There’s a new paper which quantifies the effects of the Urban Heat Island and opines on the effect of urban areas on multi-decadal surface temperature trends. It was just published yesterday in Geophysical Research Letters and is quite interesting and confirms much of what we’ve learned via the surfacestations.org project, and that is that urban areas tend to have higher trends, and the UHI effect manifests itself mostly in the overnight Tmin.
However, the authors seem to try to steer the conclusions away from urban effects being the cause, and try to use regional analysis to prove Parker (Mr. fantasy “cool parks in cities hold the thermometers”) and NCDC’s Dr. Thomas Petersen correct rather than noting that moist enthalpy related to lawn watering irrigation can have an effect on temperature as well as precipitation. More on that below. It has been noted previously on WUWT that cities can provide the elements for increased precipitation formation:
The March-April edition of WeatherWise magazine has an interesting article in it regarding UHI (Urban Heat Island) effects of enhancing thunderstorm formation in the downwind heat plume. It Stems from this paper (PDF) published in the Bulletin of the American Meteorological Society. I saw a similar study presented in August 2007 when I attended Dr. Roger Pielke’s land use conference presented by Dr. William Cotton on the enhancements modeled in St. Louis, MO. Read that paper here
In a nutshell, humans use a lot of energy and a lot of water, the two essential ingredients for convective precipitation, and both get dissipated into the atmosphere locally near their use. Cities provide a source reservoir for both elements, but even as we move to rural landscape, we find that there’s still a lot of water use related to irrigation.
The new paper is:
Mishra, V., and D. P. Lettenmaier (2011), Climatic trends in major U.S. urban areas, 1950–2009, Geophys. Res. Lett., 38, L16401, doi:10.1029/ 2011GL048255
Some highlights (emphasis mine).
In the abstract:
We evaluate changes in climatic indices for the 100 largest U.S. urban areas and paired surrounding non‐urban areas. During the period 1950–2009, we find that there were statistically significant changes in as many as half of the urban areas in temperature‐related indices, such as heating and cooling degree‐days and number of warm and cool nights, almost all of which are reflective of a general warming. Similarly, statistically significant changes (mostly increases) in indices related to extreme precipitation, such as daily maximum intensities and number of days with heavy precipitation, were detected in as many of 30% of the urban areas. A paired analysis of urban and surrounding non‐urban areas suggests that most temperature‐related trends are attributable to regional climate change, rather than to local effects of urbanization, although the picture is more mixed for precipitation.
Among the conclusions in the paper we have:
Consistent with previous studies [Easterling et al., 2000; Kalnay and Cai, 2003], trends related to temperature minima in the urban areas are generally stronger than those related to temperature maxima.
For both minimum daily temperature based climate indices and precipitation‐related trends, changes in urban and non‐urban areas are generally consistent; suggesting that the trends are dominantly a response to climate [Parker, 2004; Peterson, 2003], rather than local land cover changes during the period of analysis. However, there is somewhat less consistency in urban vs. non‐urban trends in climate indices related to daily maximum temperature, which suggests that land cover change may be at least partially responsible for those trends.
I do like this from their methodology, it is the right way to do it:
Gridding of meteorological data for the urban and non‐urban buffer regions was performed using the stations that were uniquely present only in either urban or non‐urban regions. This approach insured that data for urban and non‐urban regions were gridded with unique sets of meteorological stations to avoid any contamination that may occur due to common stations in urban and non‐urban regions.
But the statement just above it has a red flag for me:
In the interest of preserving local influences of urbanization on temperature and precipitation, we used separate NCDC‐Coop and HCN stations for urban and non‐urban areas.
What concerns me is that they didn’t make it clear what data set they used. As we know, USHCN data is heavily processed, and uses nearby COOP stations as part of FILNET to fill in missing data on the B91 reports, like this one in Marysville that is missing a lot of data:
B91 form provided by the Marysville observer (PDF format).
NCDC’s FILNET process will take data from other nearby stations and use that to interpolate the missing values, essentially mixing data from stations.
So my point is, that due to the way NCDC processes data, mixing and infilling to make every record “complete” even though Mishra and Lettenmaier went to great effort to keep rural and urban stations separate in gridding, the data they used may have been urban and rural pre-mixed anyway and the analysis may have been doomed by data pollution from the start. Until we know more about what data they used, I can’t say for sure if this is a problem or not. They make no mention of this issue in the paper that I can find, so I’m assuming they are unaware of it.
One other thing they appear not to have considered is the effect of increased humidity on Tmin, i.e moist enthalpy. Lawn watering and irrigation are common to human habitation, no matter whether you are rural or urban. And as we’ve seen, most of the COOP network stations are near dwellings, and by default near either the nice green lawn, gardens, or agricultural plots even in the far rural areas.
Yesterday, in my summary of the Susanville USHCN station, I illustrated the issue in lush detail.
Note the lush lawn. The MMTS temperature sensor is near the cattails at the right end of the ladder in this image.
The view from the air shows that there is a lot of moisture near the USHCN station.
It is a big patch of green and parking lot in the middle of an arid landscape. Does increased nighttime humidity due to watering and evapotranspiration play a role? Quite possibly.
After I pointed out the differences in USHCN data processing between 2007 and 2011 graphs as they appeared on GISS, Zeke Hausfather helpfully pointed out what NCDC has done to the data:
Zeke Hausfather says:
As far as urbanity designations for that station go for its listed lat/lon, its urban via GRUMP, impermeable surfaces, and 1930-2000 population growth, but rural via nightlights (only 19 brightness).
When run through the pairwise homogenization process, NCDC significantly reduces the 1960-present minimum trend from 0.24 C per decade to 0.03 C per decade. The max trend is mostly unchanged, going from -0.10 C per decade to -0.11 C per decade.
So, if NCDC was already tinkering with the station data by adjusting trends, is the conclusion that “…that most temperature‐related trends are attributable to regional climate change, rather than to local effects of urbanization” valid? Or is it simply an artifact of the mixing mishmash of COOP data and microsite effects like increased humidity due to irrigation that have not been considered in this paper? The authors suggest land cover change might be responsible for precipitation effects, but dismiss the issue for temperature without providing any basis for the dismissal, citing the similarity of temperature trends for rural and urban. Again we go back to the NCDC mixing of temperature data issue, which wasn’t specifically addressed.
As I understand it, NCDC does not infill missing precipitation data, due to the spotty nature of precipitation. As we know, thunderstorms often leave narrow swaths of rain, and interpolation of missing precip data would be wholly uncertain for nearby stations. So, the data mixing issue isn’t present in precip data like it is in temperature data.
The biggest downside of the COOP network is that it records mostly temperature and precipitation, agricultrual COOP stations with humidity and evapotranspiration data are few and far between, so answering the question over the long term is difficult.
The full paper Mishra, V., and D. P. Lettenmaier (2011) is here
h/t to Dr. Leif Svalgaard