Comment by crazygringo | Hacker News Reader

Comment by crazygringo | Hacker News Reader

Parent Full thread

First of all, this is super super super cool.

But also... I'm super confused about the nighttime imagery.

Looking at daytime imagery from this afternoon over the US, everything looks legit -- it looks "real", shadows from clouds get longer as the sun gets lower in the sky, shadows on mountain ranges change, etc.

But as soon as it transitions to night, all realism goes out the window. Urban areas are absurdly bright, clouds are still blindingly white on top, the pattern of lights don't change even a tiny bit throughout the night... it's like it's no longer using photos but combining a static nighttime image with radar-detected clouds drawn on top. The gradient between daytime and nighttime also seems highly artificial.

Do other people agree that the nighttime imagery is totally simulated, not photographs at all? I'd just kind of like to see a version that showed real nighttime imagery, even if it was mostly super-dark (but they could HDR it which would be fine too).

> I'd just kind of like to see a version that showed real nighttime imagery, even if it was mostly super-dark (but they could HDR it which would be fine too).

It wouldn’t just be super-dark; it would be black. The GOES visible light sensors do not produce any useful data after dark. (See it yourself: https://www.star.nesdis.noaa.gov/GOES/fulldisk_band.php?sat=... ... this is the blue channel.)

The image that’s being used here (for the US imagery) is what CIRA calls GeoColor: true-color imagery during the day (with simulated green channel; the GOES-R satellites don’t have a green channel), and multispectral IR composited on top of a static city lights image at night.

This is using GOES imagery for "live" North America which uses multispectral IR for night [1] but True Color for day. Once you pass a certain zoom level, it switches to historical imagery from Microsoft.

Side note, you can receive GOES images straight from the satellite yourself with a $30 SDR, a $30 LNA, and an antenna made from scrap (or ~$100 if you're not handy).

[1] https://www.star.nesdis.noaa.gov/GOES/fulldisk.php?sat=G16

Ah ha that explains it perfectly, thanks.

I've always been curious what a real day+night satellite image looks like. The simulated ones always show urban areas as absurdly bright.

Does anyone have links to satellite imagery that shows the actual full real transition from daytime to nighttime, at a constant camera exposure so without exaggeration?

> constant camera exposure

Sure, just take an image and color the night part completely black. Street lighting is around 10 lux -- sunlight is 100,000 lux.

This isn't constant exposure throughout but at some points you can see the transition pretty clearly such as 21:32, some great shots: https://www.youtube.com/watch?v=NM2wtte1JRE

Yes, the night side uses a combination of infra-red imagery with a static image of the Earth at night. Without that image all the landmass and cities would disappear into black, so it helps give the IR imagery context. And it looks nice. It’s a common technique called GeoColor.

Reply on news.ycombinator.com