•An improved operation-based reservoir scheme integrated with VIC is developed.•This scheme eliminates numerical calibration and uses new parametric functions.•New scheme performance was compared to two other schemes.•This scheme can remove bias caused by spatial and temporal variations. Multiyear and multipurpose reservoirs generally need to adjust operational functions for different storage levels; most existing reservoir schemes lack the pertinence to model the complex behaviors of such reservoirs. This paper proposes an improved operation-based reservoir scheme to simulate the storage and outflow variations for multiyear and multipurpose reservoirs at a daily time step. In this improved reservoir scheme, the new parametric functions to represent the weights of the outflow through considering various operation purposes (e.g., hydropower generation, flood control, and water supply) are introduced. Then, the improved reservoir scheme is integrated into a macroscale hydrological model, i.e., Variable Infiltration Capacity (VIC) model. In this study, the performance of the proposed reservoir scheme is evaluated with the observed storage and outflow of a multiyear and multipurpose reservoir (Xinfengjiang Reservoir) in South China and compared to two other reservoir schemes (i.e., a multilinear regression reservoir scheme and an operation-based reservoir scheme). In addition, through uncertainty analysis, this study shows that the time-varying operation-based parameters can reflect the changes of reservoir storage but not reservoir outflow, indicating that the parameters for establishing functional relationships need further exploration. Results reveal that the proposed reservoir scheme has a wide applicability for reservoir storage and outflow simulation, especially for the reservoirs with limited reservoir management data.