We analyzed the characteristics of three main karstic/non-karst reaches of the Lijiang River in south China to uncover the causes behind different flood behaviors by providing a better understanding of the flood formation process. Having 63 years of rainfall-runoff data and applying the Hydrologic Engineering Canter's Hydrologic Modeling System (HEC-HMS) flood forecasting model, the geo/hydrological features were investigated. The available reservoir capacity of karts (ARCK) was included in the model through soil moisture accounting loss data to assess the impact of this variable. In particular, the expected instantaneous peak discharge rates and times were found largely imbalanced with generated unit hydrographs. Moreover, there were significant gaps among the floods’ features in different subbasins in terms of required peak modifications (i.e., two to four times bigger for midstream and upstream subbasins, respectively), which were mainly associated with the unique karst structure and the initial condition due to various ARCK in rainy/dry seasons. Besides, notable dissimilarities between the wedge and prism storage volumes and the hydrograph's wave traveling/receding time were observed owing to the geomorphological conditions. Although the contribution rates of drivers in karst flood formation cannot be quantitatively modeled, however, based on our results the ARCK emerged to play a substantial role on the forecasted results compared.