A comprehensive watershed model—the Hydrological Simulation Program FORTRAN(HSPF) was set up and calibrated for the Dongjiang River basin

which was used to quantitatively predict runoff in response to possible future land use change

combining with a land use projection model(CA-Markov) and a daily rainfall stochastic simulation model. The calibration and validation of the HSPF showed a good fit and the calibrated model was applicable to simulating the monthly runoff in the Dongjiang River Basin. The future land use scenarios generated by CA-Markov model result in little change in total runoff but some noticeable changes in surface flow. Both high flow and low flow increase under the future land use scenarios

and high flow increases with the growth of urban construction land while low flow decreases. The run-off effects of land use change decreases with increasing rainfall

and increases with decreasing rainfall. The changes of total runoff due to land use changes are little under four rainfall scenarios

but while the changes of surface flow are significant. Simulation hydrological change by connecting a land use prediction model and a semi-distributed hydrological model can be a good decision support tool for water management at a basin scale and capable of long-term scenario-based assessments.