Liliane Bel

A spatial hybrid approach for downscaling of extreme precipitation fields

For a few decades, climate models are used to provide future scenarios of precipitation with increasingly higher spatial resolution. However, this resolution is not yet sufficient to describe efficiently what happens at local scale. Dynamical and statistical methods of downscaling have been developed and allow us to make the link between two levels of resolution and enable us to get values at a local scale based on large-scale information from global or regional climate models. Nevertheless, both the extreme behavior and the spatial structures are not well described by these downscaling methods. We propose a two-step methodology, called spatial hybrid downscaling (SHD), to solve this problem. The first step consists in applying a univariate (i.e., one-dimensional) statistical downscaling to link the high- and low-resolution variables at some given locations. Once this 1d-link is performed, a conditional simulation algorithm of max-stable processes adapted to the extremal t process enables us to get conditional distributions of extreme precipitation at any point of the region. An application is performed on precipitation data in the south of France where extreme (Cevenol) events have major impacts (e.g., floods). Different versions of the SHD approach are tested. Most of them show particularly good results regarding univariate and multivariate criteria and overcome classical downscaling techniques tested in comparison. Furthermore, these conclusions are robust to the choice of the 1d-link functions tested and to the choice of the conditioning points to drive the conditional local-scale simulations performed by the SHD approach.

Compound Poisson‐gamma vs. delta‐gamma to handle zero‐inflated continuous data under a variable sampling volume

Ecological data such as biomasses often present a high proportion of zeros with possible skewed positive values. The Delta-Gamma (DG) approach, which models separately the presence-absence and the positive biomass, is commonly used in ecology. A less commonly known alternative is the compound Poisson-gamma (CPG) approach, which essentially mimics the process of capturing clusters of biomass during a sampling event. Regardless of the approach, the effort involved in obtaining a sample (henceforth called the sampling volume, but could also include swept areas, sampling durations, etc.), which can potentially be quite variable between samples, needs to be taken into account when modelling the resulting sample biomass. This is achieved empirically for the DG approach (using a generalized linear model with sampling volume as a covariate), and theoretically for the CPG approach (by scaling a parameter of the model). In this study, the consequences of this disparity between approaches were explored first using theoretical arguments, then using simulations and finally by applying the approaches to catch data from a commercial groundfish trawl fishery. The simulation study results point out that the DG approach can lead to poor estimates when far from standard idealized sampling assumptions. On the contrary, the CPG approach is much more robust to variable sampling conditions, confirming theoretical predictions. These results were confirmed by the case study for which model performances were weaker for the DG. Given the results, care must be taken when choosing an approach for dealing with zero-inflated continuous data. The DG approach, which is easily implemented using standard statistical softwares, works well when the sampling volume variability is small. However, better results were obtained with the CPG model when dealing with variable sampling volumes.