The impact of climate change on weather pattern dynamics over the North Atlantic is explored through the lens of information theory. These tools allow the predictability of the succession of weather patterns and the irreversible nature of the dynamics to be clarified. It is shown that the predictability is increasing in the observations, while the opposite trend is found in model projections. The irreversibility displays an overall increase in time in both the observations and the model runs.

Understanding whether variations in ocean heat content are driven by air-sea heat fluxes or by ocean dynamics is of crucial importance to enhance climate projections. We use a relatively novel causal method to quantify interactions between ocean heat budget terms based on climate models. We find that low-resolution models overestimate the influence of ocean dynamics in the upper ocean, and that changes in ocean heat content are dominated by air-sea fluxes at high resolution.

The Madden–Julian Oscillation (MJO) has important socioeconomic impacts due to its influence on both tropical and extratropical weather extremes. In this study, we use machine learning (ML) to correct the predictions of the weather model holding the best performance, developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). We show that the ML post-processing leads to an improved prediction of the MJO geographical location and intensity.

Modeling the climate at high resolution is crucial to represent the snowfall accumulation over the complex orography of the Antarctic coast. While ice cores provide a view constrained spatially but over centuries, climate models can give insight into its spatial distribution, either at high resolution over a short period or vice versa. We downscaled snowfall accumulation from climate model historical simulations (1850–present day) over Dronning Maud Land at 5.5 km using a statistical method.

Forecasting the height of new snow is essential for avalanche hazard surveys, road and ski resort management, tourism attractiveness, etc. Météo-France operates a probabilistic forecasting system using a numerical weather prediction system and a snowpack model. It provides better forecasts than direct diagnostics but exhibits significant biases. Post-processing methods can be applied to provide automatic forecasting products from this system.

We provide a novel approach to diagnose the strength of the ocean–atmosphere coupling by using both a reduced order model and reanalysis data. Our findings suggest the ocean–atmosphere dynamics presents a rich variety of features, moving from a chaotic to a coherent coupled dynamics, mainly attributed to the atmosphere and only marginally to the ocean. Our observations suggest further investigations in characterizing the occurrence and spatial dependency of the ocean–atmosphere coupling.

Accurate models of spatial, temporal, and inter-variable dependencies are of crucial importance for many practical applications. We review and compare several methods for multivariate ensemble post-processing, where such dependencies are imposed via copula functions. Our investigations utilize simulation studies that mimic challenges occurring in practical applications and allow ready interpretation of the effects of different misspecifications of the numerical weather prediction ensemble.

Statistical post-processing of ensemble forecasts is now a well-known procedure in order to correct biased and misdispersed ensemble weather predictions. But practical application in European national weather services is in its infancy. Different applications of ensemble post-processing using machine learning at an industrial scale are presented. Forecast quality and value are improved compared to the raw ensemble, but several facilities have to be made to adjust to operational constraints.

Postprocessing schemes used to correct weather forecasts are no longer efficient when the model generating the forecasts changes. An approach based on response theory to take the change into account without having to recompute the parameters based on past forecasts is presented. It is tested on an analytical model and a simple model of atmospheric variability. We show that this approach is effective and discuss its potential application for an operational environment.

A generic methodology is developed to estimate the model error and simulate the model uncertainty related to a specific physical process. The method estimates the model error by comparing two different representations of the physical process in otherwise identical models. The found model error can then be used to perturb the model and simulate the model uncertainty. When applying this methodology to deep convection an improvement in the probabilistic skill of the ensemble forecast is found.

Statistical post-processing aims to increase the predictive skill of probabilistic ensemble weather forecasts by learning the statistical relation between historical pairs of observations and ensemble forecasts within a given training data set. This study compares four different training schemes and shows that including multiple years of data in the training set typically yields a more stable post-processing while it loses the ability to quickly adjust to temporal changes in the underlying data.

We need seasonal predictions of temperature and precipitation to prepare hydrological outlooks. Since the skill is limited, statistical correction and combination of outputs from multiple models are necessary. We use the forecasts of past situations from the EUROSIP multi-model system for 6 case studies in Western Europe and the Mediterranean Region. We identify skill for spring temperature in most areas and winter precipitation in Sweden and Greece. Sample size for training appears crucial.

We investigate the modeling of the effects of the unresolved scales on the large scales of the coupled ocean–atmosphere model MAOOAM. Two different physically based stochastic methods are considered and compared, in various configurations of the model. Both methods show remarkable performances and are able to model fundamental changes in the model dynamics. Ways to improve the parameterizations' implementation are also proposed.

The El Niño–Southern Oscillation phenomenon is a slow dynamics present in the coupled ocean–atmosphere tropical Pacific system which has important teleconnections with the northern extratropics. These teleconnections are usually believed to be the source of an enhanced predictability in the northern extratropics at seasonal to decadal timescales. This question is challenged by investigating the causality between these regions using an advanced technique known as convergent cross mapping.

The predictability of weather models is limited largely by the initial state error growth or decay rates. We have computed these rates for PUMA, a global model for the atmosphere, and MAOOAM, a more simplified, coupled model which includes the ocean. MAOOAM has processes at distinct timescales, whereas PUMA surprisingly does not. We propose a new programme to compute the natural directions along the flow that correspond to the growth or decay rates, to learn which components play a role.

Large-scale weather patterns such as the North Atlantic Oscillation, which dictates the harshness of European winters, vary over the course of years. By recreating it in a simple ocean-atmosphere model, we hope to understand what drives this slow, hard-to-predict variability. MAOOAM is such a model, in which the resolution and included physical processes can easily be modified. The modular system allowed us to show the robustness of the slow variability against changes in model resolution.