I my opinion, most of the climate research that gets published has little impact on the global warming debate. The field has become so specialized that seldom is there a finding that changes our understanding.

I think that the recent AGU poster by Anthony Watts et al. breaks this mold.

Anthony has spent years shedding light on the very real problem the thermometer network has for monitoring of temperature for climate change…most notably, local changes around the thermometer site associated with economic growth lead to spurious warming.

Back in 2010 I posted an analysis of global thermometer data which showed the very clear urban heat island (UHI) effect that exists. It averages at least 0.5 deg. C in going from completely rural to only 10 persons per km2 population density. Others have found the same thing over the years, and anyone who drives or cycles between rural and urban areas can easily notice it.

Without correcting for this, how can anyone believe long-term land-based warming trends? I’ll admit, it is not easy to correct for. UHI warming “looks like” global warming since it is more gradual than sudden, so break-point homogenization algorithms cannot correct for it.

Now, population density admittedly isn’t the best proxy for UHI…it was just an easily available one in my analysis. Even with no change in population, increasing prosperity inevitably leads to an increase in heat sources, both active and passive, around thermometer sites.

Anthony did a painstaking analysis of the USHCN thermometer network to find those thermometers which had good siting and which did not suffer from station moves and instrument changes. The result was that the NOAA analysis exaggerated the warming trend in recent decades by about 50%, compared to the trends computed from the best thermometer sites.

This is the kind of work NOAA should have done…but didn’t.

Instead, NOAA uses all of the thermometer data and employs statistical adjustments that many of us suspect forces the minority of good thermometer sites to match the majority of bad thermometer sites. In other words, it forces the rural data to match the urban data, rather than the other way around.

Watts et al. used only the best data….which I think is the best strategy. If one wants to use ALL of the thermometer data, then the bad data needs to be constrained so that it matches the good data.

As far as I know, this is not done by NOAA. And it’s a travesty that it hasn’t.