Towards Understanding Data Assimilation Impacts on Arctic Cyclone Predictability through Ensemble Sensitivity Analysis (ESA) and Observation System Experiments (OSE)

Two case studies are used to understand how well existing observations constrain analyses of processes influencing Arctic cyclone predictive skill and the potential impacts of additional observations in the Arctic polar region. The first case study of 17-20 July 2018 is characterized by Arctic cyclogenesis following a tropopause polar vortex (TPV)-induced Rossby wave initiation event, and focuses on the impact of existing observations. The second case study of 21-24 April 2020 is characterized by a series of Arctic warm air intrusions corresponding to a persistent highly amplified Rossby wave pattern within which Arctic cyclogenesis occurred, and focuses on the potential impacts of additional observations. The second case study benefits from additional radiosonde observations, beyond the current operational network, collected as part of a targeted observation period of the Year of Polar Prediction project.

In the first case study, the ESA reveals that the large-scale Rossby wave structure correlated with both Arctic cyclone track and amplitude errors. The ensemble analyses of mid-level moisture in the warm conveyor belt region were correlated with forecast cyclone amplitude, but this feature was poorly sampled in existing observations. There is also a sensitivity of Arctic cyclone forecast amplitude error to low level temperature in the air mass of the cyclogenesis region at analysis time and a sensitivity of Arctic cyclone forecast track error to low level temperature in the region of an Arctic cold front and a coastal front at the analysis time. The OSEs reveal that Arctic cyclone track error is more sensitive to denial of existing observations than amplitude error. While lower level (below 700 hPa) observations had greatest impact on the surface cyclone during the early stages, upper level (above 500 hPa) observations had the dominant impact during its later evolution. The second case study is used to focus on diagnosing the impact on forecast skill of assimilating the additional YOPP radiosondes with an emphasis on relating the physical processes affected by the observations to the physical processes contributing to the forecast uncertainty. Results for both cases will be discussed in the conference.