At sea the surface temperatures have been measured using canvas, wooden and metal buckets sampling water at different depths, hauling it up the side of ships of different heights and exposed to different winds while doing so, including that from ships travelling at different speeds. Samples were taken in conditions ranging from snowstorms to heat waves, in different currents, with thermometers read by differing untrained crew often using lantern light.

After the war, sea surface temperatures were measured by hull mounted devices, not only varying in depth according to vessel but according to load on that vessel.

More recently they have been measured by different types of buoys, recording temperatures at potentially different depths and subjected to different conversion methods to convert raw data to temperature.

But Follard & Salinger still got it right – to tenths of a degree!

I think this is a key point. The method used by NIWA is almost 100% subjective, so they would have no problem using whatever variables they might need to achieve the desired 0.91°C/century trend. But how much certainty attaches to that contrivance? Surely, statistical uncertainty must increase with the subjectivity of the method – and, if so, the uncertainty in this case must be reaching towards infinity.

This problem was foreseen at the beginning of the Review, and Minister Mapp assured Parliament last February that confidence intervals would be measured and published. This work is still in train.

On another blog, a NIWA supporter argued that certainty was increased by measuring the same thing many times and then taking an average. But others pointed out that every datapoint in a temperature series is measuring a different thing, and that averaging the numerous non-linear results added nothing to accuracy.

The incredible aspect is that NIWA (and BOM?) discarded 51 out of the 52 adjustments used for the 7SS graph.

But perhaps it is even more incredible that this raft of new adjustments came up with an identical average trend for 1909-2009 as did the discarded adjustments. Talk about “two wrongs don’t make a right”. Here, 51 wrong moves in the 7SS resulted in the right trend!!

The New Zealand data series, like that of many stations around the world, is curiously defended not individually but as a merged group, when there is nothing specific to point to. Somehow the statistical approach is given more credence than the actual data.

I’ve been struggling with the certainty issues and appearance of errors. Somehow error is perceived as lower with the group – true when errors are random and equally up and down. We’re “bracketing” the real answer. Yet NIWA adjustments are consistently in one direction. No bracketing. When adjustments are consistently in one direction, the possible error is in the style of adjustment, not in a statistical grouping. Each one would have the same potential for error, and, if anything, the error would be compounded as randomness has been removed in principle.

What now do you say about the new graph/information, if it shows the same trend?

The same as we said about the previous graph: be very sure that those adjustments are well justified, because the adjustments ARE the warming. NIWA never justified the old adjustments, but created the new series instead.

Unfortunately, it sounds as if you et al can’t disagree with the new conclusion (same as the old one). True?

Well, yes, we cannot disagree. But nor can we agree, for we must study the report first. With the old report, we never officially agreed or disagreed, because the science of it was inadequate. NIWA made adjustments, and we asked what they were and why they were made, but we could hardly judge them before we got an answer.

Now, we start again. You don’t seem to realise that NIWA have discarded the old series! Why did they do that, do you think? It certainly wasn’t because we disagreed with their conclusion of warming!

What now do you say about the new graph/information, if it shows the same trend? Could we have graph comparisons?

Unfortunately, it sounds as if you et al can’t disagree with the new conclusion (same as the old one). True?