Data assimilation in the laboratory using a rotating annulus experiment

Abstract

The thermally driven rotating annulus is a laboratory experiment important for the study of the dynamics of planetary atmospheres under controllable and reproducible conditions. We use the analysis correction method to assimilate laboratory data into an annulus model. We analyze the 2S and 3AV regular flow regimes between rotation rates of 0.75 and 0.875 rad s −1 and the 3SV chaotic flow regime between rotation rates of 2.2 and 3.1 rad s −1 . Our assimilated observations are irregularly distributed, which is more meteorologically realistic than gridded observations as used in recent applications of data assimilation to laboratory measurements. We demonstrate that data assimilation can be used successfully and accurately in this context. We examine a number of specific assimilation scenarios: a wave-number transition between two regimes, information propagation from data-rich to datapoor regions, the response of the assimilation to a strong disturbance to the flow, and a vortex-shedding instability phenomenon at high rotation rate. At the highest rotation rates we calculated the barotropic E-vectors using unobserved variables such as temperature and the vertical structure of the velocity field that are only available via the assimilation. These showed that the mean flow is weakened by the action of eddies, going some way towards explaining why vortices are shed at the very highest rotation rates but not at lower rotation. Rossby-wave stability theory suggests that the underlying instability leading to vortex shedding may be baroclinic in character.Key Words: rotating annulus; analysis correction; data assimilation; E-vector; chaos; baroclinic instability

Regimes of Axisymmetric Flow and Scaling Laws in a Rotating Annulus with Local Convective Forcing

“…The viscosity and thermal diffusivity vary similarly with temperature with coefficients given in Table 1. The fluid used is the same 17% glycerol water/83% water mixture (by volume) used by Young and Read [42].…”

A rotating annulus driven by localized convective forcing: a new atmosphere-like experiment

“…It has also proved valuable as a tractable 'test bed' within which to test numerical codes and methods (Harlander et al 2011;Vincze et al 2015) as well as to benchmark statistical-dynamical analysis methods in widespread use in meteorology, such as data assimilation (Young and Read 2013). Moreover, it is even inspiring new experiments for studying non-terrestrial planetary atmosphere dynamics in the laboratory (Read et al 2015;Yadav et al 2016).…”

Predictability of the thermally driven laboratory rotating annulus

“…In this article we addressed two questions about the rotating annulus laboratory experiment: how predictable is this system, and can common meteorological techniques be used to study its behaviour? Building on work by Young Read (, , ) the annulus simulation MORALS, ensemble generation using breeding vectors, and data assimilation using analysis correction were combined into a full framework for ensemble forecasting in the laboratory annulus. This framework was used to predict the behaviour of regular (2S, 3AV) and chaotic (3SV) rotating annulus flow, verifying forecasts against laboratory data.…”

“…During the development of FANFARE, the ensemble prediction scheme (Figure , dark grey) was first developed and used in the PMS with artificial data generated by MORALS (Young and Read, ). Second, the assimilation stage (Figure , light grey) was developed and used on its own with data from the laboratory annulus (Young and Read, ). The complete framework combines and updates these two components, adding components to produce and verify forecasts against laboratory data (Figure , black).…”

“…Flow velocities are calculated using Particle Imaging Velocimetry over a 1 s interval. The random error in each velocity component was estimated (Young and Read, ) to be 0.0057 cm s −1 , which is much smaller than the velocity variability. The sequence of observations is staggered: the first 5 s take data at the top level, followed by 5 s at each of the other heights in turn, then the cycle repeats.…”