Satellite images in the visible spectral range contain high‐resolution cloud information, but have not been assimilated directly before. This paper presents a case‐study on the assimilation of visible Meteosat SEVIRI images in a convective‐scale data assimilation system based on a local ensemble transform Kalman filter (LETKF) in a near‐operational set‐up. For this purpose, a fast look‐up table‐based forward operator is used to generated synthetic satellite images from the model state. Single‐observation experiments show that the assimilation of visible reflectances improves cloud cover under most conditions and often reduces temperature and humidity errors. In cycled experiments for two summer days with convective precipitation, the assimilation strongly reduces the errors of cloud cover and improves the precipitation forecast. While these results are promising, several issues are identified that limit the efficacy of the assimilation process. First, the linearity assumption of the LETKF can lead to errors as reflectance is a nonlinear function of the model state. Second, errors can arise from the fact that visible reflectances alone are ambiguous and only weakly sensitive to the water phase and cloud‐top height. And lastly, it is not obvious how to localise vertical covariances as visible reflectances are sensitive to clouds at all heights. For the latter reason, no vertical localisation was used in this study. To investigate the robustness of the results, the horizontal localisation scale, the assigned observation error and the spatial density of observations were varied in sensitivity experiments. The best results were obtained for an observation error close to the Desroziers estimate. High observation density combined with small localisation radii resulted in the smallest 1 hr forecast error. These settings were also beneficial for 3 hr forecasts, but forecasts at that lead time were less sensitive to the observation density and the localisation scale.

A B S T R A C TA new development in the field of reanalyses is the incorporation of uncertainty estimation capabilities. We have developed a probabilistic regional reanalysis system for the CORDEX-EUR11 domain that is based on the numerical weather prediction model COSMO at a 12-km grid spacing. The lateral boundary conditions of all ensemble members are provided by the global reanalysis ERA-Interim. In the basic implementation of the system, uncertainties due to observation errors are estimated. Atmospheric assimilation of conventional observations perturbed by means of random samples of observation error yields estimates of the reanalysis uncertainty conditioned to observation errors. The data assimilation employed is a new scheme based on observation nudging that we denote ensemble nudging. The lower boundary of the atmosphere is regularly updated by external snow depth, sea surface temperature and soil moisture analyses. One of the most important purposes of reanalyses is the estimation of so-called essential climate variables. For regional reanalyses, precipitation has been identified as one of the essential climate variables that are potentially better represented than in other climate data sets. For that reason, we assess the representation of precipitation in our system in a pilot study. Based on two experiments, each of which extends over one month, we conduct a preliminary comparison to the global reanalysis ERAInterim, a dynamical downscaling of the latter and the high-resolution regional reanalysis COSMO-REA6. In a next step, we assess our reanalysis system's probabilistic capabilities versus the ECMWF-EPS in terms of sixhourly precipitation sums. The added value of our probabilistic regional reanalysis system motivates the current production of a 5-year-long test reanalysis COSMO-EN-REA12 in the framework of the FP7-funded project Uncertainties in Ensembles of Regional Re-Analyses (UERRA).

No abstract

<p>In the framework of the project SINFONY at Deutscher Wetterdienst, we work towards seamless prediction at the very-short range blending over from observation-based nowcasting to numerical weather prediction. The key goals that we pursue in this context are:<br>1.    To deliver forecasts earlier to be displayed in the meteorological workstation NinJo of our forecasters, which is realized by hourly forecast initialization in our newly-developed Rapid Update Cycle (RUC) and shorter latency for observation arrival ahead of data assimilation. <br>2.    To provide seamlessly combined products integrating nowcasted and forecasted radar reflectivities as well as precipitation from both forecasting systems.<br>3.    To achieve a better representation of precipitation processes and convective cells in our NWP model to allow for the seamless blending with nowcasts. For this purpose, we use a two-moment microphysics scheme that predicts not only mixing ratios of hydrometeor species, but also their particle size distribution. This is also of great importance for the data assimilation of geostationary all-sky satellite data assimilation, for data assimilation of lightning data and essentially radar reflectivities.</p><p>In this presentation, we explain how data assimilation of cloudy visible satellite data can help to improve the accuracy of clouds and precipitation processes in NWP forecasts to assist a seamless blending of nowcasting and NWP in terms of radar reflectivities mentioned in 2) and 3). </p><p>Visible satellite data are directly sensitive to liquid water path, ice water path and specific humidity  which are integral quantities related to precipitation processes. Moreover, cloud positioning can be improved by deleting false alarm clouds and convective cells and introducing missing ones to the forecast. A key advantage is that visible data are particularly sensitive to water clouds, which allows to constrain convective cells already at their state of initiation in the initial conditions of our RUC forecasts.</p><p>We elaborate on the basic principles of satellite data assimilation in our ICON-D2-KENDA system making use of an ensemble Kalman filter. Case studies will be shown to demonstrate how data assimilation of all-sky satellite data reduces analysis and forecast error of clouds and precipitation. Finally, we show the impact in our rapid update pre-operational system over longer periods of time. </p>

<p>The SINFONY project at Deutscher Wetterdienst (DWD) aims to produce seamless precipitation forecast products from minutes up to 12 hours, with particular focus on convective events. While the near future predictions are typically from nowcasting procedures using radar data, the numerical weather prediction (NWP) aims at longer time scales. The lead-time in the latest available forecast is usually too long for merging both the nowcasting and NWP output to produce reliable seamless predictions.</p><p>At DWD, the current forecasts are produced by the short range numerical weather prediction (SRNWP) <span>making use of a</span> continuous assimilation cycle with relatively long cutoff times and using 1-moment microphysics. In order to reduce the differences in the precipitation to the <span>nowcasting </span>on the NWP side, we use two different approaches. First, we reduce the lead-time from the model start by running 1-hourly forecasts based on an assimilation cycle with shorter data cutoff. Secondly, we use new observational systems in the assimilation cycle, such as radar or satellite data to capture and represent strong convective activity. This procedure is called Rapid Update Cycle (RUC). As an additional measure, we introduce a 2-Moment microphysics scheme into the numerical model, resulting in a better representation of the radar reflectivities. In order to keep the model state similar to that of the SRNWP, the RUC is a time limited assimilation cycle starting from forecasts of the SRNWP at pre-defined times.</p><p>The introduction of the 2-Moment scheme leads to a spin-up affecting both the assimilation cycle and the short forecasts. The resulting effects are analysed by comparison with the corresponding assimilation cycle using the 1-Moment scheme. As a complementary approach for the analysis, the routine cycle is run with the 2-Moment scheme. The forecast quality is used as a measure to compare the results with respect to precipitation and additional observed parameters. It is shown in how far the resulting improvements are related to the assimilation and momentum scheme, or to the higher frequency of forecasts.</p>