An ensemble Kalman filter for convective-scale data assimilation (KENDA) has been developed for the COnsortium for Small-scale MOdelling (COSMO) model. The KENDA system comprises a local ensemble transform Kalman filter (LETKF) and a deterministic analysis based on the Kalman gain for the analysis ensemble mean. The KENDA software suite includes tools for adaptive localization, multiplicative covariance inflation, relaxation to prior perturbations and adaptive observation errors. In the version introduced here, conventional data (radiosonde, aircraft, wind profiler, surface station data) are assimilated. Latent heat nudging of radar precipitation has also been added to the KENDA system to be applied to the deterministic analysis only or additionally to all ensemble members. The performance of different system components is investigated in a quasi-operational setting using a basic cycling environment (BACY) for a period of six days with 24 h forecasts. For this period and an additional 28 day period, deterministic KENDA forecasts are compared with forecasts based on the observation nudging data assimilation scheme, which is currently operational at the German Weather Service (Deutscher Wetterdienst, DWD). For our experiments, lateral boundary conditions for the regional model are given by a global ensemble Kalman filter for the ICOsahedral Nonhydrostatic (ICON) model. The performance of the KENDA system proves overall to be superior to the forecast quality of the operational nudging scheme, in particular with regard to precipitation. Latent heat nudging improves precipitation forecasts in both systems and has slightly more benefit in combination with the LETKF than with observation nudging.

Data assimilation (DA) methods for convective‐scale numerical weather prediction at operational centres are surveyed. The operational methods include variational methods (3D‐Var and 4D‐Var), ensemble methods (LETKF) and hybrids between variational and ensemble methods (3DEnVar and 4DEnVar). At several operational centres, other assimilation algorithms, like latent heat nudging, are additionally applied to improve the model initial state, with emphasis on convective scales. It is demonstrated that the quality of forecasts based on initial data from convective‐scale DA is significantly better than the quality of forecasts from simple downscaling of larger‐scale initial data. However, the duration of positive impact depends on the weather situation, the size of the computational domain and the data that are assimilated. Furthermore it is shown that more advanced methods applied at convective scales provide improvements over simpler methods. This motivates continued research and development in convective‐scale DA. Challenges in research and development for improvements of convective‐scale DA are also reviewed and discussed. The difficulty of handling the wide range of spatial and temporal scales makes development of multi‐scale assimilation methods and space–time covariance localization techniques important. Improved utilization of observations is also important. In order to extract more information from existing observing systems of convective‐scale phenomena (e.g. weather radar data and satellite image data), it is necessary to provide improved statistical descriptions of the observation errors associated with these observations.

This article reviews developments towards assimilating cloud‐ and precipitation‐ affected satellite radiances at operational forecasting centres. Satellite data assimilation is moving beyond the “clear‐sky” approach that discards any observations affected by cloud. Some centres already assimilate cloud‐ and precipitation‐affected radiances operationally and the most popular approach is known as “all‐sky,” which assimilates all observations directly as radiances, whether they are clear, cloudy or precipitating, using models (for both radiative transfer and forecasting) that are capable of simulating cloud and precipitation with sufficient accuracy. Other frameworks are being tried, including the assimilation of humidity retrieved from cloudy observations using Bayesian techniques. Although the all‐sky technique is now proven for assimilation of microwave radiances, it has yet to be demonstrated operationally for infrared radiances, though several centres are getting close. Assimilating frequently available all‐sky infrared observations from geostationary satellites could give particular benefit for short‐range forecasting. More generally, assimilating cloud‐ and precipitation‐affected satellite observations improves forecasts in the medium range globally and can also improve the analysis and shorter‐range forecasting of otherwise poorly observed weather phenomena as diverse as tropical cyclones and wintertime low cloud.

ABSTRACT:To improve very-short-range forecasts particularly in convective situations, a version of the COSMO-Model (formerly known as LM) which simulates deep convection explicitly (horizontal grid length: 2.8 km) has been developed and is now run operationally at DWD. This model uses a prognostic type of precipitation scheme accounting for the horizontal drift of falling hydrometeors. To initialise convective-scale events, the latent heat nudging (LHN) approach has been adopted for the assimilation of surface precipitation rates derived from radar reflectivity data. It is found that a conventional LHN scheme designed for larger-scale models with diagnostic treatment of precipitation does not perform well and leads to strong overestimation of precipitation when applied to the convective-scale model with a prognostic treatment of precipitation. As illustrated here, surface precipitation and vertically integrated latent heating are far less correlated horizontally and temporally in such a model than with diagnostic precipitation, and this implies a violation of the basic assumption of LHN.Several revisions to the LHN scheme have therefore been developed in view of the characteristic model behaviour so as to re-enhance the validity of the basic assumption and to reduce greatly the overestimation of precipitation during assimilation. With the revised scheme, the model is able to simulate the precipitation patterns in good agreement with radar observations during the assimilation and the first hours of the forecast. The scheme also has a positive impact on screen-level parameters and on the longer-term climatology of the model. Extending the temporal impact of the radar observations further into the free forecast will be the focus of future research.

Ensemble Transform Kalman Filter (LETKF). In this study, the Efficient Modular VOlume RADar Operator is applied for the assimilation of radar reflectivity data to improve short-term predictions of precipitation. Both deterministic and ensemble forecasts have been carried out. A case-study shows that the assimilation of 3D radar reflectivity data clearly improves precipitation location in the analysis and significantly improves forecasts for lead times up to 4 h, as quantified by the Brier Score and the Continuous Ranked Probability Score. The influence of different update rates on the noise in terms of surface pressure tendencies and on the forecast quality in general is investigated. The results suggest that, while high update rates produce better analyses, forecasts with lead times of above 1 h benefit from less frequent updates. For a period of seven consecutive days, assimilation of radar reflectivity based on the LETKF is compared to that of DWD's current operational radar assimilation scheme based on latent heat nudging (LHN). It is found that the LETKF competes with LHN, although it is still in an experimental phase.