Hello All, The June 2 update follows. The final tracer release experiment is going extremely well. More later. Cheers, Dennis June 2 ------ Dye survey operations continued. Forecasting Activities ---------------------- The location of our operational activities necessitates use the gbk1 mesh in our modeling work; the northern boundary of bank150 cuts right through the region of interest. Use of the larger mesh has made the problem more challenging from a computational point of view. Wherease typical forcast simulations (3 forward runs and 2 inversions) required 2.75 hours of CPU time on bank150, they require 6.3 hours on gbk1. The additional computational load has an impact on the throughput of the modeling work. Although four simulations can be run simultaneously, the results are not available until eight hours later (allowing time for post-processing and plotting). Thus, a typical working day accomodates only one "turnaround." More specifically: in the morning, results from the previous night are plotted and analyzed, and a new set of simulations are launched. Those results are available by late afternoon, and a second set of calculations based on those results are started that night; however, the new results are not available until the following day. With respect to forecasting operations, this dictates that each day's central run must be launched in the morning, regardless of whether or not updated forcing products arrive in the 0930 email download. This is exactly the situation that arose on June 2. Because the updated atmospheric forecast was not received at 0930, the central forecast EN324NF_FC.14 was run using forcing products from the previous day. For the time period after which those products had expired, climatological surface fluxes were prescribed. A sensitivity experiment EN324NF_FC.15 was run in which ADCP data transmitted from Oceanus was assimilated. Both of these calculations were repeated later in the day when the updated atmospheric forcing products were received. The following table summarizes the results of these simulations with respect to their forecast skill in the first North Flank drifter deployment (now several days old). Note that the best forecasts from the previous three days are included for comparison. Simulated and observed drifter trajectories are plotted together in Figures 0602.1a-d. Forecast Error (km) Deep Shal --------- ---- Run ID Description 120m 150m 150m --------------------------------------------------------- EN324NF_FC.02 May 30 Central 10.1 10.1 5.0 EN324NF_FC.07f May 31 Best 5.8 4.9 11.3 EN324NF_FC.13 June 1 Best 7.5 8.3 7.4 EN324NF_FC.14 June 2 Central 8.7 9.7 5.9 EN324NF_FC.15 (14) + Oceanus ADCP 8.7 9.7 5.9 EN324NF_FC.16 (14) + updated forcing 8.9 9.9 5.8 EN324NF_FC.17 (15) + updated forcing 8.9 9.9 5.8 These results demonstrate a continuing decrease in forecast skill associated with assimilating data too far beyond the time period of the drifter observations (as noted in the May 31 / June 1 narrative). Assimilation of ADCP data from the Oceanus hass little impact on the forecast skill, due to the fact that she is conducting a local area survey some 70 miles away on the Northeast Peak. It is also worthy of note that the updated atmospheric forcing products did not significantly alter the results in this case. This is not surprising, given that this measure of skill is based on data from several days ago. More sensitivity can be expected in comparisons with the current drifter deployment, for which we expect to have data on June 3. END