Day 132 - 2.5 degree error
Today I set up my model to do forecasts over 4 months, rather than the 3 or 4 nights I've been testing before today. The average RMS error is around 2.5°C. That's not great, but I think it's already a fairly big improvement over other forecasts out there for this particular location.My goal is to reduce the error by another degree or so. The forecasts for most nights are already within about 1.5°C, but there are a few nights with errors over 5° that hurt the average a lot. I need to figure out what's happening on those nights.
I'm currently using a very bad soil scheme, and the surface properties in the model are the same in August as in December, which is not very realistic. I see those as two big areas for improvement. And I still have a lot of tweaking to do. So I think an average error below 2°C is very possible; maybe even 1°C for some months.
I suspect that the low temperature forecasts might be a little better. The numbers I'm quoting are the average error over every 5-minute period throughout the night.
1 Comments:
Matt, it is so great you are learning and working at things that really interest you. I have no doubt you will make the most of it.
Have fun and God bless!
Post a Comment
<< Home