6B.6 Assessing structural uncertainty in the U.S. Historical Climatology Version 2 adjusted temperature records

Tuesday, 25 January 2011: 4:45 PM
609 (Washington State Convention Center)
Matthew J. Menne, NOAA/NESDIS/NCEI, Asheville, NC; and C. N. Williams Jr. and P. Thorne

The challenge in recovering the climate signal from heterogeneous surface temperature records lies largely in our ability to remove the impacts of changes in the systematic bias of station records, many of which are poorly documented. Structural uncertainty refers to the uncertainty in the extracted climate signal associated with the various choices that must inevitably be made in designing a climate data homogenization procedure. Often, there is a range of reasonable options for each of the various steps required for producing a homogenized dataset, and it has become clear from work on radiosonde records that it is essential to quantify the degree to which the recovered climate signal is sensitive to the range of procedural choices. Here we describe a series of validation experiments conducted to quantify the structural uncertainty inherent in NOAA/NCDC's Pairwise Homogenization Algorithm, which is used for extracting the surface temperature signal from the U.S. Historical Climatology Network (USHCN). The experiments involved applying a set of realistic error models to homogeneous time series derived from global climate model output, which was resampled to mimic the surface temperature covariance structure as well as the record lengths present in the USHCN. A population of 100 realizations for each validation experiment was produced by varying the parameters used by the Pairwise Homogenization Algorithm in 100 different combinations. These experiments and the ways in which they allow inferences to be made about the real-world U.S. surface temperature trends are discussed.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner