Abstract: After having collected a large number of loess landslides data in Tianshui City,Gansu Province through extensive site investigation,we constructed a multivariate probability distribution model for loess landslide data via Johnson's system of distribution. We used the model to predict the width,length,and area of potential slides in an unstable loess slope. The results show that the prediction from the proposed model is reasonably accurate. The upper boundary of the 90% confidence interval of the prediction results is used as the safety value of the impact range to derive the potential impact of the unstable loess slope in the survey area. After having verified with the same type of loess landslide data outside the survey area,it is found that the prediction formula has certain applicability to the evaluation of the same type of loess landslide in other areas in Tianshui City. Only the height of the unstable slope in the prediction formula is used as independent variables,which provides a great convenience for rapid on-site assessment of the hazard and impact range of unstable loess slopes. The proposed method therefore provides a vehicle for quantifying the impact of unstable loess slopes,and hence offers basis for quantitative evaluation of loess slope hazards in the region of Tianshui City.