Toward shadowing in operational weather prediction

This is a report on a NICOP funded project to investigate the feasibility and benefits of computing shadowing trajectories with the Naval Operational Global Atmospheric Prediction System (NOGAPS). Previous results had established that an algorithm to compute shadowing trajectories was successful for a simple quasi-geostrophic (T21L3) model, without using any adjoint information. The first aim was to determine whether algorithms could be successfully applied to the significantly more complex NOGAPS. Experiments at various resolutions from T21L9 to T79L30, using a dry adjoint approximation, showed satisfactory convergence of an iterative, gradient descent based, algorithm. These experiments have provided useful information about the implementation and calibration of shadowing algorithms and an illustration of how shadowing methods could help identify model error and obtain better analyses and forecasts. The results encourage further investigation of shadowing techniques.