15R2O Microwave L-Band RFI Detection and Mitigation with Software Defined Radio Using Machine Learning

The Microwave L-band (1-2 GHz) radio frequency is vital for NOAA operations, such as satellite downlink for Earth observation data of hurricanes, and satellite missions like SMAP providing soil moisture and ocean salinity data used in weather prediction models. This data also informs irrigation decisions, and crop yield predictions. Radio frequency interference (RFI) increases the uncertainty of these measurements, making corrupted data unusable. The L-band is not only vital for Earth observation, but also for radio astronomy, as it contains emissions from cosmological and galactic hydrogen. This study investigates the 1420 MHz forbidden hydrogen transition with software defined radio, identifies RFI, and will experiment with the latest machine learning (ML) methods to distinguish and potentially remove RFI. It is known that the 1420 MHz signal emitted from the Milky Way is extremely weak, which makes it a useful target to study RFI, due to its similarity to the weak soil moisture and ocean salinity signals at a similar frequency (~1419 MHz). Using a ML method has the advantage of using advanced digital signal processing to identify and potentially remove corrupted data with AI, which can be done with large amounts of data quickly once trained. Better digital RFI cleanup methods mean less hardware having to be developed with satellites, which leads to reduced mission costs and better quality data. This method could also be applied to data from previous missions, giving us better long-term prediction and forecasting. It is predicted that RFI may become a concern for high frequency bands which may have potential impacts on microwave sounder observations in the future. Therefore, this study may provide insights for future sounder missions.