In addition to Drew Ex Machina and writing articles for various print and on-line outlets on space-related topics, I also work as a physicist specializing in the processing and analysis of remote sensing data from satellites and other platforms. One of the projects with which I have been involved over the past couple of years is called CyMISS (Tropical Cyclone intensity Measurements from the ISS). The purpose of this project is to combine information about the altitude of clouds near the eye of powerful tropical cyclones with cloud top temperature measurements to estimate the strength of these storms much more accurately than can be done using existing remote sensing methods (see “A New Satellite-Based Method to Determine Hurricane Strength”). For this particular implementation of this method, a sequence of images from a satellite in low Earth orbit is analyzed using a new pseudo-stereo technique to derive the altitude of the clouds while infrared imaging sensors provide the required temperature information.
The eventual goal of our project is to deploy a small constellation of miniature satellites to make the required observations of tropical cyclones anywhere in the tropics. As an interim step, we hope to send a prototype CyMISS instrument to the ISS sometime in the next couple of years to provide the data needed to prove the concept. But in the meantime, high resolution images of tropical cyclones acquired at high frame rates are needed to support hardware and software development. As it turns out, astronaut photography from the ISS of tropical cyclones can provide us with the required data like no other orbiting platform.
The Tropical Cyclone Experiment
While astronauts on the ISS have been performing handheld photography of tropical cyclones for years, unfortunately these images are typically of limited use for meeting our goals. They are usually acquired at irregular intervals using a zoom lens employing widely varying focal lengths over a range of poorly known or completely unknown pointing angles. Too many parameters are varying by too much and frequently in unknown ways with too wide a field of view to meet our needs. Images need to be acquire in a very specific way that maximizes their usefulness to us.
After months of working through the appropriate channels in the ISS program office at NASA to institute specific photographic protocols, the Tropical Cyclone Experiment is now an officially approved ISS experiment. When an appropriate target has been identified, the astronauts will fix-mount one of the digital cameras they have available in the ISS Cupola to acquire a sequence of images at one-second intervals for several minutes as the ISS passes over a tropical cyclone. Because of the technical limitations of such photography, performing meaningful, quantitative pseudo-stereo reconstructions of the storms observed will not be possible using these images. But these image sequences can be used to meet the following, more limited objectives:
– Characterize the scene structure near the eye of a tropical cyclone on spatial scales of 100 meters or better.
– Characterize the persistence of scene structure and how it evolves over the course of 100 to 200+ seconds with a temporal resolution on the order of one second.
– Provide realistic image sequences to support the development of pseudo-stereo reconstruction software that will be capable of absolute altitude accuracies of 100 meters.
ISS astronaut photography is currently the only practical means of providing the data we require. The resolution and imaging rate from weather satellites are too low while commercially provided satellite imagery has more resolution than we require and covers too small an area.
While we were working our way towards getting our tropical cyclone imaging procedures officially approved, we were able to use existing channels to direct the astronauts to photograph various storms during the course of this past summer on an “as-available” basis. While most of the images were not useful for our project, on occasion the astronauts were able to acquire images that were adequate to meet some of our objectives. Here is a small sample of some of those images acquired from the ISS this hurricane season.
On September 16 and 17, ISS astronauts were able to acquire images for us of Hurricane Edouard on three different passes as it churned away in the North Atlantic. Here is a typical image of Edouard taken by the astronauts. It was acquired on September 16 at 13:51:49 GMT:
The images the astronauts took were taken using two different cameras with zoom lenses with focal lengths ranging from 28 to 200 mm. A subset of the images with focal lengths longer than 40 mm and with the horizon visible to provide a pointing reference were selected for further study. In order to aid in our analysis, I had to reprocess the images to take out the huge variations in the focal length settings used as well as the changes in the viewing geometry as the ISS passed the storm. Here is a preliminary montage of a dozen images acquired over 82 seconds remapped into a common nadir-viewing projection. For this particular montage, each image uses just the red channel of the original color image and covers 100 by 100 kilometers roughly centered on the eye of Eduardo:
Such image sequences will be useful in characterizing the scene structure on spatial scales of hundreds of meters and how it changes over the course of tens of seconds.
On October 9 and 10, ISS astronauts acquired images of Typhoon Vongfong in the western Pacific Ocean as it was heading towards Asia. Of particular interest to us was a long, six-minute sequence of images acquired using a fixed-mounted camera looking out one of the windows in the Cupola. Although the images were taken with a wide-angle lens and have less than half the resolution we would like, it is proving to be an excellent series of images nonetheless. Here is a typical image in the sequence showing an oblique view of Vongfong not long after the eye of the storm became visible in the sequence. It was acquired at 8:13:51 GMT when ISS was at an altitude of about 418 kilometers with the eye of the storm at a nadir angle of about 51°:
One of the first steps in our analysis requires that all of these images be reprojected to a common perspective and scale. Here is the above oblique view of Vongfong reprojected to appear as if it is being viewed from directly above. This image covers about 2100 by 1400 kilometers.
Here is a close up view of the eye of Vongfong from the same image. This image covers an area of 210 by 140 kilometers at a pixel footprint of 200 m/pixel (close to the resolution limit of the original color image but five times better than GOES weather satellite images).
Since we have a long sequence of images, it was also possible to use two of these images to produce a very preliminary anaglyphic stereo image of the eye of Vongfong. The two images I used for this sample were acquired about 18 seconds apart over a baseline of about 125 km. Like the image above, it covers an area of 210 by 140 kilometers at a pixel footprint of 200 m/pixel. The black area in the upper right represents the part of the scene that was not viewed in both image pairs. While cloud feature motions and image distortions are mixed in with the parallax resulting from variation in the altitude of the clouds in the scene, there is enough stereo information present to provide a convincing 3D effect:
Over the course of the next several months, I will be continuing to analyze the images acquired during the 2014 hurricane/typhoon season which should help meet all three of our near-term objectives. And now that we have an officially approved ISS experiment, with luck we will be acquiring images using our new procedures as the tropical cyclone season opens in the southern hemisphere.
“A New Satellite-Based Method to Determine Hurricane Strength”, Drew Ex Machina, June 18, 2014 [Post]
“RAMOS: The Russian-American Observation Satellites”, Drew Ex Machina, June 21, 2014 [Post]
“A Little Piece of the ISS”, Drew Ex Machina, April 5, 2014 [Post]