This paper investigates the link between vegetation types and long-term water balance in catchment areas. We focus on the most widely used water balance formulas or models that relate long-term annual streamflow to long-term annual rainfall and long-term potential evapotranspiration estimates. Our investigation seeks to assess whether long-term streamflow can be explained by land cover attributes. As all but one of these formulas do not use land cover information, we develop a methodology to introduce land cover information into the models' formulations. Then, the modified formulas are compared to the original ones in terms of performance and a sensitivity analysis is performed, with a special focus on the parameters representing vegetation characteristics. In line with the global coverage of long-term water balance models, we base our work on as many basins as possible (1508) representing as large a hydroclimatic variety as possible. Results show that introducing additional degrees of freedom within the original formulas improves overall model efficiency, and that land cover information makes only a small but nonetheless significant contribution to this improvement.