Most spatially explicit hydrologic models require estimates of air temperature patterns. For these models, empirical relationships between elevation and air temperature are frequently used to upscale point measurements or downscale regional and global climate model estimates of air temperature. Mountainous environments are particularly sensitive to air temperature estimates as spatial gradients are substantial, and air temperature plays a critical role in snow-related processes. We use a distributed, coupled ecohydrologic model to compare estimates of streamflow, snowmelt, transpiration, and net primary productivity (NPP) using five temperature interpolation approaches for a forested mountain basin that is dominated by a rain-snow zone in Western Oregon, USA. We compare model estimates using a standard adiabatic lapse rate of −6.5°C km−1; basin-specific lapse rates created using daily point observations at high, middle, and low elevations; and gridded temperature estimates from the Parameter-elevation Regressions on Independent Slopes Model (PRISM) derived at 800 and 50 m resolutions. We show that temperature interpolation strategies influence model calibration. Point-based estimates using a low-elevation station or 800 m PRISM grids result in significantly fewer parameter sets that model streamflow well, suggesting a bias in parameter selection due to errors in input data. The greatest postcalibration impact of temperature lapse rate estimates occurs for model estimates of NPP. The constant temperature lapse rate results in substantially reduced NPP estimates that are more sensitive to the interannual variation in climate forcing.