Lake emissions of the climate forcing trace gas methane (CH4) are spatiotemporally variable, but biases in flux measurements arising from undersampling are poorly quantified. We use a multiyear data set (2009-2017) of ice-free CH(4)emissions from three subarctic lakes obtained with bubble traps (n = 14,677), floating chambers (n = 1,306), and surface concentrations plus a gas transfer model (n = 535) to quantify these biases and evaluate corrections. Sampling primarily in warmer summer months, as is common, overestimates the ice-free season flux by a factor 1.4-1.8. Temperature proxies based on Arrhenius functions that closely fit measured fluxes (R-2 >= 0.93) enable gap filling the colder months of the ice-free season and reduce sampling bias. Ebullition (activation energy 1.36 eV) expressed greater temperature sensitivity than diffusion (1.00 eV). Resolving seasonal and interannual variability in fluxes with proxies requires similar to 135 sampling days for ebullition, and 22 and 14 days for diffusion via models and chambers, respectively.