Search Results

You are looking at 1 - 9 of 9 items for

  • Author or Editor: Matthieu Plu x
  • Refine by Access: All Content x
Clear All Modify Search
Matthieu Plu

Abstract

The inherent predictability of tropical cyclone tracks has received much attention since the 1980s. It is still an issue because of the recent improvement of track forecasts by numerical models. The aim of this study is to assess this predictability limit globally using an approach devised by Lorenz on several up-to-date numerical models. The differences between forecasts valid at the same instant are considered to be error values; the doubling time of these small errors leads to an estimated upper bound on predictability. This method is here applied on cyclone position forecasts obtained from three different global operational models (from ECMWF, Météo-France, and the Met Office) over the main tropical cyclone basins in the world and during three recent cyclone seasons (2006–09).

The resulting estimates of predictability largely exceed the values that are commonly accepted in the literature. The doubling time of small errors is found between 30 and 50 h. An important consequence is that cyclone track forecasts have not reached their predictability limit yet. It is argued that the previous methods for computing the predictability of tropical cyclone tracks did not constrain the environment and the structure of the cyclones initially. But the Lorenz method could still underestimate the inherent predictability of tropical cyclone tracks. The sensitivity of the predictability estimates to the model characteristics is discussed. In particular, the use of wind bogus is suggested to avoid serial correlations between successive forecasts and to accelerate error growth.

Full access
Matthieu Plu
and
Philippe Arbogast

Abstract

In a nonlinear quasigeostrophic model with uniform potential vorticity, an idealized initial state sharing some features with atmospheric low-predictability situations is built. Inspired by previous work on idealized cyclogenesis, two different cyclogenesis scenarios are obtained as a result of a small change of the initial location of one structure. This behavior is interpreted by analyzing the baroclinic interaction between upper- and lower-level anomalies. The error growth mechanism is nonlinear; it does not depend on the linear stability properties of the jet, which are the same in both evolutions.

The ability of ensemble forecasts to capture these two possible evolutions is then assessed given some realistic error bounds in the knowledge of the initial conditions. First, a reference statistical distribution of each of the evolutions is obtained by means of a large Monte Carlo ensemble. Smaller ensembles with size representative of what is available in current operational implementations are then built and compared to the Monte Carlo reference: several singular-vector-based ensembles, a small Monte Carlo ensemble, and a “coherent structure”-based ensemble. This new technique relies on a sampling of the errors on the precursors of the cyclogenesis: amplitude and position errors. In this context, the precursors are handled as coherent structures that may be amplified or moved within realistic error bounds. It is shown that the singular vector ensemble fails to reproduce the bimodal distribution of the variability if the ensemble is not initially constrained, whereas it is accessible at a relatively low cost to the new coherent structures initialization.

Full access
Matthieu Plu
,
Philippe Arbogast
, and
Alain Joly

Abstract

Midlatitude cyclogenesis as interpreted in the framework of either baroclinic development or potential vorticity thinking heavily relies on the concept of synoptic-scale anomaly. Given the existence of potential vorticity inversion and attribution, what is at stake to provide a mathematical definition for this concept is a complete finite-amplitude alternative to the linear-based theory of cyclogenesis. The existence of a reasonably objective way to represent anomalies in both real and idealized flows would not only help understanding cyclogenesis, it would also have many other applications for both theory and in practical forecasts. Inspired by the recent theory of wavelet representation of coherent structures in two-dimensional fluid mechanics, a wavelet representation of three-dimensional potential vorticity anomalies is built. This algorithm relies on the selection of the appropriate two-dimensional wavelet coefficients from the stationary wavelet transform in order to guarantee the critical translation-invariance property. The sensitivity of the algorithm to the position, size, and shape of the structures is assessed.

The wavelet extraction is then applied to the upper-level precursor of a real-case storm of December 1999 and is compared to a basic monopolar extraction. Using potential vorticity inversion and forecasts with a primitive-equation model, it is found that both anomalies have similar implications on the development of the surface cyclone. However, the coherence in time of the extracted wavelet structure in the forecast and analysis sequence is more satisfactory than the extracted monopole: this suggests that the underlying mathematical description of an anomaly proposed here does, indeed, point toward the direction of an actual physical reality of the concept.

Full access
Marie-Dominique Leroux
,
Matthieu Plu
, and
Frank Roux

Abstract

This study is part of the efforts undertaken to resolve the “bad trough/good trough” issue for tropical cyclone (TC) intensity changes and to improve the prediction of these challenging events. Sensitivity experiments are run at 8-km resolution with vortex bogusing to extend the previous analysis of a real case of TC–trough interaction (Dora in 2007). The initial position and intensity of the TC are modified, leaving the trough unchanged to describe a realistic environment. Simulations are designed to analyze the sensitivity of TC prediction to both the variety of TC–trough configurations and the current uncertainty in model analysis of TC intensity and position.

Results show that TC intensification under upper-level forcing is greater for stronger vortices. The timing and geometry of the interaction between the two cyclonic potential vorticity anomalies associated with the cutoff low and the TC also play a major role in storm intensification. The intensification rate increases when the TC (initially located 12° northwest of the trough) is displaced 1° closer. By allowing a gradual deformation and equatorward tilting of the trough, both scenarios foster an extended “inflow channel” of cyclonic vorticity at midlevels toward the vortex inner core. Conversely, unfavorable interaction is found for vortices displaced 3° or 4° east or northeast. Variations in environmental forcing relative to the reference simulation illustrate that the relationship between intensity change and the 850–200-hPa wind shear is not systematic and that the 200-hPa divergence, 335–350-K mean potential vorticity, or 200-hPa relative eddy momentum fluxes may be better predictors of TC intensification during TC–trough interactions.

Full access
Thierry Dupont
,
Matthieu Plu
,
Philippe Caroff
, and
Ghislain Faure

Abstract

Several tropical cyclone forecasting centers issue uncertainty information with regard to their official track forecasts, generally using the climatological distribution of position error. However, such methods are not able to convey information that depends on the situation. The purpose of the present study is to assess the skill of the Ensemble Prediction System (EPS) from the European Centre for Medium-Range Weather Forecasts (ECMWF) at measuring the uncertainty of up to 3-day track forecasts issued by the Regional Specialized Meteorological Centre (RSMC) La Réunion in the southwestern Indian Ocean. The dispersion of cyclone positions in the EPS is extracted and translated at the RSMC forecast position. The verification relies on existing methods for probabilistic forecasts that are presently adapted to a cyclone-position metric. First, the probability distribution of forecast positions is compared to the climatological distribution using Brier scores. The probabilistic forecasts have better scores than the climatology, particularly after applying a simple calibration scheme. Second, uncertainty circles are built by fixing the probability at 75%. Their skill at detecting small and large error values is assessed. The circles have some skill for large errors up to the 3-day forecast (and maybe after); but the detection of small radii is skillful only up to 2-day forecasts. The applied methodology may be used to assess and to compare the skill of different probabilistic forecasting systems of cyclone position.

Full access
Jean-Baptiste Gilet
,
Matthieu Plu
, and
Gwendal Rivière

Abstract

Mechanisms leading a synoptic surface cyclone to cross an upper-level zonal jet and its subsequent deepening are investigated using a two-layer model on a β plane. The baroclinic interaction of a low-level circular cyclonic perturbation with an upper-level one is first studied in vertical and horizontal cyclonic or anticyclonic uniform shears. A first nonlinear effect acting on the shape and energetics of the perturbations is analyzed. If the background shear is anticyclonic, the perturbations are stretched horizontally; they lose energy barotropically but gain it baroclinically by a well-maintained westward tilt with height. Conversely, if the shear is cyclonic, perturbations remain quite isotropic, but they do not keep a favorable vertical tilt with time and the baroclinic interaction is thus only transient. The latitudinal motion of the perturbations also results from a nonlinear effect. It is found to depend strongly on the background potential vorticity (PV) gradient. This effect is a baroclinic equivalent of the so-called nonlinear barotropic “β drift” and combines the nonlinear advection and vertical stretching terms.

These results are confirmed when the anomalies are initially located south of a confined westerly jet. The poleward shift of the lower cyclonic anomaly occurs faster when the vertically averaged PV gradient is strongly positive, which happens when the jet has a large barotropic component. The lower anomaly crosses the jet from the warm to the cold side and deepens afterward. After a detailed description of this regeneration process with the help of an energy budget, it is shown that linear dynamics are not able to reproduce such behavior.

Full access
Clément Brochet
,
Laure Raynaud
,
Nicolas Thome
,
Matthieu Plu
, and
Clément Rambour

Abstract

Emulating numerical weather prediction (NWP) model outputs is important to compute large datasets of weather fields in an efficient way. The purpose of the present paper is to investigate the ability of generative adversarial networks (GANs) to emulate distributions of multivariate outputs (10-m wind and 2-m temperature) of a kilometer-scale NWP model. For that purpose, a residual GAN architecture, regularized with spectral normalization, is trained against a kilometer-scale dataset from the AROME Ensemble Prediction System (AROME-EPS). A wide range of metrics is used for quality assessment, including pixelwise and multiscale Earth-mover distances, spectral analysis, and correlation length scales. The use of wavelet-based scattering coefficients as meaningful metrics is also presented. The GAN generates samples with good distribution recovery and good skill in average spectrum reconstruction. Important local weather patterns are reproduced with a high level of detail, while the joint generation of multivariate samples matches the underlying AROME-EPS distribution. The different metrics introduced describe the GAN’s behavior in a complementary manner, highlighting the need to go beyond spectral analysis in generation quality assessment. An ablation study then shows that removing variables from the generation process is globally beneficial, pointing at the GAN limitations to leverage cross-variable correlations. The role of absolute positional bias in the training process is also characterized, explaining both accelerated learning and quality-diversity trade-off in the multivariate emulation. These results open perspectives about the use of GAN to enrich NWP ensemble approaches, provided that the aforementioned positional bias is properly controlled.

Open access
Marie-Dominique Leroux
,
Matthieu Plu
,
David Barbary
,
Frank Roux
, and
Philippe Arbogast

Abstract

The rapid intensification of Tropical Cyclone (TC) Dora (2007, southwest Indian Ocean) under upper-level trough forcing is investigated. TC–trough interaction is simulated using a limited-area operational numerical weather prediction model. The interaction between the storm and the trough involves a coupled evolution of vertical wind shear and binary vortex interaction in the horizontal and vertical dimensions. The three-dimensional potential vorticity structure associated with the trough undergoes strong deformation as it approaches the storm. Potential vorticity (PV) is advected toward the tropical cyclone core over a thick layer from 200 to 500 hPa while the TC upper-level flow turns cyclonic from the continuous import of angular momentum.

It is found that vortex intensification first occurs inside the eyewall and results from PV superposition in the thick aforementioned layer. The main pathway to further storm intensification is associated with secondary eyewall formation triggered by external forcing. Eddy angular momentum convergence and eddy PV fluxes are responsible for spinning up an outer eyewall over the entire troposphere, while spindown is observed within the primary eyewall. The 8-km-resolution model is able to reproduce the main features of the eyewall replacement cycle observed for TC Dora. The outer eyewall intensifies further through mean vertical advection under dynamically forced upward motion. The processes are illustrated and quantified using various diagnostics.

Full access
Arnaud Mounier
,
Laure Raynaud
,
Lucie Rottner
,
Matthieu Plu
,
Philippe Arbogast
,
Michaël Kreitz
,
Léo Mignan
, and
Benoît Touzé

Abstract

Bow echoes (BEs) are bow-shaped lines of convective cells that are often associated with swaths of damaging straight-line winds and small tornadoes. This paper describes a convolutional neural network (CNN) able to detect BEs directly from French kilometer-scale model outputs in order to facilitate and accelerate the operational forecasting of BEs. The detections are only based on the maximum pseudoreflectivity field predictor (“pseudo” because it is expressed in mm h−1 and not in dBZ). A preprocessing of the training database is carried out in order to reduce imbalance issues between the two classes (inside or outside bow echoes). A CNN sensitivity analysis against a set of hyperparameters is done. The selected CNN configuration has a hit rate of 86% and a false alarm rate of 39%. The strengths and weaknesses of this CNN are then emphasized with an object-oriented evaluation. The BE largest pseudoreflectivities are correctly detected by the CNN, which tends to underestimate the size of BEs. Detected BE objects have wind gusts similar to the hand-labeled BE. Most of the time, false alarm objects and missed objects are rather small (e.g., <1500 km2). Based on a cooperation with forecasters, synthesis plots are proposed that summarize the BE detections in French kilometer-scale models. A subjective evaluation of the CNN performances is also reported. The overall positive feedback from forecasters is in good agreement with the object-oriented evaluation. Forecasters perceive these products as relevant and potentially useful to handle the large amount of available data from numerical weather prediction models.

Free access