Prévisibilité des épisodes météorologiques à fort impact : sensibilité aux anomalies d'altitude

Abstract : Heavy precipitation events, which are a feature of the Mediterranean climate, may be erroneously forecast. These errors most often originate either from uncertainties related to initial conditions or the representation of physical processes. The impact of various sources of uncertainty is commonly investigated through ensemble prediction. Within this study, an ensemble methodology has been developed aiming to study forecast sensitivity to upper-level dynamical anomalies. These anomalies exhibit strong gradients which are prone to forecast errors and are furthermore precursors of meteorological disturbances. The proposed methodology is based on the advection of the conservative variable associated to upper-level anomalies: the potential vorticity (PV). The employed model, MIMOSA, performs the PV advection on an isentropic level over a high-resolution grid, which allows a better simulation of the fine-scale structures and gradients, and relaxes the large-scale component towards the analysed PV. MIMOSA PV was implemented into a primitive equation model through a PV-inversion tool. A specific algorithm was chosen, which provides a balanced solution belonging to the model solution subspace. Different advection model configurations yield a full set of initial conditions for the ARPEGE weather prediction model. The ensemble, so-called ARPEGE-MIMOSA, has been applied to 5 real case studies. The first case, also known as the "Draguignan catastrophe" over the Var region on 15 June 2010 was associated to an enhanced lack of predictability ; 4 other cases were observed in 2012 during the HyMeX field campaign and have been used as a first sample to evaluate the statistical improvement of the ARPEGE-MIMOSA ensemble prediction system. The case of 15 June 2010 has been thoroughly investigated, addressing in particular the respective role of the upper-level anomalies and the low-level anomalies. For this case, it is concluded that upper-level uncertainties only played a minor role during the first stage of the event, arguing that there is no coupling between surface and upper-level anomalies. The role of surface uncertainties is investigated through a set of perturbed experiments: manual modifications of the sea level pressure and humidity fields are each seen to improve the first stage of the event. This improvement is attributed to the triggering of a precipitating disturbance along a frontal zone. For the second stage of the event, the role of humidity is emphasized, based on a simulation which brings the situation closer to the observed event, but without yielding a correct precipitation pattern. An alternative explanation is given, highlighting model discrepancies on the evolution of an upper-level wave in a highly-diabatic environment. Finally, an objective evaluation of the ARPEGE-MIMOSA ensemble prediction system from the 4 latter cases has been carried out. Results from ARPEGE-MIMOSA are compared to these from two operational ensemble prediction systems. This comparison led us to a better quantification of the errors related to the upper-level dynamics. It is shown that the ARPEGE-MIMOSA system exhibits some skill for the heavy precipitation thresholds.
Document type :
Complete list of metadatas
Contributor : Williams Exbrayat <>
Submitted on : Friday, May 2, 2014 - 3:45:53 PM
Last modification on : Monday, April 29, 2019 - 5:33:32 PM
Long-term archiving on : Saturday, August 2, 2014 - 12:15:14 PM


  • HAL Id : tel-00986467, version 1



Simon Fresnay. Prévisibilité des épisodes météorologiques à fort impact : sensibilité aux anomalies d'altitude. Océan, Atmosphère. Université Paul Sabatier - Toulouse III, 2014. Français. ⟨tel-00986467⟩



Record views


Files downloads