Version 2 - Windy Customized Graphics

  • Thanks for the practical feedback, it's welcome input. I'll look into the brightness of the precipitation background and bright cloud.

    The present cloud layer is 20% larger and shading slightly more refined. I've expanded the size of some overlays to better explore the localized detail in short-period models.

  • It was a quick correction, try them now.


  • Sailor

    OK it looks much better, I shall try it.

  • I finally managed to get really detailed cloud layer depictions without over-cooking brightness or high-pass filtering them.

    The setting is here:

    Import the setting into all four of these cloud overlays.

    Clouds [clouds]
    Low clouds [lclouds]
    Medium clouds [mclouds]
    High clouds [hclouds]


    Hopefully that's a final version (unless feedback suggests otherwise).

  • EDIT:

    This setting was removed and the new setting posted below, the graphics have been kept for comparison.


    NAM Model

    NAM Model

    NAM Model

    NAM Model

    ICON Model

    ECMWF Model

    ECMWF Model

  • I changed the display range of temperature to show between –80C to +60C, with display increments of 0.25 degrees C.

    See the edited text within the post above for details of changes, and the new setting.

    I likewise updated the modeled graphics (see the seven images immediately above) with the new setting's current appearance. Note the shading changes around the freezing point as there's much less white or brightness near to 0 deg C.

  • This Temperature [temp] setting has been removed and an updated setting posted below at this link:

    The graphics have been retained for visual comparison.




  • NOTE: This setting has been removed and replaced with a new Pressure overlay setting that is found at this link, the graphics have been kept for comparison purposes.


    I reworked the pressure overlay to show better detail and contrast while also reducing its brightness.


    Pressure [pressure]

    0_1542376359940_NAM Pressure.png

    0_1542376375127_ICON Pressure.png

  • Sailor

    I have tried your new settings for Rain. They are better but need further improvements.

    1- The background is not dark enough to see correctly snow, lightnings and towns. Your new setting is 200,200,200, I personally use 170,170,170 and Windy settings are 111,111,111

    2- The shading for small rainfall gives wrong information. See red arrow


    This is because you have a light blue colour below 10mm which becomes darker between 10 and 20 mm and then lighter beyond 20mm.


  • @idefix37

    I disagree with you in all areas in this instance.

    The background brightness level is currently more than sufficiently dimmed to clearly depict the snow, rain and storm icons. So I won't be making the backgrounds any darker. I much prefer brighter displays and brighter colors. That’s one of the many reasons I wanted to rework the default windy overlays, they’re very dull and dim, and many are rendered very unappealing, ugly or more-or less hard to assimilate useful information from them as a result. Thus I won’t be sacrificing brightness without very good reasons too, just to end up back with dull overlays. I don’t intend to converge on Windy’s default overlay setting.

    As for that displayed shading gradient showing "wrong information", I entirely disagree with you, it's a different shade of blue, and is visually separated by a darker shade on the display in between, clearly indicating a higher rainfall intensity where your arrow is pointing. I don’t think anyone could mistake that for a lower rainfall intensity.

    Other major items involved in this situation are that Windy's default overlay depicts a mere 50mm of intensity range which for any person living in the tropics, that is a very inadequate rainfall range, per modelling forecast step.

    My current replacement overlay remedies that via increasing the display range from 50 mm up to 150mm, which is actually still much too low for the tropics (I've seen a 6 hour period where we received 760mm in five hours and 20 mins!). I would make that range much higher than 150mm if I could, but windy has a limit on how many lines of detail can be displayed within those precipitation overlays, before it ‘flips-out’. I have taken the detail level about as far as I can, due to that display limitation.

    I would like to make the display range of the three precipitation overlays about 300mm during each 3-hour model-step period, but I currently can’t do that plus maintain a desirable detail increment.

    Another major problem with the default precipitation displays within Windy is that many of them are non-linear depictions (see the graphic value increment non-linearity in the left column(s) below). The non-linearity strongly affects the resulting display overlay depiction’s intensity mapping (i.e. it’s not showing you what the forecast model output is actually indicating).

    Default Rain, thunder

    0_1542517871735_Rain thunder.png

    Default Rain accumulation

    0_1542517880935_Rain accumulation.png

    As a result I made all of my overlays entirely linear so that you see all of the events within their correct proportions and intensity levels, with the colors scaled appropriately toward that end, to show relative forecast intensities with a very quick glance at any area. Thus providing the correct perception of intensity to anyone looking at the forecast graphics. So the overlays are very much showing clear unbiased depictions of the full and correct information detail.

    I’ve also greatly reduced the linear increment sizes between changes so you can see specific geographical areas to be affected at the forecast higher intensities, much more clearly. The Accumulated Rainfall overlay for instance is now fully linear, in very close 5 mm increments all the way up to 1,500mm (have a look at Windy's default values for the same overlay … which overlay is showing the “wrong information” again?).

    Due to the above mentioned limit on the lines of display available, the only other option would be to reduce the display detail via enlarging these range size increments. But I don’t think that’s appropriate or desirable as then you’re forced back into using a less accurate non-linear display of the precipitation intensity and its proper perception and information assimilation problems come back with that.

    But using finer increments requires finer shading gradients and many more of them to show that extra detail level of changes, and the showing of relative intensity changes over time and geography, likewise within the color gradients. I'm very satisfied that the current shading does precisely that in the proper way. I reject an interpretation that they show “wrong information”, when the reverse situation is what's actually true.

    Further, your display on the left-hand side of your graphic shows red levels at such low rainfall intensities as a mere 25 mm. Which is way too much color emphasis for what is a low intensity of rainfall.

    Within the tropics 100 mm/hr is not exactly a commonplace event, but 30 to 50 mm/hr sure is. Do that for a few hours and you get up to the 250 mm to 400 mm of rainfall rather fast. Those levels are typical of tropical convergence. So where do you think red should be used if depicting intensity of an event like that, anywhere on a whole global model’s display?

    Which is why in my precipitation overlays I have the gradients of the relative colors situated approximately where they are, within the rainfall, accumulated rainfall and thunderstorm overlays for each model run's step-period.

    I’m very satisfied an altered shade of blue is appropriate for a mere 25 mm of rainfall within a global forecast model’s display of rain. The other complicating factor here is that rainfall which constitutes a flood, in one part of the world, is considered fairly insignificant rainfall within another part of the globe, with a very different drainage or geography.

    That means colors and shading have to be a (subjective) trade-off, right?

    560 mm of rain results in a significant flood anywhere that it falls within a day or two, so within my overlay of accumulated rainfalls it uses red at 560mm of accumulated rain. And for the short-period steps within the, “Rain, Thunder [rain]” overlay, I use red at 112 mm of rainfall (which is displayed linearly in 1 mm increments up to 130 mm, btw) as that's clearly a very dangerous amount of short-period rainfall being forecast. And 20 to 25 mm in three hours is generally not too dangerous.

    So I set the colors to be meaningful as relative forecasting indicators for eyes to pay attention to, and not just for depicting changes at low-end levels of rainfall. I feel that’s how the color gradient should be scaled to provide meaningful visual forecasting information of significant upcoming events. Otherwise how will the viewer discriminate or perceive what’s a significant or dangerous rainfall event, if red is used at a mere 25 mm of rainfall?

    So there’s no agreement between us on this occasion, sorry. But thank you for your critiques and your analysis, it is helpful to me to see where the different emphasis lays, and I’ll certainly be keeping those points in mind if I rework the rainfall shading at some future point.

    A practical example of the routine differences in using these overlays:

    Here Windy's display is pegging a full 86.7 mm outside of its displayable range (the large white region is outside of its scale, so displays zero intensity details above 50mm) and the geographical accuracy of the relative intensity surrounding it is also very significantly degraded as a result of both the lack of lines of increment detail, and the strong non-linearity of the display depiction, producing needless interpolation.

    Default "Rain, thunder" overlay:


    My replacement overlay of the same time-space event is displaying only 6.7 mm outside of the available displayable range for this 'extreme' end of rainfall events, within this ECMWF model run step, and the geographical depiction of the intensity is clearly much more accurate and meaningfully depicted with this overlay, due to the finer linear display increments used and due to the meaningful color shading choices used, as well, indicating where the hazard area is projected to be.


    It's very clear within this example which is conveying the "wrong (non-linear) information" to the viewer, or is likely to mislead a person about real intensities and areas worst affected by the highest rainfall levels, and their timing and transit. Which would you want to know, or would find the most relevant to you, if near that event's path? Which can you assimilate more quickly and meaningfully?

    Here's another practical example of the difference that detail and linearity of display makes:

    Windy Default Rain Accumulation

    My current "Rain Accumulation" display:


    One is displaying completely within it's scale, is linear and is very detailed and the other is not. Which is more useful? Which has color-shading that's meaningful over a whole region, at a glance? One is giving you a distorted perception of the forecast, and the other is showing you exactly what the model's forecast is predicting, and where.

    One could just dismissively say, "Ho-hum, it's just a model, it does not need that much detail". Yes, it is just a model, but it's also 'trend-mapping' delta over time, and if you can already fully see what was in the prior forecast, with all available details, then you can also see when the trend-map significantly changes, and by how much. If a trend speeds up, slows down, amplifies, decays, moves north, south, east, or west.

    One of the above overlays allows you to see those changes in trends very easily and early - the other, not so much.

  • Sailor

    Sorry for having upset you so much. I get some of your arguments, but not all of them. I do not consider that Windy colour settings are perfect as I have set my own colour scheme for Wind layer.
    So let’s leave it there.

  • Come on, you haven’t upset me, I’m just showing you something about the flaws within the way the overlays have been set.

    Take a good look at the darkest areas within this image, you’ll see that the darkest areas contain the highest levels of visual detail compared to the rest of the image.


    And it should not be doing that, as the highest visual detail area is occurring where the rainfall level information available from the model is actually almost non-existent, it’s down at the 1 mm to 5 mm level, so the detail in that area is being way over represented within that image.

    Now look at the brightest (skin-tone) pink areas within that same image. Those brightest pink areas are of course where the actual rainfall information supplied by the supercomputers at ECMWF are at their highest level, within the model’s output ‘product’. But as you can see within the default Windy overlay image that’s where the visual detail which is by far the most relevant to human beings, is actually completely absent from 500 mm increment level up in the default overlay display.

    My replacement overlay of the same area shows that rainfall as high as ~950 mm is actually being forecast within parts of that brightest pink area (see the blues near Taiwan).


    But none of that rain intensity is visible within Windy’s default overlay. The more important information for human beings is cutoff at the 500 mm level and anything above that is effectively assigned to one color and made invisible, because the next increment in the display line is set at 8,000 mm, a jump of 7,500 mm, which is 62.5 times higher than the previous increment, which was 120 mm!

    So the most important rain forecasting information for humans is thus made invisible when using Windy’s default overlays.

    This is just one of the flaws I’ve removed within the precipitation overlays. So if you say to me that you think my overlay is showing the “wrong information”, compared to the default, I’m of course going to tell you just how wretched the default display really is at displaying about 60% of the available information range it should show. The default display setting is basically heavily biased, and biased in entirely the wrong way, because where there’s the least rainfall detail available there’s the most visual details displayed, and where there‘s the most rainfall information available, there is no visual detail of it being displayed at all.

    i.e. this is why a linear display with closer increments is the way to display the entire information range available and not just the lowest 40% of it.

    As you said, “I do not consider that Windy color settings are perfect”.

    Yes, that’s understating it. Many of them are not displaying the majority of the information range available and most of them are non-linear and biasing the display to show the least details where they should be showing the most details. So I’d say they're quite a long way from perfect, and that's not to be tip-toed around.

    I said this to you at the beginning, if you remember:

    “ ... People can also edit them, as they see fit, and even use them as a template to do that if they wish, to their heart's content. If they make something even better then they can post their edit in here, or elsewhere, and all benefit. …”

    I meant that. ;-)

  • I created another shading that's designed to bring forward the otherwise hidden details within these new overlays. It still needs cleaning-up and tweaks but as you can see there's a lot of visual information there that's not normally seen. I'll make a more detailed version available soon-ish.




    Next 10 days

  • I finally figured out how the Rain, thunder [ptype] (precipitation type) overlay works, and got it working properly with a more suitable color combination:

    Place it into this overlay:

    Rain, thunder [ptype]





    Note that none of the models seem to model 'Mixed Ice', which I've set to yellow within the display overlay, so don't be concerned if you don't see any yellow on the maps.

  • The most detailed precipitation related overlays are all listed below. They use a common shading palette optimized to display maximum visual detail (screen image examples are given at the end).

    EDIT: The settings have been improved upon and replaced, they're now available from here:

    The graphics have been kept for comparison.

    Default Windy setting - Rain Accumulation
    0_1542787285241_DefaultSettingRain Accumulation.png
    New setting - Rain Accumulation
    0_1542787271244_NewSettingRain Accumulation.png

    Settings replaced and available here:

    continues below …

  • Settings replaced and available here:

  • Examples:

    Rain accumulation - Rain Shadows around Hawaiian Islands
    Rain accumulation
    0_1542714847807_Rain accumulationSA.png
    Rain accumulation
    0_1542714871795_Rain accumulationNG.png
    CAPE Index
    0_1542714924960_CAPE index.png
    Rain, thunder
    0_1542715087779_Rain, thunder.png


  • Temperature displayed at 45,000 ft (FL450) over the western Pacific's Equator (-65C).

    0_1542824026796_Screenshot_2018-11-22 Windy as forecasted.png

  • Cleaned up the visibility overlay for pilot flight-rules indications. It uses the default Windy visual range distances but provides better shading and discrimination of boundaries.

    Place into the following overlay:

    Visibility [visibility]




  • Seem a bit nippy to you for November?

    0_1542952969423_Screenshot_2018-11-23 Windy as forecasted.png

    First 'Winter' Storm (lasts for most of next week).
    0_1542876670701_First Storm 2.png
    0_1542876287188_First Storm.png

    New snow ... looks like January, in November.

    Getting an early workout of the new winter graphics.

Log in to reply

Windyty, S.E. - all rights reserved. Powered by excellent NodeBB
NodeBB & contributors, OSM & contributors, HERE maps