Abstract: The impacts of human activities on river runoff are becoming more and more serious in recent years, resulting in a difficulty of using the measured runoff in water resources assessment and hydroelectric project design because of the inconsistency of measured time series. This situation is especially common in regions where the impacts are intensive and extensive and the data for groundwater development is lacking or subject to great uncertainty. These difficulties may further cause a complication and low accuracy in estimation of the unimpaired or natural runoff time series. It is therefore necessary and urgent to find a simple, practical and accurate method of time series restoring for the regions of groundwater strongly disturbed by human activities. In this study, we have developed an improved method of natural runoff calculation that incorporates the idea of base flow separation and item-by-item investigation. Three cases were studied to investigate the feasibility of our method. The study shows that this improved method gave results quite close to those of the conventional restoring methods and it shows better applicability to regions with diverse landscapes and different degrees of groundwater disturbances. Particularly, it can complete natural runoff calculations using smaller amount of data in comparison with the conventional methods. Actually, for a restoring calculation it only needs the measured runoff data and the data of surface water consumption that can be obtained using item-by-item investigation, hence avoiding the restoring distortion caused by inadequate groundwater data. Case studies verified that the improved method is rational, reliable, and applicable to water resources engineering designs. Thus, the method is of great value in natural runoff calculation for the regions where groundwater is significantly disturbed or groundwater data is inadequate.