The commonly used success rate (SR) in evaluating cell-based landslide modelperformance is based on the ratio of successfully predicted landslide sitesover total actual landslide sites without considering the performance inpredicting stable cells. We proposed a modified SR (MSR), in which theperformance of stable cell prediction is included. The advantage of MSR isto avoid over- and under-prediction while upholding the stable sensitivitythroughout all simulated cases. Stochastic analyses are conducted by usingartificial landslide maps and simulations with a full range of performances(from worst to perfect) in both stable and unstable cell predictions.Stochastic analyses reveal mathematical responses of estimators to variousmodel results in calculating performance. The Kappa method, which iscommonly used for satellite image analysis, is improper for landslidemodeling giving inconsistent performance when landslide coverage changes. Toexamine differences among SR and MSR in real model application, we appliedthe SHALSTAB model onto a mountainous watershed in Taiwan. Case study showsthat stable and unstable cell predictions are inter-exclusive in SHALSTABmodel. The optimal estimator should compromise landslide over- andunder-prediction. According to our 4000 simulations, the best simulationgenerated by MSR projects 83 hits over 131 actual landslide sites while theunstable cells cover only 16% of the studied watershed. By contrast,despite the fact that the best simulation deduced from SR projects 120 hitsover 131 actual landslide sites, this high performance is only obtained whenunstable cells cover an incredibly high landslide cover (~75%) ofthe entire watershed exhibiting a significant landslide over-prediction.
展开▼
