Alexander Mangold

Aerosol analysis and forecast in the European Centre for Medium‐Range Weather Forecasts Integrated Forecast System: 2. Data assimilation

[1] This study presents the new aerosol assimilation system, developed at the European Centre for Medium-Range Weather Forecasts, for the Global and regional Earth-system Monitoring using Satellite and in-situ data (GEMS) project. The aerosol modeling and analysis system is fully integrated in the operational four-dimensional assimilation apparatus. Its purpose is to produce aerosol forecasts and reanalyses of aerosol fields using optical depth data from satellite sensors. This paper is the second of a series which describes the GEMS aerosol effort. It focuses on the theoretical architecture and practical implementation of the aerosol assimilation system. It also provides a discussion of the background errors and observations errors for the aerosol fields, and presents a subset of results from the 2-year reanalysis which has been run for 2003 and 2004 using data from the Moderate Resolution Imaging Spectroradiometer on the Aqua and Terra satellites. Independent data sets are used to show that despite some compromises that have been made for feasibility reasons in regards to the choice of control variable and error characteristics, the analysis is very skillful in drawing to the observations and in improving the forecasts of aerosol optical depth.

Aerosol Analysis and Forecast in the ECMWF Integrated Forecast System: Evaluation by Means of Case Studies

A near real-time assimilation and forecast system of aerosols has been developed by integration in the ECMWF IFS code within the GEMS project. The GEMS aerosol modeling system is novel as it is the first aerosol model fully coupled to a NWP model with data assimilation. Aerosol optical depth (AOD) data of the MODIS instrument on Terra and Aqua satellites was assimilated. The performance of the aerosol model was evaluated by the means of case studies. The assimilation of MODIS AOD improved the subsequent aerosol predictions when compared with observations, in particular concerning correlations and AOD peak values. The assimilation is less effective in correcting a positive or a negative bias.