A Tale of Two Datasets: Part 2 – Oh Caroline!

Part 1 started to outline my thoughts about the 154 sets of graphs in one of the Hadley leaked files (idl_cruts3_2005_vs_2008b.pdf).  Faced with such a large array of data I decided to start somewhere where there was very little data, or so I thought.  I have spent a lot of the summer looking at GIStemp output, inspired in part by E.M. Smith‘s work. I recognised in the CRU data names of islands and archipelagoes in the Pacific Ocean; in the GISS and GHCN dataset these are sparce and spread across the ocean, with individual stations on islands often a 1000km or more apart. In particular, the CRU set below (Caroline Isl.) caught my eye.

In the GIStemp data series, Yap, Ponape, Kosrae and other islands in the Federated States of Micronesia appear as surface record stations.  They are small dots of land in a vast ocean and in GIStemp, together with the Hadley sea surface data they are combined in ‘zones’ that contribute to the familiar anomaly maps. So, first, where are two of the anomaly maps with (left) and without (right) the addition of the Hadley Sea Surface Temperature data. I used 1961-90 as the base period rather than the usual GISS 1951-80, and the maps show the anomaly (difference from the base period) for the years 1998-2008.

The Caroline Islands region is but a tiny blip. However a map of the region is useful. And what has been most interesting was to locate some of the stations on the map by lattitude and longitude. I have done this for both a conventional map and a portion of the anomaly map:

Interestingly, there seem to be many more stations in the CRU set than the analogous region for GIStemp. I thought these additional stations might be for sea surface; a look at the ‘Station-list-ncep’ file, which I have pulled into Excel (if anyone would like a copy, please email me – details under profile), shows ‘Environmental Buoy labelled as such elsewhere.  Here is the relevant section of the ‘Station-list-ncep’ file (below).
The stations I recognise from the GISS list that are relevant to the CRU data in this section are:
Falalop Island
Koror, Palau
Lelu, Kosrae
Ponape (Pohnpei)
Yap, Caroline
Guam (3 stations) and Saipan appear under country ‘MY’ and Enewetak and others elsewhere. I have been working on Pacific data so I have the Station data from the six above (downloaded in September 2009).  I have not yet aggregated the monthly/seasonal data, only the annual means, so the graph below is a bit crude, but as I have been sitting on this for 10 days, while other collaborations have taken prority, I wanted to get SOMETHING out.
What I have done is to calculate GISS anomaly data for the six stations above (with the 1951-80 base period), and overlay this on one of the Caroline graphs (SON). I have then averaged the data without doing any weighting by distance, although is done by GISS for zonal means.  This is only a quick comparison anyway.  It would take some time to do this for each season of each station (as I’ve been working in Excel), but it is possible; without the appropriate weighting, and the fact that we really are only guessing with the CRU data, I’m not sure of the value of doing this.

Looking at tge average data graph and comparing it with the SON one shown and the others which I have not shown, there are differences. One thing is clear, there is data available commencing in 1924, that is not used by CRU.  Here is an anomaly plot of the six stations used by GISS:

Note that trend in Palau – that looks very man-made. My next job added to the ‘to do’ list is to start looking at the GHCN data as this is the link between both the CRU and GISS datasets.  What adjustments are done there? And are there more of those stations in the GHCN list that are not on GISS’s ‘Stations actually used’ list?  I wish I had more answers – it is all questions at the moment and that is the frustration, but the good thing is that there are plenty of people (many of them much better able than I am) looking at all the raw and adjusted data.

This entry was posted in Uncategorized. Bookmark the permalink.