Observation impact evaluation through data denial experiments in the Met Office global numerical weather prediction system

This study presents a complete assessment of the use of different observation types and their impacts in the Met Office global data assimilation system. The data denial experiments are run over three months using the observations from 15 December 2022 to 15 March 2023 and use a configuration very close to the Met Office operational global numerical weather prediction (NWP) suite, with some exceptions, mainly for reducing the computational cost. The control run uses the complete observing system assimilated operationally at the time, and data from all the main observing systems (e.g., microwave sounders, hyperspectral infrared sounders, clear‐sky geostationary sounders, and conventional data, etc.) are denied categorically for a series of experiments. Results indicate that microwave satellite radiance observations continue to be the most important data source overall, being as impactful as conventional observations in the Northern Hemisphere and, along with Global Navigation Satellite System radio occultation (GNSS‐RO), the most important source in the Tropics and Southern Hemisphere. Conventional observations from aircraft and radiosonde ascents continue to play an essential role, particularly in the Northern Hemisphere. Our investigation also suggests that GNSS‐RO provides significant benefits, especially in the upper troposphere and lower stratosphere for key atmospheric variables, especially temperature. Hyperspectral sounders show less impact than in previous studies, though still show strong positive impact in the Southern Hemisphere. Atmospheric motion vectors show impact on tropospheric wind, particularly in the Tropics. In general, all observation types show a positive impact on the mean forecast skill in at least some aspects of the global NWP system.