Disappointed with the lack of up-to-date, freely available global mosaics in true color, last month I decided to create a website that provides the latest global visible and IR composites every 30 minutes from five geostationary satellites, GOES-16, GOES-18, Himawari-9, Meteosat-9, and Meteosat-10. The site is called earth-now.
Imager data for GOES-16, GOES-18, and Himawari are provided by the National Oceanic and Atmospheric Administration (NOAA) on public AWS S3 buckets. Meteosat data must be obtained from the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) by creating an account and using their data store.
I wrote a Python script that automatically locates and downloads the latest files for any particular channel from each satellite. Data is processed using Satpy, a Python library which provides tools to create composite images from satellite data.
Full disk data is first projected into the Mercator projection with Satpy. True color images, generated using a simulated green channel, are utilized for GOES and Himawari. The Meteosat-9 and Meteosat-10 SEVIRI instruments are only capable of producing all three required visible channels for portions of the full disk. Thus, Meteosat images are "natural color" composites, which use an IR channel. I created a simple custom composite in Satpy using the built-in DayNightCompositor that generates the visible composites only on the day side. On the night side, multiple IR channels (both shortwave ~3.9 um and longwave ~10.3 um) are used to view clouds and landmasses.
To blend the images without too much computational overhead, I opted to create custom alpha masks based on satellite zenith angles. We can get satellite zenith angles for each pixel in the image using the same AreaDefinition for our images by calling the function get_satellite_zenith_angles(). We can then normalize this data between 0 and 255, with the alpha value set to 255 below a lower angle threshold and 0 above an upper angle threshold. This results in a smooth transition effect near the edges of the satellite's full disk. Through trial and error, I found a lower angle value of 70 degrees and an upper angle value of 85 degrees to provide the best results. This data is only generated once and is saved to a txt file for each satellite.
For each satellite, the visible and IR composites are combined, and the alpha mask applied to the resulting image. Every satellite's image is finally combined into a single global mosaic.
24 hour timelapse July 6/7, 2023
Overall, I am very happy with how the images have turned out. Unfortunately, there is a limit to the quality of the images due to the different composite types. At the moment, there is not much I can do to resolve this with the amount of compute available on my web server besides perhaps applying a filter to the Meteosat images.
I hope these images may be of use to you and that you find them as pleasing to look at as I do!