NSIDC 2014 Arctic SIE September minimum: August poll

[Neven]

ATTENTION: There are two polls on the ASIF. One is for NSIDC sea ice extent monthly/September average minimum, the other is for Cryosphere Today sea ice area daily minimum. Make sure you are aware of the difference before voting. You can discuss various extent/area data sets in this dedicated thread.

-----

This NSIDC extent poll will run for 8 days (until August 15th). Until then you can change your vote. This is the last poll for this melting season.

Here's how things are currently looking based on data up to August 6th:



These are the September minimums for the last 8 years (in millions km2, found here):

    2005: 5.57
    2006: 5.92
    2007: 4.30
    2008: 4.73
    2009: 5.39
    2010: 4.93
    2011: 4.63
    2012: 3.63
    2013: 5.35

The average for the June poll was 3.95 million km2.

The average for the July poll was 4.17 million km2.

You can use the comment thread below to motivate your choice, but discuss various SIE/SIA data sets in this dedicated thread.

[Neven]

PS I've added 2 bins at the top of the poll, something I forgot to do last month.

[OSweetMrMath]

My SARIMA time series forecast for July was 8.4 million sq km. The actual value was 8.25 sq km. This reduces my forecast for September to 4.7 million sq km, with 95% confidence interval of 4.2-5.4. This is slightly lower than last month's forecast of 4.8, so my forecast has shifted back into the 4.5-4.75 bin. Although there have been small variations month to month, my forecast has been essentially unchanged since the end of April.

Note 1: Extent loss at the beginning of July was relatively high, but extent loss toward the end of the month was relatively low. This means that the July average extent was low, but the extent at the end of the month was relatively high. Consequently, unless the August loss is greater than average, my current forecast will be too low.

Note 2: I use the same methodology for my NSIDC extent and my PIOMAS volume forecasts. However, they are independent, so the volume data does not enter into my extent forecast or vice versa. In the real world, however, we would expect a relationship between extent and volume.

My volume forecast for July badly missed, resulting in a major revision of my September volume forecast. However, my extent forecast for July was accurate. I do not have a sense of whether these two outcomes are consistent, or if I should expect the two forecasts to converge in the next few months, most likely by the actual extent coming in higher than my forecasts.

Note 3: Up until this month, my extent and volume forecasts for September have been consistently higher than the average poll values. Following the observations from July, the average poll values, at least so far, this month are much higher than previous predictions. My PIOMAS forecast has risen even more than the average, so my volume forecast is still higher than the poll average. On the other hand, my extent forecast is unchanged, so my extent forecast is lower, for the first time, than the poll average. Again, I do not know if there is any significance to this.

Note 4: Until this month, my extent forecast has been at the median forecast value of the SEARCH Sea Ice Outlook poll. I am very interested in whether this will continue to be true this month.

[Neven]

My June vote: Between 4.75 and 5.0 million km2
My July vote: Between 4.5 and 4.75 million km2 (expecting more of sunny weather than came about)

If it weren't for the current sunny weather that might still cause quite a bit of in situ melt, I'd go up 3 bins. As it is, I'll take a gamble and go back to my June vote: Between 4.75 and 5.0 million km2

[DavidR]

I have been consistently low (3-3.25), however the model I  am using puts the July melt as being far less than any of the last three years. As a consequence I  have moved up several notches to 4.75 to 5.0.

[Michael Hauber]

5.5 - 5.75

There has been a good run of strong melting weather and we are still behind 2013.  Assuming the weather turns poor again at some stage we will have more ice than 2013 at minimum.

My guess from July was 4-4.25 I think.

oops.

I thought that the open water in Laptev and Beaufort would overcome the apparent lack of melt ponds/open water within the ice pack, but it didn't.

[OSweetMrMath]

I assume there will not be a September poll, so I will put my update here. My prediction based on the July data was that the minimum would be 4.7 million sq. km, with 95% confidence interval of 4.1-5.4 million sq. km. However, at the time I acknowledged that the prediction for the August volume was likely too low, so the predicted September volume was also likely too low.

The actual August average was 6.22, which was just outside of my August confidence interval of 5.1-6.2. This updates the prediction for the September average to 5.1, with CI of 4.6-5.6. The point prediction seems too low, especially since this year's August average was above last year's August average, but this prediction is below last year's value for September. In addition, the CI is unnecessarily wide, as the ice extent is already below 5.6 million sq. km. Regardless, these are the current predictions for the time series model.

[crandles]

The monthly average extent figure isn't the average of the NSIDC daily values. Instead the 15% test is applied over the whole month. Nevertheless the NSIDC daily data seems most likely to be most comparable data set.

Year,  AvgSept, 1Sept, Diff
2007 4.30, 4.48236, -0.18236
2008 4.73, 4.83767, -0.10767
2009 5.39, 5.36779, +0.02221
2010 4.93, 5.13073, -0.20073
2011 4.63, 4.59970, +0.03030
2012 3.63, 3.58238, +0.04762
2013 5.35, 5.28968, +0.06032

average change over last 7 year = fall of 0.047
2014,?,5.38823

So I don't see why we shouldn't now expect monthly Sept average to be about 5.34 M Km^2
A sensible range would appear to be about +/- 0.15

(As can be seen the September average figure can be higher or lower than the NSIDC daily value on 1 Sept.)

[jdallen]

<snippage>
So I don't see why we shouldn't now expect monthly Sept average to be about 5.34 M Km^2
A sensible range would appear to be about +/- 0.15
<more snippage>

I'd disagree, for three reasons, one concrete, two hypothetical but reasonable.

1) We're already below the 5.34M KM2 value NSIDC.
2) Current conditions suggest we'll still see a further significant drop in NSIDC values
3) I consider it highly unlikely conditions will turn around fast enough for a refreeze significant enough to pull the average back to your target.

I've a hard time seeing the NSIDC average values falling in the 5.29-5.49 range.

[crandles]

1) We're already below the 5.34M KM2 value NSIDC.
2) Current conditions suggest we'll still see a further significant drop in NSIDC values

"Current conditions" if not considered typical is certainly a good reason if you believe that. I probably didn't look hard enough for this. What factors are you considering to assert current condition are atypical on the melt side? I would say temperatures are normal for time of year and we are well into bottom melt almost exclusively season. If area has receded quite rapidly recently it occurs to me that doesn't bode well for there being more warm water than usual near the remaining ice.

The latest NSIDC value that I see is 5.38823 as in the post. I am allowing for a net fall from current position. Of course there will be a larger fall but then in late September the extent increases again quite rapidly to higher than at the start of September.

[OSweetMrMath]

I had not previously noticed that the September monthly extent could be higher than the September 1st daily extent, although it does make sense. The time series I've constructed is based on year over year changes and misses weather effects. If last month's year over year change is larger or smaller than predicted, that affects the predicted change for this month, but the prediction doesn't "know" how high the observed August data is, so the September prediction isn't really adjusted appropriately.

[OSweetMrMath]

The data for the September monthly average NSIDC extent is in. The value is 5.28 million sq km. In the August poll, 31 people voted for lower values, 2 people voted for higher values, and 11 people voted for what was ultimately the correct range, 5.25-5.5 million sq km.

At the start of the season, I posted 7 different statistical prediction for the ice extent (see http://forum.arctic-sea-ice.net/index.php/topic,884.msg27385.html#msg27385) and I am pleased to report that they did quite well. The actual ice extent is within the 95% confidence interval of all of the predictions for which I computed confidence intervals. The single best predictor was based on the average year over year change, which I underrated at the time as likely to be too high.

In addition to the predictions at the start of the season, I have updated my prediction based on a monthly time series model each month. These predictions were low (4.7-4.8 ) until the August data came in, when the prediction jumped to 5.1 million sq km. The true value was within the confidence interval for the prediction every month, so the time series has performed well. The only drawback to the time series prediction is that, like my other statistical predictions, the confidence interval is too large to be particularly useful.

That said, we can now start looking forward to the maximum in March. My prediction for the March extent is 14.8 million sq km, essentially the same as this year's value of 14.76 million sq km. The 95% CI is 14.1-15.5 million sq km.

[viddaloo]

Just FYI, my latest posted 2014 prognosis based on the Five Year Cycle (5YC) ends the year at record low of 12.3 million km2, and the March 2015 maximum will be at about 14.7.