USHCN : It Is Worse Than It Seems

As mentioned earlier, I changed my USHCN methodology to match my GHCN methodology – averaging by month for all stations, and then averaging all months for the year. This fixed the discrepancy between my 2014 spike and Anthony’s.

Nick Stokes keeps talking about anomalies, which are necessary when trying to do an absolute comparison of partial years. I wasn’t attempting to do that – I was measuring temperature differences, so his point is irrelevant to what I am doing.

Here are the USHCN final and raw temperatures using the new method. It is almost identical to the old method. They are turning a 95 year cooling trend into a warming trend.

ScreenHunter_348 May. 11 07.38

But it is worse than seems. The USHCN “raw monthly data” isn’t really raw. The graph below shows the discrepancy between GHCN monthly “raw” and actual GHCN HCN daily station data – which is truly raw. As you can see, USHCN “raw” data is already adjusted to create a warming trend, and then they do their own adjustments on top of that.

ScreenHunter_349 May. 11 07.39

The next graph shows the total adjustments from GHCN HCN daily to USHCN final.

ScreenHunter_351 May. 11 07.51

But it is worse than even that seems. NCDC adds additional warming on to the USHCN adjustments. Unfortunately their web site is dead, so I can’t generate that graph right now.

About Tony Heller

Just having fun
This entry was posted in Uncategorized. Bookmark the permalink.

26 Responses to USHCN : It Is Worse Than It Seems

  1. Andy DC says:

    So, it would seem that even the knitpicking against you was not justified. Yet the alarmists still try to kill the messenger.

    • The large 2014 spike I showed originally was not the best way to do the calculation for a partial year with USHCN fabricated final data. I changed the algorithm to make a more realistic comparison between final and raw.

    • stewart pid says:

      Andy go to WUWT and read the blog & comments and Steven’s replies. It wasn’t really nitpicking but trying to unmuddy the slightly muddied waters.
      At least that is the way I see it.

      • stewart pid says:

        Also Andy for your info …. When finishing a knitting project, there are ends of yarn that stick out. “Knitpicking” is the act of removing these ends of yarn.
        There now I’m nitpicking 😉
        However maybe we are all incorrect since Merriam-Webster wants to hyphenate the word like this …. nit-picking. Usually I see it as a single word … who knows … who cares!

  2. Richard Mallett says:

    So your final (for now) graph implies that everything before 2001-2 was adjusted downwards, which makes sense if (for example) they were adjusted for UHI effects. What still doesn’t make sense to me is adjusting the earlier temperatures more than the later ones.

    • Gail Combs says:

      The ‘adjustments’ only make sense if you start from the wanted curve – a hockey stick – and plan your adjustments accordingly. They do the same with the CO2 measurements.

      At Mauna Loa we use the following data selection criteria:

      3. There is often a diurnal wind flow pattern on Mauna Loa ….. The upslope air may have CO2 that has been lowered by plants removing CO2 through photosynthesis at lower elevations on the island,…. Hours that are likely affected by local photosynthesis are indicated by a “U” flag in the hourly data file, and by the blue color in Figure 2. The selection to minimize this potential non-background bias takes place as part of step 4. At night the flow is often downslope, bringing background air. However, that air is sometimes contaminated by CO2 emissions from the crater of Mauna Loa. As the air meanders down the slope that situation is characterized by high variability of the CO2 mole fraction…..
      4. In keeping with the requirement that CO2 in background air should be steady, we apply a general “outlier rejection” step, in which we fit a curve to the preliminary daily means for each day calculated from the hours surviving step 1 and 2, and not including times with upslope winds. All hourly averages that are further than two standard deviations, calculated for every day, away from the fitted curve (“outliers”) are rejected. This step is iterated until no more rejections occur…..

      The entire hoax is based on selection of data to fit the wanted curve, whether the data is CO2 measurements or temperature measurements.

      • Sparks says:

        I pointed out the same issue about Mauna Loa CO2 measurements over 4 years ago. there was a shouting match and a lot of nagging simply for questioning it.

        And it’s interesting that there have been complaints and protests about Leif Svalgaard’s Sunspot reconstruction for increasing past sunspot numbers.

        So we have

        1. a systematic cooling of the past through “adjustments” to the temperature records.
        2. a systematic increase of past solar sunspot numbers reducing the influence of the primary driver of earths temperatures.
        3. a systematic curve fitting of an “artificial” CO2 measurement.

  3. Eric Barnes says:

    Nice Stokes. AGW apologist in chief. Any thing he comments on is surely a lie or an obfuscation of the truth.

  4. Sparks says:

    I asked Gavin Schmidt for his opinion this “data tampering”.

    “ClimateOfGavin What is your opinion on “data tampering” where the past is cooled? I understand it idea is based on time of recording.”

    He replied with this link!
    http://data.giss.nasa.gov/gistemp/FAQ.html

  5. Sundance says:

    Unfortunately Steve is unable to bully people and threaten to blackball them from being published in science journals in order to get his hockey stick (in the last graph) accepted by warmmongers. 🙂

  6. _Jim says:

    Unfortunately their web site is dead, …

    Accessible in a slightly round-about way … see thread on same further down for details. (one must use the IP address directly due to a DNS problem somewhere in “the internets”)

    The IP addresses that will get one to the ncdc website (the main page anyway) as of 10:40 AM CDT Sunday morning:

    http://205.167.25.171/
    http://205.167.25.172/

    .

  7. Hugh K says:

    The USHCN data process – Much like taking a choice cut of prime Kobe beef, marinating it for the proper time, cooking it to exact instructions and savoring every bite. In the end, it comes out as a useless turd.

  8. Nick Stokes … his point is irrelevant…

    Stokes has a hammer, & he knows how to hit things with it, so he’s pretty sure everything he sees is a nail. Try telling him otherwise & see how far you get.

  9. Nick Stokes says:

    You’re still doing it wrong. The right way is to take all stations that have both raw and final readings and for each station, get the difference. You know that difference is due to adjustment, so you can average them. This was the method suggested in the query by David A (why no answer?).

    When you average a bunch of finals and a bunch of raws, those averages include data from different times/places. Your original spike was caused by the fact that final readings were equally spread over months, but raw were weighted towards winter, because there were less in April. You got the same in Illinois. You would have got the same in any state.

    Averaging by month fixes the month issue, but not the selection of stations. If it happens that the extra stations in your final average are from warmer places, you’ll have a similar problem. You may be lucky and they will even out. But why not do it right?

    Your orange/blue plot is misconceived. It isn’t a good idea to average absolute temperatures, rather than anomalies. USHCN gets away with it because, with FILNET, they ensure that every month in “final” contains the same stations. For raw readings, that isn’t true. You’ll get variations between months depending purely on which stations report. And if there is a trend in where trends report from, you’ll get a spurious trend in temperature.

    “Nick Stokes keeps talking about anomalies:
    I explicitly said here that anomalies are not the issue. I didn’t use them.

    • My objective is to show the difference between the average final temperature and the average raw temperature for the US, and that is exactly what I have done. You are suggesting selective use of the data, and that is precisely what I want to avoid doing.

      The method you suggest would lose information about the infilling adjustments. My method retains it.

      • Nick Stokes says:

        “The method you suggest would lose information about the infilling adjustments. My method retains it.”

        There is no information about infilling adjustments to be retained. They have no corresponding raw, and there is no difference available in the data you have subtracted.

Leave a Reply

Your email address will not be published. Required fields are marked *