Wetland methane emissions at high latitudes are central to the present and future global methane budget, yet remain poorly constrained by observations. Here we combine airborne in-situ measurements from the MAGIC2021 campaign with Lagrangian transport modelling to estimate local wetland methane emissions near Kiruna, Sweden, and test whether current bottom-up wetland models reproduce the emission field that generates the observed atmospheric signal. We find no significant difference in performance between the assessed models, WetCHARTs and JSBACH-HIMMELI, and mean emissions that agree with the observation-based estimates over the targeted plumes. Emission variability, however, is underestimated by every product, with simulated standard deviations below a quarter of the observed value for half of them. Because our footprints integrate over areas larger than the model grid cells, variability misrepresentation is attributed to emissions varying on timescales shorter than model time steps, rather than to finer spatial sampling, which the method cannot test. This explains earlier findings from the same campaign, where model emissions appeared overestimated when sampling a generic atmosphere rather than plumes alone. In WetCHARTs, rescaling to a prescribed global total causes the high latitude flux to fall as the assumed temperature sensitivity rises, so the boreal contribution is set in part by a parametrisation chosen for the tropics. Agreement over emission hotspots is therefore not evidence that the regional emission field is correct, and the apparent bias of a wetland model depends on how the atmosphere is sampled, a caveat applying to tower networks, airborne campaigns and satellite retrievals alike.