Research

Six threads run through the group's work. Each is a short read.

Physical models corrected on the fly by streaming sensor data, so that a simulation and the system it mirrors stay synchronised.

Ensemble Kalman filtering for nonlinear and chaotic systems — including estimating the model error the model does not know it has.

Autoencoders, reservoir computers and transformers, built to be forecast-stable and interpretable rather than merely accurate.

Once a state can be estimated from partial observations, it can be acted upon — assimilation and reinforcement learning, combined.

Thermoacoustics, wakes, wind farms, near-wall turbulence and urban dispersion — where the methods meet real data.

Characterizing a system before modelling it: delay embeddings, Lyapunov exponents and regime classification, from the equations or from data alone.