Navigation

    Windy Community

    • Register
    • Login
    • Search
    • Unread
    • Categories
    • Groups
    • Go to windy.com

    Forcast by the time the model is refreshing...

    General Discussion
    4
    15
    264
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • B
      BPeti72 | Premium last edited by

      Hi! My English is pretty bad, but i hope you'll understand me. :) Today it was rainy here but neither of the models forcasted rain. No problem, thats life. BUT ...
      when I can see that ECMWF updated just a few minutes ago I would excpect that the weather "forcast" is pretty accurate by that time... Not this is the case. It was raining and around 24C. The model "forcasted" cloudy weather and 30C ...
      Now I dont understand what model update means and i feel premium subscription is useless... Every weather app can forcast random weathers... and most of them are accurate for the current weather. :)

      Can you describe what's happening exactly and what does update means on a model?

      idefix37 1 Reply Last reply Reply Quote 0
      • idefix37
        idefix37 Sailor Moderator @BPeti72 last edited by idefix37

        @bpeti72
        It would be useful that you say where and when you have observed these bad forecasts. And that you post screenshots.

        B 1 Reply Last reply Reply Quote 0
        • B
          BPeti72 @idefix37 | Premium last edited by

          @idefix37 I can not send screenshot from the past, but it was in Budapest (Hungary) this morning. Around 10am CET

          idefix37 1 Reply Last reply Reply Quote 0
          • idefix37
            idefix37 Sailor Moderator @BPeti72 last edited by idefix37

            @bpeti72
            You can use the Observation vs forecast tool. It shows you a comparison between a local weather station and the weather model forecast during last days.

            9E19628A-6691-43AC-A4E4-5E52881094EC.jpeg

            It seems the model of the ECMWF was predicting only 0.1 mm rain when this weather station was reporting 1mm. Another station was reporting 2mm.
            Concerning temperature the default model seems lower than the reported temperature. This tool shows that ICON-EU and METEOBLUE were better. We know that global models are not so good at predicting temperature in large towns.

            B 1 Reply Last reply Reply Quote 2
            • B
              BPeti72 @idefix37 | Premium last edited by

              @idefix37 Thank you very much, it is really helpful, I didn't even know it exists.

              My real problem is that it is said ECMWF is updated 4x a day. When the model said it was updated a minute ago (10am local time) The prediced weather was way off compared to the reality... So what a model is which can not forcast 0 minutes ahead? :o This is what i dont understand.

              Screenshot 2021-07-10 at 23.29.34.png

              idefix37 1 Reply Last reply Reply Quote 0
              • idefix37
                idefix37 Sailor Moderator @BPeti72 last edited by idefix37

                @bpeti72
                I understand your question about the model updating. But the process of updating a weather model like that of ECMWF is quite long in fact. For instance the update that you get at 10h is the update with the 00Z reference time. That means the process started roughly at 00UTC with the data assimilation, then the computation phase and finally the forecast data dissemination., 7 or 8 hours later. Then there is about one hour to get the new weather maps which is the time for processing by Windy.
                More information here:
                https://community.windy.com/topic/11361/a-difference-between-reference-time-and-update-time

                https://www.ecmwf.int/en/forecasts/documentation-and-support/data-delivery/dissemination-schedule#node-6685

                So even if weather stations report some rain at 9h this information cannot be included in the updating process. And the update that you get at 10h is based on data which have been gathered in the middle of the night.

                Locally it is possible to get forecast which include the latest weather data like radar and satellite data. It is known as Nowcasting. It involves generally very high resolution models which update every hours.
                https://en.m.wikipedia.org/wiki/Nowcasting_(meteorology)
                Windy does not offer such short update models.

                1 Reply Last reply Reply Quote 3
                • B
                  BPeti72 | Premium last edited by

                  @idefix37 thank you for all answers. The only thing which is not clear: you showed that way to compare forcast with the reality... But I can not find how long time forcast is there on the chart. If model A seams more accurate than model B in 1 day forcast that does not mean it is better in 3-5-10 days forcast. so how long term does it compare models?

                  idefix37 1 Reply Last reply Reply Quote 1
                  • idefix37
                    idefix37 Sailor Moderator @BPeti72 last edited by idefix37

                    @bpeti72
                    This chart, the Observation vs forecast tool, makes a comparison between reported values and predicted ones. But this comparison is mainly based on the first 6 hours, so you cannot measure the performance of a model at predicting 3, 5 or 10 days with this tool. It is useful for seeing if there is a bias with a model at a precise location even at analysis stage. For example as often discussed here, in mountains some locations show wrong temperatures because the weather model use a coarse topographic model and cannot see the location at the right altitude. So you see which model shows a smaller bias. But that does not mean it will be very good at forecasting the next 5 days.

                    B 1 Reply Last reply Reply Quote 2
                    • B
                      BPeti72 @idefix37 | Premium last edited by

                      @idefix37 I don't know why but there is no such a comparision of the 24-48 hours forcasted weather vs reality. Not just here, but almost nowhere... Meteorology is a strange thing... everyone is forcasting, but almost noone checks (or it is not public) how good their model is. People have to chose models according to advertisments insted of data...
                      Do you know any page where we can see how good each forcast for a longer period? I know there are forums, but it is highly personal how you feel about a model... If someone use a model for his / her wedding ceremony and it fails, they will feel this model is useless... while others reporting this model is superior than the other because it was better 3 times when he went out for a long ride... :)

                      idefix37 1 Reply Last reply Reply Quote 0
                      • idefix37
                        idefix37 Sailor Moderator @BPeti72 last edited by idefix37

                        @bpeti72
                        A document is published each year by ECMWF under control of WMO about the performance of the models of the European Center and comparing them with global models of other centers.
                        https://www.ecmwf.int/en/elibrary/19879-evaluation-ecmwf-forecasts-including-2020-upgrade
                        (You need to download the pdf)
                        However the parameters which are compared are not those which are directly interesting at ground surface (some years ago msl pressure was compared). And regional models with a higher resolution are not included.
                        Good reading !

                        B 1 Reply Last reply Reply Quote 1
                        • B
                          BPeti72 @idefix37 | Premium last edited by

                          @idefix37 said in Forcast by the time the model is refreshing...:

                          However the parameters which are compared are not those which are directly interesting at ground surface (some years ago msl pressure was compared). And regional models with a higher resolution are not included.

                          so it is not good for what i wanted to use it. :) No problem, it is not your fault. I just dont understand why every company is hiding... but maybe it is financial thing. Well... thank you for your help! :)

                          Korina 1 Reply Last reply Reply Quote 0
                          • Korina
                            Korina Administrator @BPeti72 last edited by

                            @bpeti72 Hello, it is useful to compare each model since as you said, some model can be more accurate for location A, but less accurate for location B. it is not really about personal opinion, but more about connecting the right model with a certain location.

                            Korina

                            B 1 Reply Last reply Reply Quote 2
                            • B
                              BPeti72 @Korina | Premium last edited by

                              @korina I know, you are right and of course i compare them, but it's not easy. Not too much rain here, so i can compare the last few days which model was better, but that is temperature....
                              I'd like to know which model is more accurate for the weekend, coz all of them shows rain, but they are pretty different in time. If there were a statistical publication for each model's x day accuracy, it would be easier... Ok, it is unable to get such a data for every spot, that would be more than huge... I feel weather forcast is like a fortune teller :) Btw. I'm an engineer, so i like exact things, but i know it is not easy in this case... Now It's clear for me that it is impossible to privide what I was asking :)

                              Gkikas LGPZ 1 Reply Last reply Reply Quote 0
                              • Gkikas LGPZ
                                Gkikas LGPZ Moderator @BPeti72 last edited by

                                @bpeti72
                                Meteorology is not an exact science, but ...
                                meteorologists are not fortune tellers!

                                If meteorology was an exact science,
                                it would make our job a lot more straight forward
                                but less of a challenge.

                                B 1 Reply Last reply Reply Quote 1
                                • B
                                  BPeti72 @Gkikas LGPZ | Premium last edited by

                                  @gkikas-lgpz sorry, i didnt mean that meteorologysts are fortune tellers. Not at all... The model itself is inaccurate this is why i told this and my English is not good enough to tell you what i really think. sorry!

                                  1 Reply Last reply Reply Quote 1
                                  • First post
                                    Last post
                                  Windyty, S.E. - all rights reserved. Powered by excellent NodeBB
                                  NodeBB & contributors, OSM & contributors, HERE maps
                                  Terms and Conditions     Privacy Policy