HRRR forecasts are very wrong
-
I’m in central Canada and I find that the recently added HRRR model is mostly very inaccurate. Like this morning for example, it has rained for hours, continues to rain and yet the HRRR forecasts a little bit of rain starting in a few hours. The European model has been bang on accurate showing rain all morning. I was happy to see the HRRR being added but it is completely untrustworthy.
-
@gnadenthal Hello, the final forecast also depends on the orography of the location and the used model. If the European model suits your location the best, I would recommend using it. You also use the compare feature to monitor the forecast differences.
-
@gnadenthal You noted it right. I myself faced the same issue due to which i have turned to Windy community. Found he solution here. thank you.
-
If the HRRR is not working for you, that is unfortunate. You seem to think the Euro model works. Use it if it works. I know in my area (northeast coast US), easterly winds can create low clouds, drizzle and rain that the models poorly handle. I suspect the ECMWF model does a better job with that, but I never do a proper study to determine that.
As for why the HRRR might not work in your locale, one explanation could be that you're not likely seeing the advantage of the HRRR's 3km top resolution. That resolution does not cover central Canada. In fact, HRRR's 3-km grid does not include a vast majority of Canada! Central Canada is in the 13km domain, and that might help to explain things. Also, inaccuracies might further arise possibly due to the complex interpolations required across and near-to the boundaries between the 13km and 3 km domains.
Another issue is that your forecast accuracy might also be affected by the proximity of the overall model boundary to your location. Furthermore, good satellite data for your area might also be lacking, since central Canadian weather often is affected by nearby polar areas lacking suitable coverage by continuously-overhead (geostationary) satellites (e.g. GOES).
Here are links to a NOAA source describing the model.
https://rapidrefresh.noaa.gov/hrrr/
I suggest you go about 10 lines down from top of main page section, to find a link related to the latest version HRRRv4. (Link leads to a PDF download):
http://rapidrefresh.noaa.gov/pdf/Alexander_AMS_NWP_2020.pdf
The PDF has a map of the HRRR domains as well as other information. -
@Bruzote Thanks for this. I'm on Vancouver Island, Canada, I see from the map that we are right on the border of 3km coverage. Also find that HRRR usually forecasts much less rain than most other models, and less than what actually happens. Nice to know why.
