Improved Neutral Density Predictions through Machine Learning Enabled Exospheric Temperature Model
Abstract
The community has leveraged satellite accelerometer datasets in previous years to estimate neutral mass density and subsequently exospheric temperatures. We utilize derived temperature data and optimize a nonlinear machine-learned (ML) regression model to improve upon the performance of the linear EXTEMPLAR (EXospheric TEMPeratures on a PoLyherdrAl gRid) model. The newly developed EXTEMPLAR-ML model allows for exospheric temperature predictions at any location with a single model and provides performance improvements over its predecessor. We achieve a 4.2 K reduction in mean absolute error and a 3.42 K reduction in the standard deviation of the error. Like EXTEMPLAR, our model's outputs can be utilized by the Naval Research Laboratory Mass Spectrometer and Incoherent Scatter radar Extended (NRLMSISE-00) model to more closely match satellite accelerometer-derived densities. We conducted two case studies where we compare the CHAllenging Minisatellite Payload (CHAMP) and Gravity Recovery and Climate Experiment (GRACE) accelerometer-derived temperature and density estimates to NRLMSISE-00, EXTEMPLAR, and EXTEMPALR-ML during two major storm periods. The storm-time temperature comparison showed error reductions of 7-10% and 2-5% relative to NRLMSISE-00 and EXTEMPLAR, repsectively, and the density comparison showed error reductions of 20-55% and 8-12%. We use Principal Component Analysis to identify the dominant modes of variability in the model over one solar cycle. This shows the model is dominantly driven by solar activity, and there is a strong latitudinal variation related to the Summer and Winter hemispheres.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMNG43A..04L