Best practices for estimating near‐surface air temperature lapse rates

Abstract The near‐surface air temperature lapse rate is the predominant source of spatial temperature variability in mountains and controls snowfall and snowmelt regimes, glacier mass balance, and species distributions. Lapse rates are often estimated from observational data, however there is little guidance on best practices for estimating lapse rates. We use observational and synthetic datasets to evaluate the error and uncertainty in lapse rate estimates stemming from sample size, dataset noise, covariate collinearity, domain selection, and estimation methods. We find that lapse rates estimated from small sample sizes (<5) or datasets with high noise or collinearity can have errors of several °C km −1 . Uncertainty in lapse rates due to non‐elevation related large‐scale temperature variability was reduced by correcting for spatial temperature gradients and restricting domains based on spatial clusters of stations. We generally found simple linear regression to be more robust than multiple linear regression for lapse rate estimation. Finally, lapse rates had lower error and uncertainty when estimated from a sample of topoclimatically self‐similar stations. Motivated by these results, we outline a set of best practices for lapse rate estimation that include using quality controlled temperature observations from as many locations as possible within the study domain, accounting for and minimizing non‐elevational sources of climatic gradients, and calculating lapse rates using simple linear regression across topoclimatically self‐similar samples of stations which are roughly 80% of the station population size.