A fast atmospheric turbulent parameters estimation using particle filtering. Application to LIDAR observations

Abstract

Estimating fast turbulence fluctuations in the boundary layer of the atmosphere, using remote detection instrument is an important scientific issue. Doppler LIDAR, is typically used to get this kind of information because it can make fast, distant, precise, and non-intrusive measurements of the wind field by giving the radial component in any direction. The objective of those measurements is to evaluate as precisely as possible the wind structure using the partial wind information provided, in order to estimate turbulent parameters. The approach presented in this paper, consist in coupling the remote detection system and a stochastic Lagrangian model of the atmosphere. The fluid is represented by a set of interacting particles, evolving according to an evolution system based on S.B Pope work. Data provided by the instrument are assimilated in real time in the model using a particle filtering algorithm. The purpose is to locally correct the properties of particles using measurements, to fit the real fluid observed. A precise real time estimation of the wind field, allows then to estimate turbulent parameters. The methodology has produced convincing results on simulated Doppler LIDAR measurements, in tree-dimensional modeling.