A comparison of different bias correction methods on the assimilation of high-resolution all-sky GOES-16 ABI radiances to improve Convective Initiation forecast

Convective Initiation (CI) forecast poses itself as a big challenge for NWP owing to its initially small scale and strong sensitivity to the environmental moisture. It has been hypothesized that the high space and time resolution of GOES-16 ABI radiances would enable better CI forecasts, and an accompanying increased lead time for warning that can not be obtained by assimilating radar reflectivity after storms have already developed. The operational ensemble data assimilation system has been extended to assimilate both the clear and cloudy radiances of GOES-16. Experiments have demonstrated the positive impact of the assimilation of GOES-16 ABI radiances on the CI forecast. In this study, the system is further extended to house several different bias correction schemes including a nonlinear bias correction approach and an approach that provides online estimation of bias together with the state update.

The effect of the different bias correction method is demonstrated by a case of rapidly intensifying tornadic supercell from 18 May 2017. Apart from ABI water vapor sensitive channel radiance, NEXRAD radar reflectivity observations are also assimilated every 10 minutes into the WRF model with a horizontal resolution of 3km. Experiments have been designed to reveal the effect of different bias predictors like observed brightness temperature and simulated brightness temperature for the nonlinear bias correction approach. The effect of the nonlinear bias correction approach is further compared with the linear online estimation approach. Several diagnostics highlighting these features will be discussed during the presentation.