More odd effects in HYCOM / ACNFS

[jim...@gmail.com]

Further to the SSS question posed by Peter Ellis this time last year, recent ACNFS runs have been producing some strange results. By way of example here's the 1 day sea ice thickness forecast for yesterday:

http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticictn/nowcast/ictn2016071318_2016071400_042_arcticictn.001.gif

The Kara Sea hasn't suddenly acquired a covering of sea ice:

http://go.nasa.gov/29UO7at

Is there a problem of some sort?

[Alan]

Thanks for pointing this out.

We now use the National Ice Center's IMS sea ice extent product in our assimiltion, see http://www.natice.noaa.gov/ims/.  This is manually produced every day, and generally avoids the artifacts and difficiencies of satellite sea ice concentrations.  However yesterdays IMS fields were not good, and that is where the spurious sea ice came from.  Today's IMS field is good, and we may rerun the 2016/07/13-18Z nowcast/forecast to clear this up.

Alan.

[jim...@gmail.com]

Thanks for that explanation Alan.

Yesterday's forecast was an extreme example, but I'm not the only amateur Arctic sea ice follower who has been wondering if there's something else amiss in the recent ACNFS forecasts. Endeavouring to "ground truth" ACNFS against MODIS (when clouds allow!) for example seems to be a non trivial task at the moment. Are you content that apart from the bad IMS data you refer to things are working as designed?

Jim

[Alan]

My best guess is that IMS is also the problem in the other cases you describe.  Because IMS has a person doing the analysis, it may not be as consistent as satellite data on a day to day basis.  We are looking into how to make IMS more consistent.

If you can point me to examples I can look into them further.

We also added AMSR2 (in addition to SSMIS) at the same time as IMS, see

P. G. Posey, E. J. Metzger, A. J. Wallcraft, D. A. Hebert, R. A. Allard, O. M. Smedstad, M. W. Phelps, F. Fetter, J. S. Stewart, W. N. Meier, and S. R. Helfrich, 2015:
Improving Arctic sea ice edge forecasts by assimilating high horizontal resolution sea ice concentration data into the US Navy's ice forecast systems
The Cryosphere vol 9 1735-1745

Another recent publication:

D. A. Hebert, R. A. Allard, E. J. Metzger, P. G. Posey, R. H. Preller, A. J. Wallcraft, M. W. Phelps and O. M. Smedstad, 2015:
Short-term sea ice forecasting: An assessment of ice concentration and ice drift forecasts using the U.S. Navy's Arctic Cap Nowcast/Forecast System
Journal of Geophysical Research: Oceans vol 120 doi:10.1002/2015JC011283

Alan.

[jim...@gmail.com]

Thanks Alan,

Actually I have already read those two papers, which is one of the reasons I posed my question in the first place. From the second one:

"Ice Concentration is obtained from the Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave Imager/Sounder (SSMI/S) and AMSR2. Ice concentration from SSMI/S is derived using the NASA Team 2 sea ice algorithm, while ice concentration from AMSR2 is derived from the Bootstrap Algorithm."

Here's the ACNFS sea ice concentration "nowcast" for July 18th:

http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticicen/nowcast/icen2016071718_2016071800_042_arcticicen.001.gif

whereas this is a selection from the University of Hamburg's concentration "hindcast" for July 18th, derived from AMSR2 data using their ARTIST algorithm:

http://GreatWhiteCon.info/wp-content/uploads/2016/07/Arc_20160718_res3.125_LARGE.jpg

In the sector between 80 N, 140 W and 180 W ACNFS has recently been suggesting sea ice concentrations below 50% which have been conspicuous only by their absence from Hamburg AMSR2, and indeed MODIS when clouds permit:

http://GreatWhiteCon.info/wp-content/uploads/2016/07/CAB-82-135-Terra-721-20160715-768x459.jpg

Can you see what I'm getting at?

Jim

[Alan]

I attach the ACNFS plot you referenced and the corresponding plot from our next global real-time system (GOFS 3.1), which will likely replace our current global analysis (and ACNFS) once NAVGEM 1.4 is operational (it is running here with NAVGEM 1.3, like ACNFS).  They assimilate the same observations, but ACNFS only does so near the ice edge while GOFS 3.1 assimilates sea ice concentration everywhere (but with higher error bounds in the ice interior).  The reason for ACNFS only assimilating near the edge is in part due to that being the most important area for navigation but also because SSMI satelite ice concentrations tend to "over saturate" in the summer.  We switched GOFS 3.1 to "believe" the observations because they are generally better in recent years.

The two are quite different in the interior of the sea ice and GOFS 3.1 is certainly better there.  We did not see the low concentrations near the North Pole in ACNFS this time last year, so something has happened to make the free running model produce low concentrations this summer.  We are looking into it.

Alan.

[jim...@gmail.com]

Thanks Alan,

I look forward to hearing the results of your investigations. Do you have an ETA for NAVGEM 1.4?

In the meantime it seems that I need to become an honourary member of the US Government!

Jim

[Alan]

We are testing with a 1-year hindcast from NAVGEM 1.4 now, but I don't know its exact schedule.

Once GOFS 3.1 is operational, NOAA will run the same HYCOM+CICE system from the Navy's daily initial state but with their global atmospheric forcing rather than NAVGEM.  So there will be two multi-day forecasts to compare.

Alan.