This is Lucia (The Blackboard)'s Typepad Profile.
Join Typepad and start following Lucia (The Blackboard)'s activity
Join Now!
Already a member? Sign In
Lucia (The Blackboard)
Recent Activity
Chris-- The 7 day smooth JAXA extent is lower than it was 7 days ago. Only once since 1972 did the minimum in the 7 day average occur after we'd seen the 7 day smooth extent decline relative to the previous week's value. That was in 1997. How do the weather maps compare to 1997?
Toggle Commented Sep 17, 2011 on NSIDC also calls the minimum at Arctic Sea Ice
Ned-- My 'predictometer' says the 50%50% point for the minimum in the 7-day average of JAXA is 4.443; the current value is 4.547. The daily minimum would be somewhat lower. So, based on the condition of the ice, we should expect some more drops. That said, my 95% confidence intervals are 4.318 to 4.532. (The reason for the excess decimals is merely to show people how sensitive various features are to the assumptions under my statistical model.)
Toggle Commented Sep 13, 2011 on First uptick IJIS at Arctic Sea Ice
Peter-- It's true the balance of views varies from blog to blog. But in fairness, people at my blog are also looking at every snippet of evidence snippet of data to see whether they think we might be 1 day off or 20 days off from the minimum. If you don't like the 'balance' at Neven's stop by mine-- or some other blog.
Toggle Commented Sep 12, 2011 on First uptick IJIS at Arctic Sea Ice
Seke Rob-- I haven't checked, every year, but I'm pretty sure JAXA extent has wobbled up and down before hitting the minimum every year it reported. If not, it's frequent enough. That's one of the reasons I picked a week average for betting. When the 7-day average turns up, we can be pretty confident the minimum is reached. But we're likely to see 2 or three wobble-ups and downs before the final JAXA minimum.
Kevin The melt-season stratospheric temperatures over the Chukchi plummeted after 2006. This is likely an indication of reduced ozone. Do we have relevant data going back in time? I think chris has a plausible phenomenological explanation, and it has some data support -- but the data are noisy. If we look to numerical values of AIC coefficients, we find area is a good predictor, but not so much better than other predictors that we can say area is definitively "best". It would be nice to test any alternate explanations (or even just supplemental explanations) that arise. If stratospheric temperatures are known or if ozone temperatures are known-- or there is any other relevant data available to test, that would be useful. Inclusion might 'explain' just enough of what now looks like noisy residuals to be able to really say that upcoming losses are most strongly correlated with area rather than extent or volume both of which come up as models that can't be excluded relative to area. (Mind you: area looks better. It's just the Akaike criteria don't say it's a clear winner.)
Opps! I'd bet this years minimum is reached. should read "has not been reached". I anticipate things will go down a little more.
Toggle Commented Sep 8, 2011 on New area record at Arctic Sea Ice
Neven-- I would never try to indicate what will happen tomorrow, but if I were forced to bet, I'd bet this years minimum is reached. On the graph, I indicate the level we expect if CT experiences the average loss from this day forward. It's 2.829. I haven't done anything more to see whether other features describing ice conditions ( like current area, extent, volume ) suggest we would tend to get greater than average or lesser than average area losses. (It would be quick to do-- but I haven't done it.) For extent the data describing general ice conditions suggests we'll get greater than average extent losses. I suppose if I run this for area, I'd likely get a similar result, but I don't actually know. Oh-- and I should note: even when discussing extent, none of the regressors I look at for anything include predictors about upcoming weather. (Cloudiness, storminess, wind direction, speed etc. are all missing from the regression. Any effects those have appears as "noise" in my regression. Also, the regressors don't include information that is not available as numbers. So, for example, to the extent that another reporting agency shows higher or lower values, I would suppose that information would be useful to include when guessing whether CT area is going to drop or rise. But quite a bit of interesting data are not available as numbers-- or not available over a long enough period--so I don't include those.)
Toggle Commented Sep 8, 2011 on New area record at Arctic Sea Ice
The trend lines are so close that the difference cannot be eyeballed on the graph That's why I made a zoomed in graph: (I hope that shows. If not here's the link to a larger one.) http://rankexploits.com/musings/wp-content/uploads/2011/09/AreaRecordSet.png
Toggle Commented Sep 8, 2011 on New area record at Arctic Sea Ice
Paul IJIS is equivalent to IARC-JAXA; its the same entity. Yes. Everyone knows this and were aware of it prior to your little lecture. That is why lots of people call the item that IARC-JAXA - same as IJIS labels with "IARC-JAXA" JAXA instead of calling it "IJIS". He can refer to them as JAXA, I have no problem; he knows what's going on, and who's who. I'm sure Gavin will be relieved to learn you have no problems with him calling JAXA JAXA. Your inaccurate reply shows that, yet again. What inaccurate reply? Gavin does call the data product JAXA. IJIS does lable the graph IARC-JAXA. You seem to have some gripe about me using the word JAXA. I'm going to continue to call it JAXA-- just as many others including Gavin do. I think it helps to know something about the system producing the data. For other readers, I know this sounds like a broken record, but Lucia and others can't seem to accept this. As far as I can tell noone has suggested it doesn't help to know something about the system producing the data. I haven't. I also haven't seen any of the zillions of people here discussing your various theories about the systems suggest that knowing something about the data product doesn't help. It seems to me I read a number of people suggesting that instead of fiddling trying to reverse engineer JAXA's algorithm, you just break down and read their description-- going so far as to provide you the link. So, yes. I think it helps to know something about the data, as do many people here. I most especially have no idea why you would think someone calling the product "JAXA" (as many including Gavin do) suggests they might not know something about the product.
Paul-- I'm aware there is only one earth. I am also aware that our knowledge of the actual extent comes from measurements and to the extent that we know whether a record was broken, we need to look to recorded value. I still don't know if you are actually predicting that the JAXA minimum will fall below the minimum recorded by JAXA in 2007. But I guess I can live without this knowledge. Thanks for the tutorial on the 'proper' name for the product. I'm going to continue saying JAXA just as Gavin does in his post at RC. Feel free to visit RC and post a comment explaining the proper terminology to Gavin, including your explanation of the relationship between terms like "JAXA" and "NASA". Oh, and tell the JAXA folks to start typing IJIS instead of IARC-JAXA in the right hand corner of their graphs. Heaven forbid people their choice of labels might encourage people to use the word "JAXA" to describe their product.
Paul I now believe that the chance of breaking the 2007 extent low is almost certain, since .... Do you mean you think the chance of JAXA specifically reporting a record low is virtually certain? Or do you mean the chance of at least one of (JAXA, NSIDC, Bremmen etc.) reporting a record low being almost certain? Or something else?
I look at the 7 day average minimum. If the extent loss rate and duration between now and the minimum repeats what we saw in 1984, we'll see a new extent minimum with JAXA. So a record can be achieved with extent loss rates that have been observed in the past. I'm not suggesting it's probable, but it's not out of the question based on historic extent loss rates.
Toggle Commented Sep 6, 2011 on Some more flash melting? at Arctic Sea Ice
Chris-- I just went to Jaxa: 09,02,2011,4720781 09,03,2011,4682188 I think this posted within the past 10 minutes.
Paul Someone recently used the JAXA data to squelch the idea that large extents could be lost overnight in a regional area of the ice pack. Who? Let me break that down into two questions: 1) Who has said large extent can't be lost overnight and 2) Who has the power to squelch an idea. The magnitude of recent loss rates has been discussed at my blog and my position has been that I have no reason to believe it is not possible and that as far as I can tell based on published extent loss rates there is no reason to believe that large extent loss rates of the sort we saw in August have not happened in the past and they have happened in August. (Discussion here: http://rankexploits.com/musings/2011/wednesday-nh-ice-update-recent-extent-losses/ ) In terms of overall loss rate, what we saw in August 2008 appear in historic GSCF and JAXA records. On the issue of who has power to squelch an idea-- I find it difficult to believe there is someone out there who can squelch this idea or even the contrary idea.
Paul-- I realize you've convinced yourself of all of those things. 7. The system hasn't been around as long as the NSIDC legacy system. 8. The measurement system doesn't use as small a grid size as the MASIE system. These are hardly news.
Paul You say that you must have the source data. No I didn't.
PaulK Lucia and Mosher, you are both ignoring my finding. I cannot find a set of measurements that generate the rolling 2-day averages generated by IJIS, without the measurements "blowing up" on either side of the test period. In short, it is mathematically and physically impossible that the IJIS reported data are simply 2-day averages. I am not ignoring this. I am just not commenting on it. You can confirm this easily, if you tried. Sure. So? The test is fast and easy to run. It's not a test. It's a fiddle.
Mosher-- Showing the code would qualify as method (2) above (September 02, 2011 at 21:04). I need method (1) or (2) to believe there is some sort of lag.
Paul-- The comment you reposted -- and which I had previously read--is precisely the type of discussion where you describe a sort of fiddling with numbers that does not show there is any sort of lag, smoothing, damping, or averaging beyond what JAXA already tells us they do. Not only does what you did not show it, the amount of data you describe looking at is inadequate to even beginning to test the notion. As far as I can tell nothing about what you describe yourself as having done supports your claim that "JAXA-IJIS is using some kind of algorithm to dampen large extent loss days, and 'augment' low extent loss days. Use of either an underlying trend, or longer term average in the algorithm, or some sort accumulated over/under account seems likely. I like the last idea; sort of a "slush fund" for ice extent reporting". If you want to test your theories about some sort of systematic lag between method A and method B, or some sort of, averaging, damping or any other complicated thing you speculate is going on, you are going to have to long time series of data -- multi-year--tests for lags over a long time. Otherwise, you haven't tested for these things. If you can find a comment where you describe doing either (1) or (2) in my comment above, please link it because I haven't seen it.
Paul-- I commented on your theory involving JAXA. which should have been apparent because I used the word "JAXA". A reply that does not discuss JAXA is hardly relevant. I think you should note that Ned also used the word "JAXA" when discussing your "time lag" theory in his comment "Posted by: Ned Ward | September 01, 2011 at 21:14" So your long discussion of a time lag between two series neither of which is JAXA would appear to be rather irrelevant to the issue of the time lag involving JAXA. I'll admit these discussions of your time lag theories are confusing because have more than one theory about more than one time lag. One theory seems to involve JAXA relative to others; one MASIE. ( Are there any other groups lagging? It's hard to keep track.) But no matter how "easy" you think it is to show that one group lags another, in my opinion testing any of your various theories that "A" lags "B" requires someone to either (1) post the time series for both A and B for over multiple years so that others can see the time series you are working with and then show one series lags another by a specified number of days over a span of multiple years or (2) read the description of how the data are assembled provided by the reporting agency and demonstrates that method will have some lag. (So, for example JAXA is a two day average, so, in that sense it must have a short lag-- but two day averaging doesn't cause a week lag.) As far as I am aware, no one -- including you-- has tested your various theories about lags doing (1) or (2). If you are correct and it is easy to do this, since the matter seems important to you, I think you should go ahead, do it and then present the time series and show us there is a robust lag-- or point to the comment where you showed it. Otherwise, while your speculations may pan out, as far as I can tell, they are utterly untested.
Mosher-- I for one would need to see multiple years worth of MASIE vs. IJIS data before I would consider Paul K's theory of the lag to have been tested. I'm not sure how many years-- but likely 5. If I saw one year worth of data suggesting a lag, I might start paying some attention to this. For now, my general sense is: If method A shows a greater loss than B over the past week or month, it's plausible that method B will show the larger loss in the upcoming week or month. This is because all methods have some errors and over time, if the errors are not biased, the errors will tend to correct themselves. Unless I see years worth of comparison properly processed, I will interpret behavior that looks like one method is "converging" toward the other method to tell us precisely nothing about Paul's theory of the lag. As far as I can see, if a convergence occurs when the labor day weekend is past, this will neither contradict nor confirm Paul's theory.
L.Hamilton -- Thanks! I only wish that numbers were available online so I could suck that into my weighted model which I use for blog purposes. ( But that reminds me. Steve McIntyre mentioned that R can digitize data from graphs. I'm going to have to look into that!)
Toggle Commented Sep 2, 2011 on 2011 End Zone at Arctic Sea Ice
Bfraser--Thanks!
does anyone know how to go from someones latest comment on the left of Nevens home page, to the latest comment at the end of the comments. I've been wondering the same thing. I was hoping Neven could look into setting Typepad permits to see if he can get the side links to include the /comments/page/n/ etc stuff required to get us to the comment.
Paul-- What makes you think you know what answer -- if any-- I attributed your retort about simplicity to? I will admit that when retorts are addressed to me, I tend to expect they might be responding to something I posted. That said, I am aware that this is not always the case. Thanks for pointing out that the comment you addressed to me had absolutely nothing to do with the conversation I was having with Ned. I have made, I believe, an almost overwhelming case that MASIE is dating the data six days later than taken. Well, I've read your various theories. Rest assured that I strongly suspected you are convinced you have presented an overwhelming case for your notion about the time lag. Only one of our theories can be correct. If you want to consider the full range of possibilities, you both could be wrong.