Today, I’d like to discuss an interesting problem raised recently by Joe d’Aleo here – has the temperature of New York City increased in the past 50 years? Figure 1 below is excerpted from their note, about which they observed.
Note the adjustment was a significant one (a cooling exceeding 6 degrees from the mid 1950s to the mid 1990s.) Then inexplicably the adjustment diminished to less than 2 degrees …The result is what was a flat trend for the past 50 years became one with an accelerated warming in the past 20 years. It is not clear what changes in the metropolitan area occurred in the last 20 years to warrant a major adjustment to the adjustment. The park has remained the same and there has not been a population decline but a spurt in the city’s population in the 1990s.
I’ve spent some time trying to confirm their results and, as so often, in climate science, it led into an interesting little rat’s nest of adjustments, including another interesting Karl adjustment that hasn’t been canvassed here yet.
Update (afternoon): I’ve been able to emulate the Karl adjustment. If one reverse engineers this adjustment to calculate the New York City population used in the USHCN urban adjustment, the results are, in Per’s words, gobsmacking, even by climate science standards.)
Here is the implied New York City population required to justify Karl’s “urban warming bias” adjustments.
Here is the figure from Joe D’Aleo that prompted the inquiry:
First of all, the presence of a USHCN station in New York Central Park is an interesting choice, given the characterization of the USHCN network as being mostly “rural or small town”. Obviously one exception doesn’t disprove this characterization – it’s just that this is definitely an odd choice. But it provides an interesting opportunity to examine USHCN urban adjustment procedures in the context of what even climate scientists must surely acknowledge as being an urban location.
My initial attempt at replicating this result with USHCN v2 data, GISS data and/or GHCN data was unsuccessful. I asked Joe D’Aleo where his data came from and he pointed to two NOAA locations that I had previously not used – the top chart came from ERH here and the bottom chart came from Climvis here. In each case, I had to scrape the data from webpages (since there was no convenient ASCII version.) All three USHCN v2 versions (raw, time-of-observation adjusted and adjusted) were virtually identical) and were virtually identical to the ERH version from 1912 on. There is a puzzling difference between 1871 and 1912 which remains unresolved, but which is not relevant for the present note. The Climvis version did not tie in to any USHCN v2 series however.
I exchanged a number of emails with Karin Gleason of NOAA, trying to determine the provenance of the Climvis series. She pointed to the USHCN v1 website and eventually to the following data set: http://www1.ncdc.noaa.gov/pub/data/ushcn/urban_mean_fahr.Z and, sure enough, this matched (although it only went up to 2003). Combining the Climvis values after 2003 with this USHCN v1 data set, I obtained the following results (annual instead of July in the D’Alwo post), confirming a remarkable difference between the Climvis data (using one USHCN adjustment system) and ERH using another USHCN adjustment system. As D’Aleo observed, these are not small differences: The differences in the 1960s and 1970s exceeded 3 degrees C. How is such a discrepancy possible?
USHCN v2 and USHCN v1
The Climvis information links to USHCN version 1 here: http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html . If you scroll down to the adjustments section, you see that the data package contains 4 versions (version 2 only showing 3 versions), the four versions being listed as:
1. Raw: the data in this version have been through all quality control but have no data adjustments.
2. TOB: these data have also been subjected to the time-of-observation bias adjustment.
3. Adjusted: these data have been adjusted for the time-of-observation bias, MMTS bias, and station moves, etc.
4. Urban: these data have all adjustments including the urban heat adjustments.
They describe a 6th adjustment step as follows:
6. The final adjustment is for an urban warming bias which uses the regression approach outlined in Karl, et al. (1988). The result of this adjustment is the “final” version of the data. Details on the urban warming adjustment are available in “Urbanization: Its Detection and Effect in the United States Climate Record” by Karl. T.R., et al., 1988, Journal of Climate 1:1099-1123.
However if one goes to the corresponding site for USHCN version 2 : http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ , you find that the 6th adjustment step (the Karl et al 1988 “adjustment for an urban warming bias” is not included. Climvis, for some reason, is using USHCN version 1 data with Karl’s urban warming bias adjustments, updated somehow to the present.
So we have here a nice case study with which to examine the “urban warming bias” adjustment of Karl et al 1988. The Karl adjustment results in a huge reduction in adjusted temperatures in the 1960s and 1970s relative to modern temperatures – something that makes sense only if there had been a de-urbanization of New York City in the past century. There is negligible adjustment of 1850 to 2005 however. The GISS adjustment is not as extreme as the Karl adjustment, but it also has the effect of increasing the recent trend by relative lowering of earlier data and negligible relative adjustment of older data. Without Karl’s “urban warming bias” adjustment, there is no Central Park trend since the 1960s; with Karl’s adjustment, there is a 3 degree increase in Central Park temperature in the past 30 years. What a strange “urban warming adjustment” – it increases modern temperatures relative to temperatures a40 years ago, while the expectation for a “urban warming adjustment” would be just the opposite.
We also have the interesting spectacle of two branches of NOAA seemingly unable to agree on the temperature at Central Park in the 1960s and 1970s within 3 deg C, while Mann purports to know global temperature in AD1000 within 0.2 deg C.
I’ve been able to replicate the urban adjustment between 1890 and 1990 as shown in the graphic below. The implications of this replication for the post-1990 data are, in Per’s words, gobsmacking. Karl et al 1988, the reference for the urban adjustment provides a formula for urban adjustment (in deg C) as follows:
Candidate population data used by USHCN includes the following file ftp://ftp.ncdc.noaa.gov/pub/data/ushcn/metrof_hybrid.dat which gives populations nominally in 10-year intervals from 1890 to 1990 (although 1890 is missing.) I substituted these population figures into the formula above and show the negative values as bold points in the lower frame of the above graphic – yielding a virtually exact match. So we can safely conclude that this is how the USHCN “urban warming bias” adjustment was calculated between 1890 and 1990.
But what happens after 1990? For the period from 1990 to 2006, we don’t know the population figures used by USHCN, but we do know the value of the adjustment and, by reversing the calculation, we can calculate the New York population assumed by NOAA in their calculations, which is shown in the figure below. In effect, USHCN has assumed that by 2003, the last year in the urban_mean_fahr version, New York City has reverted to the population of 1910 and that by 2006 (according to Climvis) had reverted to the population of 1850 (1.2 million).
If the population of New York is held constant after 1990, as opposed to the extermination supposed by USHCN, then the urban-adjusted values would be as shown below: