Research

Many of the dynamical systems that matter most in science and engineering are practically intractable. They may be high-dimensional, governed by complex multiscale physics, driven by stochastic processes, or accessible only through limited and noisy observations, and often several of these challenges arise at once. Direct simulation becomes prohibitively expensive, classical control design breaks down on systems we cannot fully observe, and the data we do collect is rarely sufficient to fit a fully expressive model from scratch.

For these reasons, I am motivated to develop reduced-order models and data-driven inference methods that capture the essential behavior of such systems at a fraction of the computational cost, while remaining physically interpretable and amenable to downstream control design. I see this work as building a bridge between classical mathematical structure, modern numerical and computational methods, and machine learning, so that researchers across disciplines can apply these tools to systems that are today out of reach of any one approach alone.

Publications

  1. Stevenson, S., J. Perez Cuarenta, M. Parker, S. Lilledahl, S. S. P. Shen (2025). Estimation of spatial sampling errors from non-independent meteorological sensors. Theoretical and Applied Climatology Submitted
  2. Lizerbram, A., S. Stevenson, I. Khadir, M. Tu, S. S. P. Shen (2025). Robustness test for AI forecasting of Hurricane Florence using FourCastNetv2 and random perturbations of the initial condition. Artificial Intelligence for Earth Sciences Accepted
  3. Khadir, I., S. Stevenson, H. Li, K. Krick, A. Burrows, D. Hall, S. Posey, S. S. P. Shen (2025). Democracy of AI numerical weather models: an example of global forecasting with FourCastNetv2 made by a university research lab using GPU. Artificial Intelligence for the Earth Systems arXiv:2504.17028 Submitted