Testing the performances of different fuzzy overlay methods in GIS-based landslide susceptibility mapping of Udi Province, SE Nigeria

Landslides are still wreaking havoc in many parts of the world. Several previous landslide susceptibility mapping (LSM) studies have proven that fuzzy logic methods provide reliable models for effective risk analysis and management. Various fuzzy operators are available for LSM. Globally, few studies conducted over a decade ago tested and compared the performances of several fuzzy operators in LSM and they revealed that the performances of the fuzzy operators vary due to a number of factors. Such comparative landslide studies have not been reported in Nigeria and Africa. In the current study, advanced mapping technologies were utilized in testing the performances of eight fuzzy operators in LSM of Nigeria's Udi Province, a region known for incessant soil erosion and landslide events. A total of 258 landslide locations were identified and used for area under curve (AUC) model validation performed using MATLAB and statistical analysis tools in Arc-Map software. Eleven landslide conditioning factors were chosen for the LSM. Correlations exist between the landslide locations and conditioning factors. Fuzzy PRODUCTandfuzzy ANDclassified the landslide susceptibility of the province to range from very low to moderate whereasfuzzy ORandfuzzy SUMcategorized the susceptibility to range from low to very high. However, all fourfuzzy gammas(0.8, 0.9, 0.95, and 0.975) classified the landslide risk to range from very low to high. All the fuzzy gammas and the fuzzy PRODUCT have the highest accuracy (AUC = 90.5%) compared to the fuzzy AND (AUC = 89.60%), fuzzy SUM (AUC = 89.10%), and the fuzzy OR (AUC = 86.7%). The findings and insights provided by this updated comparative study demonstrate that the fuzzy operators have global applicability. Thus, the study approach can be successfully used in other regions and related settings, for effective landslide mapping, monitoring, and management.