Urbanization and the subsequent changes in land use/cover inevitably influence the quality and even the quantity of stream water. This issue is widely studied through evaluations on land-use change scenarios or comparisons among historical patterns at the same watershed. However, observational stream discharge changes through urbanization gradient have rarely been discussed. In this study, we analyzed 5-year discharge data from 13 gauges in the Danshui River network with a wide range of urbanization gradient to explore the impacts on observational hydrological characteristics in individual catchments. The results reveal that stream discharge in pristine watersheds is characterized by a larger proportion of baseflow and is less fluctuating. When the forest coverage is <90%, the discharge fluctuation almost doubles. Meanwhile, the baseflow fraction decreases gradually with the increase of paddy area, which may concomitantly result from the increasing irrigation. Such a drop in baseflow may threaten the maintenance of the minimum flow required for the stream aquatic ecosystem. Furthermore, we simulated the stream discharges by TOPMODEL with blind land-use-independent parameters. The results show that the simulated discharges are satisfactory, particularly for the pristine catchments, but not as fitting for the paddy-intensive watersheds perhaps due to the unexpected irrigation. On the whole, the calibrated parameters are dependent with the landscape characteristics. The landscape-based parameter estimations can be applied to simulate discharge well, meaning the potential to assess the ungauged watersheds.