If we want to have accurate historic temperature data for a period longer that 28 years (satellite measurements) or 50 years (weather balloon measurements), say for 120 years, weather stations data is the best we have, and it should be rectified to become reliable source of information.

If you are asking if US weather station data (2% of surface) is indeed representative sample of global surface temperature, rough comparison of 28 years of global lower troposphere satellite measurements with 28 years of surface stations measurements indicates that yes, two data sets are highly correlated.

Let’s wait to draw conclusions (on both sides), and just follow the data where-ever it may lead. Case in point is the fact that John V. makes… the geographical representation of the rural stations is currently grossly imbalanced (not John V’s fault). And as Clayton B. pointed out, the Southeast US is only represented by 1 station out of 17 rural stations. So we have an initial graph which almost completely leaves out a region that had zero average anomalies during the period from 1976 to 2005:

The situation may swing wildly in either direction before we have conclusive data.

Steve: I’ve uploaded a script to collate NOAA lower 48 station at http://data.climateaudit.org/scripts/station/noaa/collation.us.noaa.txt .

Moving on…

There is nothing that can be done about the small number of rural CRN12 stations. I could start including the CRN3 stations for more complete geographical coverage. I would appreciate it if somebody could provide me with a list of rural CRN3 stations. (I would generate the list myself but it will be better accepted if it comes from Steve McIntyre, Anthony Watts, Kristen Byrnes, or one of the other regulars here).

Once the rural CRN3 station list is available (and assuming it has a dozen or more stations), I can do a Monte Carlo analysis to estimate the uncertainty due to the low number of stations. I hope everyone can agree that adding uncertainty bounds to the CRN12R plots would be useful.

#74 Douglas Hoyt:
I am open to removing any stations that should be removed. The list of stations was provided by Kristen Byrnes. Her assistance in choosing the best stations was helpful.

…if it only represents 2% of earth-atmosphere, is completely non representative of the thermodynamic state. Therefore, it’s irrelevant busy work.

The problem is that the 2%-of-earth data is not only the most intensely studied, it apparently forms part of the basis for calibrating an incredible amount of other data.

I don’t have links at my fingertips, but I’ve seen statements alluding to even satellite measures being calibrated off the surface station measures.

The amount of interplay between “independent” data sources is truly astounding. Someone could add great value by documenting, even generally, the connection assumptions. It is far more than “simple” 1200km remote adjustments or Paris-USA, or Sierra BristleCone Pine influence.

This is beginning to remind me of a professional project, where a dozen separate systems are manually updated when (contact) data changes. Except that few people can remember all the places that need updating.

It may seem that the CA “team” is wasting time on silly basics… yet somebody needs to do it! Who else will discover and set out unequivocally solid baseline information? You’re in the right place if you want to distinguish good from uncertain data.

I thought you were auditing the so called scientific output, not having clashes of ego’s.

If we were calculating trends for each station, then averaging the trends, then it would seem valid to eliminate a station that is not one consistent dataset. But correct me if I’m wrong, but I don’t think we’re doing that.

A different way of looking at the data is to consider all data points as stand-alone snapshots of the temperature at the particular time and place. Therefore, station moves do not make the data invalid.

Aside: It would be even more useful to weight each station with the area it represents. In this case, we don’t remove urban measurements, we simply weight them as only representing the city. Since the lower 48 have many more stations per square mile than anywhere else, this would make a single station weight quite small. As it should be.

I see that Woodstock VA (a CRN2 station) is being used. I was the one that investigated this station. It has been a CRN2 station for only a little over a year. For about 10 years before that it was a CRN3 station (under trees and fairly close to a house). Before that it may have been a CRN2 station for a while. Early in the record it was at various locations in the town proper, so it may have been lower quality then.

Bottom line – this station has moved around a lot and was only CRN2 for part of the time. I would delete it from temperature reconstructions.