Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Thursday November 15 2018, @06:49AM   Printer-friendly
from the frog-in-a-pot dept.

The San Diego Union-Tribune is one of a few sources reporting: https://www.sandiegouniontribune.com/news/environment/sd-me-climate-study-error-20181113-story.html

Researchers with UC San Diego's Scripps Institution of Oceanography and Princeton University recently walked back scientific findings published last month that showed oceans have been heating up dramatically faster than previously thought as a result of climate change.

The original paper indicated that oceans were warming 60 percent more than outlined by the IPCC and was widely published and remarked. The significantly increased warming conclusion was quickly challenged by an English mathematician looking at the methodologies used.
The authors promptly confirmed the issue thanking him for pointing it out, and have redone their calculations and submitted corrections to the journal Nature. Per one of the authors after reviewing and correcting:

"Our error margins are too big now to really weigh in on the precise amount of warming that's going on in the ocean," Keeling said. "We really muffed the error margins."

The article continues:

While papers are peer reviewed before they're published, new findings must always be reproduced before gaining widespread acceptance throughout the scientific community, said Gerald Meehl, a climate scientist at the National Center for Atmospheric Research in Boulder, Colorado.

"This is how the process works," he said. "Every paper that comes out is not bulletproof or infallible. If it doesn't stand up under scrutiny, you review the findings."

The same author indicates "the ocean is still likely warmer than the estimate used by the IPCC"


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1) by khallow on Thursday November 15 2018, @03:30PM (6 children)

    by khallow (3766) Subscriber Badge on Thursday November 15 2018, @03:30PM (#762194) Journal
    To elaborate, any deterministic method for generating pseudorandom numbers has a period. Period.
  • (Score: 1, Interesting) by Anonymous Coward on Thursday November 15 2018, @04:47PM (5 children)

    by Anonymous Coward on Thursday November 15 2018, @04:47PM (#762230)

    To elaborate, any deterministic method for generating pseudorandom numbers has a period. Period.

    This is not true in general. It is only true if the internal state of the generator is bounded (i.e., for finite state machines only). Most practical RNGs meet this condition (internal state is typically encoded as a fixed number of binary digits) because they're easy to reason about and simple to implement, but it is not an essential requirement for a deterministic generator in general.

    A deterministic generator is a set of states s, a set of output values v, an initial state (the seed) s₀s and a transition function f : ss × v. Each transition starts in a particular state and yields a new state and an output value.

    The argument is simple: if s is finite, then it has finite m = |s| elements so after m transitions it must be the case (by pigeonhole principle) that a state was repeated at least once (which, because f is a function, implies the sequence repeats from that point). Thus such a generator has finite period.

    However it is easy to construct a deterministic generator with infinite period. Such a generator necessarily has an infinite s. Here's one (this example does not produce a particularly "random looking" sequence):

    • s = ℕ (the set of natural numbers),
    • v = s,
    • f (n) = ( n+1, n+1 )

    Here's another, with a more "random looking" output sequence and generates a "true" bitstream (i.e., the output set v is finite -- one bit output per transition):

    • s = ℕ (the set of natural numbers),
    • v = {0, 1},
    • f (n) = ( n+1, t (n) ), where t (n) is defined by the following recurrence relation:

      • t (0) = 0,
      • t (2n) = t (n),
      • t (2n + 1) = 1 - t (n)

      —the Thue-Morse sequence.

    • (Score: 3, Funny) by MichaelDavidCrawford on Thursday November 15 2018, @05:05PM

      by MichaelDavidCrawford (2339) Subscriber Badge <mdcrawford@gmail.com> on Thursday November 15 2018, @05:05PM (#762240) Homepage Journal

      Consider that the Smiling Poop emoji is a Unicode character

      --
      Yes I Have No Bananas. [gofundme.com]
    • (Score: 2) by maxwell demon on Thursday November 15 2018, @07:13PM (1 child)

      by maxwell demon (1608) on Thursday November 15 2018, @07:13PM (#762302) Journal

      This is not true in general. It is only true if the internal state of the generator is bounded (i.e., for finite state machines only). Most practical RNGs meet this condition

      Not most. All. Simply because all machines we've ever built are physically finite. And a physically finite machine has only finitely many reliably distinguishable states.

      --
      The Tao of math: The numbers you can count are not the real numbers.
      • (Score: 0) by Anonymous Coward on Thursday November 15 2018, @08:36PM

        by Anonymous Coward on Thursday November 15 2018, @08:36PM (#762331)

        Not most. All. Simply because all machines we've ever built are physically finite. And a physically finite machine has only finitely many reliably distinguishable states.

        The generators described previously have unbounded state, as in the storage requirements increase without bound as the generators continue to crank out more and more numbers. However, this is different from infinite: the amount of storage is always finite, and it in practice the algorithms described can be implemented on real computers. When your disk fills up you just add another disk and crank out more numbers until that one fills up, and so on.

        Both generators are easy to program in a language with unbounded integers (basically anything modern). In both examples given, the storage and time requirements increase in proportion to the logarithm of the number of iterations so far. In reality that means the generators are fast and consume little memory proportional to the amount of output they produce, because the logarithm function increases very slowly.

        Just for fun, I hacked together an implementation of the second one in Haskell.


        -- thue_morse n computes the nth element of the Thue-Morse sequence.
        thue_morse :: Integral a => a -> Bool
        thue_morse 0 = False
        thue_morse n
                | n < 0 = error "domain error"
                | otherwise = odd n /= thue_morse (n `div` 2)

        -- Generate "random" bits by following the Thue Morse sequence.
        generator :: Integral a => Integer -> (Integer, a)
        generator n = (n+1, fromIntegral . fromEnum $ thue_morse n)

        -- Generate "random" Int values on the interval [0, 2^29-1]
        generate_int :: Integer -> (Integer, Int)
        generate_int s0 = (next, val) where
                gen = take 29 $ eval_machine generator s0
                next = fst (last gen)
                val = sum $ zipWith (*) (map snd gen) (map (2 ^) [0..])

        -- Evaluate a state machine for a given initial state by returning a list of
        -- (state, value) pairs, where each entry in the list is the result of the
        -- machine applied to the previous state.
        eval_machine :: ( a -> (a, b) ) -> a -> [(a, b)]
        eval_machine m s0 = tail $ iterate (m . fst) (s0, undefined)

        -- Print the sequence of Int values produced by generate_int. The resulting
        -- sequence has no period.
        main :: IO ()
        main = mapM_ (print . snd) $ eval_machine generate_int 0

    • (Score: 0) by Anonymous Coward on Thursday November 15 2018, @09:08PM (1 child)

      by Anonymous Coward on Thursday November 15 2018, @09:08PM (#762344)

      The Thue-Morse sequence is not only has a finite alphabet and no period, but it was famously used to construct an infinite squarefree sequence over a finite alphabet which is an even stronger property: such a sequence has no consecutive repeating subsequences whatsoever (this necessarily requires more than 2 symbols).

      For example: 0, 1, 0 is squarefree, but 0, 1, 1, 0 is not since it contains a repeating subsequence (called a square): 1 is immediately followed by 1 in the sequence. The sequence 0, 1, 0, 1 is also not squarefree, because it contains a repeating subsequence 0, 1 immediately followed by 0, 1 in the sequence (it is easy to show that there are no squarefree sequences of length 4 with just 2 symbols).

      However we can use the Thue-Morse sequence (which is not squarefree), and if we add -1 as a symbol, we can construct a new sequence by subtracting each element from the next element in the sequence, the result is squarefree. So the Thue-Morse sequence 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, ... becomes 1, 0, -1, 1, -1, 0, 1, 0, -1, 0, 1, -1, 1, 0, -1, ...

      • (Score: 1) by khallow on Friday November 16 2018, @01:30AM

        by khallow (3766) Subscriber Badge on Friday November 16 2018, @01:30AM (#762447) Journal
        That sequence can't be generated with a finite state machine.