Over the Upper Colorado River Basin (UCRB), temperatures in widely used gridded data products do not warm as much as mean temperatures from a stable set of U.S. Historical Climatology Network (USHCN) stations, located at generally lower elevations, in most months of the year. This is contrary to expectations of elevation-dependent warming, which suggests that warming increases with elevation. These findings could reflect 1) a genuine absence of elevation-dependent warming in the region, 2) systematic non-climatic influences on either the USHCN stations or high elevation stations, including known inhomogeneities related to changes in the time of observation and instrumentation, or 3) suppression of an elevation-dependent warming signal introduced by changes in the station network. While we cannot categorically dismiss the first two possibilities, we show here that over portions of the 20th century, gridded temperatures warm less than USHCN temperatures and the difference cannot be explained by accounting for known inhomogeneities. These analyses suggest that changing station coverage in the UCRB has influenced trends in gridded temperature estimates that incorporate changing suites of stations over time. Specifically, increases in the number of high-elevation stations in the UCRB may have led to an underestimation of elevation-dependent warming, particularly during the spring and summer. This phenomenon is unlikely limited to this specific basin, and may be present in other high-elevation watersheds across the western U.S.