Articles | Volume 15, issue 14
Research article
17 Jul 2015
Research article |  | 17 Jul 2015

Aerosol forecast over the Mediterranean area during July 2013 (ADRIMED/CHARMEX)

L. Menut, G. Rea, S. Mailler, D. Khvorostyanov, and S. Turquety

Abstract. The ADRIMED (Aerosol Direct Radiative Impact on the regional climate in the MEDiterranean region) project was dedicated to study the atmospheric composition during the summer 2013 in the European Mediterranean region. During its campaign experiment part, the WRF (Weather Research and Forecast Model) and CHIMERE models were used in the forecast mode in order to decide whether intensive observation periods should be triggered. Each day, a simulation of 4 days was performed, corresponding to (D-1) to (D+2) forecast leads. The goal of this study was to determine whether the model forecast spread is lower or greater than the model biases compared to observations. It is shown that the differences between observations and the model are always higher than those between the forecasts. Among all forcing types used in the chemistry-transport model, it is shown that the strong bias and other related low forecast scores are mainly due to the forecast accuracy of the wind speed, which is used both for the mineral dust emissions (a threshold process) and for the long-range transport of aerosol: the surface wind speed forecast spread can reach 50%, leading to mineral dust emission forecast spread of up to 30%. These variations are responsible for a moderate forecast spread of the surface PM10 (a few percentage points) and for a large spread (more than 50%) in the mineral dust concentration at higher altitudes, leading to a mean AOD (aerosol optical depth) forecast spread of ±10%.

Short summary
The atmospheric composition was extensively studied in the European Mediterranean region and during summer 2013 within the framework of the ADRIMED project. During the campaign experiment, the WRF and CHIMERE models were used in forecast mode in order to help scientists to decide whether intensive observation periods should be triggered or not. This study quantifies the origin of the forecast error by comparing several forecast leads to the corresponding measurements.
Final-revised paper