Research Summaries

Back Demonstration of Machine Learning Approach for Evaporation Duct NOWCAST as Part of Operational METOC Support

Fiscal Year 2020
Division Research & Sponsored Programs
Department NPS Naval Research Program
Investigator(s) Feldmeier, Joel
Wang, Qing
Sponsor NPS Naval Research Program (Navy)
Summary Artificial Intelligence/Machine Learning (AI/ML) has been increasingly used as an alternative approach
to forecasting complex stochastic processes in meteorology and oceanography (METOC). Its potential in
enhancing the Navy¿s METOC support for predictions of radar detection ranges and communication
qualities has not been assessed. The overall goal of this project will be to demonstrate and assess the skill
to be attained by building machine learning short term forecast (generally less than 6 hour, NOWCAST),
algorithms for evaporative duct properties, as well as other relevant variables for naval operational
METOC support (e.g., temperature, humidity, wind direction, wind speed). The research will utilize
datasets from the recent Office of Naval Research (ONR) sponsored Coupled Air-Sea Processes and
Electromagnetic Ducting Research (CASPER) field campaigns. These rich datasets were collected at high
frequency and span multiple weeks in multiple locations. It contains typically collected METOC data, as
well as additional information from propagation links in microwave and electro-optical wavelengths, and
various air-sea fluxes. Anticipated machine learning techniques to be tested include, but are not limited
to, usage of a Bayesian technique, regression techniques, decision tree/random forest algorithms, and
neural networks. An assessment will be made of the various techniques¿ utility over varying collection
periods, simulating a ship underway collecting data and building models constantly, and determining
what training period is necessary for different variables, as well as the effective prediction period from a
forecast model until the training dataset is updated and new predictive equations are derived. Initial
assessment will also be attempted of data storage and processing requirements, ideally using technology
approximately equivalent to current or near-future planned underway computing resources.
Keywords Electromagnetic Wave Propagation Evaporation Duct machine learning
Publications Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal
Data Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal