{"id":16719,"date":"2011-08-30T10:38:57","date_gmt":"2011-08-30T16:38:57","guid":{"rendered":"http:\/\/rankexploits.com\/musings\/?p=16719"},"modified":"2011-08-30T10:38:57","modified_gmt":"2011-08-30T16:38:57","slug":"nsdic-prediction-submission-to-search","status":"publish","type":"post","link":"https:\/\/rankexploits.com\/musings\/2011\/nsdic-prediction-submission-to-search\/","title":{"rendered":"NSDIC Prediction: Submission to SEARCH"},"content":{"rendered":"<p>In comments, crandles asked if I was going to submit my forecasts of NH Ice Extent to <a href=\"http:\/\/www.arcus.org\/search\/seaiceoutlook\/2011\/august\">SEARCH<\/a>. Though I usually discussed forecasting the 7-day JAXA average (on which we bet quatloos) I can easily tweak this for the NSDIC September NH Extent Average. I did so in August and submitted.  I was then surprised to learn they also solicit forecasts for their September report.  The forecasts were due today, which is too early to know the NSDIC August report.  But, no matter;  sent mine in making projections based on JAXA, CRYOSPHERE and PIOMAS data.  I thought some of you might want to read first sentence in my most recent submission: <\/p>\n<blockquote><p>1.\tExtent Projection: 4.47 million square kilometers  September NH sea ice extent minimum. <\/p><\/blockquote>\n<p>So, I&#8217;m predicting a lower extent than I did a month ago when the first sentence in my submission read<\/p>\n<blockquote><p>1.\tExtent Projection: 4.57 million square kilometers  September NH sea ice extent minimum. <\/p><\/blockquote>\n<p><!--more--><br \/>\nGiven the size of my uncertainty intervals, the people at Search made the reasonable decision rounded that to 4.6 when displaying the range of submissions they received:<\/p>\n<p><a href=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1-500x385.png\" alt=\"\" title=\"sio_august_fig1\" width=\"500\" height=\"385\" class=\"aligncenter size-medium wp-image-16720\" srcset=\"https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1-500x385.png 500w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1-300x231.png 300w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1-1024x789.png 1024w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/sio_august_fig1.png 1719w\" sizes=\"auto, (max-width: 500px) 100vw, 500px\" \/><\/a><\/p>\n<p>Presumably, the submitted forecasts will tend to cluster as we get closer to the minimum.  We&#8217;ll have to wait so see if they cluster sufficiently for SEARCH to post 3 digits.<\/p>\n<p>Interestingly enough, my first submission was based on a method that extrapolated based on extent only.  If I were to use that method today, I would be forecasting 4.58 millions square kilometers.  With the most recent reported values for regression parameters I use, that particular regression happened to provide the best single regression between NSDIC extent and any predictive variable I tried.  <\/p>\n<p>However, as many know, I&#8217;ve been folding in other regressors and switched to a method based on weighting multiple candidates regressions.  Today, the candidates that filtered to the top based on their low corrected Akaike criteria (AICc) were:<\/p>\n<ul>\n<li>NSDIC extent v(current JAXA Extent,current CT Area) w=58.4%<\/li>\n<li>(NSDIC extent- current JAXA Extent) v July 31, PIOMAS Volume  w=6.4%<\/li>\n<li>(NSDIC extent- current JAXA Extent) v current CT Area. w=35.2%<\/li>\n<\/ul>\n<p>I have to admit to finding the latter two a bit odd, but if I&#8217;m using a &#8220;method&#8221;, I think I have to stick with &#8220;the method&#8221;.   Afterwards, when things like &#8220;volume&#8221; or &#8220;area&#8221; pop-out as strong regressors, I can try to ponder whether I think there might be a physical reasons or whether I think these are acting as proxies for something important not included in the regression.  (FWIW: Speculating, I <i>can<\/I> think of reasons why high area could help reduce ice losses. But, I was expecting (extent-area) to matter more.)<\/p>\n<p>For now I&#8217;ll observe that it barely matters what we include in the regression: We are nearing the minimum and uncertainty intervals are narrowing. If I define &#8220;loss&#8221; as (NSDIC extent- current JAXA Extent), based on years 1979-2010, the mean &#8220;loss&#8221; between the current 7-day JAXA value and the NSDIC September extent is only 0.13 million square kilometers.  (For those eyeballing the JAXA graph thinking this sounds awfully small&#8211; it  should look small. This the Sept. NSDIC NH extent mean is generally&#8211; thought not always&#8211;higher than the minimum based on the JAXA 7-day average. This is not too surprising since the September averages includes 21 days with mostly extents larger than those that contribute to the 7-day average minimum.)  <\/p>\n<p>It turns out that if we regress this loss against Cryosphere today 7-day average area, we predict a loss of 0.53 million square kilometers.  While the difference looks large, even achieving a record loss doesn&#8217;t result in a huge difference in what I predict. For those wondering about how &#8220;loss&#8221; varies with area, here&#8217;s the graph:<\/p>\n<p><a href=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossArea.png\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossArea-500x500.png\" alt=\"\" title=\"LossArea\" width=\"500\" height=\"500\" class=\"aligncenter size-medium wp-image-16731\" srcset=\"https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossArea-500x500.png 500w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossArea-300x300.png 300w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/LossArea.png 1008w\" sizes=\"auto, (max-width: 500px) 100vw, 500px\" \/><\/a><br \/>\nThe black dot illustrates the &#8216;loss&#8217; predicted based on CT-area for the upcoming month. Notice the prediction is for a near record loss given the short amount of remaining time; this regression is included in my weighted model with a weight of 32%.  So, at least one model that survives the selection process suggests we may see record losses during September.  I can only begin to imagine the chatter at Nevens! (And WUWT, and even here!)<\/p>\n<p><b>Overall prediction<\/b><br \/>\nTurning to overall predictions: My forecast for the NSDIC September Ice Extent is 4.47 millions square km; this value is show along with values from other years below:<\/p>\n<p><a href=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/NSDIC1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/NSDIC1-500x500.png\" alt=\"\" title=\"NSDIC\" width=\"500\" height=\"500\" class=\"aligncenter size-medium wp-image-16739\" srcset=\"https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/NSDIC1-500x500.png 500w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/NSDIC1-300x300.png 300w, https:\/\/rankexploits.com\/musings\/wp-content\/uploads\/2011\/08\/NSDIC1.png 1008w\" sizes=\"auto, (max-width: 500px) 100vw, 500px\" \/><\/a><\/p>\n<p>Last years value is shown as a solid black circle. Note that it lies outside the \u00c2\u00b195% uncertainty intervals based on the weighted model (i.e. outside the slate blue horizontal dashed lines.) Assuming residuals to that fit would be gaussian, the probability of exceeding last  years September extent is estimated at less than 1%.  Meanwhile, the 2007 minimum still lies inside the uncertainty intervals and I estimate the probability of overtaking the minimum is 17.9%. That&#8217;s pretty high!<\/p>\n<p>Tomorrow, I&#8217;ll post on the horserace and tell you who is in the lead for the Quatloos.  That&#8217;s based on JAXA. <\/p>\n","protected":false},"excerpt":{"rendered":"<p>In comments, crandles asked if I was going to submit my forecasts of NH Ice Extent to SEARCH. Though I usually discussed forecasting the 7-day JAXA average (on which we bet quatloos) I can easily tweak this for the NSDIC September NH Extent Average. I did so in August and submitted. I was then surprised &hellip; <a href=\"https:\/\/rankexploits.com\/musings\/2011\/nsdic-prediction-submission-to-search\/\" class=\"more-link\">Continue reading <span class=\"screen-reader-text\">NSDIC Prediction: Submission to SEARCH<\/span> <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[49],"tags":[],"class_list":["post-16719","post","type-post","status-publish","format-standard","hentry","category-betting"],"_links":{"self":[{"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/posts\/16719","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/comments?post=16719"}],"version-history":[{"count":0,"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/posts\/16719\/revisions"}],"wp:attachment":[{"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/media?parent=16719"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/categories?post=16719"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rankexploits.com\/musings\/wp-json\/wp\/v2\/tags?post=16719"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}