Objective & Background
In order to integrate photovoltaic (PV) energy efficiently, accurate forecasts of the expected PV power production are essential. These in turn depend on the quality of the underlying numerical weather prediction (NWP) models. However, during Saharan dust outbreaks known forecast errors occur, because current NWP models rely on aerosol climatologies and do not consider the effects of the increased mineral dust concentrations in the atmosphere.
In the research project PerduS, Deutscher Wetterdienst (DWD), Karlsruhe Institute of Technology (KIT) and meteocontrol GmbH, are collaborating to improve weather and PV power forecasts during Saharan dust outbreaks. Therein, the new online coupled model system ICON-ART is used. It combines the global non-hydrostatic NWP model ICON and the ART modules for treating Aerosols and Reactive Trace gases in the atmosphere.
In PerduS a separate ICON-ART data assimilation cycle including mineral dust forecasts on a global 40 km grid has been built up at DWD and operates quasi-operational since January 2018. A large 20 km EU-NA²-nest covering Europe, North Africa and the North Atlantic region is added by a two way nesting approach. Both, the global 180 h/nested 120 h forecasts as well as the assimilation cycle account for the mineral dust radiation interactions. In the data assimilation cycle 3-hourly model forecasts of ICON-ART including the direct aerosol effects of the prognostic mineral dust are used as first guess/background in the so called “EnVar” mode. Therein, the deterministic forecasts are adapted to observations in a 3DVar and combined with flow-dependent model error information from the global ensemble Kalman filter based ICON ensemble data assimilation at DWD.
The aim of data assimilation is to determine the current state of the atmosphere, the analysis, as basis for forecast runs. So far, there is no assimilation of mineral dust observations in the data assimilation cycle of DWD. As a first step, we plan to assimilate the MODIS Level 2 aerosol optical depth (AOD) product. AOD retrievals can usually not distinguish between mineral dust and other aerosol types like sea salt or anthropogenic aerosol. Therefore, an appropriate filter has to be applied in order to isolate the information of mineral dust. In addition, clouds can affect aerosol observations, therefore the respective data points have to be identified and removed. Model results are then compared to the filtered MODIS aerosol product - the Dust Optical Depth (DOD) - and automatically adjusted by the EnVar system. In order to do so, an observation operator is used to derive the modelled DOD from the prognostic mineral dust concentration.
This contribution will give an introduction to the automatic daily mineral dust forecast system developed within PerduS. Furthermore, first steps in the assimilation of mineral dust at DWD are shown.