I have previously discussed the temperature set that I have put together for my usage here. It has been updated a bit since then and people have been asking questions so I guess this would be a good time to provide an update and compare it to other temperature sets. It is important that people understand what the temperature set I use is and where it comes from. I have always been transparent about my usage of the blended temperature set.
There are two reasons I chose this particular path. The first is that it is scientifically proper to chose one temperature source and then stick to it. Each of the different sets has different benefits and problems. Often times it is possible to determine the conclusion of a persons argument based on the temperature set they chose to use. A person that is using the UAH set is probable to be a skeptic. An argument using the GISS temperature set is likely to be a warmist. I consider that simple bias as a good reason to go a different route and stick to it.
The second reason is that each set is a different method of interpreting the available information. In what I would consider a best engineering option I decided to merge 4 sets into a single set. By choosing the duration of records from the CRU and GHCN (initially HadCRU, but I switched later) with the high resolution of the UAH and RSS modern satellite sets I decided to make a single temperature set that I would use regardless of how well it fit with what I expected. I have stuck to that except when I was discussing the behavior of a particular set. This has been important as the UAH, RSS and GHCN have tweaked themselves over the past few months.
One thing I noticed early on is that the HadCRU was a reasonably good match for the blended set. The comparison for the two sets is shown below. I am aware that CRU is a component of the HadCRU, but once past 1880 it is not a significant portion of the results. Once past 1979 which is when the main concern of global warming began there is even less of a dependence as the satellite data came on-line.
The error for the two sets is 0.084 °C. That is comparable to any two other temperature sets. A look at the same RMSD between temperature sets gives the following results:
CRU vs. GHCN: 0.120°C
CRU vs HadCRU: 0.094°C
RSS vs UAH: 0.097°C
HadCRU vs GHCN 0.138°C
In all cases the error is less between the blended set that I use and between other sets. My method for creating this set may be unorthodox, but I have created a temperature set that incorporates 4 different sources of temperature data and have achieved a result that is reasonable and usable. It could be argued that I should use the HadCRU data, but then I would be back to using a single temperature set. The purpose was to incorporate as much data as possible into a single set. It would appear that because the different sets use different periods for setting the anomaly that in the end I have a more robust data set that is less likely to be influenced by any particular bias that a single set might have.
Here is the comparison of the two sets since 1970-present.
Events like the eruption of Mt. Pinatubo or the ENSO cycle show up more clearly in the blended set than they do in the HadCRU set. This is because of the increased sensitivity and more comprehensive coverage of the included satellite temperature data.
A comparison of the four components that I use to create the blended shows why different groups tend to prefer one to another.
The two most extreme differences by set are the UAH and CRU. The error between them is an astonishing 0.381°C. Since the UAH data covers the entire Earth from 85S to 85N it is the highest resolution temperature set available with resolution as good as the RSS set. Ignoring that data is foolish, but using it exclusively when it only goes back to 1979 is also folly. The blended temperature set is my method.
I also know that each data set is used accurately from each source. That is why I keep finding tweaks to the different sets. They show up even if nothing is announced. It also highlights differences when one set responds differently than the other sources. This happens more often than I expected which is another reason why I trust this set. Finally I would like to show it by itself. It shows that the Earth has warmed about 0.7 °C in the past 100 years.