Accurate estimation of mountain vegetation gross primary productivity (GPP) at fine spatial resolutions offers opportunities to better understand mountain ecosystems' feedback to the global climate system. Eco-hydrological models have great advantages in simulating mountain vegetation photosynthesis, but their large-scale applications remain challenging at fine spatial resolutions due to the computing resources. In this work, a scheme by integrating an eco-hydrological model called Boreal Ecosystem Productivity Simulator-TerrainLab (BTL) with the linear and non-linear downscaling processes, was developed to obtain large-scale mountain vegetation GPP at the 30 m resolution over four watersheds. Firstly, two coarse spatial resolution GPP were simulated by BTL at 480 m and 120 m. Then, the 30 m resolution GPP was estimated by a linear downscaling process modelled at 120 m and a non-linear downscaling process modelled from 480 m to 120 m. The 30 m resolution BTL-simulated GPP was served as reference for evaluation. Results showed that the Root-Mean-Square-Error (RMSE) after downscaling was decreased by 110 gCm(-2)year(-1) compared to the 120 m resolution BTL-simulated GPP (500 gCm(-2) year(-1)) at the 30 m resolution, highlighting the effectiveness of the proposed scheme in recovering the topographic Variations of mountain vegetation GPP at fine spatial resolutions. Compared to the 120 m resolution BTL-simulated GPP (351 gCm(-2) year(-1)), RMSE after downscaling was decreased by 156 gCm(-2) year(-1) at the 120 m resolution, indicating that the proposed scheme is feasible in correcting GPP errors at coarse spatial resolutions. More specifically, the non-linear downscaling process was observed to effectively improve GPP estimates after linear downscaling, suggesting that the spatial scaling errors in coarse estimates should be considered in the downscaling process. Our study indicates that the scheme that runs eco-hydrological models at coarse resolutions and then downscales them by surface heterogeneity is a practical approach for obtaining large-scale mountain vegetation GPP at fine spatial resolutions.