Supermodding - A special footprint operator for mesoscale data assimilation using scatterometer winds

Satellite observation footprints may extend over many grid points of a high-resolution limited-area model, while small and fast model scales may not be traceable by the observing system. We discuss the spatial representation of scatterometer ocean surface winds for the representation of high-resolution model state variables. A prototype observation operator called supermodding is studied in the variational assimilation framework in order to avoid correcting unconstrained small scales during the assimilation procedure. The challenges connected to small scales that are represented by the mesoscale model with respect to the application of the supermodding operator are discussed through idealised experiments. These results show that the application of the supermodding operator is able to avoid correcting unconstrained small scales, putting focus on the large scales only during data assimilation. Departure-based diagnostics show that the footprint representation helps to reduce the standard deviation of observation minus background departures (4-5% reduction) while the statistics for the supermodding method show a further reduction (8-11%). The impact of the supermodding approach is discussed through a forecast sensitivity study using a moist total energy norm (MTEN)-based technique and verification of the forecasts against observations. It is shown that the impact of the supermodding method to ASCAT data assimilation on the upper-air AROME-Arctic forecasts is observed over sea and near the surface, and that it is progressively shifted towards 700-800 hPa levels. Both supermodding at 30 km and at 60 km (i.e., twice the effective resolution of the applied scatterometer observations) show significant and consistent improvement on forecasts of lower tropospheric wind and temperature compared to the operational assimilation technique, as such demonstrating the robustness of the supermodding technique.

Automated summary

This study explores the spatial representation of scatterometer ocean surface winds for high-resolution model state variables and introduces a prototype observation operator called supermodding to avoid correcting unconstrained small scales during the assimilation process. Idealized experiments are used to discuss the challenges associated with small scales represented by the mesoscale model and the application of the supermodding operator, showing significant improvements in forecast accuracy.