Embed Size (px)
DESCRIPTION
Retrieval of the turbulent and backscattering properties using a non-linear filtering technique applied to Doppler LIDAR observation. Christophe Baehr* , ** , C. Beigbeder*, F. Couvreux*, A. Dabas*, B. Piguet*. ISARS 2012 - Boulder, Colorado, USA, 5-8 June 2012. - PowerPoint PPT Presentation
Citation preview
Cliquez pour modifier le style du titre
Cliquez pour modifier le style des sous-titres du masque
Retrieval of the turbulent and backscattering properties using a
non-linear filtering technique applied to Doppler LIDAR
observation Christophe Baehr*,**, C. Beigbeder*,
F. Couvreux*, A. Dabas*, B. Piguet*
** Météo-France/CNRS – CNRM/GAME URA1357** Institut de Mathématiques de Toulouse - Université de Toulouse III
ISARS 2012 - Boulder, Colorado, USA, 5-8 June 2012
Outline
Introduction How a Doppler lidar and a particle filter can
retrieve the properties of a random medium? Some theoretical elements. Experimental results : a numerical exercise. Comparisons with Numerical products Next Steps
Introduction
Our final objective is to retrieve the characteristics of a random medium from sparse observations.
The retrieval is done locally, that is, in the vicinity of the observations.
The retrieval technique we have developed has already been applied successfully to in-situ measurements of wind and temperature performed at fixed locations.
We present here the extension of our method to Doppler lidar data for the retrieval of TKE and EDR at a fine temporal resolution.
How a Doppler lidar and a particle filter can retrieve the properties of a random medium?
Theoretical Background
are a random path and a random field
is the acquisition process of the random field along the path
is the conditional expectation according to the trajectory.
where
Theoretical Background
The probability laws of the random medium considered along a random path:
There is a time evolution that required a local model of the probed medium. Here this evolution is given by the Markovian kernel
Theoretical Background
The algorithm in term of state vectors is given by :
It is equivalent to the evolution in probability laws :
and solve the stochastic dynamical system :
The Stochastic Lagrangian Model
The local Markovian evolution needs a physical model.We have choose to use a Stochastic Lagrangian Model (SLM).
The model adapted to wind vertical profiles is excerpt from the 3D SLM we use for the atmospheric measurements :
The term is embedded in a truncated normal distribution
learned by our algorithm.
Experimental results : a numerical exercise.
Data recorded the June 19th, 2011 every 6s between 13h26 and 14h49 UTC at Lannemezan, France.
Leosphere vertical lidar involved during the BLLAST experiment.
Experimental Results
Vertical wind times series ( 6s ) black : reference, blue : observation
Experimental Results
Vertical wind times series ( 6s ) black : reference, blue : observation, red : estimated
Experimental Results
Vertical wind, black : reference, blue : observation, red : estimated
times series ( 6s )
Power Spectral Density
Experimental Results
Vertical wind PSD, black : reference, blue : observation, red : estimated
Experimental Results
Vertical wind profiles averaged on 1’. Above : reference, bottom : estimated
Comparisons with other Numerical products
Comparison with a Meso-NH simulation (for an other day and an other location) to compare the shape of the structures (above : wind, bottom TKE), especially for the TKE.
How it is possible to assess the quality of TKE and EDR estimates ?
Comparisons with other Numerical products
On June 19th, 2011 from 13h26 to 14h49 UTC, the tethered balloon flew at 60m and we compare its data with the lidar range bin 75-125m:
- Balloon wind variance ~ 0.39 m2s−2.
- lidar filtered signal variance ~ 0.42 m2s−2.
- lidar averaged TKE ~ 0.25 m2s−2.
We are waiting for other lidar data to compare with other, more representative balloon flights.
Next steps
Continue the work on the 3D estimations using hemispherical scanning lidars.
Work on the lidar observation operator. Full set of numerical comparisons. MesoNH comparisons with specific BLLAST simulations
Merge the estimation of Doppler lidar and Aerosol lidar to estimate the parameters of a full 3D atmospheric domain.
Work on the assimilation of turbulence parameters in NH models (e. g. Meso-NH).
