Purpose - To facilitate use of a robust, independent and global common benchmarking and assessment system for temperature data-product creation methodologies to aid product intercomparison and uncertainty quantification:
http://www.surfacetemperatures.org/benchmarking-and-assessment-working-group
Posting and comments are open for constructive advice and ideas.
Thursday, 10 February 2011
Review paper references
Hi All, all the comments so far have given me a lot to think about which is great. The references are really useful. I'm hoping to get chance to start the review paper this weekend so if anyone can think of anymore useful references for me to browse please can you list them here.
5 comments:
Kate, what kind of review are you thinking of. Could you be a bit more specific?
A good resource is probably the HOME bibliography. With the help of all members of the Cost Action HOME, Enric Aguilar has compiled a bibliography with (almost) all papers on homogenisation. They can be found on.
http://www.homogenisation.org/links.php
If there are any papers still missing, please inform us.
That's great, thanks Victor. In short I would like to write a review of known inhomogeneities - the causes and the effects on the temperature record of that site. This will be a reference for the creation of our benchmarks as we should be aiming to include all the known inhomogeneities in as realistic a way as possible. I'm writing a very brief description of this and posting it on the surfacetemperatures.org site.
That is good news. Such a review is very much missing and would be a great resource, especially for people new to the topic.
May be short word of caution. Only for a fraction of the homogeneities there is meta data and the meta data will often not give you sufficient information to be able to model the physical causes of the inhomogeneities. On the other hand, the statistical homogenisation methods will only give you information on the detected inhomogeneities. The smaller not-detected inhomogeneities are important for homogenisation (they lead to artificial decadal variability) and should thus also be present in the benchmark.
This makes it valuable to also include real inhomogeneous data in the benchmark dataset. By comparing the properties of the *detected* inhomogeneities in this real data and in the artificial data, you can study how realistic the artificial data was.
I have to revoke my previous comment, that a review on the causes and typical sizes of inhomogeneities would be new and interesting. I just came across the 2010 review by Blair Trewin, which has its focus exactly on these matters.
What would still be important would be a more quantitative study of the statistical properties of the inhomogeneities. Furthermore, for this global benchmark we will also need to know whether there are regional differences in the inhomogeneities.
In the Cost Action HOME we have assumed that the inhomogeneities in temperature can be modelled by a normal distribution with a width of 0.8°C. However, the histogram of the detected inhomogeneities seems to show more large breaks as expected from a normal distribution. One may want to study whether the distribution of break sizes has fat tails. These large breaks could also be due to homogenisation algorithms that detect two breaks in the same direction as one large break.
Another issue is the temporal behaviour of breaks. In HOME we have modelled the breaks as independent events (Poisson process). However, it may we be that breaks are more clustered or more uniform in time.
Furthermore, we have modelled breaks as deviations from the baseline values, i.e. as random noise. Another way to model breaks would be relative to the previous values, i.e. as a random walk. The truth may well be somewhere in the middle.
Also the direction of the break may not be random. Peter Domonkos found that raw data often has platform-like inhomogeneities, i.e. after a break there is often a break in the other direction after a short period, more often as expected in Poissonian noise.
As far as I know these kinds of features has not been studied much yet. Thus this is more a suggestion for a new study as for a review paper.
Post a Comment