Sunday, January 17, 2016

An old WUWT SST blooper back to life.

In recent days, someone who comments as Kristian, but blogs as okulaer, has been active. He commented here to draw attention to his extensive recent posts, and has been plastering the material at WUWT. The ratio of sense to graphs is small. But there is one I want to treat here, since it relates to a 2010/2 blooper by Bob Tisdale at WUWT, and on his own blog .

Kristian posted a WUWT comment here, saying that GISS land mask had failed, and showing (from here) this plot:

It shows trends (despite y-axis label) of land-masked GISS vs OIv2 SST, which GISS used at the time. The divergence near the poles is attributed to GISS "deleting data". Now I should say that this plot does not seem to have been cross-posted at WUWT. But it expresses Bob's complaint at WUWT about deleting data, and is referred to most months at WUWT (see pingback list) in Bob's updates, most recently Nov 2015 here. The boiler-plate complaint is
Unlike the UK Met Office and NCEI products, GISS masks sea surface temperature data at the poles, anywhere seasonal sea ice has existed, and they extend land surface temperature data out over the oceans in those locations, regardless of whether or not sea surface temperature observations for the polar oceans are available that month. Refer to the discussions here and here.

So what is happening? Note that the plot goes from 77.5S to 82.5N. You might reflect on what SST at the frozen ends might mean. In the 2010 WUWT post, titled "GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data", I tried to point it out here and here. The SST products OI V2 and ERSST, used by GISS then and now, adopted the somewhat annoying custom of entering the SST under sea ice as -1.8°C. They did this right up to the North Pole. But the N Pole does not have a climate at a steady -1.8°C. GISS treats this -1.8 as NA data and uses alternative, land-based measure. It's true that the extrapolation required can be over long distances. But there is a basis for it - using -1.8 for climate has none, and is clearly wrong.

So is GISS "deleting data"? Of course not. No-one actually measured -1.8°C there. It is the standard freezing point of sea water. I guess that is data in a way, but it isn't SST data measured for the Arctic Sea.

It's worth checking the origin of Bob's plots. He partly explains, saying they came from KNMI. And yes, if you go to this page, and look down to SST and "1982-now: 1/4° NOAA OI v2 SST", you'll see an interactive display inviting you to enter latitude bands. If you ask for, say, 85N to 90N, it will, by default, return an anomaly time series of band averages. Surprisingly, this isn't quite zero - apparently not all the entries are exactly -1.8. But it is close, and has no trend.

If you click on the "authoritative data", it takes you to here, with this animated GIF:

The dark blue is -1.8°C, and you'll see that both high Arctic and deep Antarctic are -1.8°C year round. Now KNMI provides a mechanical process here. It averages whatever data set you request. If it is all -1.8, that's the average it returns. St that is where the top plot comes from. Bob originally requested and plotted the latitude bands. But of course, where it is returning -1.8, there is no trend. There can't be. So the OI v2 trend goes to zero at the ends, which is nonsense. On this basis, the arctic areas just can't warm. But they do, as GISS shows.


  1. Nice clarification. So, that Bob Tisdale's SST graph is comparing sea ice interpolated from land (GISS) to a systematic NA or -1,8°C (Reyn). You bet he finds a warming bias.
    Land mask is real though when you check "land only" on GISS maps you get area with sea ice covered. But that's not a secret, one may better look at if interpolation from land is well done, which I absolutely don't know. I guess GISS is using sea ice extension data.
    UAH data shows also warming over the Arctic, about +0,85°C since 1979 vs 0,4°C for the globe. That's twice the global trend, a bit less maybe than GISS. But RSS stops at 82°C, pointing the difficulty to assess temperatures above with satellites.
    So if we get back to SST data, maybe if GISS consider -1,8°C as NA is simply because they're not trying to assess SST for itself but as a proxy for SAT. And sea ice can't be a proxy for SAT. But if you want to check SST only, it's probably ok to consider sea ice as -1,8°C.

  2. For in situ observations, the IABP gridded data go to 2004, which is enough to show the start of the recent rapid Arctic warming and falsify this claim. The grids are pretty awkward - at the time I wrote my own code to sort them out, but CDO or other standard tools might be able to do it.

    In fact I should redo that more carefully and send the data to KNMI. I'll add it to the to-do list.

    Lars Kaleshenke at Hamburg runs a masters workshop in which the students use the IABP data to evaluate the Arctic reconstruction from our original paper, the notes are here:

  3. Kevin, you are indeed answering very well to a lot of questions in your study about the coverage bias explaining the warming pause (, showing a warming trend for arctic temperatures in most data (IABP, satellites, interpolation from land, reanalysis).
    I was a bit curious of something : you find more warming than GISS with you hybrid method during the hiatus period. I doubt it's not because of the arctic trend because GISS shows a strong warming trend (maybe more than Hadcrut +UAH) over the north pole at the beginning of the 21st century. It rather seems as if the main discrepancy is at 60S where you can find sea ice around Antarctica.
    For monthly data in 2015, Antarctica is the main source of divergence between datasets like JMA and GISS. I don't know but maybe satellites and interpolation don't work so well for southern ocean when there is sea ice. If you look at GFS, NCEP and GISS, it's really different there.

  4. Yes, the big difference between us and say Berkeley is that we're hotter over Antarctica around 2010. I wrote about the Antarctic uncertainty here:
    although I hadn't made the connection with sea ice edge effects.

    Our more recent paper on comparison to models identified the blending of air and sea temperatures as a source of bias in the presence of a changing ice edge. It could be that that is what is going on here. Unfortunately the ice edge bias is very hard to pin down, because it depends on how the SST record providers handle cells which were ice covered during the baseline period. This varies from record to record and is not well documented, and so we don't actually have a very clear idea how the ice edge bias affects the observational records. One of the aims of our paper was to highlight the issue so that the record providers can revisit it. I know NOAA at least are looking into it.

  5. I looked at NOAA and NASA global temperature data thanks to Nick's graphs and it seems both datasets show good agreement during northern hemisphere winter and greatest divergence during northern hemisphere summer. So the biggest divergence is when you have a lot of sea ice around Antarctica and lower extension in the Arctic. I don't know if that points the source of the major discrepancies towards the sea ice edge in the southern ocean. But intuitively, it seems the sea ice boundary in the southern hemisphere is indeed a source of uncertainty.
    So I believe it's good news hearing you pointed that problem and that SST providers will look deeper into this.

  6. The blog has given the excellent insights about the sea surface temperature rise & fall in temperature according to circumstances.Visit the free Polar Oceans Survey & get more information of arctic.