I Have Seen the Future

One of the great things about coming to Europe to work is getting exposed to the latest and greatest modeling systems for operational weather prediction.  

Modeling systems for operational weather prediction in the United States has stagnated now for several years.  The last major upgrade was when the Global Forecast System (GFS) upgraded to the Finite–Volume Cubed Sphere (FV3) dynamical core in 2021.  Beyond that, upgrades have been  incremental, with no major changes to the grid spacing of the GFS (13-km) or the HRRR (3-km).  The GFS Ensemble (GEFS), Short-Range Ensemble Forecast System (SREF), and High Resolution Ensemble Forecast system (HREF) have aged and are no longer cutting edge. 

Meanwhile in Europe, both numerical and artificial intelligence weather prediction systems (NWP and AIWP respectively) are advancing rapidly.  Readers of this blog are likely well aware that the European Center for Medium-range Weather Forecasting (ECMWF) ensemble has the highest resolution (9 km), most members (51), best data assimilation, and best statistical performance of any operational, global, numerical modeling system in the world.  ECMWF is also also at the forefront of AIWP and now has their Artificial Intelligence Forecast System (AIFS) ensemble running four times a day.  

But European countries are also pushing the frontiers of limited-area modeling systems that forecast for a specific region.  For example, Meteo Swiss runs an operational, 11-member, 1-km ensemble eight times a day with forecasts out to 33 hours.  They also run a 21-member, 2.1-km ensemble four times a day with forecasts out to 120 hours (see https://www.meteoswiss.admin.ch/weather/warning-and-forecasting-systems/icon-forecasting-systems.html). These forecasts are produced using the ICON model, which was originally developed at the Max Planck Institute for Meteorology in Germany.  The ICON is also used by the German Weather Service.

Here at the University of Innsbruck, I have access to an experimental version of the ICON model run by the Deutscher Wetterdienst (DWD; German Weather Service) at 500-meter grid spacing.  NWP models run at such grid spacings are sometimes called large-eddy simulations (LES) rather than mesoscale simulations because they are beginning to resolve the large eddies that are found in the boundary layer, the portion of the atmosphere that interacts directly with the Earth's surface.  A model run at 500-meter grid spacing also better resolves the fine-scale terrain in mountains regions, which is particularly important in the Alps where glacier-carved mountain valleys are quite narrow.  

Let's look at last night's 6-hour forecast from the 500-m ICON valid at 0600 UTC or 8 AM local time this morning.  The plot below shows the wind (vectors with color fill for speed in meters per second) at 2000 m elevation (relative to sea level).  Terrain elevation above 2000 m is indicated by grey shading. The pattern is characteristic of what are known here as föhn, a strong wind that affects the northern Alps during southerly flow.  In the Innsbruck area, the föhn acclerates as it moves northward through Brenner Pass and down the Wipp Valley.  At this 2000 m elevation, that strong flow eventually moves over Innsbruck and encounters the Karwendel Alps, leading to flow splitting on their southern (windward) side and a wake on their northern (leeward) side.  Locally strong föhn can also be seen in and north of other Alpine Valleys in the region.  

Source: University of Innsbruck

A cross-section from the ICON taken from south (left) to north (right) down the Inn Valley illustrates the vertical structure of the föhn in this forecast.  Locations identified in the cross section include Brenner Pass (BRE) and Innsbruck (IBK).  Note in particular that the strongest föhn flow becomes elevated just upstream of Innsbruck (near EUR) and ultimately rises rapidly whe it encounters the Nordkette ridge of the Karwendal Alps to the north of Innsbruck.  

Source: University of Innsbruck

The cause for the detatchment of the föhn flow from the surface as it approaches Innsbruck is the formation of a nighttime old pool over the Inn Valley.  At night and in the morning, the föhn often rises over this cold pool, although there are times it can "break into" the cold pool, resulting in strong south winds at the surface in Innsbruck.  This is an important forecast problem for the region, including for the local airport, as it affects runway flow direction and the elevation of föhn related turbulence.

How did that forecast verify?  Quite well.  Below is a time-height section of winds over Innsbruck observed with a wind lidar.  The time-height section is for a 24-hour period (time increases to the right).  I've added a red line corresponding to the time of the forecast above.  Note how in the left half of the diagram, the föhn frequently extended to the ground.  Then, just prior to and after 0000 UTC (0200 local time), winds over Innsbruck weakened.  However, by 0600 UTC the southerlies aloft increased once again, resulting in a structure similar to that above with light flow at low levels but strong flow aloft.  

Source: University of Innsbruck

Readers of this blog with a strong interest in snow might also be interested in seeing the corresponding 500-m precipitation forecast.  Validating this is a bit more difficult given the poor radar coverage in the Alps, so I provide it simply for entertainment purposes.  

Source: University of Innsbruck

Enjoy a look at the future.