Dinghushan biosphere reserve (112°30′39″-112°33′41″ E, 23°09′21″-23°11′30″ N) is located in central Guangdong Province in southern China, about 84 km from Guangzhou city, with an area of 1 156 hm2. Due to its location on the tropic of cancer, the forest vegetation is very rich and dominated by monsoon evergreen broad-leaved forests. The dominant forest types in the Dinghushan biosphere reserve are Pinus massoniana forests (PF), mixed Pinus massoniana/broad-leaved forests (PBF), and monsoon evergreen broad-leaved forests (MBF), which form a natural successional sequence. The aim of this paper was to quantify the magnitude and annual variation of water yields in the Dinghushan Nature Reserve in the three forest types, which would be used for estimating carbon outputs in streamflow, and to discuss how hydrological processes vary at different successional stages of forest development. Climatic data were obtained from weather stations located at the Dinghushan Forest Ecosystem Research Station, Chinese Ecosystem Research Network (CERN). Runoff was monitored at three landscape levels. The first level was the entire eastern watershed. The second level referred to small catchments within the larger watershed that were dominated by the different forest types, i.e., a PF catchment, a PBF catchment, and a MBF catchment. The third level referred to three surface runoff plots placed within each of the three catchments. Stream runoff in the eastern watershed and the three smaller catchments was monitored continuously year a round by measurement weirs with streamflow recorders. The ephemeral surface runoff from the nine surface runoff plots was collected in separate plastic tanks and the water level of each tank was recorded automatically following every precipitation event. The subsurface water table depth was recorded manually at 5-day intervals in wells located in the valley of the eastern watershed at elevations of 20-30 m. Soil water content has been measured monthly using neutron probes since 1983. Soil water content was measured at 10-15 places in each of the three catchments and distributed across the elevational gradient. At each point, the water content was monitored at three soil depths, 0-15, 15-30 and 30-45 cm. Throughfall was collected at four sites in each catchment using cross-shaped troughs with a horizontal area of 2.25 m2, and was measured using a fluviograph (Zhou, 1997). Thirty trees adjacent to each throughfall site that represented the range of species and size of trees in each catchment were selected and stemflow monitored (Gash et al., 1978). Stemflow was collected in an open PVC tube wrapped around the stem of each tree that led to a tipping bucket rain gauge. Throughfall and stemflow were monitored from April 1999 to April 2000, during which time there were 61 rain events. Leaf area index and canopy cover were measured at 10 sampling sites within each catchment four times each year using a CI-110 digital plant canopy imager (CID, Inc. Vancouver). The water yield in the eastern watershed of Dinghushan biosphere reserve was 66.5% of its rainfall with the maximum outflow occurring about one day after a rain event. The subsurface water table averaged 2.22 m below the soil surface, with the deepest water levels at 2.84 m and highest at 1.14 m depth. The annual average position was 2.38, 2.27, 2.08, 2.13 and 2.11 m deep in the years 1999, 2000, 2001, 2002 and 2003, respectively. The depth of the water table in the eastern watershed was correlated with rainfall events that occurred 16 days previously. There was an abating tendency of soil water content for all the three forest communities. The tendency was statistically significant for MBF (p<0.01) and PBF (p<0.05), whereas no statistically significant effect for the PF was found. The relationship between the amount of throughfall and precipitation was linear for the three different forest communities at different successional stages, but the correlation decreased from MBF, PBF uo PF. The ratio of throughfall to precipitation also decreased with successional stage of the forest, from 83.4%, 68.3% to 59.9% for the PF, PBF and MBF, respectively. The relationship of stem flow with DBH was controlled by the effects of the whole forest community instead of a single species. The canopy structure of the forest community played a key role in the redistribution of precipitation. The canopy interception rate in the MBF was 83.3% in February when total precipitation was 28.7 mm, but was only 18.9% when precipitation was 297.8 mm in June. The canopy interception rate increased in the three forest types from PF, PBF to MBF.