Project PI: Paul Morin, University of Minnesota Polar Geospatial Center

The Surface Extraction from TIN-based Search-space Minimization (SETSM) algorithm, initially designed to extract elevation data over ice sheets, has been refined and optimized to handle stereoscopic imagery over any landcover. After an initial preprocessing step that corrects the source imagery for sensor-specific detector alignment artifacts, SETSM takes the two source images and derives increasingly detailed elevation models using its pyramid-based approach.

The poles now have better dependable topography than almost anywhere else on earth.  In addition, the ice on earth has better topography than the land surface on Earth.  The team has produced over 200,000,000 km² 2 meter posting topography covering the Arctic over 8 times.  The data has been adopted by Greenland, Iceland and Canada as their national standard elevation data.

After the polars were completed, the work has moved on to mapping the rest of the Earth’s landmass.

Project PI: Compton Tucker, NASA

Blue Waters is used to determine arid and semi-arid carbon stocks over millions of km2, from the Sahara Desert southward to the savanna belt. Satellite imagery is used to compute tree crown area and DEMs are used to estimate vertical tree heights enabling a determination of woody tree biomass or woody carbon mass by allometry.

This work involves taking a very large volume of 50 cm satellite data and assembling mosaics by UTM Zone for processing. Deep learning is then used to segment and map discrete tree crowns over millions of km2. Associated tree crown shadow length and solar position are used to determine tree crown height trigonometrically. Tree heights are compared to tree heights generated from the SETSM EarthDEM project.

Project PI: Nikolaos Pavlis, NGA

The Earth’s time-varying geomagnetic field, fundamental to studying the interior structure and evolution of the Earth, and vital for navigation and protection from solar particle radiation, is challenging to simulate.  Models must account for time scales ranging from less than a year to billions of years and spatial scales ranging from centimeters to thousands of kilometers, requiring extremely high temporal and spatial resolution.  Accurate prediction of secular variation (SV), or changes in time, can be achieved via large-ensemble assimilation of geomagnetic observations and theoretical geodynamo models, but requires petascale computational resources to reduce research time to weeks instead of years.

The main objective of this project is to investigate the convergence of assimilation with different ensemble sizes and simulation resolutions for given physical parameters. The research team found that the ensemble size of approximately 256 is optimal for assimilation, based on the computational needs and the forecast accuracies. This result is very important as it establishes a quantitative correlation among the forecast accuracy requirements, computational resource needs, and time periods for progress. Optimal ensemble sizes can greatly reduce the computational expense and research time without compromising research objectives.

Project PI: Garret Vo, NGA

This project will use the robust regression algorithm to perform object detection among millions or billions of images. The robust regression method detects objects in a single image with heavy noise and non-uniform background.  The algorithm was originally developed to detect objects in electron microscope images with heavy noise and a non-uniform background. To detect objects in the input image, the method performs image binarization to turn an image into black and white using a two-stage approach. First, the method estimates the image background and removes it from the original image. Then a global thresholding is applied to obtain the binary image.

This project will apply the method at a large scale to detect objects in millions or billions of images with the Blue Waters system. Applications include detecting biological cells in large number of electron microscope images, or isolating animals in remote sensing images with low resolution and heavy noise.  With this effort, NGA can potentially detect objects in commercial images at large-scale to inform policy makers.

Project PI: Kimberley McCormack, NGA

In this project, NGA will provide global foundational datasets to serve as inputs to the various models in the inter-agency hydrology modeling framework referred to as the Poseidon Project. This includes TanDEM-X Temporary (TDT), a global digital elevation model (DEM) with 12 meter resolution. Additionally, NGA will provide a suite of products derived from TDT, including a hydrologically conditioned DEM, flow direction and accumulation rasters, stream networks, and coupled hydrologic basins. These datasets promote consistency between Poseidon models and improve resolution compared to currently available datasets. Previously, the best available global hydrologically conditioned DEM has a resolution of 90m. The TanDEM-X hydrologically conditioned DEM will be a dramatic improvement and allow for the best-available datasets to be leveraged for a wide array of hydrologic applications.

The primary codes for this project are DEM processing algorithms built by NGA scientists, along with the open-source TauDEM and gdal software packages. The process reads in geotiff-formatted elevation tiles (1 degree x 1 degree), along with raster datasets for surface water occurrence, hydrologically conditions the DEM data, and calculates the global hydrography based on the conditioned DEM.

Emergency response to flooding following extreme weather and other events is limited by the lack of accurate and timely maps of flood extent and depth. Image-based flood mapping approaches are limited by daylight and miss areas obscured by buildings, clouds, cloud shadows, trees, and other vegetation. Alternative approaches based on hydrodynamic models depend on stream gauge data, which is available in only a limited number of locations where measuring equipment is installed.

When deployed as an automated workflow, the FIST model will decrease the delivery latency of critical wide area flood depth maps to emergency responders. Accurate and timely flood maps based on observed data will be invaluable for mobility, logistical, structural, and economic impact analysis.

The Advanced Visualization Lab used Blue Waters to create renderings of the Digital Elevation Model (DEM) data provided by the Polar Geospatial Center and processed on Blue Waters, for the above-mentioned Atlas of a Changing Earth documentary. The specific visualizations rendered using Blue Waters are of the Jakobshavn glacier in Greenland. These visualizations utilize multiple spatial resolutions of the DEM data as well as temporal timesteps. Blue Waters is also used to render the Vavilov ice cap region in Russia, and an ArcticDEM data buildup showing how the data image strips are collected over time.

The New Frontiers Initiative supports full-year graduate fellowships to advance geospatial intelligence research among PhD candidates. The fellows receive stipends, access to the Blue Waters HPC system, and technical support from the Blue Waters staff.

Education allocations on the Blue Waters system are readily available to support workshops, training events, internships, fellowships, undergraduate and graduate classroom activities, and other creative learning events offered by the national community.