Satellite data assimilation and vertical localization
The quality of numerical weather prediction (NWP) forecasts depends strongly on the accuracy of the initial conditions. Data assimilation systems provide the best estimate of the initial state by combining the latest short-range forecast with observations. Satellite observations are therefore highly valuable, as they provide globally highly resolved measurements in space and time. Different sensors across the infrared to visible spectrum provide insight into atmospheric clouds and the distribution of water vapor, thereby improving our initial conditions and leading to better forecast quality. In our group, we specifically work on the following three aspects of satellite data assimilation:
Vertical localization for ensemble systems
Ensemble forecast systems, such as the ICON-D2 system of the Deutscher Wetterdienst (DWD), use the covariances between ensemble members and simulated satellite retrievals to update the model state. Such covariances are susceptible to spurious correlations due to the limited ensemble size, which is why covariances are commonly damped with the distance between model and observation. This damping method is referred to as localization. Satellite observations, in contrast to conventional point observations, are vertically integrated measurements, which have no predetermined height from which a distance could be calculated. Furthermore, satellite measurements at different wavelengths respond differently to atmospheric variables at various heights. For these reasons, the vertical localization of satellite observations is an active area of research, which we pursue in collaboration with RIKEN (Japan) and the DWD (Germany).
References:
- Necker, T., Hinger, D., Griewank, P. J., Miyoshi, T., and Weissmann, M.: Guidance on how to improve vertical covariance localization based on a 1000-member ensemble, Nonlin. Processes Geophys., 30, 13–29, https://doi.org/10.5194/npg-30-13-2023, 2023.
- Martin Parker, Partial analysis increments: theory and application to optimisation of vertical localisation, Master Thesis, 2024.
The optimal vertical localisation for different variable combinations. See Necker et al. (2023) for details.
Estimating the potential impact of observations and optimizing their assimilation
How much benefit different satellite observations could provide to forecasts, and how best to assimilate them, are still open questions within the community. These questions can be addressed in an idealized, controlled environment using a so-called OSSE. In these Observation System Simulation Experiments (OSSEs), the NWP system, as well as a digital twin of the real atmosphere, is simulated from which satellite measurements are calculated using radiative transfer models. For example, by using an idealized forecast of deep convection, we can understand how assimilating satellite observations can improve thunderstorm forecasts.
References:
- Kugler, L., & Weissmann, M. (2025). Effects of observation operator nonlinearity on the assimilation of visible and infrared radiances in ensemble data assimilation. Quarterly Journal of the Royal Meteorological Society, 151(770), Article e4970. https://doi.org/10.1002/qj.4970
- Kugler, L., & Weissmann, M. (2025). The synergy of assimilating visible and infrared radiances and radar observations. Quarterly Journal of the Royal Meteorological Society, 151(768), e4938. Article e4938. https://doi.org/10.1002/qj.4938
- Kugler, L., Anderson, J. L., & Weissmann, M. (2023). Potential impact of all-sky assimilation of visible and infrared satellite observations compared with radar reflectivity for convective-scale numerical weather prediction. Quarterly Journal of the Royal Meteorological Society, 149(757), 3623-3644. https://doi.org/10.1002/qj.4577
Comparison of predicted probabilities of precipitation >1 mm/h without (left) and with (right) background-error inflation (α = 0.9) for the experiment VIS+WV62+WV73. Red contours show the true regions of precipitation >1 mm/h.
Improving operational forecasts in collaboration with GeoSphere Austria
State-of-the-art NWP systems are highly complex, with many interacting subsystems that can interact in unforeseen ways and require thorough optimization to run efficiently on large supercomputers. At the same time, weather centers have to thoroughly test and evaluate their systems, as forecasts and the resulting emergency warnings are critical to society. These constraints make assimilating satellite data operationally a challenging task, often taking years or even decades to operationalize methods developed by the scientific community in idealized studies like those mentioned above. As part of the Vienna Network for Atmospheric Research, we collaborate closely with GeoSphere Austria to assimilate both visible and infrared observations from the Meteosat Second and Third Generation satellites into their operational systems. This work has the potential to substantially improve GeoSphere's capabilities to forecast severe precipitation and solar power production. For further information, see the CloudyRadiances and MTGReen projects.
Infrared brightness temperatures as seen by the Meteosat Second Generation Satellite.