Environment Canada operated a thermodynamic profiling radiometer providing continuous temperature, humidity and liquid soundings during the Alpine Venue of the 2010 Winter Olympic Games. Retrievals were obtained in real time via neural networks and in post-processing via one-dimensional variational analysis (1DVAR). Background error covariance matrices for 1DVAR were calculated using brightness temperature from radiometer observations and from forward-modeled radiosonde and grid point model analysis. All radiometer, radiosonde and analysis data including all weather (rain, sleet and snow) conditions were included. 1DVAR retrieval errors, determined by comparison with radiosonde temperature and humidity soundings, are 1.6 C and 0.6 g/m3 rms or less from the surface to 10 km height. This error is smaller than the representativeness error assigned to radiosonde data when they are assimilated into numerical weather models. These results suggest that 1DVAR retrievals are ready for operational use in numerical weather modeling and prediction.
This publication was presented at the following: