Milan Kloewer (he|him)

Abstract:

Objective: To analyse associations between feel-like temperatures measured with the universal thermal climate index (UTCI) and injury rates during international athletic championships.

Methods: During 10 international outdoor athletic championships from 2007 to 2022, in-competition injuries were collected by medical teams and local organising committees. UTCI was extracted hourly from a global reanalysis of observed atmospheric conditions during each championship. We performed Poisson regressions with incidence rates (number of injuries per 1000 athlete starts) as outcomes and UTCI as a predictive variable adjusted for sex, for all and time-loss injuries, for different injured tissue types (ie, muscle, tendon, ligament, articular, bone and skin) and specific discipline (ie, sprints, hurdles, jumps, throws, middle distance, long distance, marathon and race walking).

Results: A total of 1203 in-competition injuries were reported for 29 579 athlete starts. For all in-competition injuries (ie, all injured tissue types and all disciplines), higher UTCI was associated with lower incidence rates for time-loss injuries (IRR=0.98, 95% CI 0.97 to 0.98) but not for all injuries (IRR=1.00, 95% CI 1.00 to 1.01). Based on injured tissue type with all disciplines included, higher UTCI was associated with lower incidence rates for all (IRR=0.97, 95% CI 0.97 to 0.98) and time-loss (IRR=0.96, 95% CI 0.96 to 0.96) muscle injuries. Based on the specific discipline, higher UTCI was associated with lower incidence rates for all and time-loss muscle injuries for sprints (IRR=0.95, 95% CI 0.95 to 0.96, and IRR=0.94, 95% CI 0.93 to 0.94, respectively), hurdles (IRR=0.97, 95% CI 0.96 to 97, and IRR=0.95, 95% CI 0.94 to 0.96, respectively) and throws (IRR=0.97, 95% CI 0.97 to 0.98).

Conclusions: Higher feel-like temperatures were associated with a decreased risk of time-loss and muscle injuries, particularly in sprints, hurdles and throws. Although the precise mechanism for lower injury rates with higher feel-like temperatures requires further investigation, adapting preparations such as warm-up or clothing to forecasted weather conditions may be of benefit.

Abstract:

General circulation models (GCMs) are the foundation of weather and climate prediction^1,2. GCMs are physics-based simulators that combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine-learning models trained on reanalysis data have achieved comparable or better skill than GCMs for deterministic weather forecasting^3,4. However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present a GCM that combines a differentiable solver for atmospheric dynamics with machine-learning components and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best machine-learning and physics-based methods. NeuralGCM is competitive with machine-learning models for one- to ten-day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for one- to fifteen-day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics for multiple decades, and climate forecasts with 140-kilometre resolution show emergent phenomena such as realistic frequency and trajectories of tropical cyclones. For both weather and climate, our approach offers orders of magnitude computational savings over conventional GCMs, although our model does not extrapolate to substantially different future climates. Our results show that end-to-end deep learning is compatible with tasks performed by conventional GCMs and can enhance the large-scale physical simulations that are essential for understanding and predicting the Earth system.

Abstract:

Neural operators (NOs) have emerged as effective tools for modeling complex physical systems in scientific machine learning. In NOs, a central characteristic is to learn the governing physical laws directly from data. In contrast to other machine learning applications, partial knowledge is often known a priori about the physical system at hand whereby quantities such as mass, energy and momentum are exactly conserved. Currently, NOs have to learn these conservation laws from data and can only approximately satisfy them due to finite training data and random noise. In this work, we introduce conservation law-encoded neural operators (clawNOs), a suite of NOs that endow inference with automatic satisfaction of such conservation laws. ClawNOs are built with a divergence-free prediction of the solution field, with which the continuity equation is automatically guaranteed. As a consequence, clawNOs are compliant with the most fundamental and ubiquitous conservation laws essential for correct physical consistency. As demonstrations, we consider a wide variety of scientific applications ranging from constitutive modeling of material deformation, incompressible fluid dynamics, to atmospheric simulation. ClawNOs significantly outperform the state-of-the-art NOs in learning efficacy, especially in small-data regimes. Our code and data accompanying this paper are available at https: //github.com/ningliu-iga/clawNO.