September predictions challenge 2019

[Paddy]

And it's back! Many thanks to Richard Rathbone for setting this up last year. Rules copied over from last year's thread:

This is a challenge which accumulates scores across all the monthly polls on September sea ice to come up with an overall rating of how good the predictions made are. In addition to making a prediction, entrants are required to rate their confidence in that prediction. The higher the confidence, the narrower the margin of error you are allowed, but the higher score you get if the September ice ends up where you predicted it (and the bigger penalty you take if you miss).

Currently included polls are: JAXA daily minimum area, NSIDC September average.
Other polls may be added during the course of the season if their structure fits this challenge.

Points are scored as follows:
Very High Confidence: 10 points if you pick the correct bin, -10 points for all other bins.
High Confidence: 6 points for the correct bin, 2 points if one bin out, -2 points if two bins out, -6 points for all other bins
Medium Confidence: 4 points for the correct bin, 2 points if one bin out, 1 point if two bins out, -1 if three bins out, -2 if four bins out, -4 points for all other bins
Low Confidence: 2 points for the correct bin, 1 point if within 3 bins, -1 point if 4-6 bins out, -2 points for all other bins
Very Low Confidence: 1 point if in the correct bin, no score (or penalty) for any other bin.

Note on portmanteau and end of range bins: these are excluded from the challenge, you may either select a bin within the portmanteau range, or extend the range beyond the poll endpoint for your challenge entry. If the September values fall out of the normal range, scores will be assessed by extending the bin structure. e.g. an end result of 5.9 counts as two bins out for a 5-5.5 entry.

To enter, post guess and confidence in this thread before the closing date of the poll. Editing a post to change the prediction before the closing date is allowed, editing for any reason after the closing date for a poll will result in disqualification from the challenge.

List of entries

Paddy

June
JAXA:  3.75 - 4.25, medium
NSIDC: 4.25 - 4.75, medium

Oren

June
JAXA: 3.5 - 4, medium
NSIDC: 4 - 4.5, medium

b_lumencraft

June
JAXA: 3.5 - 4, high
NSIDC: 3.75 - 4.25, high

Stephan

June
JAXA: 3.75 - 4.25, high
NSIDC:  4 - 4.5, high

jdallen

June
JAXA: 3.75 - 4.25, high
NSIDC: 4 - 4.5, high

Neven

June

JAXA: 3.5 to 4.0, medium
NSIDC: 4.0 to 4.5, medium


Juan C. García

June

JAXA: 3.5 to 4.0, médium
NSIDC: 3.75 to 4.25, médium

Richard Rathbone

June

Jaxa: 3.75-4.25, Medium
NSIDC 4.0 - 4.5, Medium

Sterks

June

Jaxa: 3.75-4.25, Medium
NSIDC 4.0 - 4.5, Medium

Aluminium

June

JAXA: 3.5-4.0 high.
NSIDC: 4.0-4.5 medium.

Brigantine

June

JAXA 3.75 - 4.25 medium
NSIDC 4.25 - 4.75 medium

slow wing

June

JAXA: 3.75 to 4.25, medium
NSIDC: 4.00 to 4.50, medium

jplotinus

June

JAXA: 4.25-4.75, low
NSIDC: 4.50-5.00, low

Gerontocrat

June

JAXA: 3.75 to 4.25, High
NSIDC: 4.00 to 4.50, High

Steven

June

JAXA: 3.5 to 4.0, medium
NSIDC: 4.0 to 4.5, medium

[oren]

Two questions:
A. I take it the bin size is 0.25 million km²?
B. Must the guess here be the same as on the poll thread? I assume not.

[Paddy]

Two questions:
A. I take it the bin size is 0.25 million km²?
B. Must the guess here be the same as on the poll thread? I assume not.

A. The bin you're guessing at is 0.5m km2 wide, but each step out is measured in 0.25m km2 increments.

So if you guess at 4 to 4.5, and the final extent is 4.9, you count as being 2 bins out. EDIT: If you had gone for Very High Confidence, you would get -10 points. If High, -2 points. If Medium or Low, 1 point. If Very Low, 0 points.

B. No, it does not have to be the same.

EDIT: My June predictions are as follows:
JAXA: 3.75 to 4.25, medium
NSIDC: 4.25 to 4.75, medium

[oren]

My June prediction: JAXA 3.5-4, medium confidence.

[b_lumenkraft]

My June prediction:

JAXA  : 3.50 - 4.00, high confidence
NSIDC : 3.75 - 4.25, high confidence

[Stephan]

My prediction (JAXA Sep minimum): 3.75-4,25 M km², high confidence
My prediction (NSIDC Sep Average): 4.00-4.50 M km², high confidence

[jdallen]

My prediction (JAXA Sep minimum): 3.75-4,25 M km², high confidence
My prediction (NSIDC Sep Average): 4.00-4.50 M km², high confidence

Ditto.

[Neven]

My June predictions are as follows:
JAXA: 3.5 to 4.0, medium
NSIDC: 4.0 to 4.5, medium

[oren]

I will copy Neven's take on NSIDC, 4-4.5 medium.

[Juan C. García]

My June predictions:

JAXA: 3.5 to 4.0, médium confidence
NSIDC: 3.75 to 4.25, médium confidence

[Richard Rathbone]

Jaxa: 3.75-4.25, Medium
NSIDC 4.0 - 4.5, Medium

I was intending to run this again, but got distracted by a friend's death just before the polls went up (I have reached the age where such things can be expected to happen). Thanks for setting it rolling Paddy.

The one change I was intending to make was to ask for entries in subsequent months to quote entries from the previous month(s) but that's not relevant this month anyway.

[b_lumenkraft]

Sorry for your loss Richard.

[Sterks]

Jaxa: 3.75-4.25, Medium
NSIDC 4.0 - 4.5, Medium

Same, but thinking will be probably around the lower end. Feels like this year can beat 2016 (2nd or 3rd) easily if picks some more melting momentum before solstice

[Steven]

To enter, post guess and confidence in this thread before the closing date of the poll.

When is the deadline for June predictions in this thread?  Last year the deadline was 16 June, 23:59 UTC, is it the same this year?

[Paddy]

Sure, why not.

[Aluminium]

JAXA: 3.5-4.0 high.
NSIDC: 4.0-4.5 medium.

[Brigantine]

June 2019

JAXA: 4.00 - 4.50 medium confidence
NSIDC: 4.25 - 4.75 medium confidence

[Brigantine]

To enter, post guess and confidence in this thread before the closing date of the poll.

When is the deadline for June predictions in this thread?  Last year the deadline was 16 June, 23:59 UTC, is it the same this year?

For context, the polls last year closed:
JAXA: 2018-06-15 23h52m49s UTC
NSIDC: 2018-06-12 22h10m19s UTC

So basically the deadline for predictions was 24 hours after the later of the polls closed.

This year:
JAXA: 2019-06-11 18h24m20s UTC
NSIDC: 2019-06-11 19h45m19s UTC

I only posted in this thread some hours after the polls closed, but linked to predictions (including confidence level) I posted in the poll threads. Those posts were last edited on June 8. Please let my prediction qualify! 

[slow wing]

June

JAXA: 3.75 to 4.25, medium
NSIDC: 4.00 to 4.50, medium

[Paddy]

Closing tomorrow, last chance time to submit or change your prediction

[jplotinus]

I had guessed there might be a June stall in extent decline, something that is now occurring. Therefore a  “high” bin was selected, which is only “high” in context of the new normal, relative to earlier decades of the satellite measurement era:

4.25-4.75 jaxa
4.50-5.00
Oh, sorry about confidence level omission. I say “low” confidence because the current stall may not last much longer and because July and August weather metrics are not foreseeable, other than temps should be above 0°.

[Brigantine]

4.25-4.75 jaxa
4.50-5.00 nsidc

OK. But what confidence level?

Since there's still time, I'm revising my JAXA prediction:

June:
JAXA 3.75 - 4.25 medium confidence (down from 4.00 - 4.50 medium confidence last week)
NSIDC 4.25 - 4.75 medium confidence (no change)

Things progressed more than expected in the last week! (though not extent data)

[Paddy]

Casual ruling: if you don't submit a confidence level, I will assume medium.

[gerontocrat]

Another hostage to fortune..

JAXA Daily minimum 3.75 to 4.25 million km2 - High Confidence
NSIDC September average 4.00 to 4.50 million km2 - High Confidence

I am going for a relatively small difference between the daily and monthly figures. This is because
- it looks like the slope to minimum is usually shallower than the increase afterwards,
- and I read somewhere (I think) that a study said their model trended to later minimum dates in the future.
This would tend to bring a September average closer to the daily minimum (I think).

What if the weather was so weird it pushed the minimum date to the 1st October?
Earliest minimum 7 Sept 2016,
Latest minimum 21 Sep 2018.
Average 13 Sep. 

[Steven]

JAXA minimum extent:       3.5 to 4.0,  medium confidence
NSIDC September extent: 4.0 to 4.5,  medium confidence

This is based on linear regression, using the NSIDC sea ice area in mid-June as a predictor for the minimum/September extent.

[Paddy]

To enter, post guess and confidence in this thread before the closing date of the poll.

When is the deadline for June predictions in this thread?  Last year the deadline was 16 June, 23:59 UTC, is it the same this year?

For context, the polls last year closed:
JAXA: 2018-06-15 23h52m49s UTC
NSIDC: 2018-06-12 22h10m19s UTC

So basically the deadline for predictions was 24 hours after the later of the polls closed.

This year:
JAXA: 2019-06-11 18h24m20s UTC
NSIDC: 2019-06-11 19h45m19s UTC

I only posted in this thread some hours after the polls closed, but linked to predictions (including confidence level) I posted in the poll threads. Those posts were last edited on June 8. Please let my prediction qualify! 

Based on what's been said before, I'll be closing this poll tonight, at 23:59 UTC. It's a bit later than precedent from last year, but I'm not too fussed.

[b_lumenkraft]

Last chance!!

Thanks, Paddy for the poll.

[Paddy]

And we're open for July entries now.  Put your guesses in anytime between now and 23:59 15/7/19 UTC.

[jdallen]

My only change is to drop JAXA  one bucket to 3.5-4.0

[Paddy]

June estimates all added to the original post.

Richard, I somehow missed your post before going back through them to update the OP.  I'm also very sorry for your loss.  Also, the suggestion you made that people should state the month's estimate they're giving when making a post is a very good one, and I'd be grateful if everyone did that. 

EDIT: Incidentally, I'm waiting until we're a bit closer to the deadline before making my own guesstimate this month, especially as we don't yet have the PIOMAS data.

[Richard Rathbone]

Jaxa: 3.75-4.25, Medium
NSIDC 4.0 - 4.5, Medium

Jaxa 3.25-3.75, M
NSIDC 3.5 - 4.0, M

Fairly spectacularly on the volume cliff, decent on the area cliff, but my preferred momentum indicator for June is the Schroeder melt pond anomaly and usually I don't see that till SIPN comes out which is too late. Having dropped 2 bins, and without my top indicator for the time of year, I'm sticking at Medium.

[oren]

For July:
Dropping JAXA to 3.25-3.75 and increasing conf to High. I think a finish above 4M has become quite improbable, and below 3M extremely difficult.
Also dropping NSIDC to 3.75-4.25, keeping Medium conf.

[Brigantine]

June:
JAXA 3.75 - 4.25 medium confidence (down from 4.00 - 4.50 medium confidence last week)
NSIDC 4.25 - 4.75 medium confidence (no change)

July: (provisional)
JAXA 3.75 - 4.25 3.50 - 4.00 high confidence
NSIDC 4.25 - 4.75 4.00 - 4.50 high confidence

Need to do some more thinking about downside risks before the deadline... but here's my back of the envelope backup prediction.

[Juan C. García]

My June predictions:

JAXA: 3.5 to 4.0, médium confidence
NSIDC: 3.75 to 4.25, médium confidence

My July predictions:

JAXA: 3.25 to 3.75, high confidence
NSIDC: 3.75 to 4.25, high confidence

[Paddy]

My June predictions are as follows:
JAXA: 3.75 to 4.25, medium
NSIDC: 4.25 to 4.75, medium

My July predictions are as follows:
JAXA: 3.75 to 4.25, medium
NSIDC: 4.0 to 4.5, medium

In terms of risk-reward, I'd be tempted to increase my confidence levels to High, but I think there's a small but significant chance that ice levels could end up quite a bit lower than these estimates.

[gerontocrat]

Dropped both one bin c.f. June. PIOMAS Volume, area and extent data all seem to point that way. So,
JAXA daily extent minimum 3.5 to 4.0                 Confidence - High**
NSIDC extent sep monthly  average 4.0 to 4.5.    Confidence - High**

** I lied

[Neven]

My June predictions are as follows:
JAXA: 3.5 to 4.0, medium
NSIDC: 4.0 to 4.5, medium

For July, it's the following, dropping two bins:

JAXA: 3.0 to 3.5, medium
NSIDC: 3.5 to 4.0, high

[Viggy]

I just cannot see this year not being a record anymore.

JAXA - 2.25 - 2.75 (High)
NSIDC - 2.75 - 3.25 (High)

[magnamentis]

I just cannot see this year not being a record anymore.

JAXA - 2.25 - 2.75 (High)
NSIDC - 2.75 - 3.25 (High)

if 2019 will reach a >300km2 gap in early august it's quite probable because then the GAC has been pre-compensated. if they reach the second week of august head-on, chance are high that 2012 will make the race one last time

[Stephan]

I stay with my guesses I made around a month ago and I hope extent will not drop further than:
JAXA Sep minimum: 3.75-4.25 M km², high confidence
NSIDC Sep Average: 4.00-4.50 M km², high confidence

[Paddy]

Droping my JAXA bin a half step further to 3.5 - 4.0.  Still Medium confidence.

[charles_oil]

Jaxa 3.25 - 3.75    Medium
NSIDC 3.5 - 4     High

[Viggy]

I just cannot see this year not being a record anymore.

JAXA - 2.25 - 2.75 (High)
NSIDC - 2.75 - 3.25 (High)

if 2019 will reach a >300km2 gap in early august it's quite probable because then the GAC has been pre-compensated. if they reach the second week of august head-on, chance are high that 2012 will make the race one last time

We are already 157k lower than 2012. I dont know if a 300k head start is enough since 2012 lost over a million km² at the start of August. But, I am also expecting 2019 to be potentially over 500k lower than 2012 in early August.

[Aluminium]

JAXA: 3.5-4.0 high.
NSIDC: 4.0-4.5 medium.

JAXA: 3.5-4.0 high.
NSIDC: 3.75-4.25 high.

[slow wing]

Missed the cutoff time again, sorry.

June

JAXA: 3.75 to 4.25, medium
NSIDC: 4.00 to 4.50, medium

July

JAXA: 3.25 to 3.75, medium
NSIDC: 3.50 to 4.00, medium

[Steven]

JAXA minimum extent:       3.5 to 4.0,  medium confidence
NSIDC September extent: 4.0 to 4.5,  medium confidence

This is based on linear regression, using the NSIDC sea ice area in mid-June as a predictor for the minimum/September extent.

July:

JAXA:     3.25 to 3.75,  medium
NSIDC:  3.75 to 4.25,  medium

This is based on linear regression, as before.  I was tempted to choose high confidence rather than medium, but I'm not willing to take the risk.

[Steven]

Missed the cutoff time again

Not really.  The deadline for July predictions in this thread is tomorrow, 15 July at 23:59 UTC, as announced by Paddy upthread.

[Richard Rathbone]

If SIPN entries had been made here: (these are a late commentary on selected June entries, not July, which is scheduled for release next week)

CPOM (Schroeder, based on May melt pond statistics) mean 4.3 sd 0.5 which is probably 4-4.5 M, but not far short of H.

UK met office (dynamic model, seasonal forecaster using ensembles from May conditions) mean 5.2 sd 0.3 which is 5-5.5 H.

US Navy (dynamic model, often referred to on these boards as HYCOM) 5.2-6.3 ensemble range, which is 5.5-6.0 M (would be H but they say the ensemble range is an under estimate in their submission)

NISDC expert group guesstimates mean 4.4 sd 0.33, which is 4.25-4.75 M (rather than H on the basis that the sd is just the variation between experts, not how confident the experts are)

NSIDC Boulder (statistics with a Slater like approach but not the actual Slater model) Mean 5.59, 95% range 3.66-7.97,  which is 5.25-5.75 L (maybe even VL)

Nico Sun a.k.a. Tealight, (statistics on albedo calculations) median 4.4 range 3.73-4.88, which is 4-4.5 H.

University of Washington (dynamic model, PIOMAS forced by a seasonal weather forecast) 4.07 +- 0.40, which is 3.75-4.25 H (might really be M but they aren't clear what the 0.4 actually is and I'm treating it as total uncertainty rather than a single standard deviation)

I expect there to be an update using June melt pond data from Schroeder but a lot of the others just make the one SIPN entry, though it would be interesting to see how an extra month of data changed the models.

[Richard Rathbone]

If SIPN entries had been made here, July edition, hot off the press.

CPOM (Schroeder, based on June melt pond statistics) mean 4.1 sd 0.5 which is probably 3.75-4.25 M, but not far short of H. Only drops 0.2 from June entry, despite record June melt pond fraction.

UK met office (dynamic model, seasonal forecaster using ensembles from June conditions) mean 5.2 sd 0.3 which is 5-5.5 H. Suspiciously similar to the previous entry.

US Navy (dynamic model, often referred to on these boards as HYCOM) 3.1-4.1 ensemble range, which is 3.25-3.75 M (would be H but they say the ensemble range is an under estimate in their submission) Massive drop from their June entry.

NSIDC Slater model 4.46, no uncertainty quantified, so 4.25-4.75 VL.

Nico Sun a.k.a. Tealight, (statistics on albedo calculations) median 3.95 range 3.61-4.17, which is 3.75-4.25 H. Down a notch, range shrunk quite a bit but stays at H because it doesn't fit the contest ranges neatly.

University of Washington (dynamic model, PIOMAS forced by a seasonal weather forecast) 4.26 +- 0.40, which is 4.0-4.5 H (might really be M but they aren't clear what the 0.4 actually is and I'm treating it as total uncertainty rather than a single standard deviation) Up a notch when almost everyone else is dropping.

[DrTskoul]

We will see how statistical models fare....