I've been able to look some more into this, with the help of Michael, who sent me the data that allowed me to create this graph:
Michael sent me the CryoSat-2 data from AWI (CryoSat-2 average volume on a combined PIOMAS-AWI grid), and I simply added the daily PIOMAS volume data for January and then divided by 31 to get an average as well.
As you can see the trend lines more or less move the same way, except for this year. According to CryoSat-2 there is slightly more volume now than last year and so the trend line goes up a bit. For PIOMAS the trend line crashes.
The same thing happens in December:
As shendric said:
Ku-Band radar (CryoSat) is sensitive to snow grain size. A larger anomaly in the snow microstructure (depth hoar, ice lenses) may result in too high freeboard/thickness values.
This is most probably the reason for this enormous divergence, as there have been so many Atlantic storms hurled into the Arctic this winter.
Here are a couple of quotes from Robert Ricker's PhD thesis paper
Radar altimetry measurements of the current satellite mission CryoSat-2 show an increase of Arctic sea-ice thickness in autumn 2013, compared to previous years but also related to March 2013. Such an increase over the melting season seems unlikely and needs to be investigated. Recent studies show that the influence of the snow cover is not negligible and can highly affect the CryoSat-2 range measurements if it is assumed that the main scattering horizon is given by the snow-ice interface. Our analysis of Arctic ice mass-balance buoy records and coincident CryoSat-2 data between 2012 and 2014 adds observational evidence to these findings. Linear trends of snow and ice freeboard measurements from buoys and nearby CryoSat-2 freeboard retrievals are calculated during accumulation events. We find a positive correlation between buoy snow-freeboard and CryoSat-2 freeboard estimates, revealing that early snow accumulation might have caused a bias in CryoSat-2 sea-ice thickness in autumn 2013.
Look at the December graph where you can clearly see a similar jump (relative to PIOMAS) in 2013.
Moreover, backscatter from both interfaces superimpose each other and cause broadened radar returns, which is largest for snow depths >20 cm (Kwok, 2014). As a result, freeboard estimates can be biased high with the presence of thick snow layers.
For wet snow at the beginning and the end of the melting season, the dielectric properties of the snow layer might even limit the physical penetration of radar waves.
We hypothesize that the snow cover significantly affects the CS-2 freeboard retrieval by snow backscatter which would affect also sea-ice thickness and volume, independently of the range retrieval method.
We find high differences of up to 45 cm (30 cm) for the 40 % threshold retrieval and up to 30 cm (20 cm) for the 80 % threshold retrieval from the comparison between November 2013 and 2012 (November 2013 and March 2013) north of Canada.
It is still difficult to quantify the snow-scatter induced bias without knowledge of the regional distribution and temporal evolution of snow depth and snow stratigraphy. Snow, accumulated early, may undergo a partial melting and subsequent freezing as well as wind compaction. This leads to a very heterogeneous snow density distribution, while for the propagation of the Ku-band signal it is widely assumed that the snow density is homogeneous. In this way formed layers may affect the location of the main reflecting horizon.
We conclude that snowfall can have a significant impact on CryoSat-2 range measurements and therefore on ice freeboard, thickness and volume. The assumption that the CryoSat-2 main scattering horizon is given by the snow-ice interface cannot be justified in regions with a thick snow layer.
My preliminary conclusions:
- PIOMAS has it more right than CryoSat-2, although probably underestimating thickness slightly.
- The discrepancy is caused by either a thick snow layer, or short melt events due to heat incursions changing the snow stratigraphy, or a combination of both.
- Bad news for the ice, as 1) snow insulates, causing the ice to thicken less, 2) snow melts more easily than ice, which can set off feedback processes earlier (melt ponding, etc), especially if it has already melted for short periods during the winter-spring transition (Stroeve published a paper about this last year).
- Is there any observational data (buoys, atmospheric data) that enables us to quantify this, or...
- Give an idea of which areas are affected most? Unfortunately PIOMAS seems to be experiencing a problem (that bulge of thick ice hovering over Fram Strait), so I don't know how useful a regional breakdown would be.
Either way, this is pretty big, IMO, as it tells us something about snow depth on the sea ice which may have consequences for the state in which the ice pack enters the melting season.
I might write about it on the blog, or mention it in the next PIOMAS update, but I thought I'd share it here first. Maybe together we can squeeze more out of this.