The 50-year NCEP atmospheric reanalysis (Kistler et al, 2001) is not sensitive to urbanization, land use and other surface effects, since land surface observations are not used in the reanalysis, and, instead, the model forecast is very sensitive to the potential vorticity and temperature of the atmosphere above the surface layer. Kistler et al (2001) noted that a comparison between the annual surface temperature in Shanghai from station observations and interpolated from reanalysis showed excellent agreement in the interannual variability. In addition, however, there was an increasingly large positive difference between station and reanalysis estimates (see their Fig. 13). They suggested that this trend in the difference could be due to urban effects.

We are exploring this idea with all US surface stations, starting with the 75 New York State stations with more than 40 years of observations and comparing them with the reanalysis estimates of surface temperature. We first subtract the 50-year monthly means from each surface temperature series, which eliminates biases unrelated to urbanization effects. We then compute the trends of the difference with respect to the 10-year mean of the 1950's.

The results are very encouraging. First of all, the two interannual series of monthly anomalies track each other extremely well. Second, station temperatures show an increasing warm trend compared to the reanalysis, and this trend occurs earlier in larger cities than in smaller ones. For example, the station on 5th Ave, NY City, shows a substantial increase in temperature (about 1C) with respect to the 1950's starting in the 1960's, increasing to 1.3C in the 1980's and 1990's. La Guardia Airport shows about 0.5C increase in the 1960-1980's, and an additional increase to 1C in the 1990's. Smaller airports show a smaller increase in temperature difference. Rural towns tend to show less trend in the difference, and Allegany State Park (forested) shows a negative trend. Of the 75 stations considered, 45 show a decadal warming with respect to the base period (1950's) of at east 0.2C during the 1960's. This number increases to 58 stations in the decade of the 1990's. 23 show no change in the 1960's and only 13 show no change in the 1990's.

Although these differences for New York State are consistent with those that could be expected from urbanization and other changes in land use, we need to check whether the methodology gives similar results for all other US stations. If the results continue to be consistent, they would suggest that this methodology could indeed be used to assess quantitatively land use impacts on the surface temperature.