Models are necessary tools for watershed management. However, applying watershed models is time consuming and requires technical knowledge, including model selection and validation. The objective of this study is to assess two commonly used watershed models and their parameter sensitivity to reduce model loadings and to gain a better understanding of the model performances. The Hydrological Simulation Program-Fortran (HSPF) model and Storm Water Management Model (SWMM) were applied to a mostly forested Taiwanese reservoir watershed with pollution from tea plantations. Statistical analysis showed that both models are suitable for the studied watershed, but the performances of the flow and water quality simulations are different. The mean flow simulated by SWMM was lower than the experimental observations. The HSPF model performed better, possibly because the soil in the study area is highly permeable and the HSPF model has more precise soil layer calculations. SWMM may underestimate the total phosphorous (TP) and suspended solid (SS) loads following small storm events in highly permeable watersheds. The Latin Hypercube-One factor At a Time (LH-OAT) method was used to determine the parameter sensitivity of the HSPF model and SWMM. In both of the models, the parameters related to infiltration and soil characteristics strongly affected the flow simulation, except when using the Horton infiltration method in the SWMM. Manning's roughness coefficient for pervious areas was more sensitive in SWMM than in the HSPF model because SWMM has fewer parameters.
