Arome 1.3 km forecast discrepancy



  • Hi all,
    I've noticed the forecast is often quite different between the meteo france app and your arome data.

    Screenshot_2020-05-17-11-24-24-208_fr.meteo.jpg
    Abobe the rain probability is given at 50%

    Screenshot_2020-05-17-11-24-15-696_com.windyty.android.jpg

    How can the temperatures be so different? Why doesn't windy expect any rain?

    Same problem this morning. In Antibes, France at 9am it was raining, confirmed on the meteo france app, but no rain forecast on windy with arome ...

    Any ideas?
    Thanks a lot for this great app!
    Alex


  • Sailor

    @rogerw
    Who says that the forecast on Météo France app is AROME 1.3km?
    It could be of an algorithm mixing AROME 1.3km, AROME 2.5km and ARPEGE.
    Or much more probably because today AROME 1.3 km predicts showers in the vicinity and so the algorithm does not exclude a high level of rain probability.
    The difference is that Windy shows only a visualization of Deterministic AROME raw data but not of Ensemble AROME.
    To get an idea of another raw AROME 1.3km visualisation, you can look at rain forecast from this model on Meteociel (select Precipitation HD). You see that today AROME 1.3km forecasts only showers near Gourdon but not exactly on this location.
    Which is similar to the Rain layer on Windy.

    60508F7A-A362-42FA-B22A-5B0E4A066BCC.jpeg



  • To provide its forecasts, a priori Meteo France uses a mix of models AROME, ARPEGE, ALADIN (deterministic/ensemble), real time observations and climatological data (see details here). I think one shouldn't compare very elaborate forecasts provided by national/regional Met Offices, to raw data from forecast models provided by Windy.


  • Sailor

    Except ALADIN which is no longer used by Météo France in this country.



  • I don't know about ALADIN, I just cited what Meteo France says. Afaik. ALADIN is still in use for overseas territories, then combined to HIRLAM for climatology/research (here). It appears also that it is still used as a climate model under the name CNRM-ALADIN, and climate models are used by Meteo France to compute forecasts (as said in the same article cited above).



  • Thanks for your answers. What I'm learning is that a tight grid doesn't mean more accuracy. In reality the grosser models on windy were more accurate.

    And about temperatures? I was expecting them to be a parameter less sensible than rain showers... Any idea on how they can be 2 3° apart just a few hours before?

    Cheers



  • Model foracasts are computed at what is called "runtime", generally at 00h, 06h, 12h, 18h UTC.

    For some models, Windy only provides runtimes every 12h, which are only available for dowvloading about 4-6 hours after each runtime. Thus, temperature forecast values for, say, 15h UTC, were computed at 00 UTC, thus 15 h ago...

    Anyway, 2°C delta values is not big stuff... Comparing model interpolated values with observed values is always tricky.
    Don't forget that forecast values are computed each 25km (e.g. for GFS 0.25°), then values of two (or more) neighbor grid cells are interpolated by Windy to get values on user's position, without taking into account local conditions (topography, vegetation urban/rural areas, valleys, etc.)... while observed values are just... always right.


  • Moderator

    @vsinceac said in Arome 1.3 km forecast discrepancy:

    Model foracasts are computed at what is called "runtime", generally at 00h, 06h, 12h, 18h UTC.

    ...while observed values are just... always right.

    Except when they are not, because of some issue with the observation equipment or the station :)

    Also the 'official' observing station isn't always close to your position, so comparing that to the forecast doesn't always make sense either. As you mentioned there are local conditions that can occur even over a relatively short distance (topography, vegetation urban/rural areas, valleys, etc.)..

    And in Antibes I often noticed differences of +2 to 4degC between the observed temperature at a local station, and the 'official' temperature from Nice Airport. The topography is different in the two locations and Antibes is normally warmer. But many 'weather apps' will use Nice Airport for Antibes' data.

    As a prudent marine navigator will tell you, never rely solely on one source of information.

    This was originally about determining the vessel's position at sea, but it applies to many other things too, including weather data, both forecast and observed.



  • A WMO station is always "right" for measurements provided at that position: all the sensors are correctly calibrated and the measurements strictly follow the same WMO (and/or ICAO) observation rules. Thus observations provided by these stations should be not "roughly" compared to observations issued from non-WMO stations (or well should be compared with "scientific" care).

    Many airports have different well calibrated stations disseminated over the airport area, and often observed temperatures are different between these stations: this is normal.

    Anyway, WMO stations should be the only ones worthy of trust (from meteorological/climatological pov.); NWP forecasted values are computed/interpolated/extrapolated/etc. (often from observations provided by WMO stations), while other types of observing stations may lack the right calibration or don't rigorously follow the same observation rules.


Log in to reply
 

Windyty, S.E. - all rights reserved. Powered by excellent NodeBB
NodeBB & contributors, OSM & contributors, HERE maps
Terms and Conditions     Privacy Policy