Improved Neutral Density Predictions Through Machine Learning Enabled Exospheric Temperature Model. Issue 12 (21st December 2021)

• Record Type: Journal Article
• Authors: Licata, Richard J.; Mehta, Piyush M.; Weimer, Daniel R.; Tobiska, W. Kent
• Is Part Of: Space weather. Volume 19:Issue 12(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-21
• Subjects: machine learning -- model development -- exospheric temperature -- thermosphere; Space environment -- Periodicals; 551.509992
• DOI: 10.1029/2021SW002918
• Languages: English
• ISSNs: 1542-7390

Abstract

Abstract: The community has leveraged satellite accelerometer data sets in previous years to estimate neutral mass density and exospheric temperatures. We utilize derived temperature data and optimize a nonlinear machine‐learned (ML) regression model to improve upon the performance of the linear EXospheric TEMPeratures on a PoLyhedrAl gRid (EXTEMPLAR) model. The newly developed EXTEMPLAR‐ML model allows for exospheric temperature predictions at any location with one model and provides performance improvements over its predecessor. We achieve reductions in mean absolute error of 2 K on an independent test set while providing similar error standard deviation values. Comparing the performance of both EXTEMPLAR models and the Naval Research Laboratory Mass Spectrometer and Incoherent Scatter radar Extended model (NRLMSISE‐00) across different solar and geomagnetic activity levels shows that EXTEMPLAR‐ML has the lowest mean absolute error across 80% of conditions tested. A study for spatial errors demonstrated that at all grid locations, EXTEMPLAR‐ML has the lowest mean absolute error for over 60% of the polyhedral grid cells on the test set. Like EXTEMPLAR, our model's outputs can be utilized by NRLMSISE‐00 (exclusively) to more closely match satellite accelerometer‐derived densities. We conducted 10 case studies where we compare the accelerometer‐derived temperature and density estimates from four satellites to NRLMSISE‐00, EXTEMPLAR, and EXTEMPALR‐ML during major storm periods. These comparisons show that EXTEMPLAR‐ML generally has the best performance of the three models during storms. We use principal component analysis on EXTEMPLAR‐ML outputs to verify the physical response of the model to its drivers. Plain Language Summary: Density in the upper atmosphere is highly variable and difficult to model. Empirical density models often rely on temperature profile predictions to determine species and mass densities. One of three key parameters in determining the temperature profiles is the asymptotic value at the top of the thermosphere called the exospheric temperature. By using temperatures derived from satellite acceleration measurements, we develop a machine‐learned global temperature model called EXospheric TEMPeratures on a PoLyhedrAl gRid Machine Learned (EXTEMPLAR‐ML). We achieve a 2 K reduction in mean absolute error on the independent test set relative to the model's predecessor. Additional analyses showed that EXTEMPLAR‐ML was more accurate than linear EXTEMPLAR across a majority of conditions and grid locations. We also look at temperatures and densities along satellite orbits during 10 major geomagnetic storms from the 21st century. In this study, we see major improvements over a significant empirical model called NRLMSISE‐00 and the linear predecessor to EXTEMPLAR‐ML. We leveraged a mathematical decomposition tool on the model outputs to assess its internal formulation. This shows that EXTEMPLAR‐ML is most heavily driven by solar activity and the seasons. Key Points: We develop a nonlinear global model for exospheric temperature prediction called EXospheric TEMPeratures on a PoLyhedrAl gRid Machine Learned (EXTEMPLAR‐ML) We leverage principal component analysis to improve our understanding of the EXTEMPLAR‐ML temperature formulation EXTEMPLAR‐ML shows increased accuracy relative to satellite observations across a majority of conditions, locations, and during geomagnetic storms