The old PG-ICON weather, v1.0.0
We take a quick look back. In the summer of 2023, during the redevelopment of the PG website, we decided to make our own weather. In this system v1.0.0 the 1099 lines long PG_ICON_Wetter_20230801.py, fed with internal functions, dictionaries and colorbars, ran daily on our weather page since the beginning of December 2023. It is embedded in the tileserver, which projects the tiled image files onto the map. This system caused us few problems: we have had a fresh forecast online early in the morning every day since then. This was an incalculable risk at the time, but thanks to the support of our web agency croox, especially Andreas Haas, a meteorologist's dream became a stable reality. Over the course of last winter, two more regions were added, the Pyrenees and Scandinavia. But PG-ICON-Wetter, v1.0.0, is going into its well-deserved retirement. Once again, cheers to this program! But now to the new system.
The new PG-Weather, v2.4.0
We have been working on a new system since mid-2024. This should be more flexible and elegant. For example, we could now create forecasts for any time step. Meteograms were to be generated in another program. Neither has been realized yet. The most obvious innovation is that the patchwork of regions (Scandinavia, Pyrenees, ...) has been dissolved and instead the entire ICON-EU area is shown, which extends from Iceland to Afghanistan, and from Norway to North Africa. We also show the global model of ICON, which covers a forecast period of 7 and a half days, for Europe as well as for North America, the Andes, Kashmir and Japan. Fears that the calculation of the tiles would take too much time have only partially materialized. After all, we are subject to our claim that the ICON forecast, which is initialized at 0 a.m. UTC and is available from around 3 a.m. UTC, is displayed from 6 a.m. CET. In the end, we had to split the program into Europe and worldwide, as remapping the non-European regions takes a lot of time. Then the question arose as to how to deal with the high-resolution Nest ICON-D2, whether we would show it as an independent model and thus leave the user the option of choosing between ICON-EU and ICON-D2, or whether we would integrate ICON-D2 into ICON-EU. We opted for the latter. Technically, this was quite challenging, but we believe it is worth it.
Valuable feedback from our community and our own reflections made us realize that the snow variables snow temperature, snow age and snow density are not very useful, and perhaps even misleading, and we no longer show them. These variables are derived from the ICON snowpack model (TERRA-ML) and this is not intended to predict snow depth and powder, it is only concerned with the water and radiation budget. Snow depth is therefore only obtained indirectly in ICON and is derived from snow water content and snow density. Other models are better suited for snow depths, especially in mountainous terrain. In return, we now offer you the variables "relative humidity (2m) [%]", "cloud cover [%]" and "zero degree line [m]", in each case at midday.
