Biostatistique et Processus Spatiaux

ABSTRACT Compound events (CEs), commonly defined as the “combination of multiple drivers and/or hazards that contributes to societal or environmental risk”, often result in amplified impacts compared to individual hazards. In order to estimate the return period of bivariate CEs, a novel nonparametric approach employing bivariate Generalized Pareto distributions (bi‐GPD) is proposed and compared to a copula‐based approach. Special attention is given to account for temporal dependencies and nonconcurrent compound events. The latter are defined as excess of variables over a threshold at a relatively close time. The return period of such bivariate events is carefully defined and closed‐form expressions are obtained for both approaches. Simulations reveal the bi‐GPD approach is effective in case of positive asymptotic dependence and should be avoided in case of asymptotic independence. The novel approach is then applied to ERA5 reanalysis data to analyze two types of compound events: a spatial CE with simultaneous floods due to accumulated precipitation across two large watersheds in France and a preconditioned CE consisting of a devastating flood triggered by extreme precipitation over a saturated soil.

The lethal control of large carnivores is criticized regarding its efficiency to prevent hotspots of attacks on livestock. Previous studies, mainly focused on North America, provided mixed results. We evaluated the effects of wolf lethal removals on the distribution of attack intensities in the French Alps between 2011 and 2020, using a Before After Control‐Impact approach with retrospective data. We built an original framework combining both continuous spatial and temporal scales and a 3D kernel estimation. We compared the attack intensities observed before and after the legal killings of wolves over a period of 90 days and a range of 10 km, and with control situations where no removal occurred. The analysis was corrected for the presence of livestock. A moderate decrease in attack intensity was the most common outcome after the lethal removal of a single wolf. This reduction was greatly amplified when removing two or three wolves. The scale of analysis also modulated this general pattern, with decreases being generally amplified at a small spatio-temporal range. Contextual factors (e.g., geographical or seasonal variations) could also lead to deviations from this general pattern. Overall, between 2011 and 2020, lethal control of wolves in France generally contributed to reducing livestock attacks, but mainly locally and to a minor extent. Our results highlight the importance of accounting for scale in such assessments and suggest that the evaluation of the effectiveness of lethal removals in reducing livestock predation might be more relevant in a local context. As a next step, we recommend to move forward from patterns to mechanisms by linking the effects of lethal control on wolves to their effects on attacks through analysis of fine-scaled data on wolves and livestock.

We examine a generalized KPP equation with a "q-diffusion", which is a framework that unifies various standard linear diffusion regimes: Fickian diffusion ( ), Stratonovich diffusion ( ), Fokker-Planck diffusion ( ), and nonstandard diffusion regimes for general . Using both analytical methods and numerical simulations, we explore how the ability of persistence (measured by some principal eigenvalue) and how the asymptotic spreading speed depend on the parameter q and on the phase shift between the growth rate r(x) and the diffusion coefficient D(x). Our results demonstrate that persistence and spreading properties generally depend on q: for example, appropriate configurations of r(x) and D(x) can be constructed such that q-diffusion either enhances or diminishes the ability of persistence and the spreading speed with respect to the traditional Fickian diffusion. We show that the spatial arrangement of r(x) with respect to D(x) has markedly different effects depending on whether , , or . The case where r is constant is an exception: persistence becomes independent of q, while the spreading speed displays a symmetry around . This work underscores the importance of carefully selecting diffusion models in ecological and epidemiological contexts, highlighting their potential implications for persistence, spreading, and control strategies.

Biological control strategies against mosquito-borne diseases--such as the sterile insect technique (SIT), RIDL, and Wolbachia-based releases--require reliable estimates of dispersal and survival of released males. We propose a mechanistic--statistical framework for mark--release--recapture (MRR) data linking an individual-based 2D diffusion model with its reaction--diffusion limit. Inference is based on solving the macroscopic system and embedding it in a Poisson observation model for daily trap counts, with uncertainty quantified via a parametric bootstrap. We validate identifiability using simulated data and apply the model to an urban MRR campaign in El Cano (Havana, Cuba) involving four weekly releases of sterile Aedes aegypti males. The best-supported model suggests a mean life expectancy of about five days and a typical displacement of about 180 m. Unlike empirical fits of survival or dispersal, our mechanistic approach jointly estimates movement, mortality, and capture, yielding biologically interpretable parameters and a principled framework for designing and evaluating SIT-based interventions.

Identifying disease risk factors, characterizing their effects, and forecasting disease risk across space and time are crucial tasks in human, animal, and plant epidemiology. Statistical and machine learning models have largely superseded purely descriptive analyses of data in handling these tasks. In addition, these models have demonstrated their full potential in the current era, characterized by an unprecedented abundance of data. However, applying these models to real-world, large-scale data sets raises critical questions: Which model should be used? Which explanatory variables should be selected? What data should be allocated for training and validation? The answers to these questions often have a significant impact on the analysis outcomes. One way to address some of these challenges is to analyze risk factors and predict risk by using an ensemble of models rather than relying on a single model. This approach is developed in this article and implemented in the case of virus yellows in sugar beet in France. Among the explanatory variables correlated with the severity of virus yellows, we identified winter and spring temperatures (positive correlation), spring humidity and precipitation (negative correlation), the proportion of cereal crops (positive correlation), the proportion of grasslands (negative correlation), and the distance to sugar beet seed production fields (negative correlation). Additionally, we found that predictions are generally more robust when using a spatial aggregation of models compared with relying on the best individual model. Our approach is highly versatile and can be applied to characterize and predict the spatiotemporal distributions of diverse diseases.

To better understand the emergence of cucumber mosaic virus (CMV) in the protected designation of origin of Espelette pepper (southwestern France), more than 7,300 samples were collected in and around 36 pepper fields in 2021 and 2022 and diagnosed using ELISA, RT-PCR, and partial Sanger sequencing of viral RNAs. This allowed the identification of five new host genera or species among the natural hosts of CMV: Arum italicum, Cerastium glomeratum, Hyacinthoides sp., Lysimachia arvensis, and Trifolium incarnatum. A CMV variant belonging to subgroup IB and presenting a low molecular diversity was highly prevalent in the pepper crops (78% of the pepper samples) as well as in naturally growing plants (8% of the non-pepper samples) within the fields. CMV isolates from group II were detected in a single pepper plant as well as in Hyacinthoides sp. (three samples), Capsella bursa-pastoris (two samples), and Stachys arvensis (one sample). To our knowledge, this is the second report of the occurrence of subgroup IB of CMV in France. Investigation of old pepper samples indicated that it has been present at least since 2009.

In many applications, it is desirable to aggregate climate model projections by combining multiple models into a single projection that aims to leverage their collective strengths, often resulting in improved performance compared to individual models. While climate models exhibit varying levels of global average bias, their local performance often displays significantly larger biases—sometimes by an order of magnitude—with each model showcasing distinct strengths and weaknesses in different regions. Aggregating models without accounting for these spatial differences can degrade the quality of projections by diluting strong regional signals from high-performing models. While many approaches ranging in complexity have been developed, including the commonly used Multi-Model Mean (MMM) and weighted MMM, these methods typically apply a global weighting to the models, overlooking the fact that certain models may excel only in specific regions.To date, the Graph Cut optimization method (Thao et al., 2022) stands out as one of the few techniques effectively leveraging the local capabilities of different models across multi-decadal periods to produce global projections. This method involves selecting the best performing model for each grid point while also ensuring the spatial consistency of the resulting fields. Despite its promising results, which surpass those of other ensemble combination techniques, it is restricted to optimizing for a single variable. This limitation causes inconsistent model selection across variables in multivariate scenarios. This leads to a loss of the multivariate relationships captured in the models. Furthermore, this technique was limited to multi-decadal averages, and is thus unable to capture the distributional characteristics of climate variables, including extreme and compound events.Here, we present significant enhancements to the Graph Cut optimization method, enabling the combination of distributions of daily values. This approach preserves multivariate relationships, better capturing the complete span of climate dynamics. By employing the Hellinger distance to assess model performance, we can identify, at each grid point, the model that most accurately represents the multivariate distribution of target variables (e.g., temperature, pressure, and precipitation), minimizing the emergence of unrealistic discontinuities in the combined fields.To demonstrate the use of our method, we combine 22 models from CMIP6 using three variables: temperature, precipitation, and sea level pressure, achieving better reproduction of ERA5 reanalysis compared to the Multi-Model Mean (MMM). Additionally, a perfect model experiment was conducted to evaluate the robustness and stability of the methodology under high climate change scenarios, such as SSP8.5, and over extended timescales reaching the end of the century. These results highlight the method's ability to maintain reliable performance and spatial consistency in challenging future conditions.REFERENCES Thao, S., Garvik, M., Mariethoz, G. et al. Combining global climate models using graph cuts. Clim Dyn 59, 2345–2361 (2022). https://doi.org/10.1007/s00382-022-06213-4