Earthshine

Metwatch Interesting how 2021 makes it there too!

Met4Cast This is another huge nail in the coffin for those few still banging on about solar activity causing warming. If the sun was outputting more energy the entire atmosphere would be warming. The fact we can observe the stratosphere in fact cooling is a sure sign of increasing carbon dioxide in the lower atmosphere.

Metwatch Very interesting! Going by highest temperatures on record for each country is also pretty fascinating: England: July 2022 Wales: July 2022 Scotland: July 2022 Northern Ireland: July 2021 Record high minimums: England: July 2022 Wales: July 2022 Scotland: July 2019 Northern Ireland: August 2001

Derecho A lot of NWP is parameterised where the scale of events is just too small to capture in the model (e.g. really local convective events). In theory AI isn't really limited to a grid, I've seen neural fields allow variables to be mapped to arbitrary resolution (e.g. your neural network maps x, y, z coordinates f(x,y, z) --> t,p,q with no reference to an actual grid). Probably won't work in NWP, but we already can already see upscaling of precipitation using AI show encouraging results: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022MS003120

Frigid Life in the UK really is relentless a winter supposedly posed to bring cold weather ended up almost seeing 20°C being reached in Scotland.

Granted I know almost nothing about this, but it would be super interesting if there was a reanalysis study to see if MJO/AAM/etc. predictions do have some medium range skill and how good they are. It would probably be a huge undertaking, but say we took the MJO forecasts from previous spells that preceded high latitude blocking events. What kind of forecasting skill could we see? I'm sure the vast density of information in the ERA5 could be used to create a pseudo forecast perhaps. It would be interesting to try to quantify how useful these driver forecasts are in a historical setting, which might inform us on how to use them now.

Relativistic Thanks! It's an interesting problem. Really the model is trained to minimise the difference in variables between the ERA5 dataset and it's own output (Pangu is four surface and 5 upper air variables - MSLP, T u, v winds + mixing ratio for upper air variables on 13 pressure levels). It infers the physics on how these variables interact with zero knowledge of the known physics, something I find extraordinary that it works!

Mike Poole You're exactly right. There is no inherent knowledge of physics, only inferred from the ERA5 training data. I'm amazed it works, it seems such an ill-posed problem yet here we are. On the problem of not respecting physical laws, this is a problem because the atmosphere is chaotic and if you aren't conserving energy the intensity of storms is underestimated. We saw this with AI forecasts for Storm Ciaran which were weaker than the physics-based IFS forecast.

The PIT That's actually a problem with "AI" models. They are trained to minimise the absolute error between the output and the training reanalysis dataset. When used operationally, the output that is an optimum solution is a "blurred" forecast with no extremes at longer lead times. You can actually see the kinetic energy just randomly drop away. I've applied for a grant to work with a software engineer to hopefully incorporate physical constraints into the training of these models. Hopefully then the output will be constrained to not only fit the observational data, but to actually force it to conserve kinetic energy, mass, etc. These data-driven models are the very first generation of operational models. I would be EXTREMELY hesitant to say that they will never overtake classic physics based models. The rate of progress has just exploded - and these are basically equivalent to the very first generation of classic NWP models. Personally I think the biggest constraints to better weather forecasts at the moment is a lack of observations - satellites are great, but the vertical resolution of sounders is seriously lacking.

A very good summer that is rarely spoken about. Every month was significantly warmer than average. Here's the season and monthly maximum CET values: Summer: 22.2°C (6th hottest on record) June: 21.3°C July: 22.0°C August: 23.3°C There were unfortunately drought conditions. Every day from the 24th July to the 28th August recorded a maximum CET of at least 20°C.

lassie23 I wouldn't be so sure, the AIFS has recently been updated by the ECMWF and has skill just as good as the physics-based IFS using attention graph neural networks: https://www.ecmwf.int/en/about/media-centre/aifs-blog/2024/first-update-aifs

I would absolutely not worry about what this summer will be like - that goes for both warm weather and cool weather fans! Many were saying 2022 will have a mediocre August because of La Nina - ended up being the third warmest on record. Last summer was supposed to be a summer of heat plumes - none except in September! Personally I'll be concerned when we get to June and we see zero warm ensemble members

My ideal climate would be 32°C during the day and 20°C at night, thinking somewhere like a Carribbean climate. I love warm nights where I can just sit outside and chill. I do struggle when temperatures get close to 40°C though.

A March like 2022, an April like 2020 and a May like 2018/2020 would be fantastic. I'd like to see gin blue skies and bone dry conditions in May, a bit like 2020 (picture below).

Potential for exceptionally mild end to the month. Long way off of course but quite a few of these are very February 2019-esque

3.7°C and 60 mm please

Word of caution - in the absence of a strong driving signal, NWP models will often drift towards the climatological mean (background covariances in data assimilation will often penalise large departures from climatological mean).

I'm surprised they are predominantly done with satellite data, I would have thought they would have been created using the analysis (i.e. all assimilated observations in that time window).

5C and 80mm please

7.4°C and 70mm please

That's amazing, won't be long until the 30 year mean for July gets to 17°C I reckon.

PanguWeather looks like maintaing a pretty mild feed out until second half of next week. Very warm this weekend. Each model run is about 10,000 times cheaper than the equivalent using a full-physics NWP model. I reckon we will see enormous ensemble runs using these, 10,000 ensemble members for the price of a single ensemble member from the ECMWF or GFS!

Makes me laugh when people say we can't reach X temperature despite it only being 1-2C off what we've seen in climatology. Have we truly sampled the entire distribution of possible temperatures in the UK records? Almost certainly not. Especially since the mean (and likely variance, although it probably isn't a normal distribution anymore) has shifted dramatically in the last few decades. Even if the mean hasn't shifted, we would probably need thousands of years to properly sample the extreme ends of the UK climate distribution (ignoring the fact that the climate rarely stays the same over those kinds of time periods)!

Definitely the most bizarre summer in my lifetime (since 2000). Could you imagine two 17°C CET months with neither being July or August? Perhaps we won't have to imagine soon!

I think we are in for a good chance to beat the all-time maximum CET record for September. 22.1C up to the 25th September compared to versus 20.9C for 1895 and 2006.