Characterizing Radio Frequency Interference Mitigation Strategies and Impacts

As portions of the electromagnetic spectrum are sold to telecommunications companies or otherwise auctioned off by the federal government, the amount of electromagnetic spectrum available for use by operational radars will decrease. This decrease will come with a commensurate increase in problems associated with radio frequency interference (RFI); such problems have already occurred at the National Weather Radar Testbed (NWRT) phased array radar, which is being used in the development of the Multi-function Phased Array Radar (MPAR). Because RFI reduces data quality, its presence could compromise the MPAR’s ability to carry out all of its missions, which include meteorological surveillance and target identification and tracking, by reducing the quality of radar observations.

Several types of RFI have been observed, but the most common type of RFI present at the NWRT was “single-hit RFI,” which generally presents itself as non-meteorological blips in observations. Several detection algorithms have been recommended by industry professionals in the Vaisala User’s Manual; all three compare the decibel power of three consecutive pulses to user defined constants to determine if one of the pulses is afflicted with RFI. Additionally, the Interference Spike Detection Algorithm (ISDA) and the Electromagnetic Interference (EMI) filter have been developed by the authors. ISDA is a constant false alarm rate-type algorithm that allows users more flexibility than the Vaisala algorithms do in determining the criteria used to identify RFI, while the EMI filter takes a statistical approach to detecting RFI. Tests to date have focused on comparing the performance of these algorithms using either subjective assessment of algorithm performance or statistical evaluation of algorithm effects on previously collected data that was artificially afflicted with RFI. These tests have delved into the performance of the algorithms but have had limited applications for operational meteorologists.

To better characterize the performance of these algorithms, weather data with known signal-to-noise ratio, radial velocity, pulse repetition interval, and spectrum width will be simulated, and the detection algorithms will be executed on this simulated data using different values of the algorithm-specific user-defined constants. This will give the radar operators a guideline for setting the algorithm constants based on desired false alarm rate, rather than having to guess. This simulated weather data will subsequently be afflicted with single-hit RFI with known interference-to-noise ratio, and the algorithms again executed; however, this time the algorithms will be executed using the algorithm constants generated using the operational guidelines. This test will allow measurement of the probability of detection of the algorithms as well as the leftover bias in the data after RFI detection and mitigation algorithms have been executed. Ultimately users will be able to determine the best algorithm for mitigating RFI depending on the current weather situation. 

Figure 1: Radar data before corruption with RFI (left), with simulated RFI (center) and after running RFI detection and mitigation algorithms (right). Future experiments would aim for good RFI mitigation performance without prior knowledge of the uncorrupted radar data.