I'm a novice software developer who's about to write some Python scripts for getting weather data from several U.S. stations, as well as a few in the Pacific (for checking Southern Oscillation Index & the like).

I've been oriented toward NOAA's site mostly, but having seen some mentions about 'European models' of weather forecasting in various articles over the past year, I wonder if it would be better to look toward mining the ECMWF.

Basically, I'm looking at current surface temperature and wind speeds, then trying to get a daily forecast out to 15-20 days forward. Haven't dug into ECMWF yet, but on NOAA's site, they do provide probabilistic outlooks, albeit on a monthly frequency (example). I could always check weather.com's 10-Day or Monthly, but again, I'm wondering about the accuracy.

Anyone with experience comparing forecasted vs actual results on a thorough basis, and can share some perspective?