An Adjoint Data Assimilation Approach to the Estimation of Pseudocalanus Population Dynamics

Dennis McGillicuddy, Daniel Lynch, Peter Wiebe, Wendy Gentleman and Cabell Davis

Our underlying scientific objective here is to determine the mechanisms that control variations in the seasonal abundance patterns of Pseudocalanus. We hypothesize that the observed distributions result from the interaction of the population dynamics with the physical circulation. This is expressed in the two dimensional advection-diffusion-reaction equation for a scalar variable C which represents the concentration of Pseudocalanus. Given initial conditions C0 at time t0, we seek the population dynamics source term R(x,y) such that integration of the forward model equation will result in predictions which minimize the sum of squares of differences between modeled and observed concentrations. An adjoint data assimilation technique has been designed for these purposes.

This approach has been used to invert for the population dynamics between the January-February and March-April climatological distributions of Pseudocalanus derived from the MARMAP data. Climatological flow fields for this time period are specified. The solution converges rapidly in the first ten iterations, reaching a minimum in the cost function by approximately the 25th iteration. The general characteristics of the source term include strong growth on the crest of the bank, and high mortality in two areas: (1) a strip just off the northwest flank of the bank extending from the northern Great South Channel almost to Georges Basin, and (2) Wilkinson Basin.