In mid-June through early July, participants on the Arctic Sea Ice (ASI) blog posted 82 individual predictions for the mean NSIDC September Arctic sea ice extent. The median value of these 82 predictions was 3.2 million km2, with an interquartile range (approximately the middle 50% of predictions) from 2.7 to 3.9 million km2.
Medians and quartiles are employed as basic summaries here because they offer high resistance to outliers, are easy for anyone to replicate, and map into more detailed analysis using graphics or quantile regression. Individual predictions ranged from 0 to 5.6 million km2. A stem-and-leaf display compactly shows the distribution:
(Predicted September mean sea ice extent, million km^2)
asi_blog rounded to nearest multiple of .1
plot in units of .1
0* | 03
0. | 89
1* | 1
1. | 7888
2* | 0002233
2. | 555678888888889999
3* | 000001222333334
3. | 5566888888889
4* | 0011233333444
4. | 55557
5* | 0
5. | 6
Grad student Matthew Cutler assisted in coding these June–July responses. Following preferences expressed by blog participants, we sent a summary and explanation of our results to the SEARCH Sea Ice Outlook (SIO) for their July report. Both the SEARCH SIO July report, and a link to the pdf file we submitted, can be seen here. This SIO round drew a total of 21 contributions, mostly from scientific sources but with a few from interested non-scientists. Our ASI blog prediction of 3.2 was the lowest one SEARCH received. Other SEARCH SIO contributions ranged as high as 5.9, with a median of 4.0 million km2:
Stem-and-leaf plot for sio (SEARCH SIO)
sio rounded to nearest multiple of .1
plot in units of .1
3* | 24
3. | 6778889
4* | 000134
4. | 677
5* | 01
5. | 9
Figure 2 compares ASI blog predictions with those compiled by SEARCH for their June and July SIO reports (which include our ASI prediction as one point in July). Box plots show median and approximate interquartile range (middle 50%) of each distribution. The ASI blog predictions are more diverse and more pessimistic, although we see a hint of convergence. Between June and July SEARCH SIO reports, the ASI median rose from 3.0 to 3.2, while the SIO median declined from 4.1 to 4.0.
SEARCH classifies SIO contributions according to type of method: modeling, statistical or heuristic. Figure 3 graphs the distributions of June and July 2013 SIO contributions by method. The overall July SIO median, 4.0 million km2, is marked by a horizontal line across the graph. Modeling and statistical predictions appear fairly stable from June to July reports, with the statistical predictions generally more pessimistic (e.g., median 3.90 compared with 4.35 for modeling in July). Heuristic predictions shifted downwards, reflecting a changing pool of contributors in June and July.
The third and final SEARCH SIO deadline for this year will be in early August. Once again ASI blog participants are invited to submit, as comments to this post, your best guess for the mean September extent of Arctic sea ice (NSIDC). Your numerical prediction should be in the first line of the comment, followed by at least a sentence or two explaining the basis for your prediction — whether pure intuition, elaborate calculations, whatever you’ve got. Predictions are not bound by what you did or did not guess last month. As always, general discussion is welcome too.
Results from this crowd-source experiment will be research data in their own right. Individuals can submit their own predictions directly to SEARCH, of course. The analysis here will focus on collective skill and uncertainty rather than individual. I will summarize the results after each cycle, possibly forming the basis for a research paper as well as future blog posts. The whole process should be completely transparent, because the raw data — predictions submitted here as comments — remain public and accessible to anyone.
The September minimums for the last 7 years (in millions km2, found here):