Seasonal forecasting

Their conclusion seems reasonable, at this stage it’s useful to at least show potential and get funding to push further with ML and AI techniques. Unfortunately it’s all or nothing in this area, if you don’t go all in with data and training you don’t get good models but that’s expensive. We saw a step change with OpenAI because they finally just used the whole dataset to train it, and training took over a year the first time.
 
I've seen work suggesting that ML techniques have value for medium-term forecasting; that is from the horizon of detailed modelling out to a month or two. That's in the context of urban resilience in a tropical marine climate, where the main danger is tropical storms or extreme rainfall. But that's in a very different climate from the UK - they actually have a climate and find it difficult to relate to the variability of UK weather 😁
 
So... .What is the best date this coming winter to look out my Skis for the attic for Skiing in the Cairngorms?
 
So... .What is the best date this coming winter to look out my Skis for the attic for Skiing in the Cairngorms?
I can tell you that without reference to any AI / weather data - it will be the days when I have immovable work commitments and am unable to go. All days when I am available will be 40+ mph winds with hail, rain, or the snow will have melted. This is based on years of evidence!
 
But that's in a very different climate from the UK - they actually have a climate and find it difficult to relate to the variability of UK weather
I think this is the reason ML/AI has to be the answer here. We can’t model with enough parameters within a reasonable time and cost using physics.
 
Top