In somewhat of a surprising and good move, NOAA showed this morning that they are willing to step up to the plate and at least try to match their European counterparts in terms of computer modeling!
— Nate Johnson (@nsj) January 5, 2015
The US desperately needs to catch up to Europe in this department. The major European model (the ECMWF) has a substantial lead on the GFS in terms of verification statistics, and this is an issue that has led the US to finally show some sort of interest in investing in their models outside of just mesoscale models.
Global weather model skill scores: 500-hPa Z NHEMI 2014: ECMWF: 0.905 — GFS: 0.875 2013: ECMWF: 0.904 —- GFS: 0.880
— Ryan Maue (@RyanMaue) January 5, 2015
The official NOAA report indicates that, “the capacity of each of NOAA’s two operational supercomputers will jump to 2.5 petaflops, for a total of 5 petaflops – a nearly tenfold increase from the current capacity.” This will help to increase the forecasting capibility of the major US models, and will help to better mid and longer range forecasts out of the US. In our minds, we like to believe that this is a major step forward in catching Europe, but in reality that isn’t necessarily true.
The US still isn’t investing enough. Europe has such a wide lead that even this will make it hard to catch up, as the ECMWF is going to continually invest more into their computer models and the UKMET is investing a whopping $156 million and according to BBC, the system will be able to, process at, “16 petaflops – meaning it can perform 16 quadrillion calculations every second.” That is 5.2 times the power that the US will have to compute forecast models in.
I wrote about this issue back in October, and even with this investment, nothing has changed. The NWS needs to be receiving more money so that it can upgrade its various services, and until that happens, none of this will change. It is an exciting thing that NOAA is trying to catch up to Europe, but if we are really going to catch up anytime soon (not happening), the NWS is gonna need more funding. And from that perspective, American computer modeling looks like it may have a bleak future.