![\"\"](\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/SeaIceMinima.png\" "\\"SeaIceMinima\\"")
<\/a>\nThe JAXA<\/a> data has been supplemented with gsfc.nasa.gov<\/a> data prior to 2002. The method of merging was: Find the first day for which JAXA has data; this was in June 2002. Trim off all gsfc data prior to that day. Merge the two sets. I made no attempt to smooth, find shifts or do anything fancy.<\/p>\n
After that, I removed all missing data and smoothed over 7 adjacent remaining days. (This means, if a day is missing, the average is over 7 entries, but they may span more than 7 days. This type of average is useful for betting<\/i> when data might drop out but it’s odd for science.) The current level is now only the 2nd lowest in the JAXA record and also the 2nd lowest in the merged record.<\/p>\n
I’m sure that Andrea will be pleased to point out that the black line indicating the current 7 day smooth ice-extent no longer falls below the 2007 ice extent for the similar time in 2007. <\/p>\n
Also, although the legend doesn’t fit in the image, some of you will notice the colors were created by moving along the color spectrum with time. The pink\/magenta colors are all in the lower region indicating a general trend toward lower minima. I could show a graph with the trend with time. Maybe next week.<\/p>\n
Instead, I’ll show the loss rate computed using weekly changes in the 7-day smooth ice extent:<\/p>\n
Bill Illis commented that the loss rate is at record minimum levels for this time of year using his slightly different method of computing loss rate; so we are both getting this. Bill suggested the possibility of a sensor error. That’s something that should always be suspected when data are fresh and a value is at the extremes of the range observed. I wouldn’t be shocked if we learned of some sort of sensor failure. By the same token, the loss rate is not so<\/i> far out as to be unbelievable. <\/p>\n For now all I can say is: This is the loss rate based on reported numbers. If this low loss rate is real and it remains low, we may even see the ice minimum remain above last years level. That said: if betting was still open, I wouldn’t bet on that! <\/p>\n On to Betting!<\/b> The orange circle represents the current 7-day ice extent and the corresponding prediction for the minimum that will occur this year. Last years value is highlighted in black. <\/p>\n Using this method of prediction the current best estimate for the 2011 ice minimum is 4.8 millions square kilometers. This exceeds my bet of 4.569 million square km placed on July 31. (My bet was computed using the smoothed value from July 28, and a fit based on JAXA only data.) Using the more recent data and the fit based on the merged data and neglecting uncertainty in the fit parameters, I estimate the \u00c2\u00b195% confidence interval to be 4.14 to 5.45 millions square kilometers. ( Including uncertainty in the fit parameters would increase the spread of my confidence intervals. )<\/p>\n I’ve also been reporting my estimate of the probability we will break the 2007 record. Using my simple linear fit, and assuming residuals are normally distributed, the probability of breaking the record is supposedly 6%. However, including uncertainty in the fit parameters would increase the estimate of that probability. Moreover, the method assumes no instrument error. This means that based on this method, I estimate the probability of breaking the record is more<\/em> than 6%.<\/p>\n Meanwhile, I estimate the probability the 7 day minimum ice extent falls below last<\/I> years minimum roughly 59%. For this prediction, considering the uncertainty due to sensor error or statistical uncertainty in the fitting parameters would reduce my estimate of the probability, moving it toward 50%. I’d judge that this method is telling us that beating last years ice extent minimum is a statistical dead heat.<\/p>\n Purple!<\/b> Blackboard bettors exhibit a wider spread in predicting the sea ice than my curve fit would suggest we should expect based on the current 7day ice extent. The Blackboard betters spread may very well be more realistic since a) I didn’t include uncertainty in the fit coefficients when computing the uncertainty based on the curve fit, b) Blackboard bettors may have bet early when they should<\/i> have a greater level of doubt, c) it’s not clear that my method of predicting is “true”. That is, even given current ice extent, it may be that I should use a non-linear fit of some sort or there may be a factor left out that would suggest the 2011 ice ought to be higher or lower. Finally d) my method doesn’t consider the possibility of sensor malfunction.<\/p>\n I find it interesting the average Blackboard bet corresponds to a value just a hair below last year’s 7 day minimum, particularly since the ice extent was rather low near the time bets closed. <\/p>\n If you would like to see your bets, visit: the betting page<\/a> which now displays the bets. I’ll be fixing iv0p’s bet for him. If your bet is missing or entered in error in a way that is pretty obvious, let me know. (I don’t know if the bet for 8.69 millions sq. km is ‘obviously wrong’, but iv0p’s is obviously wrong.)<\/p>\n I’ll try to post again mid week until the contest closes. Because less “weather” can happen in 4 weeks compared to 8, and the orange error bars are based on the most recent observation, the error bars should<\/I> tighten over time. Within a few weeks we may be declaring some people are “already losers” while other’s will begin to be in contention. Of course, I hope I win. We’ll see! <\/p>\n","protected":false},"excerpt":{"rendered":" I’ve been planning to post updates on Wed., plus “Andrea” is agitating to have us all see that the JAXA NH sea ice extent is no longer below the 2007 minimum. So, today, I bring you the supplemented JAXA NH se ice extent figure: The JAXA data has been supplemented with gsfc.nasa.gov data prior to … Continue reading Wed NH Ice update.<\/span> ![\"\"](\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossRate-500x500.png\" "\\"LossRate\\"")
<\/a><\/p>\n
\nBelow, I’ve posted a scatter plot of 7-day smooth ice extents from the merged files vs. the minima (black circles) for a given year along with a simple linear fit in blue and the \u00c2\u00b195% confidence intervals computed neglecting the uncertaity in the coefficients of the fit(blue and blue dashed):
\n![\"\"](\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/BetsBlackboard-500x500.png\" "\\"BetsBlackboard\\"")
<\/a><\/p>\n
\nAre any of you wondering what the purple lines are? The solid purple line is the average<\/I> bet by Blackboard bettors computed by utterly ignoring Dr. Jay\/Shoosh’s bet (which exceeds the surface area of the earth) and fixing iv0p’s error in decimal place. The dashed lines contain 95% of the bets placed. (The full range was 3.872 to 8.69 million sq. km). <\/p>\n