Dr Roy W. Spencer, did a fine analysis of the bias (positive effect) of the measured surface temperature caused by population density of cities where those thermometers are placed.
Here's the complete study
My conclusions are:
- The influence on temperature of cities start very fast even with low densities cities.
- The Bias get quite large, > 2 deg C for large cities
- Using those temperature records from when the density was low in the past to high today for those records, will show a temperature increase, caused by population increase.