On data losses -vs- geography in Cowtan and Way 2013: data dropout may account for most of the recent observed differences

At Climate Audit, Roman M. has a very interesting analysis that shows the surface grid cell losses from HadCRUT4 in C&W. It hones in on the issue of why the temperature differences from 2005 are much more significant than the earlier part of the record. Data loss seems to be recently seasonally large, especially in the Antarctic, something C&W seems not to have considered important enough to dedicate more than a passing note about. Roman’s analysis was made possible by the fact that C&W provided data and code here, something not so common these days, so some props are in order.

He writes: 

The cells lost as a percentage of all possible cell values (missing or not) is 2.32%, approximately the value given in the paper. However, the percentage of cells containing temperatures (representing the real reduction of information) lost through anomalisation is 3.34%. The monthly distribution of the losses can be seen in the following graph. The change in the pattern shortly before 2005 could be related to Steve’s observations in his earlier post on CA.

[CA Figure 2]

Which grid cells are losing the information?

[CA Figure 3]

It is quite obvious that the losses occurred in those areas which were already short of temperature information thereby exacerbating the problems inherent in trying to create a reliable reconstruction of the gridded temperatures in the polar regions. I would suggest that the paper’s observation that “coverage is very slightly reduced” understates the impact of these reductions by ignoring the geography of where they occur.

Based on the two graphs shown above, and in particular the plot of frequency of grid cell loss, I left some comments at CA which I think give insight into the reason for the loss and I repeat them here for wider consideration.

===============================================================

From CA Figure 2 above, it appears that the breakpoint observed in this graph Steve produced a few days ago,

Figure 2. Delta between CW Hybrid (basis 1961-1990) and HadCRUT4.

Owes its 2005 up shift primarily to an increased loss of cells starting about that time.

===============================================================

In CA Figure 3 above, there are two red bands of missing cells. The Arctic band is weaker than the Antarctic one, which is quite pronounced. Note also how the southern limit of the Antarctic band follows the shoreline of the continent almost perfectly. Roman asked if there was a seasonal component.

I think the geography difference and sea ice can explain much of this. Observe this graph from NSIDC:

Above: Sea ice climatologies: Arctic and Antarctic sea ice concentration climatology from 1981-2010, at the approximate seasonal maximum and minimum levels based on passive microwave satellite data. Image provided by National Snow and Ice Data Center, University of Colorado, Boulder.

Note the difference in seasonality of sea ice extent between poles. It is far more dramatic in Antarctica.

We know that for HadCRUT is both land and ocean data.

For the Arctic, there are few manned bases not on land, and a lot of the data in the Arctic ocean is brought in from buoys. The manned Arctic bases on land are supplied almost exclusively by airplane, and they operate 24/7/365 thanks to being supplied. There are no significant fuel/distance limits to aircraft operations since they can hopscotch north. There are no surface readings taken during transport by aircraft. The sea ice extent doesn’t affect the operation of the bases, the aircraft, and mostly doesn’t affect the buoys, though some losses do occur. Sea ice does affect some data gathered by ships though. Only a few icebreakers operate in winter.

In the southern hemisphere, it is an entirely different story.

The sea ice when it reaches maximum in September, creates an nearly impenetrable barrier between the open sea and bases on the periphery of the Antarctic continent. Ships generally can’t operate in this zone during winter. And what little supply occurs (if it occurs) is by aircraft. The bases hunker down for the winter and continue to operate, as do the AWS stations on the continent.

So IMHO, that bright red ring in Roman’s figure 3 above showing frequency of grid cell loss circumventing Antarctica represents the sea ice barrier and loss during those months when ships cannot operate.

As we know, sea ice extent in the Arctic is decreasing, while increasing in the Antarctic. See this graph from NSIC:

Above Arctic and Antarctic Sea Ice Extent Anomalies, 1979-2012: Arctic sea ice extent underwent a strong decline from 1979 to 2012, but Antarctic sea ice underwent a slight increase, although some regions of the Antarctic experienced strong declining trends in sea ice extent. Thick lines indicate 12-month running means, and thin lines indicate monthly anomalies. Image provided by National Snow and Ice Data Center, University of Colorado, Boulder.

So the Arctic has become more accessible to ships on the periphery of seasonal sea ice boundary, while the Antarctic has become less accessible to ships during the last decade.

As for the seasonal component note the differences in the minimums vs the maximums:

Sea ice extent for the month when its at its minimum (i.e. the end of local summer) Source: James Hansen http://www.columbia.edu/~mhs119/UpdatedFigures/

This looks like a simple case of access to surface measurement having increasing bipolar difference, something the satellites don’t have a problem with beyond 80N and 80S.

===============================================================

As for the high missing grid cell frequencies in Africa and Asia, I’d wager they are late/missing reports. The plot of frequency for those regions center on many countries where political unrest in common. The one black cell in Russia is an oddity. Roman do you know what stations are under that cell?

[RomanM: No, that information is not contained in the paper’s data and it would be a tough task to chase it down on my computer, Sorry.]

===============================================================

As Bob Tisdale points out in theis WUWT post:

This breakpoint may be related to loss of seasonal measurements in Antarctica, note that in this Cryosphere Today graph, most of the ice extent anomalies post 2005 are positive.

This increase in Antarctic sea ice may translate into less seasonal ship traffic (and less measurements) in the Antarctic coastal zone during that time.