This project will deliver a purely data driven global specification of the high energy electron environment near earth that includes retrospective real time and forecast outputs using machine learning…

This project will deliver a purely data driven global specification of the high energy electron environment near earth that includes retrospective real time and forecast outputs using machine learning techniques. The proposed work will improve the existing shells electron radiation belt model that was created to demonstrate the feasibility of the neural network technique. The demonstration model uses a neural network to learn the mapping from low altitude NOAA poes electron flux measurements to near equatorial van allen probes measurements. Once the mapping is learned the electron flux can be specified throughout the magnetosphere from l=2-6 using only the KP index and measured poes electron fluxes. The proposed work will build upon this initial success and expand the time energy and spatial coverage of the demonstration model. The work will update the model to use deep learning techniques that can capture more complicated relationships and the time cadence of the output will be increased. Most importantly the architecture will be modified so that off equatorial regions can be modeled to provide a global flux mapping. The final improvement will be to add forecast capabilities. As each improvement is added the model will be assessed and validated by comparing to GPS crres and out of sample van allen probes observations. These comparisons will produce metrics to evaluate whether the completed model satisfies user requirements and is ready for transition. The metrics will also provide a baseline for evaluating other current and future models. Once the model has been validated the project will produce a publicly accessible longterm history of the environment driven by past poes and KP data as inputs. Lastly the project will create a real-time and forecast framework that can be implemented and run operationally that will include displays for some select orbits inside of GEO that could be used to drive alerts. The improved model and applications will fulfill the O2R objectives by giving satellite industry stakeholders a tool to more effectively diagnose and anticipate satellite anomalies related to internal charging. IT will give NOAA the ability to assess and warn users of global space weather threats to the satellite infrastructure and establish data driven baseline performance metrics for evaluating current and future modeling efforts. And IT will provide the scientific research community with a long term history of the electron radiation environment for analysis that complements in-situ data from the NASA heliophysics observatory.