A Comparison of Gravity Wave Drag Parameterizations using Inverse Techniques

SCHEFFLER, G. AND M. PULIDO

Honolulu Hawaii, USA

Conferencia; AGU Chapman Conference on Atmospheric Gravity Waves and Their Effects on General Circulation and Climate; 2011

American Geophysical Union

This work attempts to make a comparison between different gravity wave drag parameterizations. To perform a throughout comparison two inverse techniques are used. The inverse techniques offer a robust framework to perform the drag comparison: for a given ´observed´ gravity wave drag, the inverse techniques estimate the optimal parameters and therefore the optimal gravity wave drag given by each parameterization. In this way the error of each parameterization  can be inferred.The estimation of optimal parameters is performed with two different inverse techniques: the genetic algorithm and the ensemble Kalman filter. We show that the parameter estimation can be carried out efficiently using both inverse techniques. An hydrostatic non-rotational version of Scinocca gravity wave drag scheme and a version of the Warner & McIntyre ultra-simple spectral gravity wave drag scheme are used as parameterization schemes. The ´observed´ gravity wave drag used in this work is the one estimated in Pulido & Thuburn 2008. For both parameterizations, the impact in the drag field of using optimal parameters compared to the use of standard tuned values is shown. In general, Scinocca parameterization gets a closer match to the ´observed´ gravity wave drag using optimal parameters. Both parameterizations do not present a realistic drag in the lower tropical stratosphere. The inverse technique based on the ensemble Kalman filter is a feasible data assimilation technique to be implemented in a full GCM due to its low computational cost. An in depth comparison between the techniques and how does each of them developes under each gravity wave drag scheme is performed, paying special attention to drag estimation on critical latitudes like tropics. Both techniques are likely to be implemented in a GWD parameterization in a GCM, but only the one based in EnKF may result feasible and effective due to its low computational cost.