The Truth about Extreme Average Temperatures In the US Region

One of the common concerns about the alleged doom and gloom from Anthropogenic Global Warming, is that the climate will become more “extreme” in terms of temperature, with both more extreme cold events and extreme warm events. At least when averaged over the region of the world that encompasses the contiguous US, I have just done an extensive analysis that shows that this idea is flat out wrong. In seeking to test the hypothesis that extremes of both warmth and cold would increase in the region of the US during warming, I decided I first needed daily data covering the region. This is available from the NCEP reanalysis dataset, and that data downloadable at the KNMI Climate Explorer. Now, reanalyses have their problems but there performance in characterizing weather and climate should be quite good in a region like the US wherein there are extensive observations to determine their behavior. I would be much more circumspect about using this if A) I were examining a more sparcely observed area and B) the result seemed inconsistent with other observational datasets in the region. Investigations I have done so far suggest that the NCEP data is consistent with other data in the region. Anyway, to define the US region, I’ve had to pretend that the US has annexed small parts of Canada and Mexico, and also this region will include some of the Atlantic and Pacific Oceans, as well as parts of the Gulf. The region is defined as from 24 N latitude to 49 North Latitude, just a little to the South of Western Dry Rocks, Florida, the Southernmost point in the contiguous 48 states that is occasionally above water, and just South of Northwest Angle/Angle Township in Lake of the Woods, Minnesota, Northernmost point in the contiguous US, and from 235 E longtitude to 293 E longtitude, these approximate the longitudes of Cape Alava, Washington and Sail Rock, Maine.

Anyway, after getting daily data averaged over this whole region, I separated data out into calendar years, and ranked each day in each year from coldest to warmest. Subsequently, for leap years, I copied the 183 coldest day (183 warmest in those cases, as well) to be a fake 184 coldest day, so that leap years would have the same number of days as non leap years. I am not entirely satisfied with doing things this way, but it matters very little, from what I’ve been able to tell. I will try to redo the analysis with any other method people suggest if they are interested. Anyway, I then calculated trends for day ranks over the last 32 years (approximately one half the length of the dataset) from 1979-2010 as this corresponds to the satellite period, the period when anthropogenic warming is supposed to be strongest, and in the thirty previous years the global climate appears to have changed little and temperatures cooled in the US. I will also calculate the trends for the first thirty two years for anyone interested if they ask (of course, those trends will have the central year in common in this dataset) but my primary interest is in the satellite period for now. So what did I find? Well, if the temperatures were getting more extreme in both cold and warm days, the coldest days should be slightly cooling and the warmest days strongly warming. If there is even just a net increase in extremes of warmth with no loss of cold extremes, then there should be more hot day warming than cold day, and if there is no change in over all extremes, then all days should show about the same trend. In fact, what I’ve found is that the most warming occurs on the very coldest days of the year, indicating that any increase in warm extremes is more than offset by much larger loss of cold extremes!


The X axis the rank of temperature within each year, y axis the slope of the linear trend line for that set of days from 1979-2010, in degrees Celsius per year.

This finding is consistent with earlier work by Knappenberger et al. Who came to the same conclusion examining different data (temperature stations, as opposed to mainly radiosonde derived products from the reanalysis) and slightly different periods of time. If this cold day warming is a signature of anthropogenic warming, it seems like a pretty nice thing to me. It certainly doesn’t support ideas about the US climate getting more extreme, at least in terms of temperatures. Once again, the results of testing a hypothesis about climate doom and gloom result in the reality being revealed: the world isn’t coming to an end. That’s information worth knowing, I think.


5 Responses to “The Truth about Extreme Average Temperatures In the US Region”

  1. steven mosher Says:

    Why not just use ghcn daily as opposed to data from a weather model?

  2. timetochooseagain Says:

    Steven-fair point, I want to say it’s because it will be a lot more work and I haven’t had time yet. After downloading the data from KNMI, I had to do the sorting one year at a time in Excel (to be honest I still can’t figure out R to save my life) Knappenberger et al. looked at station data and came to a similar conclusion.

    I’ll look into it, however, to see what kind of results I can get.

  3. timetochooseagain Says:

    So far it looks like when I look for stations in GHCN-D for Max temps (I searched for Min yet) there are 6967 stations in the region I previous mentioned with at least 32 years of data. Naturally a bunch of those are in Canada and other places. This is good, in a way, because I can narrow down to truly US data. I think I’ll start by doing that, then requiring that the station not end before 2010. At that point I am going to guess there will still be a huge amount of data. This looks like quite the undertaking!

  4. timetochooseagain Says:

    I’ve started by examining a few stations at a time. Arcadia, Florida was interesting since I live in Florida. One problem is that I can demand 32 years of data, and deliberately select stations afterward that end in 2010 or later, but apparently this just means “some” data for any calendar year, and not necessarily continuous data or complete data. I started preparing sort the days for this station, and instantly realized something was wrong, as before I got to sorting, some years had less than 365 days. I figure, okay, if one or two days are missing a year, I can work with that. Well, I started to determine which days were missing and, it turns out, that for 1998 and 1999, Arcadia went on a long winter vacation from reporting data. Literally from Christmas Eve to the twelfth of February, the data is all missing. For an analysis of extremes, considering January is one of the coldest, if not the coldest month of the year, having that entire month, and many days surrounding it, completely missing, is very problematic.

    I hate to say this, but having to examine individual stations, only to find I can’t use them after already checking for missing data in half the station’s years, this looks like a colossal waste of time. I must confess to lacking the motivation to examine thousands more stations in this manner. Especially since it probably doesn’t make a bit of difference. Chastise me if you must, but I give, this looks like too much work.

  5. Lessing Colding | Hypothesis Testing Says:

    […] an earlier analysis, now with data through 2013: I can ask two questions: ranking days within the year by temperature, […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: