The NERC Earth Observation Data Acquisition and Analysis Service uses NVIDIA DGX systems to reduce training time from months to days.
Climate change is a major issue, and dealing with major issues necessitates the use of large amounts of data.
The NERC Earth Observation Data Acquisition and Analysis Service is one of the few research centers that looks at environmental science from a broader perspective (NEODAAS). Since the 1990s, the service, which is supervised by the National Centre for Earth Observation (NCEO) and is part of the United Kingdom’s Natural Environment Research Council, has made Earth observation data gathered by hundreds of satellites easily available to researchers.
NVIDIA DGX devices are used by the NEODAAS team at the Plymouth Marine Laboratory mostly in United Kingdom as part of the Broad Graphical Processing Device Cluster for Earth Observation (MAGEO) to enable cutting-edge research that utilizes deep learning to open up new ways of looking at Earth observation outcomes.
They can now analyze these troves of data faster than previously thought possible thanks to NVIDIA’s accelerated computing platform.
Earth Under Observation
Sensors on more than 150 satellites circling the globe gather more than 10TB of Earth observation data every day. This would take a lot of computing resources to process and analyze.
NEODAAS installed MAGEO to make deep learning easier to apply to this data and obtain useful insights into the planet’s wellbeing. The wide accelerated computing cluster consists of five NVIDIA DGX-1 systems linked to 0.5PB of dedicated storage through NVIDIA Mellanox InfiniBand networking.
MAGEO was funded by a transformational capital bid from the Natural Environment Research Council during 2019 in order to provide NEODAAS the ability to use a large number of NVIDIA GPU cores to apply deep learning & other algorithms towards Earth observation results.Researchers will use the compute capacity and skills of NEODAAS workers since the cluster is managed as a utility.
Stephen Goult, a computer scientist at Plymouth Marine Laboratory, said, “MAGEO offers a wonderful opportunity to advance artificial intelligence and environmental intelligence science.” “Because of its similarity to the NEODAAS database, it allows for accelerated prototyping and training utilizing vast volumes of satellite data, which can change how we utilize and interpret Earth observation data in the long run.”
The NEODAAS team will conduct forms of research that would otherwise be impossible using NVIDIA DGX systems. It also encourages the team to greatly intensify their studies, reducing preparation time from months to days.
NEODAAS was also awarded funding to finance the running of an NVIDIA Deep Learning Institute course, which was made accessible to representatives of the National Centre for Earth Observation in March and aims to promote AI growth and training in the environmental and Earth observation fields.
“The course was a huge success; participants left feeling well-informed and excited about applying AI to their research,” Goult said. “As a result of the discussions held during the course, several new projects leveraging AI to solve problems in the Earth observation space have been developed.”
Transforming Chlorophyll Detection
The NEODAAS team has used MAGEO to work on innovative methods that have shown important insights into the essence of Earth observation results.
One such achievement is the creation of a modern chlorophyll detector to aid in the monitoring of phytoplankton concentrations in the world’s oceans.
Phytoplankton, which is microscopic in size, is a food supply for a vast range of aquatic creatures, from small zooplankton to massive blue whales. However, they often have a function that is beneficial to the planet’s environment.
They use chlorophyll to absorb sunlight, which they then convert into chemical energy through photosynthesis, just like any other land plant. Phytoplankton absorb carbon dioxide during photosynthesis. When phytoplankton dies, the carbon byproduct is transported to the ocean’s floor, and when phytoplankton is absorbed, it is carried to other levels.
Phytoplankton is responsible for the movement of around 10 gigatonnes of carbon from the atmosphere to the deep ocean per year. Phytoplankton are critical in mitigating ambient CO2 and the impacts of climate change, since elevated CO2 levels are a significant contributor to climate change. Also a minor decline in phytoplankton development may have disastrous effects.
NEODAAS collaborated with scientists to create and train a neural network that allowed a new kind of chlorophyll detector for studying phytoplankton abundance on a global scale using MAGEO. The system employs statistics on particulate beam-attenuation coefficients, which are determined dependent on the energy expended by a light beam passing through seawater owing to the existence of suspended particles.
Scientists can now produce reliable chlorophyll measurements with a lot less money and with a lot more data than they did with the former lab-based process, which was once called the “gold standard” in high-performance liquid chromatography.
“Thanks to the extremely parallel environment and computational efficiency powered by NVIDIA NVLink & the Tensor Core design in the NVIDIA DGX systems, what might have taken at least 16 months on a single GPU took just 10 days on MAGEO,” said Sebastian Graban an industrial placement student at Plymouth Marine Laboratory. “The qualified neural network that emerges has a strong degree of precision in predicting chlorophyll, giving experts a simpler, quicker way of tracking phytoplankton.”