f. mercier , l. barthes , a. chazottes , c. malletmercier.page.latmos.ipsl.fr/poster_egu.pdf ·...
TRANSCRIPT
F. Mercier*1, L. Barthes1, A. Chazottes1, C. Mallet1
1 LATMOS/IPSL/UVSQ/CNRS
Results
4D-VAR data assimilation algorithm
Context and objectives
Data
Conclusions / Perspectives
Fig. 2 : Spectropluviometre
Fig. 3 : Micro-rain radar (MRR)
Optical disdrometer
Micro Rain Radar (MRR)
• Owner : LATMOS
• Type of measurements :
Water fluxes
• Location :
- Palaiseau (2010 – 2013)
- Ardèche (2013)
• Owner : Météo France
• Type of measurements :
Water volumes
• Location :
- Ardèche (2013)
Abstract
The goal of this work is to have a better understanding of the structure of rain and of the spatiotemporal variations of
the drop size distribution (DSD). For this purpose, we use two different kinds of data at different scales : firstly, a
Doppler vertically pointing radar which gives volumetric information on the DSDs at different heights ; secondly, a
disdrometer which gives flux information on the ground DSDs. In order to link these observations, we use the physics
of the DSDs evolution, which is mathematically represented by partial differential equations able to model different
phenomena, including advection of droplets by wind and gravity, collisions, or evaporation. As a first step, the model
takes only into account the advection (fall of droplets). The linking of observations thanks to the model is made by a
4D-VAR data assimilation algorithm. This algorithm appears able to retrieve the characteristics of the DSD fields on
simulated data and to produce realistic DSDs and vertical wind fields for quiet stratiform real rain events.
The droplet size greatly affects the interactions between
the rain drop and radio waves (Rayleigh / Mie scattering
theories).
Hence the study of the drop size distribution (DSD) is of
great importance for both radar (and microwave in
general) meteorology and telecommunications.
Moreover, a better understanding of the phenomena
acting during the fall of droplets could help to model the
rain in numerical weather forecasts.
A large number of models have been proposed
for the parameterization of DSD and its vertical
evolution, including a number of physical
phenomena (see beyond). But these models are
sometimes incompatible and hard to validate.
In this work, we want to control these models
with observations by assimilating different kinds
of data providing information at different scales.
Fig. 1 : Example of
auto break-up of a
large drop.
Extracted from
Villermaux and
Bossa 2009.
We use a numerical model of evolution of the DSDs to link these
heterogeneous observations.
Partial differential equations (PDE) more general model taking
account of different phenomena is of the form :
With � �, , , , � the number of drops by unit of volume and
diameter (�−3��−1).
As a first step, we limit the model to advection in a time/height
plane : We look for the DSDs just under
the freezing level.
We use the numerical model to
propagate these top DSDs at the
different heights and times.
We compare these DSDs and the
observations (Doppler spectra +
disdrometer data).
Evolution model
Fig. 5 : Diagram
describing the
assimilation
algorithm used
in this work.
- A new method for studying evolution and variability of DSDs from heterogeneous radar and disdrometer data.
- Appears quite robust on simulated data.
- Appears to satisfactorily retrieve both measurements and give realistic results on a case study of a quiet rain event.
- Needs to be validated with other independant measurements.
- Possibly underconstrained, especially in case of strong vertical wind field.
Prospects :
- Addition of microphysics phenomena parameterization.
- Use of Doppler cloud radar and/or wind profilers to get a good first
guest for the vertical wind field (to add constraints in the problem) and
to valid the results.
Here we focus on an
event which occurred
on the 12/10/2013 in
Ardèche.
This is a quiet event
(mean rainfall rate
around 2��/ℎ, see
figure 6).
Fig. 6 : Rainfall rates on the ground, as measured by the optical
disdrometer (red) and retrieved from data of both the
disdrometer and the radar by the assimilation algorithm (blue).
Fig. 7 : Doppler spectra at 900m height measured by the MRR
(top) on the 12/10/2013. Doppler spectra estimated from the
DSD retrieved by the assimilation algorithm (bottom).
Fig. 9 : 3 integrated
parameters of the
DSDs retrieved by the
algorithm : mean
diameter Dm (top),
normalised number
of drops No* (middle)
and slope of the DSD
calculated by linear
regression on the
logarithm of the DSD
(bottom). Fig. 8 : Vertical wind field retrieved by the algorithm for the 12/10/2013
event. Also are presented the mean and variance of this wind field for
each vertical layer (100m/layer).
Some conclusions for this event :
- the algorithm is able to very satisfactorily explain the observations (both radar doppler spectra
and ground disdrometer observations).
- The corresponding DSD and vertical wind fields are quite realistic, with patterns in accord with
what we could expect for such a stratiform event.
- It is hard to valid the results, all the available observations being used to produce them.
Fig. 4 : Location of the two
experimental sites. In Ardèche,
the instruments are deployed as
part of the HyMeX campaign.
This is a simplified model implying to limit the study to stratiform
rain events.
With � the drop falling velocity (gravity + vertical wind)
• We work both on simulated data (simulation of cloud base DSDs and wind fields
time series) and on real data.
���� �, , � = � �, ��� �, , �
���� = ���� ��� � + ���� � � � + ���� �� � �
Fig. 5 : Diagram summarizing the 4D-VAR
theoretical algorithm.
Retrievals of observations Vertical wind and DSD retrievals