Watts Up With That?

USHCN2 unadjusted and adjusted CONUS max and min temperatures – click for larger image

Background for new readers: The US Historical Climatological Network is a hand picked set of 1221 weather stations around the USA used to monitor climate. The network is currently being surveyed by a team of volunteers at www.surfacestations.org

When I visited The National Climatic Data Center a couple of weeks ago, one of the presentations made to me was on the upcoming implementation of the USHCN2 adjustments to the NOAA surface temperature record. Many have been critical of USHCN1 adjustments and for good reason – it completely misses some documented and undocumented station moves and biases that we have shown to be present via photography from the www.surfacestations.org project.

The goal of USHCN2 appears to be mostly in finding undocumented change points in the data, since from my discussions at NCDC it became clear to me that the metadata they have on hand is not always accurate (or even submitted by NWS personnel to NCDC in some instances). The weak link is the field reporting of metadata. They recognize that.

NCDC is thus faced with the task of finding and correcting such change points so that the bias from a  site move to a warmer or cooler measurement environment doesn’t show up as a false trend in data. In some cases, such moves are easy to spot in the data, such as the USHCN station in Lampasas, TX that got moved from an observer’s back yard to a parking lot location 30 feet from the main highway through downtown. The changepoint made it all the way through the NOAA data into GISS as shown by the GISS graph below:

Click to see the full sized GISS record

Matt Menne of NOAA is leading the USHCN2 implementation, and was kind enough to present a slide show showing how it works for me.

You can see the PowerPoint here watts-visit (PPT 6.1MB)

To get an idea of the differences, here is a summary from the slide show:

USHCN1 Originally released in 1987, 1221 stations

Addresses:

• Time of observation bias (Karl et al. 1986; Vose et al. 2003)
• Station History Changes (Karl and Williams 1987)
• MMTS instrument change (Quayle et al. 1991)
• Urbanization (Karl et al. 1988 )

USHCN2, To be released in 2008, 1218 stations (actually more stations have been closed than this)
Addresses:

• Time of observation bias (Karl et al. 1986; Vose et al. 2003)
• Station History and Undocumented Changes (Menne and Williams, Journal of Climate, in review)

While it seems that USHCN2 will be an improvement to detecting and correcting undocumented station moves and biases, it remains to be tested in a real world scenario where an undocumented station move is known to occur, but hasn’t been reported by NWS COOP managers for inclusion into the NCDC MMS metadatabase.

Fortunately, during my week long survey trip in NC and TN that coincided with my NCDC visit, I found two such stations that have in fact been moved, with significant differences in their surroundings, but have not been reported to NCDC.

Matt Menne has graciously agreed to run a blind test on the data for these two stations I’ve located with undocumented changes to see if the new USHCN2 algorithms can in fact detect a changepoint and correct for it. I’ll keep you posted.

A good portion of the changepoint detection work can be traced back to:

Lund, R., and J. Reeves, 2002: Detection of undocumented changepoints: a revision of the two-phase regression model. J. Climate, 15, 2547-2554.

Abstract:
Changepoints (inhomogeneities) are present in many climatic time series. Changepoints are physically plausible whenever a station location is moved, a recording instrument is changed, a new method of data collection is employed, an observer changes, etc. If the time of the changepoint is known, it is usually a straightforward task to adjust the series for the inhomogeneity. However, an undocumented changepoint time greatly complicates the analysis. This paper examines detection and adjustment of climatic series for undocumented changepoint times, primarily from single site data. The two-phase regression model techniques currently used are demonstrated to be biased toward the conclusion of an excessive number of unobserved changepoint times. A simple and easily applicable revision of this statistical method is introduced.

I talked with Robert Lund at length last summer at Roger Pielke’s conference about USHCN2 and his view then was that the NOAA implementation wasn’t as “robust as he’d hoped” (my description based on the conversation).

Station moves are only part of the potential biases that can creep into the surface record of a station. USHCN2 will not catch and correct for things like:

• Gradual UHI increase in the surrounding area
• Tree shading/vegetation growth/loss near the sensor increasing/decreasing gradually
• A gradual buildup of surface elements around the sensor, such as buildings, asphalt, concrete etc. Though is an asphalt parking lot suddenly went up close by that would likely show as a step which may be detected.
• Drift of the temperature sensor +/- over time.
• Other low frequency changes that don’t show a step function in the data

When I queried Matt Menne during the presentation at NCDC about the sensitivity of the new algorithm to detect changepoints, he suggested it would have about a 0.5°C step threshold. I had hoped it would be more sensitive.

While I look forward to seeing the improvements that USHCN2 has to offer, I feel that what also needs to be done is a historic time-line reconstruction for each USHCN site to help locate other possible change points that may be missed by the USHCN2 algorithms. I feel that a qualitative analysis can help zero in on potential errors in the station record. NCDC’s claim that they can statistically adjust the poorly sited locations so as to add skillful trend information is contradicted by some of Roger Pielke’s papers, so it will be interesting to see if the test cases I’m submitting will be detected or not.

Here are the cites for several of Pielke’s papers on the problems with the land surface temperature record as applied to long term trends:

That is part of the reason NCDC invited me for a visit, they feel the USHCN station surveys are a valuable tool to help them judge the health of the sites and network. They look forward to the completion of the surfacestations.org USHCN survey.