Skip to main content

View Diary: Morning Feature: Governing Science, Part II - Climate Mysteries and Myths (129 comments)

Comment Preferences

  •  The truth is the 95% is not a true probability. (12+ / 0-)

    It's a number made up by a group of experts to reflect their expert judgment. It's not a probability determined by statistical evidence.

    And that's just fine.

    FWIW Venus has extremely high clouds and is extremely hot on the surface. The cloud "iris" hypothesis has been effectively disproved by recent research reports.

    Radiation physics tells us that the more CO2 we put into the atmosphere, all else held equal, the warmer it gets.

    And the geologic record confirms radiation physics.

    look for my DK Greenroots diary series Thursday evening. "It's the planet, stupid."

    by FishOutofWater on Fri Sep 17, 2010 at 04:15:38 AM PDT

    [ Parent ]

    •  I am sorry to disabuse your notion of probability (6+ / 0-)

      While probability and statistics are strongly related probability is not derived from statistics. Coin flips are coin flips. Dice rolls are dice rolls. One can determine the likelihood the dice are loaded from statistics but not the probability of fair dice.

      Practice tolerance, kindness and charity.

      by LWelsch on Fri Sep 17, 2010 at 04:30:32 AM PDT

      [ Parent ]

      •  I don't think you understand FOOW's point (7+ / 0-)

        For a coin flip there is a predetermined reason for assigning probabilities.  You don't need a test to determine the probability of getting three heads in a row on a fair coin.  

        His point is that the numbers the IPCC came up with are not based on any calculation, statistical or otherwise.  Statistically you could compare the current rise in temperatures relative background levels of change and based on NCrissieB's comment below that seems to be where the 95% comes in.  The 90% probability of human involvement would seem to be a purely educated (and conservatively biased) guess.  There is no way that I can think of you could actually calculate such a value.

        "We are normal and we want our freedom" - Bonzos

        by matching mole on Fri Sep 17, 2010 at 05:12:47 AM PDT

        [ Parent ]

    •  As Dr. Muller explained it ... (14+ / 0-)

      ... some people say the 95% and 90% probabilities were too conservative (giving in to industry or government pressure), and others said those were too dramatic (overhyping inconclusive data). He takes the statements at face value, based on his evaluation of the data as a peer reviewer.

      As I understood him, the 95% probability expressed the 5% chance that the warming measured in recent years is an anomaly, based on statistical analysis of historical temperature fluctuations. But I'm not a physicist and I may have misunderstood him.

      As for the 90% probability that human activity is a major contributing cause, he presents that as their best estimate of what they don't know about the formation and temperature effects of clouds.

      Good morning! ::hugggggs::

      •  I don't know how they calculated the 95% (6+ / 0-)

        but it would seem difficult to do it in a rigorous fashion.  What you are talking about is comparing the temperature change over the last, say, 100 years to the 'background' rate of change.  If you were to take a random 100 year period pre-industrial revolution, what is the probability it would show a similar or greater rate of temperature change?

        To calculate that probability you need to generate a frequency distribution of 100 year temperature changes.  I don't think there is any way to do that from first principles so the best way to proceed would be to sample random 100 year periods over and over again to get the distribution.  One problem is that the level of detail in our knowledge of the global climate is not constant over time.  Another problem is that you would ideally want to sample a lot of 100 year periods to get an accurate distribution.  To avoid having them completely overlap one another you would need to sample over a really long time period (say 100,000 years).  Over this time the earth's climate has radically changed (ice ages) making the older samples fairly useless in generating a probability for a modern event.

        I'm not claiming there isn't compelling evidence - just that given the constraints of the system (mostly that we have a sample size of one planet) there is no way to rigorously come up with these kinds of figures.  I will say that I am not a climatologist so I stand ready to be corrected.

        "We are normal and we want our freedom" - Bonzos

        by matching mole on Fri Sep 17, 2010 at 06:17:54 AM PDT

        [ Parent ]

        •  If I understand it correctly ... (4+ / 0-)
          Recommended by:
          Jim W, DBunn, matching mole, FarWestGirl

          ... and that's a big "if" because I'm not a physicist or climate scientist ...

          ... they plotted temperature changes since the last ice age based on ice cores and other sources that can be dated. The data on that are not as exact as they wish, and they use statistical interpolation methods that are common in working with such data. (There are some legitimate disputes about how to best do that; see the comments at the bottom of the thread.) Based on that, they seem to have derived means and standard deviations. In a normal curve model, 95% confidence means the target data - current temperatures - are at or more than two standard deviations from the mean.

          Again, I'm not a physicist or climate scientist, but I think that's how they arrived at the 95% likelihood that increased current temperatures are a real event (statistically significant) rather than a passing weather anomaly (background noise).

          Good morning! ::hugggggs::

          •  Thanks (2+ / 0-)
            Recommended by:
            Jim W, NCrissieB

            that explanation implies looking at mean temperatures rather than trends over time periods.  That would seem to be more statistically tractable although IMHO less compelling.

            I think the really compelling data is over the last 150 years - measures of CO2 concentration and rising temperatures.  Good data on both of those and a straightforward experimentally verified model of how CO2 would affect temperature.

            "We are normal and we want our freedom" - Bonzos

            by matching mole on Fri Sep 17, 2010 at 04:52:09 PM PDT

            [ Parent ]

            •  Why we need both.... (1+ / 0-)
              Recommended by:
              Jim W

              Before we look for causes of an event - the kinds of measurements you suggest and which they are doing - we must first be confident the event is not simply statistical noise.

              For example, if you roll a six-sided die and get a sample set where the mean is 3.5 and the standard deviation is 1.87 ... you would not look for causes as those are exactly the outcomes you would expect from random chance. In 100 consecutive rolls, you'd expect a series of three 6s to happen about 3 times. So long as the mean hovers at 3.5 and the standard deviation at 1.87, a series of three 6s would have no meaningful cause. It's just statistical noise.

              When climate scientists say they are 95% confident that current warming temperatures are a real event, they mean it's only 5% likely those "three 6s" are statistical noise. There is a meaningful cause.

              Their 90% confident that higher C02 levels due to human activity is a major contributing cause ... is from calculations like you describe. But in science you first have to prove there's a need to do those calculations.

              Good morning! ::hugggggs::

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site