Chevalier, Clément

2018-7, Rullière, Didier, Durrande, Nicolas, Bachoc, François, Chevalier, Clément

2016-1-7, Barton, Yannick, Giannakaki, Paraskevi, von Waldow, Harald, Chevalier, Clément, Pfahl, Stephan, Martius, Olivia

Temporal clustering of extreme precipitation events on sub-seasonal time-scales is of crucial importance for the formation of large-scale flood events. Here, the temporal clustering of regional-scale extreme precipitation events in southern Switzerland is studied. These precipitation events are relevant for the flooding of lakes in southern Switzerland and northern Italy. This research determines whether temporal clustering is present and then identifies the dynamics that are responsible for the clustering. An observation-based gridded precipitation data set of Swiss daily rainfall sums and ECMWF reanalysis data sets are used. We use a modified version of Ripley’s K-function, which determines the average number of extreme events in a time period, to characterize temporal clustering on sub-seasonal time-scales and to determine the statistical significance of the clustering. Significant clustering of regional-scale precipitation extremes is found on sub-seasonal time-scales during the fall season. Four high-impact clustering episodes are then selected and the dynamics responsible for the clustering are examined. During the four clustering episodes, all heavy precipitation events were associated with an upper-level breaking Rossby wave over Western Europe and in most cases strong diabatic processes upstream over the Atlantic played a role in the amplification of these breaking waves. Atmospheric blocking downstream over Eastern Europe supported this wave breaking during two of the clustering episodes. During one of the clustering periods, several extratropical transitions of tropical cyclones in the Atlantic contributed to the formation of high-amplitude ridges over the Atlantic basin and downstream wave breaking. During another event, blocking over Alaska assisted the phase locking of the Rossby waves downstream over the Atlantic.

2014, Chevalier, Clément, Picheny, Victor, Ginsbourger, David

Several strategies relying on kriging have recently been proposed for adaptively estimating contour lines and excursion sets of functions under severely limited evaluation budget. The recently released R package KrigInv 3 is presented and offers a sound implementation of various sampling criteria for those kinds of inverse problems. KrigInv is based on the DiceKriging package, and thus benefits from a number of options concerning the underlying kriging models. Six implemented sampling criteria are detailed in a tutorial and illustrated with graphical examples. Different functionalities of KrigInv are gradually explained. Additionally, two recently proposed criteria for batch-sequential inversion are presented, enabling advanced users to distribute function evaluations in parallel on clusters or clouds of machines. Finally, auxiliary problems are discussed. These include the fine tuning of numerical integration and optimization procedures used within the computation and the optimization of the considered criteria.

2012, Mahlstein, Irina, Martius, Olivia, Chevalier, Clément, Ginsbourger, David

In this study we establish a link between the position of the extratropical jet over the Atlantic and extreme temperature events (below the local seasonal 5% quantile or above the 95% quantile respectively) and wind events (above the local 98% quantile) in the Atlantic basin for the present climate. The strongest link between temperature extremes and the jet is found for the winter season. Extended areas in Europe, Greenland and North America see an increased or decreased chance of winter cold extremes depending on the jet position, even up to a week after the jet occupied the position of interest. Chances of warm temperature extremes in summer in southeastern Europe are higher several days after the jet was located in the central Atlantic. Chances of wind extremes are higher in Western Europe at the latitude of the jet and reduced to the north and south.

2016-8-2, Azzimonti, Dario, Bect, Julien, Chevalier, Clément, Ginsbourger, David

2015, Chevalier, Clément, Emery, Xavier, Ginsbourger, David

Gaussian random field (GRF) conditional simulation is a key ingredient in many spatial statistics problems for computing Monte-Carlo estimators and quantifying uncertainties on non-linear functionals of GRFs conditional on data. Conditional simulations are known to often be computer intensive, especially when appealing to matrix decomposition approaches with a large number of simulation points. This work studies settings where conditioning observations are assimilated batch sequentially, with one point or a batch of points at each stage. Assuming that conditional simulations have been performed at a previous stage, the goal is to take advantage of already available sample paths and by-products to produce updated conditional simulations at minimal cost. Explicit formulae are provided, which allow updating an ensemble of sample paths conditioned on n≥0 observations to an ensemble conditioned on n+q observations, for arbitrary q≥1. Compared to direct approaches, the proposed formulae prove to substantially reduce computational complexity. Moreover, these formulae explicitly exhibit how the q new observations are updating the old sample paths. Detailed complexity calculations highlighting the benefits of this approach with respect to state-of-the-art algorithms are provided and are complemented by numerical experiments.

2014, Chevalier, Clément, Bect, Julien, Ginsbourger, David, Vazquez, Emmanuel, Picheny, Victor, Richet, Yann

2016-7-9, Martius, Olivia, Pfahl, Stephan, Chevalier, Clément

2015, Froidevaux, Paul, Schwanbeck, Jan, Weingartner, Rolf, Chevalier, Clément, Martius, Olivia

Determining the role of different precipitation periods for peak discharge generation is crucial for both projecting future changes in flood probability and for short- and medium-range flood forecasting. In this study, catchment-averaged daily precipitation time series are analyzed prior to annual peak discharge events (floods) in Switzerland. The high number of floods considered – more than 4000 events from 101 catchments have been analyzed – allows to derive significant information about the role of antecedent precipitation for peak discharge generation. Based on the analysis of precipitation times series, a new separation of flood-related precipitation periods is proposed: (i) the period 0 to 1 day before flood days, when the maximum flood-triggering precipitation rates are generally observed, (ii) the period 2 to 3 days before flood days, when longer-lasting synoptic situations generate "significantly higher than normal" precipitation amounts, and (iii) the period from 4 days to 1 month before flood days when previous wet episodes may have already preconditioned the catchment. The novelty of this study lies in the separation of antecedent precipitation into the precursor antecedent precipitation (4 days before floods or earlier, called PRE-AP) and the short range precipitation (0 to 3 days before floods, a period when precipitation is often driven by one persistent weather situation like e.g., a stationary low-pressure system). A precise separation of "antecedent" and "peak-triggering" precipitation is not attempted. Instead, the strict definition of antecedent precipitation periods permits a direct comparison of all catchments. The precipitation accumulating 0 to 3 days before an event is the most relevant for floods in Switzerland. PRE-AP precipitation has only a weak and region-specific influence on flood probability. Floods were significantly more frequent after wet PRE-AP periods only in the Jura Mountains, in the western and eastern Swiss plateau, and at the outlet of large lakes. As a general rule, wet PRE-AP periods enhance the flood probability in catchments with gentle topography, high infiltration rates, and large storage capacity (karstic cavities, deep soils, large reservoirs). In contrast, floods were significantly less frequent after wet PRE-AP periods in glacial catchments because of reduced melt. For the majority of catchments however, no significant correlation between precipitation amounts and flood occurrences is found when the last 3 days before floods are omitted in the precipitation amounts. Moreover, the PRE-AP was not higher for extreme floods than for annual floods with a high frequency and was very close to climatology for all floods. The fact that floods are not significantly more frequent nor more intense after wet PRE-AP is a clear indicator of a short discharge memory of Pre-Alpine, Alpine and South Alpine Swiss catchments. Our study poses the question whether the impact of long-term precursory precipitation for floods in such catchments is not overestimated in the general perception. The results suggest that the consideration of a 3–4 days precipitation period should be sufficient to represent (understand, reconstruct, model, project) Swiss Alpine floods.

2014, Ginsbourger, David, Baccou, Jean, Chevalier, Clément, Perales, Frédéric, Garland, Nicolas, Monerie, Yann

Given a function depending both on decision parameters and nuisance variables, we consider the issue of estimating and quantifying uncertainty on profile optima and/or optimal points as functions of the nuisance variables. The proposed methods are based on interpolations of the objective function constructed from a finite set of evaluations. Here the functions of interest are reconstructed relying on a kriging model but also using Gaussian random field conditional simulations that allow a quantification of uncertainties in the Bayesian framework. Besides this, we introduce a variant of the expected improvement criterion, which proves efficient for adaptively learning the set of profile optima and optimizers. The results are illustrated with a toy example and through a physics case study on the optimal packing of polydisperse frictionless spheres.