This issue also extends to your Satellite layers. By applying the same smoothing technique to satellite imagery, you are creating ghosting artifacts. For instance, when clouds dissipate or form between two frames, the interpolation shows a slow fading effect that doesn't exist. It masks the actual dynamics of cloud genesis and dissipation. Just like with the radar, professional users need to see the exact structure of the clouds, not a morphed animation.

Subject: Feedback regarding radar layer interpolation

Hello,

I am a user of your Windy application. Lately, I have noticed that you are using a visual smoothing technique for the precipitation radar layers. The issue is that you interpolate the data points to make the image more visually appealing.

Consequently, this creates a problem where, from one minute to the next, it appears as though it is raining so hard that we expect significant rainfall totals, which is not the case at all. For example, currently in Carpentras, France, when I place the information picker, it displays very high precipitation rates. It looks like it’s pouring, but in reality, it isn't.

May I ask/suggest that you give your users the choice between viewing the pixelated version or the interpolated version? From a "scientific" standpoint, the current display does not reflect reality and can even be misleading.

I hope you will take my feedback into consideration.

Wishing you a great day. 

@idefix37

Thank you for your response. However, I believe there is a misunderstanding regarding the nature of my request. I am not referring to the smoothing of values (averaging), but to the temporal morphing and motion advection used for your animations.

From a meteorological standpoint, precipitation is often cellular and highly variable. Your current interpolation algorithm creates a "smearing effect" between two radar scans. By trying to achieve a fluid visual transition, you are artificially spreading high-intensity pixels over areas that are actually dry or experiencing much lighter rain as the picture shows discrete pixels.

This leads to a violation of the Mass Conservation principle:
If a pixel represents 50 mm/h at Time T at coordinates (x,y), and moves to (x+1, y) at Time T+1, your interpolation displays a continuous path of high intensity between these two points. For the user, the Picker tool then displays a high rain rate where, in reality, there is no water falling at that micro-moment.

This visual candy masks the true spatial variability and creates misleading ghost rain. Scientifically, the pixelated raw data is the only ground truth. A striking example occurred during the recent snowfall in Paris. Your map displayed high-intensity radar echoes over my specific area for several minutes, yet not a single snowflake was falling.

This was a direct result of spatial smearing : there were high-intensity pixels (likely amplified by the size of frozen hydrometeors) on both sides of my location. Your interpolation algorithm filled the gap between these cells, creating a bridge of heavy snow on the screen that did not exist in reality. This proves that your smoothing logic creates false positives by artificially connecting distinct precipitation cells.

Could you please consider adding a "Disable Interpolation" or "Raw Data" toggle in the layers settings? This would allow users to see the actual radar grid and the true structure of convective cells.

@idefix37

Ah OK I thought you was.

Thank you

@idefix37

Look here and the reality

some places are not under precipitations. You create artificial datas on area where nothing occurs.

That's what I try to explain.

This issue also extends to your Satellite layers. By applying the same smoothing technique to satellite imagery, you are creating ghosting artifacts. For instance, when clouds dissipate or form between two frames, the interpolation shows a slow fading effect that doesn't exist. It masks the actual dynamics of cloud genesis and dissipation. Just like with the radar, professional users need to see the exact structure of the clouds, not a morphed animation.

@idefix37

Thank you for your response. However, I believe there is a misunderstanding regarding the nature of my request. I am not referring to the smoothing of values (averaging), but to the temporal morphing and motion advection used for your animations.

From a meteorological standpoint, precipitation is often cellular and highly variable. Your current interpolation algorithm creates a "smearing effect" between two radar scans. By trying to achieve a fluid visual transition, you are artificially spreading high-intensity pixels over areas that are actually dry or experiencing much lighter rain as the picture shows discrete pixels.

This leads to a violation of the Mass Conservation principle:
If a pixel represents 50 mm/h at Time T at coordinates (x,y), and moves to (x+1, y) at Time T+1, your interpolation displays a continuous path of high intensity between these two points. For the user, the Picker tool then displays a high rain rate where, in reality, there is no water falling at that micro-moment.

This visual candy masks the true spatial variability and creates misleading ghost rain. Scientifically, the pixelated raw data is the only ground truth. A striking example occurred during the recent snowfall in Paris. Your map displayed high-intensity radar echoes over my specific area for several minutes, yet not a single snowflake was falling.

This was a direct result of spatial smearing : there were high-intensity pixels (likely amplified by the size of frozen hydrometeors) on both sides of my location. Your interpolation algorithm filled the gap between these cells, creating a bridge of heavy snow on the screen that did not exist in reality. This proves that your smoothing logic creates false positives by artificially connecting distinct precipitation cells.

Could you please consider adding a "Disable Interpolation" or "Raw Data" toggle in the layers settings? This would allow users to see the actual radar grid and the true structure of convective cells.

Subject: Feedback regarding radar layer interpolation

Hello,

I am a user of your Windy application. Lately, I have noticed that you are using a visual smoothing technique for the precipitation radar layers. The issue is that you interpolate the data points to make the image more visually appealing.

Consequently, this creates a problem where, from one minute to the next, it appears as though it is raining so hard that we expect significant rainfall totals, which is not the case at all. For example, currently in Carpentras, France, when I place the information picker, it displays very high precipitation rates. It looks like it’s pouring, but in reality, it isn't.

May I ask/suggest that you give your users the choice between viewing the pixelated version or the interpolated version? From a "scientific" standpoint, the current display does not reflect reality and can even be misleading.

I hope you will take my feedback into consideration.

Wishing you a great day. 

@idefix37

Bonsoir,

C'est clair j'ai demandé directement à Meteo France. Données ouvertes mais pas tant que ça tellement ils décrivent mal ou partiellement ceux-ci.

Cordialement

Bonjour la communauté,

Quelqu'un sait-il qu'elle est la projection des fichier hdf5 de lame d'eau sur la metropole ainsi que la couverture ?
J'ai le sentiment que Meteo France ne donne pas toutes les informations. Par exemple leurs fichier .BUFR sur les intensités de précipitation ne sont pas des formats BUFR classiques décodables avec eccodes.

merci pour votre aide éventuelle.

Stéfan

@Ondřej-Šutera Hi, could you tell me where do you source the Meteo France radar data please ?

@Ondřej-Šutera

Hi,

thank you for your reply. But the timeline display map each 5 minutes and you say we get data every 15 min

Hi Windy team,

it seems have differences between Meteo France and you

Have a look on this for today at 12:40 local time for close to Corse Island. The radar echo seems stronger on Meteo France than you. Is it possible to display these as pixels and no as interpolation ?

Best regards.

Stéfan