arXiv subjects

Lesley De Cruz

Publications and source records attributed to Lesley De Cruz.

At least 19 recordsLinked to original sources

arXiv2026

Spread/Error relationship and spatial error structure of precipitation ensemble nowcasting: Comparison of STEPS and generative AI

The predictability of the generative AI-based nowcasting model LDCast (trained on another region) is evaluated over Belgium, together with the pysteps implementation of the nowcasting algorithm STEPS. STEPS and LDCast are slightly underdispersive, but the ensemble spread provides an estimation of the error at almost all scales. Both models adapt the properties of their ensembles to the type of event, either convective or stratiform. The spatial scores of the STEPS and LDCast ensembles are compared with those of surrogate ensembles having some key properties, revealing that both STEPS and LDCast have very little ability to spatially localise the ensemble mean error vector through their ensemble members. This suggests that the content of STEPS and LDCast ensembles is informative in terms of statistics, but not in terms of dynamics.

physics.ao-ph↗

arXiv2025

GINTRIP: Interpretable Temporal Graph Regression using Information bottleneck and Prototype-based method

Deep neural networks (DNNs) have demonstrated remarkable performance across various domains, but their inherent complexity makes them challenging to interpret. This is especially true for temporal graph regression tasks due to the complex underlying spatio-temporal patterns in the graph. While interpretability concerns in Graph Neural Networks (GNNs) mirror those of DNNs, no notable work has addressed the interpretability of temporal GNNs to the best of our knowledge. Innovative methods, such as prototypes, aim to make DNN models more interpretable. However, a combined approach based on prototype-based methods and Information Bottleneck (IB) principles has not yet been developed for temporal GNNs. Our research introduces a novel approach that uniquely integrates these techniques to enhance the interpretability of temporal graph regression models. The key contributions of our work are threefold: We introduce the Graph INterpretability in Temporal Regression task using Information bottleneck and Prototype (GINTRIP) framework, the first combined application of IB and prototype-based methods for interpretable temporal graph tasks. We derive a novel theoretical bound on mutual information (MI), extending the applicability of IB principles to graph regression tasks. We incorporate an unsupervised auxiliary classification head, fostering diverse concept representation using multi-task learning, which enhances the model's interpretability. Our model is evaluated on real-world datasets like traffic and crime, outperforming existing methods in both forecasting accuracy and interpretability-related metrics such as MAE, RMSE, MAPE, and fidelity.

cs.LG↗

arXiv2025

DIRESA, a distance-preserving nonlinear dimension reduction technique based on regularized autoencoders

In meteorology, finding similar weather patterns or analogs in historical datasets can be useful for data assimilation, forecasting, and postprocessing. In climate science, analogs in historical and climate projection data are used for attribution and impact studies. However, most of the time, those large weather and climate datasets are nearline. This means that they must be downloaded, which takes a lot of bandwidth and disk space, before the computationally expensive search can be executed. We propose a dimension reduction technique based on autoencoder (AE) neural networks to compress the datasets and perform the search in an interpretable, compressed latent space. A distance-regularized Siamese twin autoencoder (DIRESA) architecture is designed to preserve distance in latent space while capturing the nonlinearities in the datasets. Using conceptual climate models of different complexities, we show that the latent components thus obtained provide physical insight into the dominant modes of variability in the system. Compressing datasets with DIRESA reduces the online storage and keeps the latent components uncorrelated, while the distance (ordering) preservation and reconstruction fidelity robustly outperform Principal Component Analysis (PCA) and other dimension reduction techniques such as UMAP or variational autoencoders.

cs.LG↗

arXiv2020

Statistical Postprocessing for Weather Forecasts -- Review, Challenges and Avenues in a Big Data World

Statistical postprocessing techniques are nowadays key components of the forecasting suites in many National Meteorological Services (NMS), with for most of them, the objective of correcting the impact of different types of errors on the forecasts. The final aim is to provide optimal, automated, seamless forecasts for end users. Many techniques are now flourishing in the statistical, meteorological, climatological, hydrological, and engineering communities. The methods range in complexity from simple bias corrections to very sophisticated distribution-adjusting techniques that incorporate correlations among the prognostic variables. The paper is an attempt to summarize the main activities going on this area from theoretical developments to operational applications, with a focus on the current challenges and potential avenues in the field. Among these challenges is the shift in NMS towards running ensemble Numerical Weather Prediction (NWP) systems at the kilometer scale that produce very large datasets and require high-density high-quality observations; the necessity to preserve space time correlation of high-dimensional corrected fields; the need to reduce the impact of model changes affecting the parameters of the corrections; the necessity for techniques to merge different types of forecasts and ensembles with different behaviors; and finally the ability to transfer research on statistical postprocessing to operations. Potential new avenues will also be discussed.

physics.ao-ph↗

arXiv2019

Routes to long-term atmospheric predictability in reduced-order coupled ocean-atmosphere systems -- Impact of the ocean basin boundary conditions

The predictability of the atmosphere at short and long time scales, associated with the coupling to the ocean, is explored in a new version of the Modular Arbitrary-Order Ocean-Atmosphere Model (MAOOAM), based on a 2-layer quasi-geostrophic atmosphere and a 1-layer reduced-gravity quasi-geostrophic ocean. This version features a new ocean basin geometry with periodic boundary conditions in the zonal direction. The analysis presented in this paper considers a low-order version of the model with 40 dynamical variables. First the increase of surface friction (and the associated heat flux) with the ocean can either induce chaos when the aspect ratio between the meridional and zonal directions of the domain of integration is small, or suppress chaos when it is large. This reflects the potentially counter-intuitive role that the ocean can play in the coupled dynamics. Second, and perhaps more importantly, the emergence of long-term predictability within the atmosphere for specific values of the friction coefficient occurs through intermittent excursions in the vicinity of a (long-period) unstable periodic solution. Once close to this solution the system is predictable for long times, i.e. a few years. The intermittent transition close to this orbit is, however, erratic and probably hard to predict. This new route to long-term predictability contrasts with the one found in the closed ocean-basin low-order version of MAOOAM, in which the chaotic solution is permanently wandering in the vicinity of an unstable periodic orbit for specific values of the friction coefficient. The model solution is thus at any time influenced by the unstable periodic orbit and inherits from its long-term predictability.

physics.ao-ph↗

arXiv2018

Exploring the Lyapunov instability properties of high-dimensional atmospheric and climate models

The stability properties of intermediate-order climate models are investigated by computing their Lyapunov exponents (LEs). The two models considered are PUMA (Portable University Model of the Atmosphere), a primitive-equation simple general circulation model, and MAOOAM (Modular Arbitrary-Order Ocean-Atmosphere Model), a quasi-geostrophic coupled ocean-atmosphere model on a beta-plane. We wish to investigate the effect of the different levels of filtering on the instabilities and dynamics of the atmospheric flows. Moreover, we assess the impact of the oceanic coupling, the dissipation scheme and the resolution on the spectra of LEs. The PUMA Lyapunov spectrum is computed for two different values of the meridional temperature gradient defining the Newtonian forcing. The increase of the gradient gives rise to a higher baroclinicity and stronger instabilities, corresponding to a larger dimension of the unstable manifold and a larger first LE. The convergence rate of the rate functional for the large deviation law of the finite-time Lyapunov exponents (FTLEs) is fast for all exponents, which can be interpreted as resulting from the absence of a clear-cut atmospheric time-scale separation in such a model. The MAOOAM spectra show that the dominant atmospheric instability is correctly represented even at low resolutions. However, the dynamics of the central manifold, which is mostly associated to the ocean dynamics, is not fully resolved because of its associated long time scales, even at intermediate orders. This paper highlights the need to investigate the natural variability of the atmosphere-ocean coupled dynamics by associating rate of growth and decay of perturbations to the physical modes described using the formalism of the covariant Lyapunov vectors and to consider long integrations in order to disentangle the dynamical processes occurring at all time scales.

physics.ao-ph↗

arXiv2017

The Modular Arbitrary-Order Ocean-Atmosphere Model: MAOOAM v1.0

This paper describes a reduced-order quasi-geostrophic coupled ocean-atmosphere model that allows for an arbitrary number of atmospheric and oceanic modes to be retained in the spectral decomposition. The modularity of this new model allows one to easily modify the model physics. Using this new model, coined the "Modular Arbitrary-Order Ocean-Atmosphere Model" (MAOOAM), we analyse the dependence of the model dynamics on the truncation level of the spectral expansion, and unveil spurious behaviour that may exist at low resolution by a comparison with the higher-resolution configurations. In particular, we assess the robustness of the coupled low-frequency variability when the number of modes is increased. An "optimal" configuration is proposed for which the ocean resolution is sufficiently high, while the total number of modes is small enough to allow for a tractable and extensive analysis of the dynamics.

physics.ao-ph↗

arXiv2015

Low-frequency variability and heat transport in a low-order nonlinear coupled ocean-atmosphere model

We formulate and study a low-order nonlinear coupled ocean-atmosphere model with an emphasis on the impact of radiative and heat fluxes and of the frictional coupling between the two components. This model version extends a previous 24-variable version by adding a dynamical equation for the passive advection of temperature in the ocean, together with an energy balance model. The bifurcation analysis and the numerical integration of the model reveal the presence of low-frequency variability (LFV) concentrated on and near a long-periodic, attracting orbit. This orbit combines atmospheric and oceanic modes, and it arises for large values of the meridional gradient of radiative input and of frictional coupling. Chaotic behavior develops around this orbit as it loses its stability; this behavior is still dominated by the LFV on decadal and multi-decadal time scales that is typical of oceanic processes. Atmospheric diagnostics also reveals the presence of predominant low- and high-pressure zones, as well as of a subtropical jet; these features recall realistic climatological properties of the oceanic atmosphere. Finally, a predictability analysis is performed. Once the decadal-scale periodic orbits develop, the coupled system's short-term instabilities --- as measured by its Lyapunov exponents --- are drastically reduced, indicating the ocean's stabilizing role on the atmospheric dynamics. On decadal time scales, the recurrence of the solution in a certain region of the invariant subspace associated with slow modes displays some extended predictability, as reflected by the oscillatory behavior of the error for the atmospheric variables at long lead times.

nlin.CD↗

arXiv2013

The $p(γ,K^+)Λ$ reaction: consistent high-spin interactions and Bayesian inference of its resonance content

A Bayesian analysis of the world's $p(γ,K^+)Λ$ data is presented. We adopt a Regge-plus-resonance framework featuring consistent interactions for nucleon resonances up to spin $J = 5/2$. The power of the momentum dependence of the consistent interaction structure rises with the spin of the resonance. This leads to unphysical structures in the energy dependence of the computed cross sections when the short-distance physics is cut off with standard hadronic form factors. A plausible, spin-dependent modification of the hadronic form factor is proposed which suppresses the unphysical artifacts. Next, we evaluate all possible combinations of 11 candidate resonances. The best model is selected from the 2048 model variants by calculating the Bayesian evidence values against the world's $p(γ,K^+)Λ$ data. From the proposed selection of 11 resonances, we find that the following nucleon resonances have the highest probability of contributing to the reaction: $S_{11}(1535)$, $S_{11}(1650)$, $F_{15}(1680)$, $P_{13}(1720)$, $D_{13}(1900)$, $P_{13}(1900)$, $P_{11}(1900)$, and $F_{15}(2000)$.

nucl-th↗

arXiv2012

Kaon photoproduction from the deuteron in a Regge-plus-resonance approach

We present a Regge-inspired effective-Lagrangian framework for kaon photoproduction from the deuteron. Quasi-free kaon production is investigated using the Regge-plus-resonance (RPR) elementary operator within the relativistic plane-wave impulse approximation. The RPR model was developed to describe photoinduced and electroinduced charged-kaon production off protons. We show how this elementary operator can be transformed in order to account for the production of neutral kaons from both protons and neutrons. The model results for kaon photoproduction from the deuteron compare favourably to the 2H(g,K)YN data published to date.

nucl-th↗

arXiv2012

Bayesian inference of the resonance content of p(gamma,K+)Lambda

A Bayesian analysis of the world's p(gamma,K+)Lambda data is presented. We adopt a Regge-plus-resonance framework featuring consistent couplings for nucleon resonances up to spin J=5/2, and evaluate 2048 model variants considering all possible combinations of 11 candidate resonances. The best model, labeled RPR-2011, is discussed with special emphasis on nucleon resonances in the 1900-MeV mass region.

nucl-th↗

arXiv2012

A Bayesian analysis of kaon photoproduction with the Regge-plus-resonance model

We address the issue of unbiased model selection and propose a methodology based on Bayesian inference to extract physical information from kaon photoproduction $p(γ,K^+)Λ$ data. We use the single-channel Regge-plus-resonance (RPR) framework for $p(γ,K^+)Λ$ to illustrate the proposed strategy. The Bayesian evidence Z is a quantitative measure for the model's fitness given the world's data. We present a numerical method for performing the multidimensional integrals in the expression for the Bayesian evidence. We use the $p(γ,K^+)Λ$ data with an invariant energy W > 2.6 GeV in order to constrain the background contributions in the RPR framework with Bayesian inference. Next, the resonance information is extracted from the analysis of differential cross sections, single and double polarization observables. This background and resonance content constitutes the basis of a model which is coined RPR-2011. It is shown that RPR-2011 yields a comprehensive account of the kaon photoproduction data and provides reasonable predictions for $p(e,e' K^+)Λ$ observables.

nucl-th↗

arXiv2012

Bayesian inference of the resonance content of p(gamma,K^+)Lambda

A Bayesian analysis of the world's p(gamma,K^+)Lambda data is presented. From the proposed selection of 11 resonances, we find that the following nucleon resonances have the highest probability of contributing to the reaction: S11(1535), S11(1650), F15(1680), P13(1720), D13(1900), P13(1900), P11(1900), and F15(2000). We adopt a Regge-plus-resonance framework featuring consistent couplings for nucleon resonances up to spin J=5/2. We evaluate all possible combinations of 11 candidate resonances. The best model is selected from the 2048 model variants by calculating the Bayesian evidence values against the world's p(gamma,K^+)Lambda data.

nucl-th↗

arXiv2011

Consistent interactions for high-spin fermion fields

We address the issue of consistent interactions for off-shell fermion fields of arbitrary spin. These interactions play a crucial role in the quantum hadrodynamical description of high-spin baryon resonances in hadronic processes. The Rarita-Schwinger description of high-spin fermion fields involves unphysical degrees of freedom, associated with their lower-spin content. These enter the interaction if not eliminated outright. The invariance condition of the interaction under the unconstrained Rarita-Schwinger gauge removes the lower-spin content of the fermion propagator and leads to a consistent description of the interaction. We develop the most general, consistent interaction structure for high-spin fermions. We find that the power of the momentum dependence of a consistent interaction rises with the spin of the fermion field. This leads to unphysical structures in the energy dependence of the computed cross sections when the short-distance physics is cut off with standard hadronic form factors. A novel, spin-dependent hadronic form factor is proposed that suppresses the unphysical artifacts.

nucl-th↗

arXiv2011

Consistent interactions for high-spin baryons

Consistent interactions for off-shell fermion fields of arbitrary spin are constructed from the gauge-invariance requirement of the interaction Lagrangians. These interactions play a crucial role in the quantum hadrodynamical description of high-spin baryon resonances in hadronic processes. We find that the power of the momentum dependence of a consistent interaction rises with the spin of the fermion field. This leads to unphysical structures in the energy dependence of the computed tree-level cross sections when the short-distance physics is cut off with standard hadronic form factors. A novel, spin-dependent hadronic form factor is proposed that suppresses the unphysical artifacts.

nucl-th↗

arXiv2010

Regge-plus-resonance predictions for neutral-kaon photoproduction from the deuteron

We present a Regge-inspired effective-Lagrangian framework for neutral-kaon photoproduction from the deuteron. Quasi-free kaon production is investigated using the Regge-plus-resonance elementary operator within the relativistic plane-wave impulse approximation. The Regge-plus-resonance model was developed to describe photoinduced and electroinduced charged-kaon production off protons. We show how this elementary operator can be transformed to account for the production of neutral kaons from both protons and neutrons. Our results compare favourably to the sole 2H(gamma,K0)YN dataset published to date.

nucl-th↗

arXiv2009

Creating conditions of anomalous self-diffusion in a liquid with molecular dynamics

We propose a computational method to simulate anomalous self-diffusion in a simple liquid. The method is based on a molecular dynamics simulation on which we impose the following two conditions: firstly, the inter-particle interaction is described by a soft-core potential and secondly, the system is forced out of equilibrium. The latter can be achieved by subjecting the system to changes in the length scale at intermittent times. In many respects, our simulation system bears resemblance to slowly driven sandpile models displaying self-organised criticality. We find non-Gaussian single time step displacement distributions during the out-of-equilibrium time periods of the simulation.

cond-mat.stat-mech↗

arXiv2009

Regge-plus-resonance predictions for charged-kaon photoproduction from the deuteron

We present a Regge-inspired effective-Lagrangian framework for charged-kaon photoproduction from the deuteron. Quasi-free kaon production is investigated using the Regge-plus-resonance elementary operator within the non-relativistic plane-wave impulse approximation. The Regge-plus-resonance model was developed to describe photoinduced and electroinduced kaon production off protons and can be extended to strangeness production off neutrons. The non-resonant contributions to the amplitude are modelled in terms of K+(494) and K*+(892) Regge-trajectory exchange in the t-channel. This amplitude is supplemented with a selection of s-channel resonance-exchange diagrams. We investigate several sources of theoretical uncertainties on the semi-inclusive charged-kaon production cross section. The experimental error bars on the photocoupling helicity amplitudes turn out to put severe limits on the predictive power when considering quasi-free kaon production on a bound neutron.

nucl-th↗