Tropical Cyclone Prediction: Initial Condition Sensitivity

The numerical prediction of tropical cyclones is critically dependent on the initial structure of the TC circulation in the numerical model. Defining this structure can be difficult, due to the lack of in-situ observations, particularly in the eastern and western Pacific basins. Often, we must rely on satellite imagery and remotely-sensed observations to define the key parameters of the TC structure. A variety of methods have been, and are currently used, to incorporate these parameters into the initial conditions of tropical cyclone models. Within the version of the Coupled Ocean/Atmosphere Mesoscale Prediction System used for tropical cyclone prediction (COAMPS-TC), we create synthetic observations, resembling rawinsonde data, using available observations and/or observed features of tropical cyclones; and then we assimilate these observations into the COAMPS-TC initial conditions using the 3-dimensional NRL Atmospheric Variational Data Assimilation System (NAVDAS). We have found that there can be a large sensitivity to the predicted track of tropical cyclones when using different methods for generating and placing synthetic observations for the specification of the tropical cyclone initial structure.

Our current method for the generation of synthetic observations is to define observations of pressure, temperature, and wind at specific points around the TC, as represented by a Rankine wind vortex. The details of the Rankine vortex are defined by the parameters in the warning message: the location of the TC; the radii of the 34, 50, and 65 kt winds in each of four quadrants; and the radius and intensity of the maximum wind. Synthetic observations are placed at 41 points around the TC, with 1 point at the storm center, and then 8 points on each of 5 different radii out from the storm, at ½-, 1-, 2-, 3-, and 4-degrees from the center of the storm. For weak systems, an additional 8 points are included at 6-degrees from the TC center. At each of these locations, synthetic wind observations are defined at 1000, 965, 925, 850, 775, 700, 600, 500, and 400 hPa; with the height and temperature specified only at 1000 hPa. Prior to inserting the observations within NAVDAS, the TC circulation in the COAMPS-TC first-guess field is relocated to the observed TC location. During the NAVDAS analysis, the geostrophic constraint is relaxed and the correlation lengths are shortened in the vicinity of the tropical cyclone. By initializing the tropical cyclone circulation in this manner, we have found that while the initial conditions created often exhibit the general structure of the observed TC, our ability to more precisely fit the observed structure was occasionally hampered by specifying the synthetic observations at fixed locations relative to the TC, and that the initial circulation within the TC vortex was often unbalanced.