bdgwx

It might be important to note that my predictions are based on GISTEMP lagging ENSO by 3 months. I know other analysis say it is 4-5 months and that is what I get for the satellite datasets, but my modeling hones in on 3 months for the surface datasets. I'm also using an equal weighting of the statistical and dynamic model averages from the IRI ensemble. That is [statistical_average] + [dynamic_average] / 2. I debated on whether to use the "Average, all Models" line in the table and decided against it since it weights the dynamic models more heavily because there are more of them. Based on what I've seen dynamic models are superior Jan-May (through the spring predictability barrier), but then statistical models have similar skill from June onward. For that reason I wanted to equally weight them. That may end up causing my global average temperatures to be conservative since statistical models basically say this ENSO cycle has peaked which seems unlikely.

One last thing...Hansen's monthly email came through today. He does not mince words. He says the warming rate has likely accelerated due to the extremely high Earth energy imbalance. If that is the case then that may contribute to me underestimating the global average temperature increase as well.

1 hour ago, GaWx said:

@bdgwx @chubbs and others,

 Did any of you see what the ERA5/Copernicus peaked at in early July? I'd like to compare it to the CFSR peak. I'd be shocked if it weren't slightly cooler as ERA5 generally runs slightly cooler than CFSR (~~0.1 C cooler recently).

I don't have an exact value since I don't typically download daily ERA5 data (it's a lot of work). But here is the twitter post.

1 hour ago, chubbs said:

Looks like GISS lowered some of the monthly values in 2016 and 2020, now 1.01 is the record instead of 1.02. Since Kalshi defines a record as 1.03, possible to break the record and not get paid.

Yeah. That is interesting. My guess is that there was a larger than usual data upload into GHCN and ERSST this month. I know there are ongoing record digitization efforts so it's possible some newly digitized records got uploaded.

I had to deal with this situation in 2020 as well. The 2016 value was hovering very near a 0.005 value causing near monthly rounding changes. That market was based on 2020 being higher than 2016 when rounded to 2 decimal places. So I had to track both 2016 and 2020 values down to 3 decimal places which required running a modified version of GISTEMP on my own machine. Fortunately like you say Kalshi is strictly defined as >= 1.03 regardless of what 2016 and 2020 are doing. 

One strategy I was thinking of was to cash out my position on the >= 1.03 contract and take my gains and instead try to play the new 1.05 - 1.07 contract. The 1.05 - 1.07 contract is undervalued according to my analysis. The problem is that there isn't a lot of market depth on that contract. A mere $99 will take it from $0.20 to $0.69 instantly. I show a fair value around $0.60. So that play isn't going to work right now. And I only show maybe another $0.10 gain on the >= 1.03 contract before it reaches fair value. I think the opportunities via Kalshi are limited right now.

My initial impression of the Kalshi is meh. I don't think it has been correctly valuing the contracts. It seems to be very reactionary leading me to believe that there are very few participants using predictive tools. The other markets I've tracked in the past had more participants and more educated/sophisticated participants at that resulting in the market being a leading indicator instead of a lagging indicator like what Kalshi is. So be it I guess. It's easier to get an edge this way.

GISS published the June update for GISTEMP. It came in at 1.07 C.

Here are the changes on this update cycle.

Jan: 0.87 => 0.86

Feb: 0.98 => 0.97

Mar: 1.21 => 1.20

Apr: 1.00 => 1.00

May: 0.94 => 0.93

Here is my updated analysis.

Jan: 0.86 ± 0.01 C (3m lagged ENSO -0.99)

Feb: 0.97 ± 0.01 C (3m lagged ENSO -0.90)

Mar: 1.20 ± 0.01 C (3m lagged ENSO -0.86)

Apr: 1.00 ± 0.01 C (3m lagged ENSO -0.71)

May: 0.93 ± 0.01 C (3m lagged ENSO -0.46)

Jun: 1.07 ± 0.02 C (3m lagged ENSO -0.11)

Jul: 1.11 ± 0.16 C (3m lagged ENSO +0.14)

Aug: 1.05 ± 0.22 C (3m lagged ENSO +0.46)

Sep: 1.09 ± 0.23 C (3m lagged ENSO +0.86)

Oct: 1.15 ± 0.24 C (predicted 3m lagged ENSO +1.03)

Nov: 1.17 ± 0.25 C (predicted 3m lagged ENSO +1.17)

Dec: 1.17 ± 0.26 C (predicted 3m lagged ENSO +1.26)

Jan - Jun average: 1.01 C

Jul - Dec predicted average: 1.12 C

2023 Average: 1.062 ± 0.07 with 80% chance of >= 1.03 (new record) and 60% chance >= 1.05.

1 hour ago, SnoSki14 said:

I remember when the climate change skeptics would use above normal Antarctic sea ice as a way to push back against the scientific consensus. 

What's their excuse now? 

And keep in mind that the expectation was that sea ice in the SH would actually remain level or even increase through 2030 - 2060 before beginning to decline. So observing a decline so soon and so dramatically obviously raises eyebrows.

Berkeley Earth is now saying there is an 81% chance that 2023 will be the warmest in their dataset.

https://berkeleyearth.org/june-2023-temperature-update/

What does 1.10 ± 0.16 C mean? Is this the Average Global Temp anomaly for the month of June?

6 minutes ago, GaWx said:

 Thank you. Just comparing the CFSR and ERA5 curves shows the very strong correlation between the two. The main difference appears to be that ERA5 has often been ~0.1C cooler, including the last few days. So, CFSR may be warm biased though only slightly (0.1C) assuming ERA5 isn't cool biased. Otherwise, it appears that the daily change of CFSR for a particular date can probably be used as a pretty good proxy for the change that ERA5 later shows for the same date. Thus I agree with your feeling that the daily CFSR still has value since it comes out earlier and thus can be used to predict ERA5 daily moves.

Exactly. It's like how we use CDAS to track ENSO. It's not the best metric. But it's close enough to give us a pretty good idea of what ERSST is going to show. And since CDAS is near real-time it is still incredibly useful.

3 hours ago, GaWx said:

Here's a pro met's Tweet addressing this source that I just was pointed to by an AmericanWx member met. posting at another board. He suggests caution in using it as an accurate source for climate analysis and instead using ERA5 from Copernicus ECMWF as the most trusted source. He says this is based on a 2009 version of the GFS:

Any opinions about using ERA5 from Copernicus ECMWF instead of this University of Maine source? How is the global temperature looking on ERA5 compared to this?

I was actually getting ready to respond to your question from this morning. I don't use CFSR for my own modeling because it's skill is subpar. Of the reanalysis datasets I generally use ERA5 and/or JRA55. That doesn't mean CFSR should be dismissed though. It's correlation with ERA and JRA is still very high.

I use ERA5 a lot. In fact, it is one of my go-to sources for getting an early lead on what GISTEMP is going to report since it correlates at R^2 = 0.89. Note that Copernicus confirmed the new record via ERA5. And it's likely July 4th eclipsed that as well. ERA5 has a 2 day lag so we don't know for sure yet.

Nick Stokes just processed the June GHCN-M and ERSST files. I estimate the GISTEMP value for June at 1.09 ± 0.06 C.

There is a lot of a information coming out today. Sorry for the rapid posting. Anyway, the April CERES Earth Energy Imbalance came in at +1.81 W/m2. 

I realize the CERES EEI calculations have high uncertainty, but if the EEI truly is this high then there is more than enough warming in the pipeline to go well beyond 1.5 C. 

Copernicus won't publish the official numbers from ERA5 for another couple days. However, based on Hausfather's twitter post we might expect GISTEMP to publish around 1.10 C for June. ERA5 correlates with GISTEMP at R^2 = 0.89. Note that the previous record was 0.92 C in 2022. It is all but guaranteed at this point that June 2023 will be a new record in the GISTEMP dataset as well.

For the first time in recorded history the global average temperature breached 17 C; at least according to CFSR anyway.

10 hours ago, chubbs said:

Very warm (and stable recently) forecast by gfs for first week of July. Pushing daily reanalysis temperatures further into record-breaking territory. Could break the 17C barrier on the chart posted above.

I was actually getting ready to post the GFS forecast as well. You scooped me.

Anyway, yeah, July is already forecasted to start off quite warm. I noticed that there was a small blip down in the global SST last week. I wonder if that means excess heat is transferring from the ocean into the atmosphere now.

As of this moment my June expectation is a tick up to 1.08 ± 0.10. Once the June data starts rolling in I can get that uncertainty envelop down to ± 0.06 prior to the GISTEMP update. By all indications June 2023 is going to be the warmest June on record.

BTW...the page is buried pretty well. But you can get the probability density function corrected forecast (bias corrected) for the CFS from the following link. It took me forever to find it a couple of weeks ago and Google isn't much help. 

https://www.cpc.ncep.noaa.gov/products/people/wwang/cfsv2fcst/CFSv2SST8210.html

And the bias corrected CFS forecast is even lower at maybe +1.4.

On 6/26/2023 at 10:55 AM, chubbs said:

Think the odds will only increase as the year progresses.

I cannot disagree. If I have understood the Brown & Caldeira method correctly they only make predictions based on the current state of the climate system. In other words they do not take into account the expectation of future ENSO (or other highly correlated climatic element) states. This may explain why I get higher odds of a new record. My model is simpler (more like a multiple regression that minimizes RMSE), but incorporates the future expectation of ENSO and to a lesser extent total solar irradiance (which I do find to be at least minimally correlated with global temperatures). Both of which can be predicted 6 months in advance with reasonable skill. And yes, your point about global SST is well taken. My expectation is that the atmosphere will catch up to the higher SSTs in the next couple of months.  It may be interesting to note that my model does not use SST has an input right now. In that regard one might argue that even I may be underestimating the warming potential in the later half of the year, but I'm going to remain more guarded on that matter as I also think a reversion to the trend may also be on the horizon. Afterall highly deviant increases/decreases tend to reverse some eventually. BTW...my current June expectation for GISTEMP is 1.07 ± 0.10 C. To put that into perspective even taking the ~2.5% chance that it comes in -0.10 C below the expectation at 0.97 C it will still easily surpass the previous record of 0.92 C set all the back in 2022. That should raise some eyebrows.

Berkeley Earth says there is a 54% chance of new record in their dataset. 

@GaWx I bet you're right. I've not been able to find any daily ERSST updates. And I can't disagree, I see a very high correlation between NCEP/NCAR vs ERSST in ENSO 3.4 region at R^2 = 0.98. So while they don't match exactly it is very close. I certainly wouldn't dismiss it.

Someone can correct me if I'm wrong but it is my understanding that CDAS, which stands for Climate Data Assimilation System, is actually the model core for the NCEP/NCAR reanalysis. Or at least that is what the NCEP/NCAR built upon. The NCEP/NCAR reanalysis is known to be subpar. But I don't think that is justification for dismissing it outright. I do wonder if it wouldn't be better for Levi Cowan to use a more accepted product like OISST or ERSST though.

Several months back I suggested that if the shenanigans kept up in the Antarctic then perhaps it might be time for a dedicated thread. The data collected by the IPCC suggested that sea ice extents may increase through 2030 at the very least. Yet here we are with record lows. Perhaps the time for a dedicated thread as come.

Here is my latest expectation for GISTEMP which includes the June IRI ENSO ensemble forecast.

Jan: 0.87 ± 0.01 C (3m lagged ENSO -0.99)

Feb: 0.98 ± 0.01 C (3m lagged ENSO -0.90)

Mar: 1.21 ± 0.01 C (3m lagged ENSO -0.86)

Apr: 1.00 ± 0.01 C (3m lagged ENSO -0.71)

May: 0.94 ± 0.02 C (3m lagged ENSO -0.46)

Jun: 1.05 ± 0.12 C (3m lagged ENSO -0.11)

Jul: 1.03 ± 0.22 C (3m lagged ENSO +0.13)

Aug: 1.06 ± 0.23 C (3m lagged ENSO +0.39)

Sep: 1.10 ± 0.24 C

Oct: 1.13 ± 0.25 C

Nov: 1.16 ± 0.26 C

Dec: 1.17 ± 0.26 C

2023 Average: 1.06 ± 0.08 with 75% chance of a new record (>= 1.03)

The June IRI ensemble suite just got published. The statistical+dynamical average peak jumped up 0.2 to 1.5 on this update.

https://iri.columbia.edu/our-expertise/climate/forecasts/enso/current/?enso_tab=enso-sst_table

IRI June Seasons (2023 – 2024)                                    
Model       JJA    JAS    ASO    SON    OND    NDJ    DJF    JFM    FMA
Dynamical   1.295  1.539  1.679  1.761  1.746  1.598  1.473  1.264  1.009
Statistical 0.749  0.785  0.825  0.840  0.846  0.812  0.734  0.579  0.415
All         1.120  1.298  1.406  1.466  1.403  1.248  1.125  0.899  0.670

1 hour ago, FPizz said:

The bias corrected CFS peak has come down about 0.5 in the last few weeks.