Guest analysis by Gary Boden
A little history
Poorly sited stations in the U.S. Historical Climate Network (USHCN) have been documented extensively at WUWT in many posts and by the Surface Stations project Anthony started over a decade ago.
The USHCN is a 1,219 station subset of the much larger NOAA Cooperative Observer Program (COOP) Network that was selected for baseline monitoring of long-term temperature trends in the contiguous United States. See https://www.ncdc.noaa.gov/ushcn/introduction for more information.
Between 2007 and 2012 dozens of Surface Stations project http://www.surfacestations.org/ volunteers surveyed 82.5% of the network to document just how poorly it conformed to specifications. The effort resulted in publication (Fall, et al. 2011) of a strong critique of siting deficiencies that bias reported temperatures and call into question the suitability of USHCN data for climate research.
CRN 1 – a clear flat surface with sensors located at least 100 m from artificial heating and vegetation ground cover <10 cm high.
CRN 2 – same as CRN 1 with surrounding vegetation <25 cm and artificial heating sources within 30 m.
CRN 3 – same as CRN 2, except no artificial heating sources within 10 m.
CRN 4 – artificial heating sources <10 m.
CRN 5 – sensor located next to/above an artificial heating source.
Even with so many poorly sited stations, is it possible to use the well-sited ones to evaluate the temperature record? A comprehensive analysis of the environment immediately around the sensors using on-site station photography, satellite and aerial imaging, street level Google Earth imagery, and detailed evaluation of station metadata found only 410 USHCN stations with minimal artificial influences during the 1979-2008 period. The study by Anthony Watts and Evan Jones of surfacestations.org, John Nielsen-Gammon of Texas A&M, and John R. Christy of the University of Alabama – Huntsville was presented at the 2015 AGU meeting in San Francisco, CA. Among other conclusions, the data suggest
“that the divergence between well and poorly sited stations is gradual, not a result of spurious step change due to poor metadata.”
A better network
Upon the realization that the quality of essential data was suspect even before Surface Stations made it glaringly obvious, NOAA undertook an effort to build a network that would be contaminated minimally by microsite influences. Thorne et al. (2018) gives a brief introduction to the history and challenges of meteorological observations at ground stations. It also proposes a conceptual framework for a truly fiducial (i.e., trustworthy) network, saying it “would provide measurements that are meteorologically traceable, with full metadata. It would not only provide unambiguous high-quality time series but serve to validate and enhance trust in the quality of the other networks…”
The U.S. Climate Reference Network (USCRN), deployed beginning in 2002, is given as an example of a national system of well-sited stations with modern, redundant, and carefully-maintained data collection equipment. https://www.ncdc.noaa.gov/crn/. The network consists of 114 stations distributed throughout the contiguous United States. Seven sites have paired stations for providing local redundancy; the remaining 100 are single station sites.
The goal is to provide the best data possible for answering a question in the future: how does climate changed over the decades? For reliability, sites were carefully selected to be in open rural areas, likely to be free from human development and vegetation encroachment for many years. Photographs of the stations can be found at www.ncdc .noaa.gov/crn/photos.html.
Diamond, et al. (2013) describes the USCRN in a report after ten years of operation.
“USCRN has historically maintained a data record availability exceeding 99.8% annually since 2007; this equates to missing about 18 hourly observations of data per station per year: many of which occur during the AMVs” (Annual Maintenance Visit).
Some instruments have triple redundancy so that problems with individual units can be discovered by comparison and resolved quickly. Hourly data is transmitted from station via GOES satellites to the National Climate Data Center (NCDC) for processing and eventually public release.
Early data from five of the seven paired sites was examined by Gallo (2005) for station representativeness. He found significant differences
“between the paired stations in the annual minimum, maximum, and mean temperatures for all five pairs of stations” and concluded that “results suggest that microclimate influences on temperatures observed at nearby … stations are potentially much greater than influences that might be due to latitude or elevation differences between the stations.”
One paired set Gallo did not look at is in Kingston, Rhode Island. The two stations stand 1.4 km apart with only about a 3-meter difference in elevation. Plains Road is situated closer to developed land on the campus of the University of Rhode Island — mostly athletic facilities and sports fields — near small agricultural test plots and turf fields. The installation was placed about 50 meters north of an existing USHCN station placed at this location in 1971, no doubt to run parallel for cross-checking. Peckham Farm lies to the south in open fields surrounded by woodlands. Both stations would have rated a 1 or 2 on the CRN siting criteria scale at installation.
As the Surface Stations project volunteer who documented the old USHCN station, I was curious about what the new instrumentation was recording; but more than that, concerned that development might be encroaching on a “pristine” site. Over the years, the University has expanded development in the direction of the Plains Road site as the following sequence of dated Google Earth aerial images illustrates.
Plains road on 04/04/2001 before installation of the USCRN station. The old USHCN station is located near the buildings slightly left of center. Note the otherwise open fields around the site.
The recently installed Plains Road station is indicated by the yellow arrow in this image from 04/29/2003. From the air temperature sensor array it is 55 meters west to a dirt road, 48 meters southeast to a greenhouse, 95 meters southwest to nearest pavement (small parking area), and 145 meters south to nearest paved road. Note the recent construction of large parking lots 212 meters to the east in what was open fields.
By 07/28/2007 the parking lot had been extended to within 182 meters of the station.
By 04/02/2012 three new greenhouses had been constructed only 23 meters west of the instrument arrays.
Even more construction is evident in this image from 09/11/2014. A new road comes within 68 meters and another parking lot extension lies only 93 meters away. The lone greenhouse was removed to make way for the road, but the other three remain.
In contrast, Peckham Farm has remained unchanged from installation as represented in this image from 09/11/2014. There is a narrow dirt access road about 20 meters away on the west side. Distance to the trees is 126 meters east.
Looking at the data
The primary criteria for siting USCRN stations are 1) spacial representativeness of regional climate, and 2) temporal stability with the likelihood the site will not experience major change, especially from human encroachment. Peckham Farm fits these criteria quite well, but Plains Road fails on the second point. Documented encroachment of a type known to affect temperature measurement (paved surfaces) has occurred already.
Does this mean Plains Road should be moved? Not necessarily; there may be a benefit in disguise. The situation now provides an excellent opportunity to evaluate the effects of encroachment. After sixteen years of data collection we probably have enough information for a reasonable evaluation.
To its credit NCDC makes the data for all USCRN stations publicly available at its website https://www.ncdc.noaa.gov/crn/qcdatasets.html so I downloaded files for both stations from January 1, 2002 through December 31, 2017. (see spreadsheet below as well)
Data is collected every hour as temperatures rise and fall throughout the day. For this analysis I’m looking at the monthly averages of daily mean, maximum, and minimum temperatures. The README.TXT file says mean air temperature in degrees C is calculated using the traditional formula of (T_MONTHLY_MAX + T_MONTHLY_MIN) / 2. According to standards,
“Monthly maximum/minimum/average temperatures are the average of all available daily max/min/averages. To be considered valid, there must be fewer than 4 consecutive daily values missing, and no more than 5 total [daily] values missing.”
Because readings are taken hourly there is no concern about time-of-observation adjustments.
I calculated the simple difference between stations (Plains minus Peckham) and plotted it against time, omitting months with either station having a missing value. A plot of the mean temperature difference between sites is quite revealing. First, the vertical axis shows that overall Plains Road was a bit warmer than Peckham Farm as indicated by most values being on the positive side of zero. This might be expected because of proximity to artificial structures at Plains Road.
The next thing to observe is that data from the initial three years (2002-2005) is problematic. It’s quite spotty and erratic compared to what comes later. 2005 is especially strange. Closer examination of the daily and hourly readings reveals some clearly erroneous numbers where the sensors became unstable. Why they remain in the NCDC data sets is unclear, but fortunately, we can exclude anything before 2006 without affecting the rest of the analysis.
Monthly mean temperature differences from 2006 through part of 2013 (blue) behave better. Most lie in a range of about 0.5 degrees C. There is some suggestion of an annual cycle but not too much should be made of it. The dotted blue line marks the average for the period and indicates that Plains Road was about 0.13 degrees warmer than Peckham Farm. Note that this is not a trend line which would be sensitive to endpoints; it’s merely the average over the eight years.
About May 2013 the data (red) suddenly step up and consistently stay more positive by 0.2 degrees over the previous period. Plains Road experienced a warming effect that Peckham Farm did not. The red dotted line marks the average of value (0.34 degrees) of the latest period.
What happened to cause the jump? The prime candidate is a road and parking lot extension shown in the 9/11/2014 aerial image. It appears that this occurred in summer 2013, dramatically altering microsite conditions. Earlier parking lot construction doesn’t seem to have produced discernable changes in the data, probably because of distance beyond the 100 m siting standard. The newest lot and road now lie within that boundary and are causing an effect.
Just how is the encroachment influencing temperatures? There is information to answer that question. A plot of differences in the monthly average maximum temperatures shows no difference before and after the May 2013 step jump that shows in the mean temperature plot. After excluding the suspect data before 2006, neither site is warmer than the other.
Not so with minimum monthly temperatures where Plains Road measures about 0.35 degrees warmer in the latest period. Since minimums usually occur at night, it’s reasonable to suspect that the encroaching blacktop at Plains Road is retaining the day’s heat into the evenings and artificially raising the numbers. This finding is consistent with observations at so many of the USHCN sites.
In spite of painstaking efforts to site USCRN stations to be representative of regional climatic conditions with small chance of contamination by local factors either at construction time or in the future, encroachment is always a potential threat. The paired stations at Kingston, RI, provide a clear example of even best intentions being confounded.
Yet it would be unfair to dismiss the Kingston USCRN paired stations as a failure. Actually, the site provides a unique opportunity to investigate in detail the effects of encroachment. Peckham Farm still is unaffected by development and can serve as the source of regional data as intended. Plains Road can be monitored carefully and analyzed using daily and hourly data to understand better the subtle effects of local contamination.
NCDC would do well to conduct a thorough evaluation and document what has taken place. The possibility of encroachment at the 106 other sites remains very real. Vigilance is required to keep the USCRN from suffering the kinds of degradation that has occurred at many USHCN sites. A compromised data set will not serve the purpose for which the USCRN was built – the recording of accurate and reliable information for necessary for climate research.
Fall, S., Watts, A., Nielsen‐Gammon, J., Jones, E., Niyogi, D., Christy, J.R., and Pielke Sr., R.A. Analysis of the impacts of station exposure on the U.S. Historical Climatology Network temperatures and temperature trends. J. Geophys. Res, Vol. 116, D14120, doi:10.1029/2010JD015146, 2011
Thorne PW, Diamond HJ, Goodison B, et al. Towards a global land surface climate fiducial reference measurements network. Int. J. Climatol. 2018;1–15. https://doi.org/10.1002/joc.5458
Diamond, H. J., T. R. Karl, M. A. Palecki, C. B. Baker, J. E. Bell, R. D. Leeper, D. R. Easterling, J. H. Lawrimore, T. P. Meyers, M. R. Helfert, G. Goodge, and P. W. Thorne, 2013: U.S. Climate Reference Network after one decade of operations: status and assessment. Bull. Amer. Meteor. Soc., 94, 489-498. doi: 10.1175/BAMS-D-12-00170.1
Gallo, K. P., 2005: Evaluation of temperature differences for paired stations of the U.S. Climate Reference Network. J. Climate, 18, 1629–1636.
USCRN MONTHLY DATA – 2002-2017 (Excel .xlxs)