Using Deep Learning to Develop a High Resolution Planetary Boundary Layer Model for Infrasound Propagation.
Year: 2022
Type: conference-paper
Abstract: Previous research has shown the utility of deep learning in predicting atmospheric structure for regional infrasound propagation.A Long-Short Term Memory network can be used to predict temperature, wind speed, and wind direction up to altitudes of ~40 km with performance equal to or better than state-of-the-art reanalysis models.However, these regional models do not resolve the planetary boundary layer (PBL), the lowermost part of the atmosphere, well.An accurate model of the PBL is important at local distances for acoustic waveform modeling.This is often accomplished by using a suite of robust and expensive meteorological sensors.A method to predict atmospheric structure at this resolution would be very beneficial for low yield monitoring scenarios where the ability to field sensors is limited or the timing of the event is not known beforehand.Here we show progress towards a high resolution PBL model developed via deep learning techniques using an ambient urban infrasound data set coupled with weather measurements in Las Vegas, Nevada, USA.We anticipate results will show the ability to use a data-driven method to predict, interpolate, and extrapolate an atmospheric model that can be used in acoustic waveform modeling.
DOI: 10.2172/2002308
Authors: Sarah Albert, Daniel Bowman, Douglas Seastrand, Melissa Wright
Institutions: Sandia National Laboratories, Nevada National Security Site
Language: English
Subfield: Atmospheric Science
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4388301752