TY - GEN
AU - Niels Bormann
AU - Linus Magnusson
AU - David Duncan
AU - Mohamed Dahoui
AB - This study investigates orbital biases in passive microwave (MW) temperature-sounding observations in the 50-60 GHz band by comparing observations against the ECMWF background. Methods to correct for the biases by modifying the model used in the variational bias correction are also examined. The analysis suggests that there are considerable orbital biases in MW temperature-sounding observations from AMSU-A and ATMS, with a distinct seasonal cycle, and magnitudes that can reach levels comparable in size to the instrument noise (0.1-0.2 K). The largest orbital biases occur during June-October, and they lead to systematic diurnal increments in the ECMWF analysis over the Southern Hemisphere. All AMSU-A instruments exhibit similar bias patterns, despite different equator crossing times, albeit with some variation in the magnitude. NOAA-18 AMSU-A shows the strongest biases. Further analysis suggests that the bias is observation-related, rather than a bias in the background. Assimilation experiments show that the orbital biases can be addressed within the variational bias correction by adding terms based on a Fourier-series in the orbital angle to the bias correction model, following Booton et al (2013). This leads to a significant reduction of the systematic diurnal increments, with minor benefits in the short-range forecasts up to day 2. Care has to be taken to limit the additional degrees of freedom introduced through the modified bias correction, to avoid over-fitting of the biases and interactions with model biases in the stratosphere. To limit these effects, the use of an “alternating Fourier series” as bias model is proposed, together with not applying the modified bias model to the highest stratospheric channels
BT - ECMWF Technical Memoranda
DA - 10/2023
DO - 10.21957/d281dc221a
M1 - 912
N2 - This study investigates orbital biases in passive microwave (MW) temperature-sounding observations in the 50-60 GHz band by comparing observations against the ECMWF background. Methods to correct for the biases by modifying the model used in the variational bias correction are also examined. The analysis suggests that there are considerable orbital biases in MW temperature-sounding observations from AMSU-A and ATMS, with a distinct seasonal cycle, and magnitudes that can reach levels comparable in size to the instrument noise (0.1-0.2 K). The largest orbital biases occur during June-October, and they lead to systematic diurnal increments in the ECMWF analysis over the Southern Hemisphere. All AMSU-A instruments exhibit similar bias patterns, despite different equator crossing times, albeit with some variation in the magnitude. NOAA-18 AMSU-A shows the strongest biases. Further analysis suggests that the bias is observation-related, rather than a bias in the background. Assimilation experiments show that the orbital biases can be addressed within the variational bias correction by adding terms based on a Fourier-series in the orbital angle to the bias correction model, following Booton et al (2013). This leads to a significant reduction of the systematic diurnal increments, with minor benefits in the short-range forecasts up to day 2. Care has to be taken to limit the additional degrees of freedom introduced through the modified bias correction, to avoid over-fitting of the biases and interactions with model biases in the stratosphere. To limit these effects, the use of an “alternating Fourier series” as bias model is proposed, together with not applying the modified bias model to the highest stratospheric channels
PB - ECMWF
PY - 2023
T2 - ECMWF Technical Memoranda
TI - Characterisation and correction of orbital biases in AMSU-A and ATMS observations in the ECMWF system
UR -  
ER -