 Matthias Morzfeld

Matthias Morzfeld is a researcher in computational and mathematical geophysics whose work combines models with observations. His Scripps Oceanography profile identifies research on data assimilation, inverse problems, Monte Carlo methods and uncertainty, and records a mechanical-engineering doctorate at UC Berkeley in2011. His author-maintained UC San Diego homepage describes a particular interest in how sampling methods behave in high-dimensional spaces.^[ref]

His earlier work with Fai Ma and Ali Imam studied decoupling of viscously damped linear systems in oscillatory free vibration. Their2009 paper constructs a time-varying coordinate transformation by synchronizing phases of damped modes, extending classical modal analysis beyond its usual classical-damping setting.^[ref] This addresses how a damping matrix can keep motion coupled even when undamped normal modes have been identified.

An original Bayesian inference example illustrates the model–observation combination underlying his later research interests. Suppose a scalar state has prior distribution $X\sim \mathcal{N}(0,4)$ and an independent measurement error has a Gaussian distribution $\epsilon \sim \mathcal{N}(0,1)$, with observed $Y=X+\epsilon =3$. The posterior distribution has density proportional toThus $X\mid Y=3\sim \mathcal{N}(2.4,0.8)$. The estimate moves toward the measurement while retaining the information and uncertainty from the prior. The posterior variance is lower than either the prior variance4 or the measurement-error variance1 in this stipulated independent Gaussian model.

These chosen numbers are not a result from Morzfeld's papers. They isolate a simple update mechanism; nonlinear dynamics, sparse observations and large state spaces can make the corresponding computation much harder. His work develops and studies numerical tools for such model–data problems rather than assuming that a small Gaussian calculation suffices for an Earth-science application.^[ref]

The author-maintained homepage provides a primary biographical reference and links a profile image.