With a veritable army of robots scanning its surface, Mars is the most surveilled planet after Earth. Those rovers and orbiters collect a mountain of image data, but very little is processed into an easily viewable format. The US Geological Survey Astrogeology Science Center is looking to change that. Leveraging supercomputers and cloud computing, the USGS has released 4,800 digital terrain models (DTMs) and 155,000 high-resolution images of the red planet.
The newly released data comes from the Mars Reconnaissance Orbiter (MRO), a satellite observing Mars for almost 20 years. The satellite has several vital instruments, including the Context Camera and HiRISE camera, and they collect a great deal of data on the Martian surface. However, scientists don't have time to analyze and process every frame manually. The USGS hopes that by releasing the data in a usable format, researchers and the public can use them. "Now anyone on the planet with a smartphone can search, use, and marvel at these data," says Jay Laura from the USGS Astrogeology Science Center.
The digital terrain models come from the Context Camera (CTX), designed to capture background images around smaller points of interest. It has a resolution of 6 meters per pixel and can capture areas as large as 30 kilometers (18.6 miles) wide and 160 kilometers (100 miles) long. To produce a DTM with accurate terrain height, the Astrogeology Science Center combined two overlapping images using the NASA Ames Stereo Pipeline tools.
It wasn't as easy as pressing a button and waiting, though. The team had to align the DTMs to existing low-resolution topography maps to align the images. Then that data was fed back into the computer for processing into DTMs. Using a regular PC would have added years to the project, but the USGS has its own supercomputer for this kind of thing. The Denali system has 9280 CPU cores with a peak speed of 660 Tflop/s. It was able to process hundreds of DTMs at a time.
The high-resolution images from the HiRISE camera will have a bit more general appeal than the DTMs. HiRISE is designed to capture fine details with a resolution of 25 cm (9.8 in) per pixel. However, it scans a smaller swatch of the landscape, measuring 6 kilometers (3.7 miles) by 60 kilometers (37 miles). Each file can be more than 1 gigabyte, making releasing the full catalog challenging. The Astrogeology Science Center developed a processing pipeline with an HTCondor High Throughput Computing environment and an Amazon AWS slot. The data was streamed from the NASA Planetary Data System servers in batches of 4,000. The team processed all 114 terabytes of image data in a mere four hours.
The result of all this data processing magic is a collection of thousands of new Mars images, ready for your perusal. Some of the files are huge, which will be ideal for research. There are also smaller thumbnail images that are still thousands of pixels wide. That's more than enough to marvel at the alien landscape.
