Improving Wildlife Tracking via HPC
Improvements in digital biotelemetry tracking devices (biologgers) have enabled researchers to collect accurate, long-term datasets on the movements of animals that would be prohibitively difficult to observe directly in the wild.
Global Positioning System (GPS) biologgers have dramatically reduced in size/weight and can record an animal’s location at accuracies of ~2m for durations of more than a year, or even longer when equipped with a solar panel. For example, the California condors reintroduced to their former habitat in Mexico by San Diego Zoo Global (SDZG) have a <50g solar-powered GPS biologger attached to their wings that provides hourly locations that can be downloaded directly from the Internet.
Biotelemetry has contributed to major advances in understanding key concepts of animal ecology, including resource use, home range, dispersal, and population dynamics. Biotelemetry is also spawning powerful tools for informing strategies for conserving endangered species and habitats. For example, information on animal movements can be matched to the environmental attributes to build a biologically realistic picture of its ranging patterns and habitat use. Conservation managers and regulatory agencies can use this information to gauge and improve the effectiveness of existing and proposed measures to protect animal populations, such as habitat conservation zoning, reserve boundaries and wildlife corridors.
This collaboration between the San Diego Zoo, the US Geological Survey and the San Diego Supercomputer Center brought together wildlife experts and HPC computational scientists to make major advances in the algorithms for processing biologger data, with an initial application to track endangered California condors. We developed an efficient implementation of a 3D movement based kernel density estimator for determining animal space use from discrete GPS measurements. Through a combination of code restructuring, parallelisation and performance optimisation, we were able to reduce the time to solution by up to a factor of 1000x, thereby greatly improving the applicability of the method, and making feasible full 3D estimates rather than just 2D projections. Full 3D results are a critical advance, particularly for species that make large vertical excursions (e.g. birds, aquatic mammals and arboreal species). For example in this case, it was critical to accurately assess the condor home range, including elevation, in order to evaluate the potential impact of a proposed wind farm near the condor habitat.
In addition to enabling 3D estimation, the large improvements in code performance enable science previously not feasible. Obviously it will be easier to process larger amounts of data from greater numbers of animals and more frequent observations. A very fast algorithm allows calculations to be launched and have results returned in just seconds. This might be an extremely valuable capability for a conservation biologist in the field who wants immediate information on the recent range of tracked animals. Finally, researchers are beginning to consider calculations involving multiple animals and overlaps in the space use. These applications are just a beginning and end users will drive new use cases based on the enhanced capabilities enabled by HPC.
Dr Robert Sinkovits is Director of the Scientific Computing Applications group at the San Diego Supercomputer Center (SDSC).
Over the past 20 years, he has collaborated with researchers spanning a wide range of fields, including physics, chemistry, astronomy, structural biology, flow cytometry, finance, ecology, text analysis and graph theory. In all of these projects, the common theme has been the development of efficient software to make optimal use of high end computing resources.
He was the primary developer of the AUTO3DEM and IHRSR++ software packages used for solving the structures of icosahedral and helical macromolecular structures, respectively.
Robert has been actively involved in HPC educational activities and has chaired or co-chaired SDSC’s Summer Institute in recent years.
He was Technology Track Chair for the XSEDE14 conference and will serve as XSEDE15 Program Chair.
Prior to joining SDSC, he was a postdoc and staff scientist in the Laboratory for Computational Physics & Fluid Dynamics (LCP&FD) at the Naval Research Laboratory in Washington DC.
He obtained his undergraduate degree in Engineering Physics at Lehigh University and his PhD in Physics from the University of Connecticut.
In his spare time, Robert is an avid cyclist and mountain climber.