Recorded presentation4.2
On the predictability of coastal winds using an adjoint modeling system
James D. Doyle, NRL, Monterey, CA; and C. M. Amerault
In this study, we investigate the predictability of summertime coastal wind flow along the U.S. West Coast through application of the recently developed adjoint and tangent linear models for the atmospheric portion of the nonhydrostatic Coupled Atmosphere/Ocean Mesoscale Prediction System (COAMPS) over the Northwestern Pacific. The adjoint response function in this case is the kinetic energy centered on Monterey Bay. Strong winds in the atmospheric boundary layer, directed from a northwesterly direction, are prevalent along the California coast during the spring and summer months. As this strong northwesterly flow impinges on the complex coastal terrain, which includes a series of coastal bays and capes along with several prominent coastal mountain ranges, sharp gradients are forced in the near surface along-shore wind speed, which contribute to forcing an upwelling response in the upper ocean circulation. In order to improve our understanding of the predictability of these atmospheric conditions, we apply the adjoint system to a typical case during the summer months over a 12-h period initialized at 1200 UTC 25 July 2007. The nonlinear model forecast wind speed valid at 0000 UTC 26 July 2007 exhibits several strong maxima along the northern California coast and extends southward across the Monterey Bay. At 500-m, the u-wind component sensitivity is comprised of a complex pattern along the coast and over the coastal ranges to the north of Monterey Bay. Some of the sensitivity maxima and minima extend as far as ~550 km to the north of the cost function box, implying a phase speed of ~15 m/s, considerably faster than the mean advective speed of 10 m/s within the boundary layer. At an altitude of 3 km, well above the boundary layer, the sensitive regions extend farther offshore and are less in magnitude. The adjoint model is used to create optimal perturbations for the tangent linear model based on a scaling of the sensitivities. In this manner, the accuracy of the tangent linear approximation can be tested through comparison of the optimal perturbations evolved in the tangent linear and nonlinear models. The tangent linear perturbation growth is a maximum just downwind of the Santa Cruz headlines near the cost function region and in excellent agreement with the nonlinear model. This preliminary study underscores the complex nature of atmospheric predictability in the coastal environment. The most sensitive structures are dispersed over large distances upstream of the region of interest along the northern California coast with a wave-like character to the structure, which implies rapid energy propagation in the low-levels.
Session 4, Advances in Modeling and Forecasting-II
Monday, 12 January 2009, 4:00 PM-5:30 PM, Room 126A
Previous paper Next paper