The invention discloses an ESDA (exploratory spatial data analysis) method of a spatial-temporal differentiation law of regional shallow landslide. The ESDA method comprises steps as follows: firstly, multi-source spatial-temporal data of a hazardous area of landslide is acquired, spatial-temporal evolution traces of landslide and landslide disaster formation factor information in a changing environment are extracted, and spatial-temporal distribution of factors is subjected to ESDA; secondly, a spatial-temporal comprehensive susceptibility index is calculated, and an optimal semi-variable function model is selected for performing spatial valuation and surface fitting of the index; finally, ESDA method based spatial-temporal differentiation law atlas analysis of the susceptibility index is realized. With the adoption of the ESDA method, the bottleneck problems of low accuracy, unclear landslide differentiation mechanism, insufficient evaluation for susceptibility influence and the like in the field of susceptibility evaluation with existing methods are solved; a spatial-temporal differentiation phenomenon and law mining are taken into full consideration, the hazardous area and frequent occurrence points of landslide are determined more accurately, the complexity of the method is lower, and susceptibility of multi-scale space landslide, especially large and meso scale landslide is analyzed easily.