An Application of Lagrangian Data Assimilation to Katama Bay, MA

Data assimilation is the process of combining predictions from numerical models with observations of the system. Traditionally, Lagrangian data from drifters is assimilated into circulation models by first converting the position data into velocities via, for example, finite differences. These approximate velocities are then assimilated directly into the model. Fully Lagrangian data assimilation seeks to assimilate the drifter trajectories directly. Here, we test whether assimilating trajectories from surface drifters into a model of a small, two-inlet bay can yield an improved estimate of a spatially-dependent model parameter. We focus on the Manning’s n coefficient of friction, a parameter that generally must be tuned by hand, in the narrow, time-varying southern inlet of the bay. Synthetic experiments show that Lagrangian data assimilation can successfully estimate this parameter, regardless of whether the drifters were located in the inlet or elsewhere. Experiments with real data from 2013 show that assimilating Lagrangian trajectories from surface drifters, released for a time period on the order of an hour, can improve upon the original tuned value of this parameter. However, this improvement is dependent on the initial offset between modeled and observed velocities. We judge the performance of the assimilation in the real experiments by comparing velocities measured by current meters at several locations in the bay to those estimated by the model with the assimilated friction parameter.