> Last time they decided we needed more supercomputing power, and they got it. Did it help? Per a few sources I've seen, no, it didn't. One big cause of the G

> Last time they decided we needed more supercomputing power, and they got it. Did it help? Per a few sources I've seen, no, it didn't. One big cause of the GFS (USA) model's inaccuracy is lack the of computing power, which we've addressed a few times with new supercomputing clusters. The bigger problem is that the GFS model uses bad underlying physics. Decades ago we made several choices to optimize around the current architecture and processing ability, which was probably okay in the mid 1990s. Today, it just results in greatly inferior predictive ability compared to the European models. When you combine the poor model physics with lower resolution compared to the competitors, it becomes clear that we need to reboot the whole thing. https://cliffmass.blogspot.com/2016/06/us-numerical-weather-prediction-is.html https://cliffmass.blogspot.com/2016/06/us-numerical-weather-...