[Objective] This study seeks to assess flood risk to provide reference for regional disaster reduction and socio-economic development in upstream areas of the Ganjiang River.[Methods] The temporal and spatial features data of a rainstorm flood occurring under natural conditions were collected and analyzed. The weights of 14 indicators of 4 factors among risk of disaster-causing factors, sensitivity of disaster-bearing environments, vulnerability of disaster-bearing bodies and the capacity of disaster controls were determined using the analytic hierarchy process (AHP), and then risk factor analysis and comprehensive risk assessment were carried out in GIS.[Results] The floods were mostly concentrated from March to July in the study area. The rainstorms in spring and summer account for 80.4% of the whole year, while the floods accounted for 95.9%. The floods mainly occurred in the northeast. The hazard levels appear higher in the northeast and lower in the west. The sensitivity of the flood inducing environments bands from the northwest to the northeast. The sensitivity in the middle was a little higher than in surrounding areas. The vulnerability levels were higher in the central and western regions but lower in the surrounding areas. The anti-disaster capacity was decreasing from the southwest to the northeast, higher in the west and the south, lower in the east and the north. The comprehensive risk levels were higher in the north central and eastern regions, but lower in the southeast and west. Areas of each risk level was very different in the proportion. The extremely high and high risk accounted for 37.3%.[Conclusion] Floods in the upstream areas of the Ganjiang River are characterized by floods occurring in both the hilly and plain areas. There is a significant positive correlation between rainstorms and floods. These risk assessment results are consistent with the actual situations observed in the study area in most cases, which verifies indirectly the rationality of the indicators selected and the evaluation model.