The influences of complex landscape-multipond system on water resource, the dynamics of nutrients in the complex landscape, and the factors on the area of multipond system, and the retention of multipond systems on surface runoff and pollutants were studied in Liuchahe watershed, southern China. The nonpoint source pollution (NPS) model was also developed. The results were given as follows. As a traditional irrigation system, influences of the multipond system on the water resources were simulated according to crop water requirement, surface runoff, precipitation and ponds storage capacity of Liuchahe watershed. The results showed that the system could offset water shortage between crop water requirement and precipitation. Water influence by the ponds were amounted to 277 mm in a normal year, 239 mm in a wet year, and 195 mm in a dry year, and the corresponding annual precipitation was 914.6 mm, 1796 mm, and 658 mm, respectively. The system effectively influenced the hydrological process of surface runoff, including water amount and peak discharge. Most runoff events became intermittent flow, and peak discharge reduced in a large amount. The multipond system held 90% runoff for a daily rainfall 141 mm, and peak discharge reduced from 2.6 mV1 to 0.3 mV. The studied stream was divided into 4 channelized reaches (1.3 km), a pond reach (0.15 km) and 3 estuary reaches (0.36 km). It was found that retention of TP, DRP, TN, N03~-N, NH4+-N, and TSS predominantly occurred in the pond reach and estuary reaches. Pollutants retained in the pond reach and estuary reach accounted for more than 50% of those retained in the whole stream. The retention mostly happened in the rain-runoff events and it was 13 to 27 times than that in baseflow. The results showed that the channelized reach was the most important source for pollutants release under either runoff or baseflow conditions, and it accounted for more than 90% of whole stream release. There was a high spatial variability of nutrient dynamics in different channelized reaches. The channelized reach directly discharging into the pond reach did always retain pollutants under base flow and runoff conditions, whereas the other channelized reaches performed differently in different hydrological conditions. The high spatial dynamics of nutrients and TSS in the stream indicated that anthropogenic disturbance of the agricultural headwater stream, such as channelization and excavation, would be expected to decrease the capacity of pollutants retention in the stream. (3) Landscape pattern, fertilizer application, and land management played an important role in pollutants transportation and retention, especially for complex landscape structure, multipond system. Five types of land uses (fertilized paddy, paddy, seeding paddy, pond, waste land) were chosen to study the relationship between landscape pattern and phosphorus, nitrogen and TSS, and the land uses on soil nutrients variability was also investigated. The results were shown as follows. (1) The multipond system influenced the NPS production, transportation and nutrients form. (2) There was no obvious difference of TP, TN among waste land, paddy and seeding paddy, whereas significant difference of N03~-N and NH4+-N was discovered, and there was significant difference of NO3~-N and NH4+-N between the upper and lower soil. (3) The fertilized paddy was the most pollution source in the watershed exporting much pollutants during rain-runoff periods, and the pond, waste land and unfertilized paddy could retain pollutants, and the river-channel system as the main transport channel for pollutants. (4) The amount and forms of nutrients were influenced by the landscape pattern of watershed, and multipond system showed stable retention characteristics during runoff periods, especially for particulate pollutants. The spatial distribution of source and sink landscape types in the watershed has a strong impact on the pollutants transportation and forms in surface runoff, and the human interference accelerated the spatial variability. (4) Constructed wetlands were being used to intercept surface runoff, in order to reduce the pollutants loads and to improve the water resource utilization efficiency. A model on wetlands area and retention rate was developed, which considered hydrological, hydraulic characteristics of wetland. In wetland's area/efficiency sub-model, main hydrological parameters including precipitation, infiltration, evaporation, transpiration and outflow were considered. As runoff flow in wetland as unsteady one-dimensional, the diffusive wave approximation to the Saint-Venant equations was used to simulate the nutrients flow and retention in the wetland, and inverse Guss distribution function was utilized as analytical solutions for diffuse approximation to present hydrological and hydraulic characteristics of inflow. The model fits different types wetlands, with its parameters easy to confirm. (5) The NPS model, including rain-runoff sub-model, pollutants sub-model, surface runoff and pollutants flow concentration sub-model, and retention sub-model, was developed. In the rain-runoff and pollutants sub-model, a distributed group response unit approach was used to calculate water quantity and quality of watershed. That was runoff, sediment yield, and nutrient concentration was calculated separately for each land cover class, weighted by area. In the surface runoff and pollutants flow concentration sub-model, according to the contributing area lied in the sides of stream, the watershed was treated as the drainage system to simulate runoff and pollutants routing in the watershed. And analytical solutions for diffuse approximation to the Saint-Venant equations in one dimensions was used to model the runoff and pollutants routing. In the retention sub-model considering the hydrological and hydraulic characteristics effects of runoff, the advection equation was employed to simulate runoff and pollutants retention in the ponds, and inverse gauss distribution function was used to represent the effects of ponds on pollutants. The two important model parameters were calculated either by measured data or watershed hydromorphology, which based on the rainfall-runoff and rainfall-pollutants relationship that was related to rainfall excess, land use and the geographic information of the basin.