Rapid urbanization coupled with climate change poses a severe threat to the sustainability of water resources in megacities located within arid and semi-arid regions. Accordingly, this study aims to evaluate the spatio-temporal dynamics of the "water yield" ecosystem service driven by the interactive effects of Land Use/Land Cover (LULC) modifications and climate variations in the urban-mountainous Kan-Tehran watershed. To achieve this objective, the InVEST model—grounded in the Budyko curve framework—was employed over a 32-year statistical period (1990–2022). Input data, including LULC maps derived from Landsat 5 and 8 satellite imagery, alongside meteorological data from synoptic stations (comprising precipitation and reference evapotranspiration), were processed and modeled within a GIS environment. The findings indicate a twofold expansion in built-up areas, increasing from approximately 20% in the 1990s to 40% in the 2000s. Modeling results revealed a strong correlation between water yield and fluctuations in precipitation and evapotranspiration; specifically, during the 1990s, the highest water yield depth was recorded in the central Kan watershed (378.41 mm), with short vegetation contributing the largest volumetric share (3,013.11 thousand cubic meters). Following a surge in precipitation during the 2000s (reaching up to 558 mm in the northern reaches), water yield peaked. Consequently, surfaces with lower permeability, such as rock outcrops, accounted for the highest share, generating 44,329.92 thousand cubic meters of runoff. Conversely, the 2010s witnessed a substantial decline in the water balance, coinciding with a severe deficit in precipitation (dropping to 173 mm in southwestern Tehran) and an increase in evapotranspiration. This resulted in a significant volumetric reduction in water yield across rock outcrops and tree canopies, plummeting to 27,218.40 and 1,378.42 thousand cubic meters, respectively..