Variability and combination as an ensemble of mineral dust forecasts during the 2021 CADDIWA experiment using the WRF 3.7.1 and CHIMERE v2020r3 models

• Published in: Geoscientific Model Development
• Main Author: Menut, Laurent
• Other Authors: Laboratoire de Météorologie Dynamique (UMR 8539) (LMD), Institut national des sciences de l'Univers (INSU - CNRS)-École polytechnique (X)-École des Ponts ParisTech (ENPC)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Département des Géosciences - ENS Paris, École normale supérieure - Paris (ENS-PSL), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École normale supérieure - Paris (ENS-PSL), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)
• Format: Article in Journal/Newspaper
• Language: English
• Published: HAL CCSD 2023
• Subjects: [SDU]Sciences of the Universe [physics]; Aerosol Robotic Network
• Online Access: https://insu.hal.science/insu-04195501; https://insu.hal.science/insu-04195501/document; https://insu.hal.science/insu-04195501/file/gmd-16-4265-2023.pdf; https://doi.org/10.5194/gmd-16-4265-2023

International audience As operational support to define the Clouds-Atmospheric Dynamics-Dust Interactions in West Africa (CADDIWA) field campaign which took place in the Cape Verde area, the coupled regional model WRF-CHIMERE is deployed in forecast mode during the summer 2021. The simulation domain covers West Africa and the eastern Atlantic and allows the modeling of dust emissions and their transport to the Atlantic. On this route, we find Cape Verde, which was used as a base for measurements during the CADDIWA campaign. Meteorological variables and mineral dust concentrations are forecasted on a horizontal grid with a 30 km resolution and from the surface to 200 hPa. For a given day D, simulations are initialized from D-1 analyses and run for 4 d until D+4, yielding up to six available simulations on a given day. For each day, we thus have six different calculations, with better precision expected the closer we get to the analysis (lead D-1). In this study, a quantification of the forecast variability of wind, temperature, precipitation and mineral dust concentrations according to the modeled lead is presented. It is shown that the forecast quality does not decrease with time, and the high variability observed on some days for some variables (wind, temperature) does not explain the behavior of other dependent and downwind variables (mineral dust concentrations). A new method is also tested to create an ensemble without perturbing input data, but considering six forecast leads available for each date as members of an ensemble forecast. It has been shown that this new forecast based on this ensemble is able to give better results for two AErosol RObotic NETwork (AERONET) stations than the four available for aerosol optical depth observations. This could open the door to further testing with more complex operational systems.