Expert Forecasts for the 2024 North Atlantic Hurricane Season

[ChrisH-UK]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

[zzzh]

TSR forecast will be released on the 8th.

[blp]

Eagerly awaiting the ECMWF forecast which I think will be out tomorrow. That will include October and I suspect a very active W. Caribbean given the trends.

[USTropics]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

[ChrisH-UK]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

Thanks, so 20 times out of 30 they was under with named storms. Also another thing to note every time they they predicted ACE over 160 the actual ACE has been way lower.

[wxman57]

Analogs: 1878, 1926, 1998, 2010, 2020

We thought 1998 was the best analog. I don't know how 1878 and 1926 can be considered, as little upper air data was available back then and SSTs were not as well documented. I told him we were going 21/11/5.

[Blown Away]

Analogs: 1878, 1926, 1998, 2010, 2020

We thought 1998 was the best analog. I don't know how 1878 and 1926 can be considered, as little upper air data was available back then and SSTs were not as well documented. I told him we were going 21/11/5.

Will you post your hot spots for landfalls?

[StPeteMike]

Analogs: 1878, 1926, 1998, 2010, 2020

We thought 1998 was the best analog. I don't know how 1878 and 1926 can be considered, as little upper air data was available back then and SSTs were not as well documented. I told him we were going 21/11/5.

What prevented you guys with going with 2010 or 2020, or even a combination of the two seasons? The breakdown from El Niño to La Niña fairly resembles what developed between Dec 2009 to April 2010, though the transition from El Niño to neutral happened a little faster this year, I believe, than in 2010.

[Teban54]

Is this their first forecast going into the auxiliary list?

A bit pedantic here: it's the highest forecast issued in April, yes. 2020 did have an August forecast of 24/12/5 with 200 ACE, which would have ended at Gamma, but that forecast was issued when the season was already at 9/2/0. This means April 2024 is the second CSU forecast ever that enters the auxiliary list territory (22+ TS).

Meanwhile, the April 2024 forecast (23/11/5 with 210 ACE) having even higher ACE than the August 2020 forecast, despite one fewer TS and H each, is quite telling. This year's forecast calls for 2 more MH days than they did in Aug '20, and 4.25 more MH days that what 2020 actually got. It suggests CSU is expecting storms to be much more long-tracking than 2020.

[Teban54]

Also, this is quite alarming, especially in the face of people saying "things can still change":

The early April forecast is the earliest seasonal forecast issued by Colorado State University and has modest long-term skill when evaluated in hindcast mode. While the skill of this prediction is low, our confidence is higher than normal this year for an early April forecast given how hurricane-favorable the large-scale conditions appear to be.

Furthermore:

We find that 52% of basinwide ACE occurs west of 60°W in El Niño years, while 60% of basinwide ACE occurs west of 60°W in La Niña years. In neutral ENSO years, 59% of basinwide ACE occurs west of 60°W. Given that we are favoring La Niña with this outlook, we are estimating ~60% of basinwide ACE to occur west of 60°W in 2024.

Figure 28 shows the forecast for the next ~2 weeks of low-level winds across the Atlantic. In general, trade winds are forecast to be near to slightly weaker than average, indicating that extremely warm SST anomalies are likely to continue. Overall, the current SST anomaly pattern correlates very well with what is typically seen in active Atlantic hurricane seasons (Figure 29).

[USTropics]

Here are track plots and geodensity plots for CSU's April forecast analog years:

[JetFuel_SE]

Here are track plots and geodensity plots for CSU's April forecast analog years:

https://i.imgur.com/2thPEyo.jpeg

https://i.imgur.com/SfBdOPY.jpeg

I don't like that little blob riiight next to South Florida.
A repeat of the 1926 Miami Hurricane would easily be the most destructive Atlantic hurricane by far, and 1926 is one of the analogs.

[LarryWx]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

https://i.imgur.com/KzUOnHG.png
https://i.imgur.com/5bqZefm.png
https://i.imgur.com/dfrCYsH.png
https://i.imgur.com/wvldsuT.png
https://i.imgur.com/yXx92bm.png
https://i.imgur.com/9DufBL5.png
https://i.imgur.com/nE5GxYk.png

Great stuff, thank you for posting! Very scary!
I see a pretty clear pattern when they’ve gone very active in April:

1. # NS: They’re predicting 23. Prior to this the highest they predicted in April was 17-19 (five times). Four of those five progs ended up too high. For those 5, they averaged 2.8 too high.

2. NS days: They’re predicting 115. Prior to this the highest they predicted in April was 85-95 (four times). All four of those progs ended up too high. The highest actual of those four was only 58 and they averaged a whopping 40.5 too high!

3. # H: They’re predicting 11. The prior highest predicted in April was 9 (six times). All of those 6 progs ended up too high by an average of 3.

4. H days: They’re predicting 45. The prior highest predicted in April were 40-45 (five times). Of these five, four progs came in too high. The five averaged a whopping 19 too high!

5. # MH: They’re predicting 5, which they’ve predicted three other times in April. All three of those progs came in too high by an average of 2.33.

6. MH days: They’re predicting 13, which is tied for the highest ever progged in April. They’ve progged 10-13 five times. Of those five, three progs were too high. The five averaged 3.25 too high.

7. ACE: They’re predicting 210. The prior highest predicted in April were 160-183 (five times). Of these five, all progs ended up too high with even the closest still being 34 too high! The five averaged a whopping 85.8 too high!

Conclusion: If I were a betting man, I’d short CSU’s April ‘24’s forecast overall, especially NS Days, # H, and ACE.

[Teban54]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

https://i.imgur.com/KzUOnHG.png
https://i.imgur.com/5bqZefm.png
https://i.imgur.com/dfrCYsH.png
https://i.imgur.com/wvldsuT.png
https://i.imgur.com/yXx92bm.png
https://i.imgur.com/9DufBL5.png
https://i.imgur.com/nE5GxYk.png

Great stuff, thank you for posting!
I see a pretty clear pattern when they’ve gone very active in April:

1. # NS: They’re predicting 23. Prior to this the highest they predicted was 17-19 (five times). Four of those five progs ended up too high. For those 5, they averaged 2.8 too high.

2. NS days: They’re predicting 115. Prior to this the highest they predicted was 85-95 (four times). All four of those progs ended up too high. The highest actual of those four was only 58 and they averaged a whopping 40.5 too high!

3. # H: They’re predicting 11. The prior highest predicted was 9 (six times). All of those 6 progs ended up too high by an average of 3.

4. H days: They’re predicting 45. The prior highest predicted were 40-45 (five times). Of these five, four progs came in too high. The five averaged a whopping 19 too high!

5. # MH: They’re predicting 5, which they’ve predicted three other times. All three of those progs came in too high by an average of 2.33.

6. MH days: They’re predicting 13, which is tied for the highest ever progged in April. They’ve progged 10-13 five times. Of those five, three progs were too high. The five averaged 3.25 too high.

7. ACE: They’re predicting 210. The prior highest predicted inApril were 160-183 (five times). Of these five, all progs ended up too high with even the closest still being 34 too high! The five averaged a whopping 85.8 too high!

Conclusion: If I were a betting man, I’d short CSU’s April ‘24’s forecast overall, especially NS Days, # H, and ACE.

Here are the years that I believe you're referring to, i.e. the years where CSU historically gave bullish forecasts in April:

• NS: 2022, 2013, 2006, 2007, 2021
• NS Days: 2013, 2022, 2006, 2007
• H: 1999, 2006, 2007, 2011, 2013, 2022
• H Days: 2006, 1999, 2007, 2008, 2013
• MH: 2006, 2007, 2011
• MH Days: 2006, 2007, 1999, 2010, 2011
• ACE: 2006, 2007, 2013, 2011

The fact that 2006, 2007 and 2013 were repeated so many times should be a red flag, and IMO, skews the dataset and methodology too much. In particular, 2006 and 2007 show up in all 7 categories, and 2013 shows up 5 times. (The remaining years are: 2011*4, 2022*3, 1999*2, 2008*1, 2010*1, 2021*1.)

In practice, we know that 2013 was extremely anomalous, and 2006 was a surprise El Nino that will absolutely not happen this year. Among the other frequently mentioned years, 2022 was probably as close to 2013 as we'll get, 1999 still turned out hyperactive, and 2011 was still above-average. 1999, 2011 and 2022 were also returning La Ninas, thus also less likely to be applicable this year.

This leaves 2007 (and arguably 2011) as the only possible "overly bullish" parallels. Indeed, there's a general belief that if 2024 were to fail, a 2007-like situation is the most likely. (See these two tweets from Andy Hazelton.) That year was when a persistent +NAO in spring (+1.44 in March) wiped out the preseason +AMO pattern. However, we haven't seen nearly as strong and long-lasting of a +NAO as of late, and in fact we're mostly around zero on average (-0.21 in March). While we have another weak +NAO episode coming up, it's not expected to persist, and we're exactly a month past when Andy said "should be able to somewhat rule it out within a month".

It should also be noted that 2007 was 17 years ago. Both models and human forecasts have improved significantly since then: this figure shows that CSU's historical April forecasts went from negative skill in 1995-2013 to positive skill in 2014-2023. Note that only 4 of the 32 entries above are from the "positive skill" period (2022*3, 2021*1), whereas the period consists of 10 of the 29 years. To me, this suggests CSU's April forecasts have actually trended more conservative over time, which makes the aggressiveness of 2024 even more notable.

So, while kudos to you for elegantly showing that the most bullish CSU April forecasts have busted too high in absolute numbers, the combination of their heavy inclusion of extreme events, lack of analogs to this year, and the reduced occurrence of such events in recent years paints a much more complex picture upon a deep dive. Personally, I remain unconvinced that just because these years' forecasts busted means 2024 will likely experience the same. At least, it's not as simple as the impression of "CSU predicting hyperactive means it will be just average" that one may get when just looking at the numbers.

[SFLcane]

Here are track plots and geodensity plots for CSU's April forecast analog years:

https://i.imgur.com/2thPEyo.jpeg

https://i.imgur.com/SfBdOPY.jpeg

Based on Phil's analogs i will be watching closely here in SFL this season. Everyone is fair game surely.

[LarryWx]

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

https://i.imgur.com/KzUOnHG.png
https://i.imgur.com/5bqZefm.png
https://i.imgur.com/dfrCYsH.png
https://i.imgur.com/wvldsuT.png
https://i.imgur.com/yXx92bm.png
https://i.imgur.com/9DufBL5.png
https://i.imgur.com/nE5GxYk.png

Great stuff, thank you for posting!
I see a pretty clear pattern when they’ve gone very active in April:

1. # NS: They’re predicting 23. Prior to this the highest they predicted was 17-19 (five times). Four of those five progs ended up too high. For those 5, they averaged 2.8 too high.

2. NS days: They’re predicting 115. Prior to this the highest they predicted was 85-95 (four times). All four of those progs ended up too high. The highest actual of those four was only 58 and they averaged a whopping 40.5 too high!

3. # H: They’re predicting 11. The prior highest predicted was 9 (six times). All of those 6 progs ended up too high by an average of 3.

4. H days: They’re predicting 45. The prior highest predicted were 40-45 (five times). Of these five, four progs came in too high. The five averaged a whopping 19 too high!

5. # MH: They’re predicting 5, which they’ve predicted three other times. All three of those progs came in too high by an average of 2.33.

6. MH days: They’re predicting 13, which is tied for the highest ever progged in April. They’ve progged 10-13 five times. Of those five, three progs were too high. The five averaged 3.25 too high.

7. ACE: They’re predicting 210. The prior highest predicted inApril were 160-183 (five times). Of these five, all progs ended up too high with even the closest still being 34 too high! The five averaged a whopping 85.8 too high!

Conclusion: If I were a betting man, I’d short CSU’s April ‘24’s forecast overall, especially NS Days, # H, and ACE.

Here are the years that I believe you're referring to, i.e. the years where CSU historically gave bullish forecasts in April:

• NS: 2022, 2013, 2006, 2007, 2021
• NS Days: 2013, 2022, 2006, 2007
• H: 1999, 2006, 2007, 2011, 2013, 2022
• H Days: 2006, 1999, 2007, 2008, 2013
• MH: 2006, 2007, 2011
• MH Days: 2006, 2007, 1999, 2010, 2011
• ACE: 2006, 2007, 2013, 2011

The fact that 2006, 2007 and 2013 were repeated so many times should be a red flag, and IMO, skews the dataset and methodology too much. In particular, 2006 and 2007 show up in all 7 categories, and 2013 shows up 5 times. (The remaining years are: 2011*4, 2022*3, 1999*2, 2008*1, 2010*1, 2021*1.)

In practice, we know that 2013 was extremely anomalous, and 2006 was a surprise El Nino that will absolutely not happen this year. Among the other frequently mentioned years, 2022 was probably as close to 2013 as we'll get, 1999 still turned out hyperactive, and 2011 was still above-average. 1999, 2011 and 2022 were also returning La Ninas, thus also less likely to be applicable this year.

This leaves 2007 (and arguably 2011) as the only possible "overly bullish" parallels. Indeed, there's a general belief that if 2024 were to fail, a 2007-like situation is the most likely. (See these two tweets from Andy Hazelton.) That year was when a persistent +NAO in spring (+1.44 in March) wiped out the preseason +AMO pattern. However, we haven't seen nearly as strong and long-lasting of a +NAO as of late, and in fact we're mostly around zero on average (-0.21 in March). While we have another weak +NAO episode coming up, it's not expected to persist, and we're exactly a month past when Andy said "should be able to somewhat rule it out within a month".

It should also be noted that 2007 was 17 years ago. Both models and human forecasts have improved significantly since then: this figure shows that CSU's historical April forecasts went from negative skill in 1995-2013 to positive skill in 2014-2023. Note that only 4 of the 32 entries above are from the "positive skill" period (2022*3, 2021*1), whereas the period consists of 10 of the 29 years. To me, this suggests CSU's April forecasts have actually trended more conservative over time, which makes the aggressiveness of 2024 even more notable.

So, while kudos to you for elegantly showing that the most bullish CSU April forecasts have busted too high in absolute numbers, the combination of their heavy inclusion of extreme events, lack of analogs to this year, and the reduced occurrence of such events in recent years paints a much more complex picture upon a deep dive. Personally, I remain unconvinced that just because these years' forecasts busted means 2024 will likely experience the same. At least, it's not as simple as the impression of "CSU predicting hyperactive means it will be just average" that one may get when just looking at the numbers.

Nice analysis!
To clarify in case anyone might be wondering, I don’t think CSU is trying to sensationalize. I fully respect their objectivity, knowledge, analytical abilities, and clear communication showing how they derived their numbers. I can see the reasoning in going so high when looking at the recent past and considering how ridiculously warm the Atlantic is in combo with the expectation for a moderate+ La Niña. Thus I still think the season will be very active (as per my just made Storm 2K preliminary forecast) even if this April CSU forecast ends up being a bit too bullish.


I wasn’t trying to say they’ve ended up too bullish on average with their April forecasts overall as they, indeed, have averaged too bearish with them as other posts show. It is only when they’ve been their most bullish in April that they’ve averaged too bullish.

[chaser1]

Kudos to both Teban and Larry on a very interesting, insightful, and respectful "point/counterpoint" consideration of CSU's April forecast!
It made for an enjoyable thought-provoking Sunday morning read

[wwizard]

The April CSU predictions for that years hurricane season tend to be right (+/- 1) for Hurricanes and Majors But they tend to be under of the named storms. So something like 28 named storms is more likely.

Now I have been looking for the information about predicted forecasts and actual numbers that I have seen before but I can't find it now if anyone knows where it is.

Here are April forecasts vs. observed for that year going back to when April forecasts began at CSU (1995). Some takeaways:

This is the highest forecast in every forecast parameter.

23 named storms (previous high was 19 in 2022)
115 named storm days (previous high was 95 in 2013)
11 hurricanes (previous high was 9 in 1999,2006,2007,2011,2013,2022)
45 hurricane days (tied with 2006 forecast of 45)
5 major hurricanes (tied with 2006, 2007, 2011)
13 major hurricane days (tied with 2006)
210 forecasted ACE (previous high was 183 in 2006)

https://i.imgur.com/KzUOnHG.png
https://i.imgur.com/5bqZefm.png
https://i.imgur.com/dfrCYsH.png
https://i.imgur.com/wvldsuT.png
https://i.imgur.com/yXx92bm.png
https://i.imgur.com/9DufBL5.png
https://i.imgur.com/nE5GxYk.png

Great stuff, thank you for posting! Very scary!
I see a pretty clear pattern when they’ve gone very active in April:

1. # NS: They’re predicting 23. Prior to this the highest they predicted in April was 17-19 (five times). Four of those five progs ended up too high. For those 5, they averaged 2.8 too high.

2. NS days: They’re predicting 115. Prior to this the highest they predicted in April was 85-95 (four times). All four of those progs ended up too high. The highest actual of those four was only 58 and they averaged a whopping 40.5 too high!

3. # H: They’re predicting 11. The prior highest predicted in April was 9 (six times). All of those 6 progs ended up too high by an average of 3.

4. H days: They’re predicting 45. The prior highest predicted in April were 40-45 (five times). Of these five, four progs came in too high. The five averaged a whopping 19 too high!

5. # MH: They’re predicting 5, which they’ve predicted three other times in April. All three of those progs came in too high by an average of 2.33.

6. MH days: They’re predicting 13, which is tied for the highest ever progged in April. They’ve progged 10-13 five times. Of those five, three progs were too high. The five averaged 3.25 too high.

7. ACE: They’re predicting 210. The prior highest predicted in April were 160-183 (five times). Of these five, all progs ended up too high with even the closest still being 34 too high! The five averaged a whopping 85.8 too high!

Conclusion: If I were a betting man, I’d short CSU’s April ‘24’s forecast overall, especially NS Days, # H, and ACE.

I think it should be pointed out as well that CSU's April predictions have come short of actual numbers by an average of 3.4 NS, 1.4 hurricanes and 0.4 majors over the last 14 years. Just 3 times have they been over on named storms and hurricanes. Sure, 2 of those times were their most bullish forecasts ever, but still.

[ScottNAtlanta]

I think a takeaway from this would be that forecast busts don't occur in a vacuum, and something is learned and molded into future forecasting. 2013 taught many people that just having a warm Atlantic isn't the end all when it is, in fact, warm everywhere. This year (so far) there is a clear distinction between a very warm MDR and a not so warm subtropics. That's a pattern that is highlighted in CSU's analysis. It is reminiscent of 2017 most recently which many people at the time in April of that year here on Storm2K were talking about. The tripole seems to be poking it's head up looking to influence 2024.
These in depth posts that people are making are not something they write off the top of their heads. I, for one, appreciate the time they take to do them. They often answer questions I might wonder about but don't have the time to research. So here is one more thank you.

[cycloneye]

The usually conservative private firm Tropical Storm Risk (TSR) will be up on Monday. A hint has been revealed by Dr Adam Lea in his thread.