The picture below compares my graph of the Berkeley Earth data with their graph. Since my graph goes farther back, I have marked three common points. From this you can see, other than a minor Y axis scaling difference and a different ending date, the two graphs are identical. I have made no alterations to the data of any kind. At the end of this piece I will provide precise information so any who wish to do so can verify my work.
Now, let's begin deconstructing the data from which both charts above were made.
When organizations such as Berkeley Earth presents a graph of data to the public they are asking you to believe their data. The proof, as they say, is in the data. Yet, they are in fact lying to you by omission. They are omitting important details concerning not only where their data comes from but when. The illusion they are creating is you are looking at a continuous, consistent record of temperature over a very long span of time. That is absolutely not the case. From 1950 to 2004 a total of 2789 temperature reporting stations were opened while an additional 2349 were shuttered. That is a very important detail.
In order to properly assess data trends over a long period of time it is absolutely essential to establish consistent, stable measuring points without bias over the entire length of the time period being studied. If you move or change a measuring point or if you add or subtract measuring points you introduce additional variability into the data not created by the system being studied.
Take for example measuring the annual fluctuations of the water level in a large reservoir. To do that you would establish one or more stable locations from which to measure. Such as driving a pylon down into the lake bed until a firm under structure is reached and attaching a measuring device to the stable platform. What you would not do is drive around in a boat every year taking a number of depth readings at random locations. That would give you a lot of very useless data.
My conclusion, which is well warranted, is the complete Berkeley Data set is worthless for trend analysis and forecasting. As I stated above, in order to be useful in that manner a data set must consist of continuous, stable measuring points. The Berkeley data set, taken as a whole, is anything but that. There is, however, data with in the complete set which is valuable.
In 1951 there were 576 stations which came on line. Of those stations 211 maintained continuous, reasonably consistent measurements for 50 years. The graph below is the result of those 50 years of measurements from those stations.
This is a drastically different result from that of the complete data set. Instead of a pronounced warming trend as in the complete data set, we see a slight cooling trend of less than .25° C. Essentially this is no change as the year to year variability is many times greater than the total projected change.
The conclusion from this study is no evidence of any warming. None.
You can replicate my data for this time period or for any time period within the data set. As long as you follow the guidelines for a proper time based study of temperature trends, namely you can only use data from stations reporting for the entire time period being studied, you will get similar results. No warming, slight warming, or slight cooling. What you will not find is any warming trend even remotely resembling the chart put forth by Berkeley Earth. That is fact.
The methodology I have followed using their data is the correct one. The results I have derived are therefore valid. The methodology Berkeley Earth uses is, as I discussed above, very much incorrect and therefore their results are not valid.
Update: I follow a number of climate blogs and engage in conversation with a number of people. One of those blogs is the Deplorable Climate Science Blog by Steve Goddard. A reader on there passed this youtube clip on to me. This clip shows how the number and locations of climate stations have changed over time. During the period where the "experts" say warming has increased dangerously what you see is exactly what I stated. Large numbers of stations in colder climates disappear while stations in hotter parts of the world appear.
Additional information concerning data and source
My data source. I used the raw, unadjusted data from land based measurements: http://berkeleyearth.org/data/
Below is the break down of the 211 stations used for my study based upon the long term annual average temperature for each station.
Below are the stations numbers I used for my study.