If you look at the more detailed source materials of various surface temperature reconstructions, you will see error bars, and that they are bigger as you go ba

If you look at the more detailed source materials of various surface temperature reconstructions, you will see error bars, and that they are bigger as you go back in time, for instance, NASA's data, 2nd chart on this page (click on "Global Annual Mean Surface Air Temperature Change"), the grey error is the error range, and is thicker as you go back in time: https://data.giss.nasa.gov/gistemp/graphs_v4/ https://data.giss.nasa.gov/gistemp/graphs_v4/ it doesn't take a very large number of sensing stations to get a reasonable global mean, surprisingly. You get pretty much the same result as the serious efforts if you just pick ~300 random weather stations.