RADIO FREQUENCY INTERFERENCE RESILIENT PASSIVE MICROWAVE SENSOR

A. Madrid¹, M. Eble¹, B. Backus^2,3,4, M. Klein¹, A. Naqvi¹

¹Boulder Environmental Sciences and Technology, ²Johns Hopkins University Applied Physics Laboratory, ³National Oceanic and Atmospheric Administration, National Environmental Satellite, Data, and Information Service , ⁴IEEE/GRSS

Passive microwave sensors in orbit are critical tools for observing Earth’s climate, as well as the water and energy cycles. Their ability to contribute to all-weather observations makes them invaluable for weather forecasting models, e.g., (Lupu 2019). These sensors will have a lot more to offer if their spatial, temporal, and spectral resolutions are improved. Radio Frequency Interference (RFI) is a growing challenge for these remote passive sensors, and there is a continual risk of observational data corruption via said interference. This is a critical and challenging problem within the microwave portion of the electromagnetic spectrum, and taking measures to minimize the presence of RFI within passive microwave observations has become a necessity.

There are many solutions for detecting interference within a microwave channel, or a narrow bandwidth (e.g., up to tens or hundreds of MHz). Such algorithms are already being used for the Soil Moisture Active Passive (SMAP) microwave radiometer, e.g., (Piepmeier, et al. 2014), and many other papers describe applications of such algorithms to future sensors, (Kummerow, et al. 2022). These solutions provide detection within a relatively narrow radiometer band by digitizing the channel output and applying sophisticated algorithms to detect an interfering signal. These solutions are important and can be effective, but they are often also power hungry.

An alternative approach to handle passive band RFI is to widen the observational spectrum and utilize redundant channels observing similar environmental variables. Radiation from natural sources in the frequency domain is a continuous curve that has no breaks, sharp corners, or cusps outside of the known absorption lines. By creating an algorithm that leverages the expected smoothness of this curve, and bolstering it with redundant spectrally adjacent channels, we can isolate RFI contaminated channels even in high-contamination environments. Boulder Environmental Sciences and Technology (BEST) has been developing and evaluating such an approach spanning the 50-70 GHz oxygen absorption band and surrounding windows, as part of a project study for NOAA/NESDIS Joint Venture Partnerships. We use a simple analog multiplexer with 41 channels to cover this spectrum.

For simulation of the instrument’s observed brightness temperatures, we use an in-house radiative transfer model in conjunction with a database of atmospheric profiles from the ECO1280 nature run (Hoffman, Malardel and Peevey 2018). The ECO1280 nature run offers an extensive set of environmental variables with a 9 km average spatial resolution and 137 vertical layers, covering 14 months starting on September 30, 2015.

A “Wideband Associative RFI Detector” (WARD) algorithm is being developed. This is a hybrid model neural network, utilizing a 1D convolutional neural network with several fully connected layers. The RFI detection algorithm leverages the spectral response of the redundant 41 channels, along with locational and temporal context, to detect anomalous channel behavior indicative of RFI. In tests using synthetic single-channel RFI events of ~1.5K magnitude, the current version of WARD reliably detects over 92%-95% of events across a wide range of climatologies, achieving an Equitable Threat Score of ~0.8.

In parallel with the algorithm development, an analog multiplexer is also being prototyped and simulated. We will present both the hardware design and simulation results, as well as demonstrate WARD’s effectiveness under various cases of RFI contamination and across different climatologies and environmental conditions.

References

Hoffman, R. N., S. Malardel, and T. Peevey. 2018. “ECMWF Nature Run.” Cooperative Institute for Research in the Atmosphere. November 3. Accessed June 30, 2023. https://www.cira.colostate.edu/imagery-data/ecmwf-nature-run/.

Kummerow, C. D., J. C. Poczatek, S. Almond, W. Berg, 0. Jarrett, A. Jones, M. Kantner, and

C. P. Kuo. 2022. “Hyperspectral Microwave Sensors – Advantages and Limitations.” IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing 764-775. doi:10.1109/JSTARS.2021.3133382.

Lupu, C. 2019. “Data assimilation diagnostics: Assessing the observations impact in the forecast.” ECMWF Data assimilation training course 51.

Piepmeier, J. R., J. T. Johnson, P. N. Mohammed, D. Bradley, C. Ruf, M. Aksoy, R. Garcia, D. Hudson, L. Miles, and M. Wong. 2014. “Radio-frequency Interference Mitigation for the Soil Moisture Active Passive Microwave Radiometer.” IEEE Transactions on Geoscience and Remote Sensing 761-775. doi: 10.1109/TGRS.2013.2281266.