arameter Estimation for Gravity Wave Schemes using a Genetic Algorithm

M. PULIDO1, S. POLAVARAPU1, T. SHEPHERD1, J. THUBURN2

Montreal Canada

Otro; MOCA Joint Assembly; 2009

International Association of Meteorology and Atmospheric Sciences (IAMAS), IAPSO and IACS

<!-- @page { size: 8.5in 11in; margin: 0.79in } P { margin-bottom: 0.08in } --> There is a current need to constrain the parameters of gravity wave drag schemes of climate models using observational information instead of tuning them arbitrarily. In this work, a technique is developed using data assimilation principles to estimate parameters from gravity wave schemes. We define a cost function that measures the differences between the zonal and meridional components of the 'observed' gravity wave drag field and the gravity wave drag calculated with a scheme. Because most gravity wave drag schemes assume instantaneous vertical propagation of gravity waves in a column, observations in a single column can be used to obtain information about input parameters to a gravity wave drag scheme so that a one-dimensional problem is formulated. The geometry of the cost function presents some difficulties, including multiple minima and ill-conditioning. To overcome these dificulties we propose a genetic algorithm to minimize the cost function. In this work the performance of this algorithm to estimate the parameters is shown using twin experiments, i.e. gravity wave drag 'observations' are generated with the gravity wave drag scheme and known 'true' parameters. The minimization is constrained to a parameter domain that is based on a physically realistic range of wave parameters. We conclude that the parameter estimation using a genetic algorithm is robust over a broad range of prescribed 'true' parameters.