search

georgemilosh / Climate-Learning

How to predict extreme events in climate using rare event algorithms and modern tools of machine learning
https://georgemilosh.github.io/Climate-Learning/
MIT License
19 stars 5 forks source link

Try subtracting the average yearly temperature for computing the labels #25

Open georgemilosh opened 2 years ago

georgemilosh commented 2 years ago

Currently we are mixing years which start very cold and very warm in summer. Probably it is easier and more interesting to predict if there is 14 day anomaly above this level

AlessandroLovo commented 1 year ago

What level? This issue is not clear

georgemilosh commented 1 year ago

So currently heatwaves are defined as events that are above 95 percentile relative to the climatological anomaly. For practical applications it could important to look at the absolute temperatures, but another possibility is to subtract the seasonal temperature of the previous month, or the temperature of May. I guess that's what I meant by level, but perhaps the title and the description is not very clear.