Sorry for the forum being offline some hours, guys! DM
If you look at this again: http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss_nowcast_anim30d.gifThere is a perfect, circled boundary in the salinity near the end of the simulation. No way that this is caused by some natural current [...]
Plinius first mentioned it here about a week ago:Quote from: plinius on July 07, 2015, 10:54:18 AMIf you look at this again: http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss_nowcast_anim30d.gifThere is a perfect, circled boundary in the salinity near the end of the simulation. No way that this is caused by some natural current [...]It's still there, it's now in the nowcasts as well as the forecasts, and it affects both versions of the Navy's models, see here:http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsssnowcast.gifhttp://www7320.nrlssc.navy.mil/GLBhycomcice1-12/navo/arcticsssnowcast.gifDoes anyone have any contacts with the people running these models so we can ask what's going on? As it stands, I don't think either of the versions of the model can be remotely trusted since they are both giving grossly unphysical outputs.
On Thursday, 16 July 2015 14:25:25 UTC+1, Alan wrote:Thanks for letting us know. Reports like this are very useful in catching issues early. As you say, this one seems to have been going on since the middle of June and is clearly an artifact.We recently noticed the related drop in ice thickness in the global model, and it initially looked like it was due to day to day variability in AMSR2 satellite sea ice concentration (which we only started using this year). However, ACNFS is less senstitive to these observations (because it only assimilates them near the ice edge), but it has the SSS issue too. It did not occur to me to look at SSS, but it is effected whenever sea ice freezes or melts so I should have done so.I will report back once we have found and fixed the problem.Alan.QuoteOn Wednesday, July 15, 2015 at 4:31:07 PM UTC-5, Peter Ellis wrote:Hello,Not sure if this is the right forum to ask, but there is something very odd about the sea surface salinity modelling in the current version of ACNFS (which uses HYCOM), here:http://www7320.nrlssc.navy.mil/hycomARC/arctic.htmlhttp://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsssnowcast.gifFor some reason the surface salinity shows a very sharp circular discontinuity, almost precisely at the 80th parallel. It's been consistent for at least the last two weeks, and arguably before.http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss/nowcast/sss2015070118_2015070200_040_arcticsss.001.gifhttp://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss/nowcast/sss2015061818_2015061900_040_arcticsss.001.gif(In that last, although it's not as pronounced, there still looks to be a signficant discontinuity at many points along the 80th parallel, although the colour scheme can of course sometimes be misleading.It shows up even more clearly in the global version of the model here (may be password-protected).http://www7320.nrlssc.navy.mil/GLBhycomcice1-12/navo/arcticsss/nowcast/sss2015071412_2015071500_924_arcticsss.001.gifI cannot think of any plausible mechanism which would lead to such an effect, and wondered if this is indicative of something going astray "under the hood" in terms of either the model itself or these two specific implementations of it.Can anyone shed any light on this?Thanks,Peter Ellis
On Wednesday, July 15, 2015 at 4:31:07 PM UTC-5, Peter Ellis wrote:Hello,Not sure if this is the right forum to ask, but there is something very odd about the sea surface salinity modelling in the current version of ACNFS (which uses HYCOM), here:http://www7320.nrlssc.navy.mil/hycomARC/arctic.htmlhttp://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsssnowcast.gifFor some reason the surface salinity shows a very sharp circular discontinuity, almost precisely at the 80th parallel. It's been consistent for at least the last two weeks, and arguably before.http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss/nowcast/sss2015070118_2015070200_040_arcticsss.001.gifhttp://www7320.nrlssc.navy.mil/hycomARC/navo/arcticsss/nowcast/sss2015061818_2015061900_040_arcticsss.001.gif(In that last, although it's not as pronounced, there still looks to be a signficant discontinuity at many points along the 80th parallel, although the colour scheme can of course sometimes be misleading.It shows up even more clearly in the global version of the model here (may be password-protected).http://www7320.nrlssc.navy.mil/GLBhycomcice1-12/navo/arcticsss/nowcast/sss2015071412_2015071500_924_arcticsss.001.gifI cannot think of any plausible mechanism which would lead to such an effect, and wondered if this is indicative of something going astray "under the hood" in terms of either the model itself or these two specific implementations of it.Can anyone shed any light on this?Thanks,Peter Ellis
The issue was spurious "SST" observations from the VIIRS satellite under sea ice, due to a new cloud detection scheme at NAVOCEANO that allowed more sun-glint areas (to get more SST observations) but which failed to reject sea ice regions in summer. The model "believes" the data and so SST and sea ice concentration were fighting each other, leading to massive ice melt. The bad SSTs started late in May, so we are going to rerun without VIIRS from May 1st to the present. This is going to take a while to complete.Alan.
What? No conspiracy - just science at work including the occasional misstep? Phaw, how boring Nice catch guys. Glad to see they were able to find the problem fairly quickly once they were aware of it.
The GLB HYCOM+CICE model seems to be fixed, working as in Mayhttp://www7320.nrlssc.navy.mil/GLBhycomcice1-12/arctic.html
