|From||Roger Brugge <email@example.com>|
|Date||Tue, 23 Jan 2018 17:01:36 +0000|
Forwarded from CLIMLIST...
Postdoctoral Research Associate position
“Laboratoire des Sciences du Climat et de l’Environnement”
(LSCE, Gif-sur-Yvette, France)
funded by CEA on
Event Attribution of Climate Changes
with dynamically driven-Stochastic Weather Generators
Expected starting date_: Spring 2018.
Duration_: 16 months
Context of the position:
There is an increasing interest worldwide in assessing the extent to
which recent extreme weather and climate events can be solely linked to
natural climate variability or be significantly altered in frequency or
intensity by human-induced climate change (Stott, 2016). Examples are
numerous: Paris flood in June 2016, drought in Central Europe in Summer
2015, or the extremely cold European winter of 2009/10. The usual
question is to know to what extent such extremes are linked to climate
change, and if they are becoming (or will become) more or less frequent.
Therefore, the science of “event attribution” is evolving rapidly.
This postdoctoral position is part of the European ERA4CS “EUPHEME”
project, whose one of the main objectives is to develop state-of-the-art
event attribution methods for a range of timescales, and new techniques
for evaluating their reliability. The vision of the EUPHEME project is
to place extreme weather events in the context of climate variability
and change, thereby helping European citizens adapt to a changing
climate and mitigate its worst effects.
Event attribution relies on comparisons of various statistics of extreme
events between a “factual world” (i.e., the world as we observe it or at
least with the real physical forcings) and a “counter-factual world”
(i.e., the world as it would be without one or several given forcings,
as human-induced greenhouse gas emissions). Those comparisons are
usually performed based on a very high number (1000s of runs) of climate
model simulations. One issue is that General circulation models are
generally too expensive to be run as long as needed to compute such
statistics and at the demanded resolution. Stochastic weather generators
(SWGs) are therefore a valid alternative: they are statistical models
calibrated on past observations to simulate as many datasets as desired
with the same statistical properties [Wilks, 2010, 2012].
Moreover, the dynamics of atmospheric motions depends on the observed
atmospheric states. Recently, a technique to measure such
state-dependent dynamical properties have been found [Faranda et al.,
2017] by the computation of the so-called local dimensions (a measure of
the state disorder) and the local persistence. Preliminary studies have
shown that such information can be lumped with the statistical model to
obtain “conditional SWGs” (e.g., Carreau & Vrac, 2011) providing
simulations with a realistic dynamics.
The goal of the postdoctoral research is to provide both a theoretical
framework and a numerical tool to build conditional SWGs driven by
dynamical systems properties.
-The first step will be to introduce such properties and indicators in
simple autoregressive models to mimic the dynamical properties of
univariate time series of climate variables and simple dynamical systems.
-Those state-dependent indicators will then be used to condition
progressively more sophisticated SWGs (e.g., Vrac et al., 2007), up to
spatial models (i.e., simulating fields) for climate variables as wind,
temperature or precipitation.
-At each stage, the different conditional SWGs developed will be applied
in an event attribution purpose to evaluate the improvements achieved or
still needed, depending on the extreme events of interest.
Skills & diploma requested:
The successful candidate should possess:
-A PhD in Statistics, Applied Mathematics, Statistical Physics, Climate
sciences or related field;
-Strong bases on stochastic models including their statistical inference;
-A basic knowledge of climate dynamics;
-An excellent knowledge of R (recommended) or Python (second choice)
-A basic knowledge of dynamical systems framework will be a plus but is
-English proficiency and attitude to work in an international research
Geographical location & scientific team:
This postdoctoral position will be located at Gif-sur-Yvette (France),
in the “Extremes – Statistics – Impacts – Regionalization” (ESTIMR)
scientific research team of the “Laboratoire des Sciences du Climat et
de l’Environnement” (LSCE). The ESTIMR team develops a methodological
research aiming to better understand the climate data: statistical
analyses of observations and simulations in order to investigate the
variability and identify the trends, modelling of extreme events,
detection and attribution of their changes, downscaling, bias adjustment
of simulations, uncertainty modelling of climate projections, etc. The
ESTIMR team leads and participates to international projects, from pure
to more applied science project. The main activity of the team relies on
the use and development of advanced statistical models via a strong
multidisciplinary interaction among climatology, modelling and statistics.
How to apply:_Applications will be open until February 15, 2018 (or
until the position is filled)and have to be submitted by e-mail to M.
Vrac (mathieu.vrac[at]lsce.ipsl.fr) and D. Faranda
(davide.faranda[at]lsce.ipsl.fr)as soon as possible and must include:
-a CV (max 2 pages + Publication list),
-A statement of research interests describing why the candidate fits the
position (max 2 pages),
-The names of at least two references including e-mail addresses and
More information on the “Extremes – Statistics – Impacts –
Regionalization” (ESTIMR) team:
More information on the “Laboratoire des Sciences du Climat et de
Carreau, J., Vrac, M. (2011) "Stochastic downscaling of precipitation
with neural network conditional mixture models". Water Resources
Research, 47, W10502, doi:10.1029/2010WR010128
Faranda, D., Messori, G. and Yiou, P. (2017) Dynamical proxies of North
Atlantic predictability and extremes. /Scientific Reports/, 7-41278
Vrac, M., M. Stein, K. Hayhoe. Statistical downscaling of precipitation
through nonhomogeneous stochastic weather typing. Climate Research,
2007, 34: 169-184, doi: 10.3354/cr00696
Wilks DS. Use of stochastic weather generators for precipitation
downscaling. /WIRES Clim Change/ 2010, 1(6):898–907
Wilks, D. S. (2012), Stochastic weather generators for climate-change
downscaling, part II: multivariable and spatially coherent multisite
downscaling. WIREs Clim Change, 3: 267–278. doi:10.1002/wcc.167
Go to: Periods · List Information · Index by: Date (or Reverse Date), Thread, Subject or Author.