4b.3 Assessing the urban heat island signal in the U.S. Historical Climatology Network monthly temperature data

Tuesday, 19 July 2011: 9:00 AM
Salon C2 (Asheville Renaissance)
Zeke Hausfather, Efficiency 2.0, San Francisco, CA; and M. J. Menne, D. Jones, R. Broberg, T. Masters, and C. N. Williams Jr.

Urbanization over the past century in the United States has contributed to a warming bias in some U.S. Historical Climatology Network (USHCN) temperature records. While the impact of the urban warming bias is removed to some degree through data homogenization procedures, the extent to which the overall urban bias is corrected remains largely un-quantified. As a result, additional urban-specific corrections are sometimes applied to the homogenized data. In order to quantify the magnitude of urbanization bias in the dataset, and to quantify the extent to which the USHCN version 2 homogenization procedures correct for it, we examine minimum and maximum temperature trends from stations classified using four different proxies for urbanity--urban boundaries, satellite nightlights, population growth, and percent of impermeable surface, each created from publically available high-resolution GIS datasets. These urbanity proxies are used to segment stations into separate urban and rural sets, and temperature differences between the two are calculated using both spatial gridding and station pairing approaches. The analysis is performed on the USHCN version 2 time-of-observation-bias adjusted data and on data homogenized using NCDC's pairwise homogenization algorithm. Homogenized data that have been further adjusted using NASA GISS's Satellite Nightlight urban adjusted data are also evaluated. In addition, the pairwise homogenization algorithm is used to generate multiple versions of homogenized USHCN temperature data using comparisons solely with other COOP stations classified as rural according the four urbanity proxies and compared to homogenization results from the full COOP network. The magnitude of the urbanization bias in the un-homogenized (TOB-adjusted) data and the degree to which this bias is mitigated with homogenization is discussed.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner