How odd is the Palmyra Atoll Coral Proxy?

In the last post I noted that there was something odd about the Palmyra proxy used in the Gergis paper, particularly in the late 20^th century. This is at 5°52′ N, 162°06′ W.

There are four other coral proxies in the Mid-Pacific area. There are two proxy studies from
Rarotonga in the Cook Islands at 21° 14′ 0″ S, 159° 47′ 0″ W and two from the Fiji. For all five proxies I calculated a nine year centred moving average.

Palmyra shows a late 20^th century warming trend more than twice that of the other series. Unless there is a locally recorded temperature anomaly on the atoll, then this is clearly wrong. If there is a local temperature spike, then it one should question why it is included in a reconstruction for which it is over 2000km outside the boundary. Either way it should be deleted from the study.

So how reliable are coral proxies. Here we have two pairs. If they are a good proxy for temperature, then they should be a good proxy for each other. On Fiji, they studies be less than 150km apart and on Rarotonga less than 10km apart, meaning they should be near identical. So I have plotted the differences between the moving averages.

It is not a statistically sound method, but indicative of the real issues with the proxy data sets. It also seems that the further back, the greater the consistency. The Palmyra study has four sections, the oldest of which starts in the 12^th Century. Although Gergis claims to have done a series of tests for robustness, there is no correlation test over the known temperature record. Given that a central conclusion is:-

The average reconstructed temperature anomaly in Australasia during A.D. 1238-1267, the warmest 30-year pre-instrumental period, is 0.09°C (±0.19°C) below 1961-1990 levels.

Given that there is some question of the selection of the ice core studies at Vostok in preference for the closer and more robust studies at Ice Dome, then central conclusion of the study is not credible on such a small number of proxies.