Radio Frequency Interference Mitigation Based on Low-Rank Sparse Decomposition for Polarimetric Weather Radar

In the context of escalating frequency spectrum congestion, the prevalence of radio frequency interference (RFI) poses a growing challenge for weather radars, compromising data quality and adversely affecting the accuracy of variable estimations. This article proposes a novel algorithm for the separation of precipitation and RFI, based on low-rank sparse decomposition (LRSD). When precipitation and RFI overlap, this method effectively filters out RFI while minimizing its impact on precipitation by analyzing the property difference between precipitation and RFI in the time domain. The proposed algorithm operates on the assumption that precipitation exhibits low-rank properties, whereas RFI manifests as sparse signals. This assumption is grounded in the observation that RFI in weather radars is typically unintentional, occupying a limited number of pixels in the range-time power image, while precipitation demonstrates approximate stationarity with a slow speed within a coherent processing interval (CPI). This algorithm is intended to alleviate the adverse effects of RFI, and the efficacy of the approach is validated using data collected by polarimetric Doppler weather radar systems at the Royal Netherlands Meteorological Institute. This article systematically evaluates the impact of the proposed LRSD method in comparison to two traditional RFI mitigation strategies (e.g., the Vaisala-3 method and 2-D filter) on the quality of meteorological data. The results demonstrate that the proposed LRSD method outperforms the alternatives in terms of RFI suppression and precipitation retention performance. At present, the LRSD method loses phase information, which may impact the accuracy of measuring polarization parameters of precipitation.