Evaluation on spatial accuracy and validation of geological hazard susceptibility based on a multi-factor combination

The occurrence of geological hazards is the result of the interaction, constraints, and triggers of various factors. For a long time, researchers have predicted the potential locations and time (or frequency) of future geological hazards in areas prone to historical geological disasters based on environmental factors such as geology, topography, and hydrology, known as geological hazard susceptibility assessment. A prerequisite for geological hazard susceptibility assessment is the selection of impact factors, and there are dozens of factors contributing to the occurrence of geological hazards. Does the accuracy of assessment model increase with the addition of more factors? Is there an “optimal number of factors”? This simple yet crucial question is worth exploring. Taking Wenchuan County, Sichuan Province, as an example, this study selects 11 commonly used influencing factors in geological hazard susceptibility assessment and arranges them into four different combination, superimposing information from 3 to 11 factors to obtain corresponding distribution maps of geological hazard susceptibility indexes. The area under curve (AUC) value was used to evaluate the predictive accuracy of the results. Experimental results show that the prediction accuracy of the model reaches its maximum value when the number of superposition factors added to the model reaches 8 according to the preliminary set four combinations. However, during the process of factor superposition, it is found that there are certain difference between the control of susceptibility by each factor and the control determined by individual experience. When the actual factor control is arranged from largest to smallest or from smallest to largest, the predictive accuracy of the model reaches its peak value after superposition of multiple key factors. The research findings indicates that the more factors added to the geological hazard susceptibility assessment, the higher the predictive accuracy of the model. If key factors are not included during the superimposition process, the predictive accuracy of the model will not reach its peak, indicating that there is no “optimal number of factors” in geological hazard susceptibility assessment.