The question Is the sunlight on the GOES-16 Imagery reaching from the east? shows a visible light image and so in the night areas the Earth appears dark and in the day areas the Earth is a bit brighter and the clouds are the brightest.

We can have 24 hour imaging of clouds and weather in thermal infrared satellite images. Below is a a thermal infrared image at about mid day in Japan and Australia. I'd expect the tops of the clouds to be cold and the outback to be hot, and yet it's just the opposite!

  • The clouds are nearly white
  • Central Australia is nearly black
  • The ocean and all of the Asian continent are dull, almost featureless gray; if there weren't a superimposed outline it might be hard to even tell where the water ends and the land begins in some places.

Why is this? What's going on?

Himawari-8 IR1 image from Central Weather Bureau click for full size

infrared satellite image from Himawari-8 at about 12:00 PM Taiwan time


I had a very similar question in a job interview! The only difference is that it was an image from SEVIRI on Meteosat.

The imager on HIMAWARI is called the Advanced Himawari Imager (AHI). The AHI IR1 channel is actually channel 13 with a central wavelength of 10.4 µm, which is in the window region (apparently it's called IR1 in reference to an older satellite). The clear-sky atmosphere is mostly transparent in this region of the electromagnetic spectrum, so we can see the surface or clouds mostly unobstructed by water vapour, ozone, or other gases.

In an infrared image such as this one, high clouds are cold. Forecasters who look at such images for a living like to see clouds as white. I'm a physicist and I prefer to map values of high intensity to white, which would make clouds black, but forecasters are the most important users so they get their way. :)

To display cold areas as white, hot areas must logically be displayed as black. So what we are seeing in this image:

  • Central Australia is hot. This is expected, because it is a hot desert with high emissivity. The image is apparently taken at 12:00Z, which should be between 20:00 and 22:00 legal standard time in mainland Australia, so it's not long after sunset, when the land is still hot. Australia should appear less "black" at an image taken 8 hours later. The image is apparently taken at 12:00 "local" (140°E) time (03:00 UTC). It's February and summer in Australia, and the northern hemisphere looks a lot less hot (it's also earlier in the morning there, compared to sunrise).
  • Cloud tops are cold. This is expected, because they are in the upper troposphere. Those white clouds are ice clouds.
  • There are grey-ish clouds too with little contrast to the ocean. Those are lower, liquid clouds, they are warmer. The lowest clouds may be difficult to identify in infrared images, which is why (for example) night time fog detection from satellites is difficult.

(I got the job!)

  • 1
    $\begingroup$ Beat me to it! But when I looked on JAMSTEC this image matched up with their 0300 UTC image, so 1200 at Himawari longitude (140 E), as per the image text. It might also be worth adding that Asia is earlier in the morning and in the winter hemisphere, which is one reason the land/sea contrast is less pronounced there. $\endgroup$ – Deditos Feb 17 '20 at 15:00
  • 5
    $\begingroup$ @Deditos Huh. Thanks, corrected. I thought everyone in satellite meteorology used UTC. Not using UTC AND not stating the timezone is confusing! :) They also call their full disk image "world wide", which Americans and Europeans may disagree with... $\endgroup$ – gerrit Feb 17 '20 at 15:07
  • 1
    $\begingroup$ In short, the image essentially is a negative. By way of analogy, astronomers also like to look at negative images. Some say the preference is just tradition. But others say that it truly is easier to see dark blobs against a light background (a negative) versus light blobs against a dark background (a positive). $\endgroup$ – David Hammen Feb 18 '20 at 16:16
  • 1
    $\begingroup$ Congrats on the job! $\endgroup$ – David Hammen Feb 18 '20 at 16:20

It is likely because there is already a conversion from the raw data to the grayscale image posted on the CWB website. From this online course (emphasis mine):

[...] using the mathematics behind the laws of radiation, computers can convert the amount of infrared radiation received by the satellite to a temperature (formally called a "brightness temperature" even though it has nothing to do with how bright an object looks to human eyes). Finally, these temperatures are converted to a shade of gray or white (or a color, as you're about to see), to create an infrared satellite image. Conventionally, lower temperatures are represented by brighter shades of gray and white, while higher temperatures are represented by darker shades of gray.

Note that the images you're probably thinking about, where hot = bright and cold = dark (like the first image on the right in the course), also use an arbitrary color scale called pseudo-color. These colors do not correspond to any physical parameter, they are just a representation of temperature made easy for the human eye.

Now I guess the question becomes: Why were these conventions chosen, when they go against each other? This I don't know...

  • $\begingroup$ Thank you for your speedy answer! I'm not thinking of temperature necessarily, I'm just thinking that satellite images for the 0.4 to 0.7 micron band show higher intensity as lighter, as do thermal IR cameras, so Himawari-8's IR1 (channel 13) at 10.4 microns would as well. Maybe the inversion making the clouds white just makes them more intuitive. $\endgroup$ – uhoh Feb 17 '20 at 13:14
  • 6
    $\begingroup$ I can confirm that these conventions were chosen because meteorologists looking at the pictures want that clouds look white, regardless of channel (source: personal communication with forecasters). I'm not sure if representing a single channel as a greyscale image is technically pseudo-color. By that reasoning any narrowband greyscale image would be pseudo-color, even if taken in the visible range, wouldn't it? $\endgroup$ – gerrit Feb 17 '20 at 13:30
  • $\begingroup$ @gerrit I was thinking of images like this: en.wikipedia.org/wiki/False_color#/media/… Apparently they are computed from the grayscale images, and there are lots of different algorithms, meaning that a given shade of yellow (or red...) does not always represent the same temperature: diva-portal.org/smash/get/diva2:797457/FULLTEXT01.pdf $\endgroup$ – Jean-Marie Prival Feb 17 '20 at 14:02
  • $\begingroup$ @Jean-MariePrival Sure. I don't know if there is a scientific definition of false color. Either RGB's (putting three satellite bands or band differences in the R, G, B, channels respectively regardless of actual wavelength) or applying colormaps to single-channel images is commonly done. At least in satellite meteorology, it's unusual to call them false color images. $\endgroup$ – gerrit Feb 17 '20 at 14:07
  • $\begingroup$ The image shown in the question is a gray scale image, so this isn't quite pseudo color or false color. It's a negative. $\endgroup$ – David Hammen Feb 18 '20 at 16:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.