A potential Sara now looks like a bust in terms of adding a large amount of ACE given latest ensembles… so hedging once again on a post season downgrade

I've excluded 1970 as that is an extreme outlier (-40 ACE adjustment); some statistics on above:

Mean: -0.57
Median: 0.00
RMSE: 2.51
Standard Deviation: 2.44
Minimum: -7.10
Maximum: 10.56

Looks like post-season downgrades (with respect to ACE) are indeed rare (positive values in graph), with only 2 years out of 50+ years having significant enough downgrades. (Implying a probability of 2-5%)

Given Sara is very likely at this point (and expects to add ~ 6 ACE (or more) from EPS/00Z at this point), this further justifies me selling the remaining amount in Above Normal bin even if it ends up going to post-season (not pushing over the 10 ACE required).

159.4 as of yesterday.

https://tropical.atmos.colostate.edu/Realtime/index.php?loc=northatlantic

@NicoDelon Why post yesterday? Today is 159.8 (now that Rafael is done…)

https://tropical.atmos.colostate.edu/Realtime/

Anyway this will go to post season (next year) most likely still at this point per resolution details…

@parhizj Last I checked the data was from yesterday. Even better if we have today.

147.3 ACE now that Patty is finished

I think more than 10 ACE is unlikely from 18L/future-Rafael (maybe 6-7 ACE) while not expecting a Sara either to develop in the rest of the month (NHC has 20% for another TC in the same area but it seems less likely to me).

Since I wrote that yesterday, the NHC has twice upgraded the advisory forecast for Rafael; additionally, NHC is now favoring a track more along the ECMWF models (more westerly tracks) ; these westerly tracks in EPS favor possibilities of a longer duration spent traversing the GoM instead of making landfall early on.

Unfortunately this will keep this market a virtual coin toss for the next ~ 5 days (unless the RI amount/duration is significantly off in the latest NHC forecast).

The latest advisory brings my estimate using OFCL (interpolating VMax linearly) for Rafael's storm lifetime to ~12 ACE, which would put the season ACE (barring any Sara) at 159.3.

At this point, such an estimate is going to be crude in predicting the result as it will be sensitive to even slight changes in the forecast length, intensity, as well as a result of the interpolation (during the landfall over Cuba).

Back to a potential Sara (as mentioned in the TWO in the SW Atlantic)... if I recall GEFS/12Z was a bit more aggressive in predicting a TS (40%?) than EPS/GEPS which I believe were about half that or lower (yielding an comprehensible 30% in the TWO). It's development does not seem particularly likely to me and the tracks that do develop showed a minimal TS; at this point it seems moot until genesis is nearly certain as the uncertainty in Rafael's ACE should swamp it's contributions, if any, given it looks to be a rather weak system with dim prospects from the genesis tracks. If it does have end up getting higher chances of development, then even some small contributions could tip the amount of ACE if it ends up close.

Based on last couple of advisories Rafael now looks like it will add (roughly) 14 ACE over its lifetime (or higher if it lasts a bit longer in the GoM), which puts it in a fairly exclusive club as far as November storms...

Relevant discussion from last advisory:

"The apparent beginning of an eyewall replacement cycle could ease the recent rate of rapid intensification, but it appears very likely that Rafael will become a major hurricane before it makes landfall in western Cuba later today. Some weakening is forecast when the storm crosses Cuba, but Rafael is likely to remain a hurricane over the southeastern and southern Gulf of Mexico during the next few days. After that time, increasing southwesterly shear and significantly drier air are likely to result in weakening, however a more southerly track over the Gulf could result in less hostile conditions, and there is larger-than-normal uncertainty regarding Rafael's intensity later in the forecast period. "

@parhizj If we end up finishing slightly over the cutoff, hyperactive is still not a lock. Per the market description I'll wait for reanalysis to resolve, and that could potentially cut down the ACE total a little bit if storms are adjusted down.

@SaviorofPlant Based on what I remember so far in the season, I think it's likely there will only be more ACE added not less, but this is just a guess.

More than expected (rapid) weakening in the last day for Rafael has brought down the expected life-time ACE to ~12.7 ACE, which should bring the total to ~160 now.

This again makes it a near coin toss for the (operational) total ACE given the rapid weakening; Barring a Hurricane Sara, it seems it will end up going to the post season but there is still some weeks left for a disturbance to form in the Caribbean and SSTs remain very high in the area.

My guess is to believe it will only increase; the guess of mine is that NHC is rather conservative in its operational upgrades of intensity and classifications, and such cases are going to contribute more as compared to rapid weakening (by frequency and accuracy), or so my reasoning goes.

To get an empirical idea for initial estimate of the uncertainty of the operational vs. post-season distributions I would have to compare operational b-decks against the post-season best tracks to get a starting idea (I expect the the spread would likely be larger and correlated with such an active season as this one though). If I have time I'll try to do so before the market closes.

I'll go out on a limb and say hyperactive is favored at this point.

@ScottSupak I missed all the drama with Oscar intensifying into a hurricane.

Future contributions (up until 2024-10-23 18:00:00) of cyclone ACE values in AL basin (including INVESTs)
    from latest forecast using OFCL (excluding 00h) by interpolating intensities
  ACE(future) for AL15 = 0.28
  ACE(future) for AL16 = 6.88
  Total ACE(future): 7.16

Real-time ACE: 141.9
Total ACE (real-time + forecast) up until 2024-10-23 18:00:00: 149.06

It's a bit iffy to interpolate on OFCL since it's points are 12 hours apart and the intensity is very uncertain but I don't think any of the other models are going to be close to getting Oscar's ACE right. (I was expecting only 2 ACE earlier from the first NHC forecast).

Looks like we'll only need 1 more moderately lived hurricane or a strong, longer-lived tropical storm to push it up into hyperactive, which might come as early as end of the month. As ECMWF extended range seems to be predicting an outsized amount of ACE for ~ early November from such storms I've increased my bet per my climatology notebook that it won't be quiet early November.

My climatology notebook says its likely now. RMSE is about 14 however for the first (exp) model (middle value of 162 (151 - 174)). Simple seasonal calculation conditioning on above past ACE from current real time (15 samples) yields a median of ~176, mean of 181 ACE. Weighting the seasonal higher I get only a 21% chance of it being above normal now.

Will have to substantially revise downwards if the NHC forecast for Oscar does not roughly verify though.

Future contributions (up until 2024-10-13 18:00:00) of cyclone ACE values in AL basin (including INVESTs)
    from latest forecast using IVCN (excluding 00h) by interpolating intensities
  ACE(future) for AL12 = 1.01
  ACE(future) for AL13 = 9.1
  ACE(future) for AL14 = 19.23
  Total ACE(future): 29.34

Real-time ACE: 107.3
Total ACE (real-time + forecast) up until 2024-10-13 18:00:00: 136.64

(Approximate) as they may revise Milton upwards even further...

@parhizj Still an unlikely scenario, but given the very high October activity so far, hyperactive isn't looking impossible anymore.

@SaviorofPlant Of the two predictions I ran I get (139 (exp., with RMSE of 15 to 20), and 152). Weathertiger is slightly above. Given it's supposed to be relatively normal (quiet) for the next month this seems less likely... This is in part the reason I created the Nadine market:

Uhh I have no clue how he gets 100% but this seems wrong... we should be in the neighborhood of ~ 138 ACE in ~ 3 days (when they are forecast to become post/tropical) if the forecasts for Leslie and Milton hold. (Milton adding a whopping ~25 ACE).

Unfortunately for me the higher the ACE the less robust my model gets as it is based on climatology (+ forecast ACE for current storms) and the samples that are useful are starting to dwindle...

I've created a relevant question for TD12 as it seems to me it will impact this question's probabilities as it develops:

Currently at ~ 77 ACE ... haven't run the models for Joyce and Isaac but based on analysis, they'lll add 1 or 2 ACE more from the current total (both are forecast to become non-factors as Isaac is soon to become post/extratropical and Joyce is weakening).

With TD12 we can expect much more...

OFCI from 00Z for TD12 only adds ~ 6 ACE but doesn't cover the full expected life time for the storm..

Likewise for the regional models which only go out to Oct 5th, which give a mean of ~10 ACE.

GEFS 18Z which goes out a couple days further, suggests a mean of about 11 ACE (max 18.6) but also does not seem will cover the full life of the storm, especially if does not recurve along the mean and moves further west...

More dated deterministic (GFS/CMC/ECM) from 12Z suggest ACE of ~13 (max 21 from GFS) with a very speculative GFS going out to the Oct 11th; keep in mind the same GFS has a lower peak intensity (90 kt) for the storm strength compared to the other regional models (upwards of 129 kt and OFCL which also has it become a major hurricane) and if it takes a more westerly track or takes longer to recurve this will both lead it's ACE to be an underestimate...

All of these are still carry a good amount of uncertainty as it is not even a named storm yet.

It looks like it is reasonable to expect at least 10-20 ACE from TD12 (soon to be Kirk) if it develops according to the models, as it will be around for some days in the Atlantic....

Speculatively, I believe 25 ACE might be plausible for Kirk at this point despite it being very early, and this is not accounting for the other disturbances given 50/50%, this suggests above-normal is still under-valued...

Based on 00Z guidance if the high end of these model predictions come true we should end up with above 73 ACE from PTC AL09 (soon to be Helene). (IVCN gives 5 ACE, but HMNI yields ~13 ACE, and this is still lower than the late tracks).

@parhizj Getting to above average hinges on the MDR storm that ensembles are showing generating a lot of ACE. The MJO is imminently moving past the Atlantic and probably won't return until November, so if we still need to make up 40 ACE by like October 7, it becomes very unlikely.

Haven't updated in ~ 10 days. Currently at 61 ACE per Colostate.

For my own trivial climatological model I get a couple estimates starting from today -- point wise ACE ~87 +- 23 (weighting season analogs (exponentially) by ACE at this point of time), and a much simpler point estimate of ~126 conditionalizing on seasons with >= current ACE ). If you assume no named storms for rest of September only the first point estimates drops a tiny bit to 84.

Weathertiger has the 50% value at 128 as of yesterday.

Climatologically (1990-2023) there is upwards of about 70% chance there should be a storm present sometime in the remaining part of September.

For next genesis, TWO puts a TC in the Caribbean/GoM at 40% for next 7 days.

The Helene market (https://manifold.markets/SaviorofPlant/when-will-a-tropical-or-subtropical-8b0q8g76za) is sitting right now at 55% for a TC by the 26th (this may be generous). Speculatively, ACE from the deterministic CMC, GFS 12Z looks to add about 5 to 10 ACE for the Caribbean/GoM disturbance if it develops. High potential heat content right now in the area if develops east of the Yucatan and tracks north:

Above-normal looks too cheap... so I'm betting that up a bit.

A season ACE where we end up < 73 I'd still assign as a very unlikely (maybe ~1%) given how much time is left, so I'm not willing to hedge above the current value of 4%. I've put in a limit order for the rest of September though for 1% but I doubt it will ever reach there given there is not enough bettors.

@parhizj The next week or two will be important in determining whether we end up near-normal or above-normal, since we have the MJO amplifying Atlantic activity.

From IVCN for Isaac, Joyce:

  Total ACE(future): 5.51

Wherein I consider 4 scenarios based loosely on deterministic models and ensemble forecasts:

For rest of these numbers I put model/ensemble ACE for genesis candiates, with model ACE for Isaac, Joyce in parentheses (the difference shows roughly how far off they are from IVCN)

weighted_ensemble_agg_mean (from deterministic ECM,CMC,GFS/12Z: up to +240h,+240,+384):

storm_ace (10^-4): 15.4 (16.5 including Isaac, Joyce)

weighted_ensemble_agg_mean (EPS, up to +10days):

storm_ace (10^-4): 8.5 (12.7 including Isaac, Joyce)

weighted_ensemble_agg_mean: (GEFS up to +5 days):

storm_ace (10^-4): 2.9 (6.7 including Isaac, Joyce)

weighted_ensemble_agg_mean: (GEPS up to +5 days)

storm_ace (10^-4): 0.9 (3.1 including Isaac, Joyce)

If any storms form that become hurricanes it will also likely be in an underestimate for these are global models. In that sense these may likely be low ball numbers.

We are presently at ~ 73.8 ACE .. (combining GEFS+GEPS for 5 days). Relative to this:

+1.9 ACE in 5 days

+8.5 ACE in 10 days

+15.4 ACE in 16 days

(scenario A)

Simply IVCN + real time ACE (~ next 5 days)

73.8 + 5.5 = 79.3 ACE

If we add this in with IVCN for Isaac, Joyce..

(scenario B)

73.8 + 15.4 + 5.5 = 94.7 at October 14.

This scenario doesn't seem too implausible....

(GFS/12Z adds 29 ACE (32.2 ACE including Joyce, Isaac) by itself; this might be an unrealistic high scenario)

(scenario C)

73.8 + 29 + 5.5 = 108.3 at October 14 under this high scenario.

(scenario D) (EPS through October 8 12Z) (this might be the best balance of cautious and speculative)

73.8 + 8.5 + 5.5 = 87.8 ACE

For reference, colostate's graph shows 105.9 as the climatological mean (1991-2022) for October 14.

(scenario A)

Running just the plain IVCN forecast on my simple ACE notebook gives two figures:

(Exponential decay): 89.9 (with an RMSE of perhaps up to 29), and a much simpler climatological estimate

(Simple): 135.5 (using 1996-2023 climatology).

I weight these by (contributing) sample size (20/80% in this run), as the exponentential method seems less reliable given many of the samples end up with miniscule contributions...

So this gives a CDF of ~ 65% of an ACE reaching 127, implying 35% chance of ACE <= 127.

(scenario B)

Same analysis for scenario B) except using the ACE value for October 14, 12Z

Exponential: 102.1 (RMSE: 17)

Simple (realtime): 152

(ACE=126) Weighting the two predictors: (weighting Exp: 0.3, Realtime: 0.7): 74.30%

Implying 26%, ACE <= 127

(scenario C)

Exponential: 127.7 (RMSE: 19)

Simple (realtime): 160.1

(ACE=126) Weighting the two predictors: (weighting Exp: 0.3, Realtime: 0.7): 89.69%

Implying ~10%, ACE <= 127

(scenario D)

Exponential: 97.6 (RMSE: 22)

Simple (realtime): 146.1

(ACE=126) Weighting the two predictors: (weighting Exp: 0.2, Realtime: 0.8): 74.99%

Implying ~25%, ACE <= 127

Weighting based on how far out we are going, first averaging (B), (C) to get 18% (October 14, 16 days out). Then we are left with (D) for October 8 (10 days out), and (A) for October 3 (5 days out). Using a (temporal) inverse weighting based on the days out (0.5514, 0.2759, 0.1727) for (5, 10 , 16) days out (D, A, (B&C)).:

(0.35 * 0.5514) + (0.25 * 0.2759) + (0.18 * 0.1727) ~= 29 % <= 127 ACE

Weathertiger is currently at 135 (eyeballing > 127 at around 55-60%, implying 40-45% <= 127 ACE).

As a check, what does it say about the scenarios for >= ACE value (exceedance) (eyeballing CDF) from my two models:

A) 89.9 ACE: ~100% , 135.5 ACE: ~52%

B) 102.1 ACE: ~100%, 152 ACE: ~9 %

C) 127.7 ACE: 55-60%, 160.1 ACE: ~5 %

D) 97.6 ACE: ~100%,146.1 ACE,: 30-40%

This implies for the exponential model, only (C) plausibly agrees.... for the simple model, only (D) and (A) plausibly agree. (This matches my guess that (D) and (A) seem the most plausible). C is plausible if GFS correctly forecasts a strong storm (a Kirk in the MDR) to match the current environment (SST wise)...

This gives me some confidence that Near-Normal might be presently overvalued at 78% based on the ramp up expected by the models.... If the current forecast genesis for the Central/East Atlantic doesn't come in the next 7 days (currently 70%), I will reassess.

We've bent the curve and reached the climatological mean for 1991-2020...

Francine looks to add ~ a meager 5 ACE.

TWO has next TC 70% in 7 days (I put a TS well below that though around 50%). In short my ACE forecast has dropped a bit into the normal range...