I’ve previously played with 3-d modelling tools to take advantage of the lo-fi, lego-like blocks that make up the HadCRUT4 data set, but there are lots of lovely hi-res datasets out there to work with.

Since October last year there’s been a long run of record low ice extents* in the Arctic. Monthly Arctic ice extents were record low in October, November and January, which suggests there was some kind of relapse during December. The daily extents tell a somewhat different story. In the 100 days to the end of January, only 10 didn’t set a new daily record for Arctic ice extent.


That’s quite a long run of records or near records. December wasn’t a monthly record because of the way that the monthly extents are calculated – it’s not an average of daily extents, but the extent of the monthly average sea ice field which is quite a different beast. NSIDC have, of course, a brilliant write up of the January situation (and in their archive, every month for quite a while).

A time series is all very well, but the sea ice is a beautifully mobile wobbly thing. The wind pushes it around and stirs it up and it reacts in a wonderfully organic manner, retreating ahead of a southerly wind like the sensitive outer fringes of a snail when you poke it with a stick. As you mull that incongruous image, lets move on.

One thing in the NSIDC write-up caught my eye which was that sea-surface temperatures round Svalbard were keeping the ice off it. If there are two things that go together well in climate data its sea ice and SST, so I made a joint animation of the two which runs from 1 September 2016 to 17 February 2017.

The data come from the OSTIA data set which is presented on a 0.05° lat-lon grid. It has SST and sea-ice concentration for each of the grid cells. I set a threshold of 10% sea-ice concentration to define an ice edge. The landscape is the NASA Blue Marble and the topography over the land is from the same source. The four data sources are wrapped on a globe and rendered out into video. For the sea-ice and SST data, I produce grayscale png files and use Blender to control the colour scales. These can be changed and re-rendered very quickly so its easy to play around with different choices to get something that looks OK.

This next video shows the same data but with the SSTs expressed as anomalies from a modern climatology which runs from the mid 80s to 2007 from the MyOcean OSTIA reanalysis. I’ve switched the sea ice to greyscale to indicate varying ice concentrations within the ice pack. In the video you can see the high SST anomalies throughout the period, around the ice edge (red bits) and later on, how the ice can quickly grow into areas where the SSTs are close to normal (at this time of year, those that are usually ice covered) and retreat from them again depending on the winds.

And finally, because we can (and because the swirling eddies are so pretty), the same for the Antarctic** which also saw record low sea ice extents in the latter months of 2016.

* Ice extent is a beautifully weird measure of ice area. It’s calculated by dividing the earth up into grid cells, identifying gridcells in which the concentration of sea ice is >15% and then adding together the areas of all the gridcells so identified. The weird thing is that the extent then depends on the resolution of your data set. If you have a finer or coarser mesh of grid cells, then the extent will change. This can make it quite hard to reconcile estimates of sea-ice extent (not that it isn’t quite hard already).

** I made a version that rotated as well (spirals being the thing) but it made me feel sick.