An alliance of government agencies is taking a deep dive, as it were, into the world’s oceans as part of a larger project to develop a comprehensive environmental Earth model that could more accurately make predictions about weather and climate. The new model could enable forecasting events ahead of time, by days or even decades.
The Naval Research Laboratory last week awarded Vencore a five-year, $23 million contract to provide scientific analysis of ocean data gathered on-site and via remote sensors, and to provide ocean model programming and validation, according to an announcement. In addition to supporting the Navy’s next-generation ocean models, work by Vencore–a Chantilly, Va.-based company that provides analytics to the U.S. government–will feed into the multi-agency effort to build the National Earth System Predication Capability (ESPC) model.
The National ESPC is a collaboration among the National Oceanic and Atmospheric Administration, the Navy, Air Force, Department of Energy, NASA, and National Science Foundation (NSF) to combine weather and ocean data into a prediction model capable of identifying trends in climate, improve predictions, and “help decision makers address critical policy and planning issues by extending the national predictive capability from hours and days to seasonal, annual and decadal time periods through improved, coupled global environmental prediction.” In addition to ocean studies, the work will also include the atmosphere, land, cryosphere (frozen water), and space.
Beyond deploying ships, sensors, and other devices to gather information, gaining such a detailed global perspective is also a computing challenge. The National ESPC is developing a High Performance Computing (HPC) architecture to develop its Earth system models. New techniques in machine learning, which is something Vencore employs in deep data analytics, are also a way to draw useful conclusions from the high volume of gathered data.
Machine learning is relatively new in environmental science, but it is making headway. NSF uses machine learning in its data-driven approach to understanding climate change. In a paper presented early this year at an American Meteorological Society conference, researchers said that using advanced machine learning techniques for both average conditions and extreme events were comparable to or superior than traditional methods. Machine learning has been able to predict earthquakes in a laboratory setting. A New Mexico startup is using Department of Agriculture data and machine learning to predict crop yields.
A global model that could reliably predict weather and/or climate–including, possibly, natural disasters–in the near- and long-term would be a boost to both government and business, as well as the public at large, which is why the government has been working on it for almost 10 years. The project started in 2008 as the National Unified Operational Prediction Capability (NUOPC), made up of NOAA, Navy, and Air Force. ESPC kicked off separately in 2010 with the additions of NASA, DoE, and NSF. In 2016 it received an updated charter to officially merge the groups into the National ESPC.