In any study which tries to scale grounded field data to a regional perspective, using (as-much-as-possible) unbiased methods is necessary so that data points are comparable across regions. A good recent paper experimentally looked at the best air temperature sensor design:
Check out the recent paper by Terando et al. It’s open access, available here: http://onlinelibrary.wiley.com/doi/10.1002/ece3.3499/epdf
They experimentally compared 11 different sensor brand and shield combinations to weather station data, and found that quite a few methods had a positive bias. It’s not something we haven’t known is a danger (positive bias due to inadequate/improperly designed shielding), but this is a nice way to standardize methods with some experimental backing behind it.