SBIR/STTR Award attributes
This project, titlednbsp;Machine learning Enabled Thermosphere Advanced by HASDMnbsp;(META-HASDM), will provide publicly, for the first time, the operationally derivednbsp;SET HASDM databasenbsp;as a national benchmark for improving thermospheric density forecasting for LEO satellites. To aid rapid, public access we will develop machine learning (ML) algorithms to characterizenbsp;the terabyte-sized two-solar cycle SET HASDM database by using JB2008 indices and other physical drivers; this algorithm is callednbsp;HASDM-ML. Thenbsp;HASDM-ML algorithm will be validated against a subset ofnbsp;the SET HASDM database. In addition, the team will use HASDM-ML densities with Two-Line Elements (TLEs) of a few representative satellites to create more accurate ballistic coefficients,nbsp;B*, for a few satellites in a historic atmosphere under known solar irradiance and geomagnetic conditions. Thenbsp;uncertainty will be quantified for density and for forecast satellite drag. We will extend orbit estimates to current and near-term epochs to improve orbit forecasting relevant to LEO space traffic management by using the newnbsp;B*nbsp;values.nbsp;Finally, we will improve JB2008 indicesrsquo; forecast over the next 72 hours using ML techniques; this willnbsp;be complemented by the capture of high speed streams (HSS) not currently represented for the Dst geomagnetic index as well as by extending forecast accuracy out to 7 days using the existing ADAPT model values for F10 and S10.
