TY - RPRT AU - David Duncan AU - Niels Bormann AB - Radio frequency interference (RFI) is a critical issue for numerical weather prediction (NWP), as a significant fraction of overall forecast skill comes directly from assimilation of passive microwave radiances. Underpinning the assimilation of microwave radiances is an assumption that observed signals are natural in origin, rather than man-made. As the microwave spectrum becomes more crowded, particularly at lower frequencies exploited for telecommunications, there is an increased risk that previously pristine parts of spectrum used for Earth observation contain interference, and consequently that forecast skill could be affected.
This document evaluates flagging of RFI provided by Zenithal Blue Technologies (ZBT) and Research and Development in Aerospace (RDA) for passive microwave bands of consequence for data assimilation in NWP. Specifically, frequencies from 6.9 to 89 GHz are analysed from the satellite-borne radiometers AMSR2 and AMSU-A. The RFI flagging is assessed using simulated radiances from the ECMWF model background, comparing differences of observed and simulated brightness temperatures (i.e. O-Bs) to indicate whether flagged data appear contaminated by unnatural emission signals not simulated by the model, assumed to be RFI. In order to do this, analysis of simulated radiances needs to be restricted to scenes where the forward modelling is of a consistently high quality, screening out difficult surfaces such as sea-ice and mountainous regions. Two one-month periods in 2022 have been evaluated.
In line with previous studies, significant RFI is seen at the lower frequencies observed by AMSR2, namely at 6.925, 7.3, and 10.65 GHz. RFI is also observed at 18.7 GHz, primarily around the coast of the United States. These identified regions of RFI include both direct sources and signals reflected off the ocean surface. No areas of RFI have been identified and corroborated by departure-based analysis at 23.8 GHz or above for the periods studied, with no significant RFI seen in the AMSU-A radiances studied here. RFI detection is perhaps most difficult at 10.65 GHz over sea, with relatively weak but frequent RFI observed primarily in the seas around Europe as reflected signals from direct broadcast geostationary satellites. The issue of false positives (RFI flagged for a clean scene) is not a serious concern in this analysis, with at most 0.05 to 0.2% of the likely RFI-free observations being flagged as potentially contaminated. However, false positives due to geophysical signals are prevalent in some areas such as coastlines and sea-ice edges, and a known antenna pattern bias in AMSU-A radiances causes false positives in stratospheric channels. Further work is recommended for identifying low-level RFI over sea at C- and X-bands, as these are crucial channels for SST sensitivity and more affected by interference than currently assimilated frequencies. In particular, newly identified sources of 6.925 GHz RFI over sea will require mitigation for this channel’s effective use in an assimilation system. BT - ESA Contract Report CY - Reading DA - 10/2024 DO - 10.21957/eefd6f0954 M1 - ESA AO/1-11605/22/NL/SD M3 - ESA Contract Report N2 - Radio frequency interference (RFI) is a critical issue for numerical weather prediction (NWP), as a significant fraction of overall forecast skill comes directly from assimilation of passive microwave radiances. Underpinning the assimilation of microwave radiances is an assumption that observed signals are natural in origin, rather than man-made. As the microwave spectrum becomes more crowded, particularly at lower frequencies exploited for telecommunications, there is an increased risk that previously pristine parts of spectrum used for Earth observation contain interference, and consequently that forecast skill could be affected.
This document evaluates flagging of RFI provided by Zenithal Blue Technologies (ZBT) and Research and Development in Aerospace (RDA) for passive microwave bands of consequence for data assimilation in NWP. Specifically, frequencies from 6.9 to 89 GHz are analysed from the satellite-borne radiometers AMSR2 and AMSU-A. The RFI flagging is assessed using simulated radiances from the ECMWF model background, comparing differences of observed and simulated brightness temperatures (i.e. O-Bs) to indicate whether flagged data appear contaminated by unnatural emission signals not simulated by the model, assumed to be RFI. In order to do this, analysis of simulated radiances needs to be restricted to scenes where the forward modelling is of a consistently high quality, screening out difficult surfaces such as sea-ice and mountainous regions. Two one-month periods in 2022 have been evaluated.
In line with previous studies, significant RFI is seen at the lower frequencies observed by AMSR2, namely at 6.925, 7.3, and 10.65 GHz. RFI is also observed at 18.7 GHz, primarily around the coast of the United States. These identified regions of RFI include both direct sources and signals reflected off the ocean surface. No areas of RFI have been identified and corroborated by departure-based analysis at 23.8 GHz or above for the periods studied, with no significant RFI seen in the AMSU-A radiances studied here. RFI detection is perhaps most difficult at 10.65 GHz over sea, with relatively weak but frequent RFI observed primarily in the seas around Europe as reflected signals from direct broadcast geostationary satellites. The issue of false positives (RFI flagged for a clean scene) is not a serious concern in this analysis, with at most 0.05 to 0.2% of the likely RFI-free observations being flagged as potentially contaminated. However, false positives due to geophysical signals are prevalent in some areas such as coastlines and sea-ice edges, and a known antenna pattern bias in AMSU-A radiances causes false positives in stratospheric channels. Further work is recommended for identifying low-level RFI over sea at C- and X-bands, as these are crucial channels for SST sensitivity and more affected by interference than currently assimilated frequencies. In particular, newly identified sources of 6.925 GHz RFI over sea will require mitigation for this channel’s effective use in an assimilation system. PB - ECMWF PP - Reading PY - 2024 T2 - ESA Contract Report TI - Assessing RFI flags at passive microwave bands with an NWP model UR -   ER -