Animated satellite maps
This example maps several days of cloud/storm movement during Winter Storm Eboni, from Dec. 26-28, 2018, using GOES-16 satellite imagery.
Similar code could be used to generate animations of interesting storms or weather patterns. Satellite data is most easily downloaded here.
Usage
First run get_data.sh to download satellite images using rclone. This will take a while, and you'll need to set things up using these instructions if it's your first time. As it is, the script is set up to download readings taken every 15 minutes (even though the satellite takes them every 5).
Once the data is downloaded, make.sh will stitch a bunch of images into a rough movie of the storm. It could use some adjustments in terms of contrast, gamma, etc., but the results should be reasonably clean for a first pass.
Dependencies
Ooooooh boy. Get ready for it.
The big dependency is GDAL. But not just any GDAL. You need GDAL that can also work with NetCDF files, which can only be installed by following obscure incantations.
You'll also need ffmpeg to make movies (brew install ffmpeg) and imagemagick to handle image annotation and conversion.
Notes
Working with satellite images seems to require knowing obscure factoids about how satellites work. Some random useful things to know are:
- Data is natively represented in raster form using NetCDF files, which are like special raster files for scientists. These can easily be converted to GeoTIFFs using GDAL.
- The native satellite projection is weird because images are taken from spaaaaaace. Directions for projecting the images into a more familiar form are here. Once you project the data into something sensible, you can manipulate it as you would other GIS files — such as, say, overlaying vectors.
- The data we want is from the CONUS (Continental US) dataset, not the Full Disk dataset, which is huge.
- Examples here work with multi-band files (AKA "ABI L2 Cloud and Moisture Imagery: Multi-Band Format"), which encode all of GOES-16's individual bands into a single file. Descriptions of those bands are here, but the most important ones for this are Blue (Band 2), Red (Band 1) and veggie (Band 3) — which can be converted into a faux green band for the purposes of creating true-color imagery (see instructions in
_plot_image.py).
Todo
These examples are the barest possible representations of animated satellite data, so future work could be focused on making it look sharper. This includes:
- Making gamma and contrast adjustments and otherwise sharpening the results
- Potentially trying to use higher resolution imagery so the zoomed-in version doesn't look so blocky. (Test :Rad files and see if this helps?)
- Labels
Resources
In no particular order:
- This amazing code from OCC
- GOES-16 band reference guide
- Manipulated GOES-16 data with GDAL
- GOES-16 S3 explorer
- Conquering the Earth with cron
- GOES-16 Product Manipulation Using Free Software Tools
- Mapping GOES-16 data (notebook)
- GOES imagery viewer
- Source: How we made Billions of Birds Migrate
- NICAR 18 data blitz
- Manipulating GOES-16 data with Python
- Some useful stuff from Iowa State
Questions?
Contact chase.davis@gmail.com