Knowledge of the dominant water sources and their relative contributions to streams in time is important for understanding the underlying hydrological processes as well as managing the quantity and quality of water resources. In many subtropical regions, the complexity of mixed agricultural land and water use in combination with lack of data further inhibits such understanding of the dominant catchment scale runoff generation processes. This study provides new insights into the time-variable interactions of natural and anthropogenic influences on the catchment response through integrated hydrometric and multi-tracer (stable water isotopes, Mg2+, Na+, Si4+, Cl−, and Electricity Conductivity) analyses. The combined diagnostic tools of mixing models (DTMM) and end-member mixing analysis (EMMA) were successfully used to evaluate the spatiotemporal variability in key water sources of a subtropical catchment in China. This study site is characterized by rain-fed uplands and irrigated water paddy fields. The EMMA results for one year of data showed that irrigation water, rainwater and ground water were the three main sources, which contributed to 64%, 19% and 17% of the streamflow on average, respectively. However, temporal patterns in rainfall and irrigation practices did cause significant variability in these relative contributions. Overall, we found that routine agricultural practices to optimize crop growth (especially during paddy growth periods) was a more important factor than hydro-meteorological conditions in controlling the regime and properties of water sources. The relatively simple but successful application of EMMA and DTMM in a complex environment demonstrates that it is a valuable approach for understanding water sources and hydrologic processes concerning agricultural or mixed-land use catchments.