Hello All, I finally figured out why the residual velocity errors were so high in the forecast runs. The short answer is that the ADCP transducer was reinstalled in January, so the correction factor we had been using for its angular offset was obsolete. Through experimentation I came up with a new value for the offset, and everything is back to normal. Fortunately, this problem had little impact on the forecast drifter trajectories. Our initial forecasts seem to match the drifter data quite well. See details below. Charlie: THANK YOU FOR COMING TO CAMP OSSE IN DECMEBER! There is no way I would have ever figured this out had it not been for the training you provided in ADCP data processing. I am sending an m3d file containing the ADCP data from the first VPR survey to the Edwin Link. Data from the second survey to follow later today. Cheers, Dennis May 8 ----- The fact that the residual velocity error was so high in assimilation run EN323_FC.06 was troubling. Closer examination of the residuals revealed that they were consistently ninety degrees to the right of the ship's alongtrack direction (Figure XX). This suggested that there may be a problem with the rotation angle used to correct the alignment of the ADCP (parameter PHI in Charlie Flagg's processing software). A calibration experiment was conducted in which PHI was systematically varied to minimize the RMS velocity observed in a sample data set (en323_dr1) consisting of the first portion of the first VPR survey. It was found that the optimal rotation angle was 9.01 degrees: Angle RMS Velocity -------------------- 5.00 0.5072 6.63 0.4547 * Old PHI 7.00 0.4459 8.00 0.4289 8.50 0.4245 8.75 0.4233 8.88 0.4230 8.94 0.4229 8.97 0.4229 9.00 0.4228 9.01 0.4228 * Optimal PHI 9.03 0.4228 9.13 0.4229 9.25 0.4231 9.50 0.4240 10.00 0.4280 A clear difference between the old and new calibrations is evident in the processed data (Figure XX). With the old calibration, there is clear vertical striping in the alongtrack velocity (now obvious in retrospect). This striping is not present when the new calibration is used. Run EN323_FC.06 was then re-run (as EN323_FC.06b) with the calibrated ADCP data. The residuals dropped down to levels consistent with prior experiments: Prior Calibration New Calibration EN323_FC.06 EN323_FC.06b ----------------------------------------------- BPE: 22.3 cm/sec 14.9 cm/sec 2nd Q4 run: 17.3 cm/sec 7.9 cm/sec 3rd Q4 run: 17.0 cm/sec 6.9 cm/sec It is interesting to note that while this calibration significantly reduced the residual velocity error, it did not have a particularly adverse impact on the forecast drifter trajectories; they were very similar to those in the run using the uncalibrated data (Figure XX). This demonstrates how effective the inversion in frequency space is at ignoring noise which is not associated with the signal of interest. After it was determined that this new calibration of the ADCP data was needed, the ship's tech Bill Fanning was consulted as to why there would have been any change in its configuration. He said that the transducer had been out for service last year and it was re-installed in January. Eureka! Regular forecasting operations were continued using the new ADCP calibration. The May 8 central forecast EN323_FC.11 included updated atmospheric forcing as well as TDBCs received in the 1530 email download. ADCP data from the first survey is assimilated. Duration of the run is now extended to day 129.5. Results show that the drifters continue their westward movement. This appears to be consistent with a preliminary look at the drifter observations; the data will be available in digital form on May 9 for quantitative comparisons.