Towards operational assimilation of surface‐sensitive microwave radiances over land at Environment and Climate Change Canada : Investigation of key factors using a 1D ‐ EnVar framework

Abstract The present study investigates several factors related to the assimilation of surface‐sensitive radiance observations over land from the Advanced Microwave Sounding Unit‐A within a non‐cycling one‐dimensional Ensemble‐Variational (1D‐EnVar) framework, with the aim of informing future improvements to numerical weather prediction (NWP) at Environment and Climate Change Canada (ECCC). These observations are currently not assimilated in ECCC's operational systems, since it requires accurate knowledge of both surface emissivity and skin temperature. Typically, surface emissivity is retrieved using a surface‐sensitive channel observation combined with prior knowledge of the atmospheric state and skin temperature through an inverse radiative transfer calculation. The present study examines the impact of assimilating multiple surface‐sensitive channels to simultaneously estimate surface emissivity, together with the standard set of state variables (including skin temperature), by including surface emissivity as an additional analysis variable in a 1D‐EnVar system. In addition, other related factors are examined, including modifications to the topographic criteria in the observation quality control to allow more channel 6 observations over higher‐elevation surfaces to be assimilated as surface‐sensitive channels. Although these changes lead to analysis improvements in the mid‐troposphere, they also result in an increased analysis error in the lower troposphere, caused by unexpectedly large lower‐tropospheric air temperature and skin temperature increments. This increase in error is mitigated by modifying the ensemble‐based background error covariances. Lastly, given the limited knowledge on the true surface emissivity uncertainty, the impact of varying the emissivity background error covariances is examined. The combined impact of integrating the modifications from all factors while assimilating channels 4–14 resulted in up to a 3.4% reduction in the normalized root‐mean‐squared error for the mid‐ and lower‐tropospheric temperature compared with assimilating only channels 6–14 and without considering the discussed factors. These findings provide a basis for future advancements in assimilating surface‐sensitive microwave radiances over land in 4D‐EnVar in the full ECCC NWP system.