A frustrating aspect of observing climate change is the lack of uniformity within the major global data sets. Variations of instrumentation, infrastructure, methods of observation, automation, time of observation, standards, effects of human habitation and numerous other discrepancies pollute data sets within and among comparisons.
I was recently confronted with these issues doing a simple comparison of the RATPAC radiosonde data and the UAH LT temperature trends. Below is a sample of the RATPAC trends at 700 mb versus the UAH-LT (v6) for 1979 through 2014. The RATPAC values are calculated by an ordinary least squares linear regression of the monthly anomalies.
Some regions ( North America, Eastern Pacific, Australia ) indicate confidence inspiring correlation. Other sonde locations appear wildly divergent ( both positively and negatively ).
This is only a graphical representation, of course, but one I didn’t otherwise find by googling. It reminds us more of the constituency of the data, rather than the single global trend series which can cloud thinking.