Adjoint-Based Targeting of Observations for Fastex Cyclones,

Abstract : Recent development of adjoint versions of numerical weather prediction models makes possible a number of new diagnostic and interpretive techniques that can be applied to atmospheric forecasting. For example, adjoint models can be used to evaluate conceptual ideas of atmospheric processes, including considerations of potential vorticity and the Charney and Eady problems that relate to extratropical cyclone development (Langland et al. 1995). In addition, the sensitivity of forecasts to initial condition error can be studied with adjoint methods, providing insight into atmospheric predictability. A related problem is the idea of using an adjoint model to 'target' specific upstream locations where additional observations are likely to have maximum impact on forecast skill. The work described here is performed with the goal of assisting the Fronts and Atlantic Storm Track Experiment (FASTEX), which has a field phase in January and February 1997. Patterns of initial condition sensitivity in four North Atlantic frontal cyclones are examined using adjoint and tangent linear versions of the Navy Operational Global Atmospheric Prediction System (NOGAPS, Hogan and Rosmond 1991), and the Mesoscale Adjoint Modeling System Version 1 (MAMS1, Errico et al. 1994). In MAMS1, moist processes, including convective and nonconvective precipitation are included. NOGAPS is run at T79L18 resolution, MAMS1 with 60km grid spacing and 14 sigma levels.