arXiv subjects

Rodrigo Almeida

Publications and source records attributed to Rodrigo Almeida.

5 recordsLinked to original sources

arXiv2026

Uncertainty-Aware End-to-End AI Weather Forecasting: Disentangling Observation and Model Contributions

End-to-end weather forecasting systems produce skillful global gridded and station forecasts directly from raw Earth observations, replacing the numerical weather prediction pipeline, including data assimilation, at a fraction of its cost. These systems are deterministic and issue no uncertainty. Here we render the Aardvark Weather model probabilistic by attaching one stochastic mechanism to each component: learned, input-dependent noise at the observation encoder, capturing aleatoric uncertainty inherited from the observing system, and Monte Carlo dropout in the processor, capturing epistemic uncertainty in the learned dynamics. The resulting nested ensemble attributes forecast spread to the two sources through a law-of-total-variance decomposition, cross-checked by withholding observation streams. Probabilistic finetuning significantly improves the mean forecast, by 4.2% on average across variables and lead times. The ensemble is calibrated against ERA5 through the medium range (spread-skill ratio 0.98), keeps station RMSE within 2.4% of the deterministic model while beating it in CRPS at every lead time, and trails the operational ECMWF ensemble. The encoder branch behaves as observation-driven uncertainty. Component-attributed uncertainty makes end-to-end forecasts more transparent, a step toward observation-driven digital twins of the atmosphere.

physics.ao-ph↗

arXiv2025

On the Predictive Skill of Artificial Intelligence-based Weather Models for Extreme Events using Uncertainty Quantification

Accurate prediction of extreme weather events remains a major challenge for artificial intelligence-based weather prediction systems. While deterministic models such as FuXi, GraphCast, and SFNO have achieved competitive forecast skill relative to numerical weather prediction, their ability to represent uncertainty and capture extremes is still limited. This study investigates how state-of-the-art deterministic artificial intelligence-based models respond to initial-condition perturbations and evaluates the resulting ensembles in forecasting extremes. Using four perturbation strategies (Gaussian, Perlin noise, Hemispheric Centered Bred Vectors, and Huge Ensembles), we generate 50 member ensembles for the August 2022 Pakistan floods and China heatwave, and complement these case studies with a global threshold-based evaluation. Ensemble skill is assessed against ERA5 and compared with IFS ENS and the AIFS ENS probabilistic model using deterministic and probabilistic metrics. Results show that simpler perturbations like Gaussian and Perlin noise produce similarly realistic ensemble spread and probabilistic skill as flow-based approaches like HCBV and HENS, narrowing but not closing the performance gap with numerical weather prediction ensembles, or native probabilistic models which retain the highest probabilistic skill across variables. Model choice is the dominant factor for ensemble performance, not perturbation method. Across variables, models capture temperature extremes more effectively than precipitation. These findings demonstrate that simple input perturbations can extend deterministic models toward probabilistic forecasting in hardware-constrained settings, supporting artificial intelligence-driven early warning systems.

physics.ao-ph↗

arXiv2025

A biophysical approach to the design of networks of communication systems

Inspired by the growth dynamics of the protist \textit{Physarum polycephalum}, we employ a formalism that describes adaptive, incompressible Hagen-Poiseuille flows on channel networks to identify graphs connecting different nodes within Euclidean space. These graphs are either suboptimal or optimal with respect to their length. Occasionally, we derive graph tree configurations that are topologically equivalent to Steiner trees. This methodology can be utilised to assist in making decisions regarding the design of communication networks, such as fibre webs, motorways, or railway networks. As a demonstration of the practicality of this approach, we explicitly apply this framework to the Portuguese railway network.

physics.soc-ph↗

arXiv2023

Formation and Optimisation of Vein Networks in Physarum

Physarum polycephalum is an acellular slime mould that grows as a highly adaptive network of veins filled with protoplasm. As it forages, Physarum dynamically rearranges its network structure as a response to local stimuli information, optimising the connection between food sources. This high-level behaviour was already exploited to solve numerous optimisation problems. We develop a flow-based model for the adaptive network formation of Physarum, which solves some inconsistencies of previous models. We first derive a general class of equations describing the adaptation and flow dynamics of a static network comprised of elastic channels filled with an incompressible fluid undergoing a Hagen-Poiseuille flow. An explicit form of the model is obtained by minimising the total power dissipated by the network. Considering a more general functional form of the adaptive equations, a phase transition in the system is also found. The model is used for maze-solving and to build efficient and resilient networks in an arena mimicking mainland Portugal. By comparing the resulting networks with the real Portuguese railway system, we found that the model produced networks with a better overall performance when considering fluctuations in the network flows. Finally, the adaption model is extended to incorporate the network growth in the presence of multiple food sources. The coupling of both processes produces networks with similar traits to several network systems found in nature. We found that when the food sources operate alternately, the model can replicate the direct connections between the food sources observed in Physarum.

physics.flu-dyn↗

arXiv2021

Adaptive Hagen-Poiseuille flows on graphs

We derive a class of equations describing low Reynolds number steady flows of incompressible and viscous fluids in networks made of straight channels, with several sources and sinks, and adaptive conductivities. The flow is controlled by the fluxes at sources and sinks. The network is represented by a graph and the adaptive conductivities describe the transverse channel elasticities, mirroring several network structures found in physics and biology. Minimising the dissipated energy per unit time, we have found an explicit form for the adaptation equations and, asymptotically in time, a steady state tree geometry for the graph connecting sources and sinks is reached. A phase transition tuned by an order parameter for the adapted steady sate graph has been found.

physics.flu-dyn↗