A discussion on WUWT prompted me to take another look at the GHCNv3 Quality Adjusted Data I had downloaded on earlier this month.
I plotted the average temperature for the entire U.S. from 1881 to July 2012. Earlier years have too much noise due to the small number of observations.
I did this because I hate 1) looking at charts produced by the GISS people at NASA where the vertical axis is compressed; and 2) I hate looking at so called temperature anomalies.
In the plot below, you'll notice a tick up in the temperature and a leap down in the observation count. That's because that data point corresponds to a partial year. The average would go down once more observations covering the rest of the year are added.
Note that some stations have multiple series associated with them. To avoid over-representing those stations, I first averaged those multiple series into a single one for each station.
Note the declining number of observations in the recent decades.
And here's a plot of the average July temperature for the entire U.S. This makes it easier to make year-to-year comparisons as some years may have more observations from the colder months and vice versa. It seems we've had a few warm Julys recently. But, I want to attract your attention to the middle part of the graph, with the most observations. Notice how the period with the most observations has the least variability? Food for thought.