Ocean Ensemble Forecasting and Adaptive Sampling - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 400

the upwelling period, respectively. The ensemble spread is found to be maximized
near the upwelled cold water transport across the mouth of the Monterey Bay.
An ocean adaptive sampling system derived from the ETKF technique is
illustrated using the data collected during the AOSN II field campaign. For a
large number of possible adaptive observations, a signal variance summary map
provides an overview of the predicted reduction in forecast error variance within the
verification region as a function of the location of a plausible future observation. The
predicted reduction in forecast error variance for a large number of possible glider
tracks is summarized and displayed in a bar chart for each feasible deployment.
The real glider tracks from the AOSN II field campaign are used to derive a
signal variance bar chart with 13 possible glider deployments. The ETKF adaptive
sampling distinguishes one path with a large summarized signal variance near the
verification area. The use of this path, in our view, would have been most likely to
reduce the forecast error within the verification region.
As discussed in Majumdar et al. ( 2002 ), the quantitative assessments of the
accuracy of ETKF signal variance predictions require a large number of events.
Unfortunately, the limited events during the AOSN II do not provide enough cases
for such quantitative assessments to be made. Nevertheless, the aforementioned
experiment indicates that the adaptive sampling locations selected using the tech-
nique presented here are, at the very least, consistent with the group velocity of
wave packets of ocean forecast errors that are unlikely to propagate very far over
a 24 h period in the ocean. For the future work, we hope to use a large number
of cases to quantitatively measure the accuracy of the ETKF prediction of forecast
error variance reduction in the ocean prediction.
Acknowledgements The support of the sponsors, the Office of Naval Research, Ocean Modeling
and Prediction Program, through program element N0001405WX20669 is gratefully acknowl-
edged. Computations were performed on zornig, which is a SGI ORIGIN 3800 with IRIX 6.5
OS and 512 R12000 400 MHz PEs and is located at the U.S. Army Research Laboratory (ARL)
DoD Supercomputing Resource Center (DSRC), Aberdeen Proving ground, MD.
References
A OSN: Autonomous ocean sampling network (2003) http://www.mbari.org/aosn/Accessed 30
May 2012
Bishop CH, Toth Z (1999) Ensemble transformation and adaptive observations. J Atmos Sci
56:1748-1765
Bishop CH, Etherton BJ, Majumdar SJ (2001) Adaptive sampling with the ensemble transform
Kalman filter part I: Theoretical aspects. Mon Wea Rev 129:420-435
Bishop CH, Etherton BJ, Majumdar SJ (2006) Verification region selection in adaptive sampling.
Quart J Roy Met Soc 132:915-933
Bishop CH, Holt T, Nachamkin J, Chen S, McLay JG, Doyle JD, Thompson WT (2009) Regional
ensemble forecasts using the ensemble transform technique. Mon Wea Rev 137:288-298
Buizza R, Richardson DS, Palmer TN (2003) Benefits of increased resolution in the ECMWF
ensemble system and comparison with poor-man's ensembles. Quart J Roy Meteor Soc
129:1269-1288