We will contribute to a more realistic incorporation of the middle atmosphere in weather models, providing a path towards more accurate medium-range forecasts. We use techniques having the advantage of providing continuous observations over long timescales with wide geographical coverage. Our data-driven tools to derive the dynamic properties of the middle atmosphere at 30-60 km altitudes rely on time series of infrasound (very low-frequency, inaudible sound) recorded at ground-based Arctic stations. In particular, we consider infrasonic ambient noise from ocean surface wave interaction - microbaroms - where oceanic wave-action models will be used to identify active sources. These data will provide constraints on probabilistic medium-range weather forecasts and climate models. The middle atmosphere is relatively inaccessible to continuous measurements, particularly for winds above 40 km altitude. The region is poorly studied and not well understood compared with the thermosphere aloft from which satellites and radars provide data, and with altitudes below probed by weather balloons and other technologies. Established forecast models assimilate very little middle atmospheric data, so modelling is less accurate above ~50 km altitude. Still, this region is coupled to the troposphere below, with events like Sudden Stratospheric Warmings (SSWs) known to influence the conditions at ground level. Infrasound data will also be interpreted in the context of: a) Meteor radar wind measurements at 70-100 km: We will obtain new understanding of stratosphere-mesosphere coupling, especially for regional and global events like SSWs. b) The role of middle atmosphere dynamics in probabilistic medium-range weather prediction and sub-seasonal prediction: This will be based on combined meteor radar and infrasound data. We will design diagnostic tools to interpret the data in terms of the middle atmospheric circumpolar vortex circulation.