Air temperature is a key control of processes affecting snow and glaciers in high-elevation catchments, including melt, snowfall and sublimation. It is therefore a key input variable to models of land–surface–atmosphere interaction. Despite this importance, its spatial variability is poorly understood and simple assumptions are made to extrapolate it from point observations to the catchment scale. We use a dataset of 2.75 years of air temperature measurements (from May 2012 to November 2014) at a network of up to 27 locations in the Langtang River, Nepal, catchment to investigate air temperature seasonality and consistency between years. We use observations from high elevations and from the easternmost section of the basin to corroborate previous findings of shallow lapse rates. Seasonal variability is strong, with shallowest lapse rates during the monsoon season. Diurnal variability is also strong and should be taken into account since processes such as melt have a pronounced diurnal variability. Use of seasonal lapse rates seems crucial for glacio-hydrological modelling, but seasonal lapse rates seem stable over the 2–3 years investigated. Lateral variability at transects across valley is high and dominated by aspect, with south-facing sites being warmer than north-facing sites and deviations from the fitted lapse rates of up to several degrees. Local factors (e.g. topographic shading) can reduce or enhance this effect. The interplay of radiation, aspect and elevation should be further investigated with high-elevation transects.