Sensitivity of mountain wave drag estimates on separation methods and proposed improvements

Internal gravity waves (GWs) are ubiquitous in the atmosphere, making significant contributions to the me-soscale motions. Since the majority of their spectrum is unresolved in global circulation models, their effects need to be pa-rameterized. In recent decades GWs have been increasingly studied in high-resolution simulations, which, unlike direct observations, allow us to explore full spatiotemporal variations of the resolved wave field. In our study we analyze and refine a traditional method for GW analysis in a high-resolution simulation on a regional domain around the Drake Passage. We show that GW momentum drag estimates based on the Gaussian high-pass filter method applied to sepa-rate GW perturbations from the background are sensitive to the choice of a cutoff parameter. The impact of the cutoff parameter is higher for horizontal fluxes of horizontal momentum, which indicates higher sensitivity for horizontally propagating waves. Two modified methods, which choose the parameter value from spectral information, are pro-posed. The dynamically determined cutoff is mostly higher than the traditional cutoff values around 500 km, leading to larger GW fluxes and drag, and varies with time and altitude. The differences between the traditional and the modi-fied methods are especially pronounced during events with significant drag contributions from horizontal momentum fluxes. SIGNIFICANCE STATEMENT: In this study, we highlight that the analysis of gravity wave activity from high -resolution datasets is a complex task with a pronounced sensitivity to the methodology, and we propose modified versions of a classical statistical gravity wave detection method enhanced by the spectral information. Although no optimal methodology exists to date, we show that the modified methods improve the accuracy of the gravity wave activity estimates, especially when oblique propagation plays a role.