Forest ecosystems are critical for adjusting the dynamic balance of the hydrological cycle. This balance is affected by vegetation community types, phenology, and forest density. Previous long-term catchment-scale model studies have focused on changes in forest areas while ignoring the above factors. Since the 1980s, climate change caused by increases in atmospheric CO2 levels has enhanced forest growth. Moreover, amendments to forest management policies, including intermediate cuttings caused by economic factors, have yielded unprecedented changes in forest ecosystems. In this study, we designed a methodology and created a credible model using the Soil and Water Assessment Tool (SWAT) that can precisely reflect water balance variations caused by different ecosystem situations during long-term changes in forest density. We focused on the Yamato River catchment in Western Japan, which includes three planted forests and one primeval forest, each markedly different with respect to vegetation community composition and management policy. In the process, we examined the ratio of coniferous vegetation and broad-leaved vegetation in different forest areas, used remote sensing methods to quantify the maximum and minimum leaf area index (LAI) of each forest region over 40 years, and calibrated the model by comparing the LAI growth curve, evapotranspiration, and streamflow with observed data. Moreover, we separated the decadal canopy evaporation, transpiration, and soil evaporation from the SWAT output results. We found that (1) forest evapotranspiration has increased in recent decades because of the above reasons; (2) in young or well-managed forests, the forest water balance may have changed significantly with forest growth. For long-term studies, it is necessary to distinguish the growth characteristics of different forests during different periods, and a detailed definition of a mixed forest is required. The forest parameters and growth characteristics are critical for understanding forest ecosystems and cannot be ignored at catchment-scale.