with the help of
nineteen geospatial thematic layers using the Geospatial Frequency Ratio (GSFR)
model at Idukki district of Kerala, India. The landslide inventory dataset of 1,850
landslide points was identified from historical records (NASA-Co-operative Open
Online Landslide Repository (COOLR) and Google Earth dataset divided into training
(1,295-70%) and testing (555-30%).
The inventory data and landslide conditioning parameters are used to establish a
prediction model of landslide susceptibility. The results showed that 6.22% of the
district area falls under very high susceptibility while 17.9 % is categorised as
having high susceptibility. Receiver operating characteristic curve (ROC) and area
under the curve (AUC) are used to validate the success rate and prediction rate
of landslide susceptibility. The FR model achieved the accuracy with a success rate
of 0.827 and a prediction rate of 0.835 in the current study.
