Physical scientists

observes, records and collates data on atmospheric conditions from weather stations, satellites, and observation vessels to plot and forecast weather conditions

AI-driven numerical weather prediction systems like Google DeepMind's GraphCast and NVIDIA's FourCastNet now match or exceed traditional models in forecast accuracy, significantly automating data collation, pattern recognition, and short-range forecasting.

applies mathematical models and techniques to assist in the solution of scientific problems in industry and commerce and seeks out new applications of mathematical analysis

AI tools such as Wolfram Alpha, MATLAB AI extensions, and symbolic computation platforms can rapidly solve and optimise mathematical models, though identifying the right model formulation and translating results into applied contexts still requires expert judgement.

conducts experiments and tests and uses mathematical models and theories to investigate the structure and properties of matter, transformations and propagations of energy, the behaviour of particles and their interaction with various forms of energy

AI tools like DeepMind's AlphaFold and physics-informed neural networks can accelerate hypothesis generation and pattern recognition in experimental data, but human scientists still design experiments, interpret anomalies, and validate theoretical frameworks.

uses surveys, seismology and other methods to determine the earth’s mantle, crust, rock structure and type, and to analyse and predict the occurrence of seismological activity

Machine learning models (e.g., those used by USGS and seismic monitoring networks) can detect and classify seismic signals faster than humans, but geophysicists are still needed to interpret complex subsurface structures and contextualise findings.