SPGBRBT - Special Project Routine Back Trajectories - Interim Report 1996 Paul Berrisford - 5th June 1996 The aim of the project is to provide a database of routine back trajectories. During the whole of 1995 252 5 day back trajectories were computed on a daily basis from a time of 12z. The arrival points were chosen to be on three regular latitude/longitude grids centred over the U.K., the mid-Atlantic storm track region and the eastern U.S.A. and were all located at 900hPa. The trajectories are computed for clusters of arrival points in order to estimate how sensitive the results are to small changes in the arrival position. This knowledge is useful for determining how accurate the results are likely to be. The computations were performed using the ECMWF trajectory package and the 6 hourly spectral T106 uninitialised operational analyses of the three components of the wind and ln(surface pressure). This data is interpolated to a 1.5x1.5 degree regular latitude/longitude grid. The package uses a simple mid-point type of algorithm with linear interpolations in both time and space and a timestep of 15 minutes. The output gzipped ASCII files contain information giving latitude, longitude and pressure at each timestep. The dataset is available both at the ECMWF (in ecfile directory /rgp/traj) and at the BADC at RAL. For 1996 several changes were made to the basic system. Firstly the algorithm was changed: a 4th order Runge-Kutta scheme was introduced and the vertical interpolations were made to be cubic. When a timestep of 30 minutes is used this scheme consumes the same amount of computer resources as the mid-point type scheme mentioned above. The difference in horizontal position between calculations performed with the old and the new algorithm are usually quite small, though sometimes they can differ markedly. The pressure difference however is typically a few hPa, sometimes 20-30hPa and occasionally as large as 200hPa. In addition to the arrival points used in 1995 112 extra points were added in order to provide data for the ACSOE community. These points are located in the vicinity of the U.K. and the eastern Atlantic and their pressure's range between 1000 and 700hPa. The package has been run four times daily (for 00z, 06z, 12z and 18z) starting from 1st March 1996. On 30th January 1996 the ECMWF analysis system was changed to 3D variational analysis. Hence, there are no separate uninitialised/initialised analyses from this time. The data gap of January and February 1996 (caused when making the changes to the system) will be filled in the near future using initialised analyses. In 1996 trajectories computed from forecast data are also being produced for various observational campaigns that are running over the summer. The ACSOE community requires 5 day forecast back trajectories from the 112 arrival points mentioned above. During June to mid-August 1996 six such computations are performed daily from the midday forecast with trajectory arrival times at the following forecast time steps (hours): 24, 36, 48, 72, 96, 120. The TOASTE community require 3 day forward trajectories for 77 points around Europe centred at 500hPa. During June 1996 three such computations are performed daily from the midday forecast with trajectory departure times at the following forecast time steps (hours): 12, 24, 36. Implementing the above means that at the peak of production, in June 1996, 13 sets of trajectory computations (4 actual, 6 back forecast, 3 forward forecast) are computed daily. Job submission to the Cray has been automated by means of cron jobs. Again, the results are stored at the ECMWF and are also sent to the BADC at RAL for analysis and dissemination to the various communities. FILENAMES --------- --------- These contain the trajectory base date in the form YYMMDD e.g. 960105 meaning 5th January 1996. This date indicates the arrival (departure) time for back (forward) trajectories. 1995 Actual Back Trajectory Files (trajectory base time 12z) ------------------------------------------------------------ atYYMMDD_1.gz 1996 Actual Back Trajectory Files --------------------------------- aYYMMDDx.gz where x=a indicates trajectory base time 00z x=b 06z x=c 12z x=d 18z 1996 Forecast Back Trajectory Files (forecast base time 12z) ------------------------------------------------------------ fYYMMDDx.gz where x=a indicates forecast time step 24 hours and trajectory base time 12z x=b 36 00z x=c 48 12z x=d 72 12z x=e 96 12z x=f 120 12z 1996 Forecast Forward Trajectory Files (forecast base time 12z) --------------------------------------------------------------- mYYMMDDx.gz where x=a indicates forecast time step 12 hours and trajectory base time 00z x=b 24 12z x=c 36 00z