WAP-1D-VAR v1.0: A One-Dimensional Variational Data Assimilation Model for the West Antarctic Peninsula

WAP-1D-VAR (v1.0) is a variational data assimilation 1-D marine ecosystem model combining the forward model simulation and the backward model simulation for model optimization. The backward model simulation is a tangent linear adjoint version of the forward model and is used in a variational adjoint method that optimizes model parameters to adjust model outputs towards observations.

The variational adjoint method requires four components for data assimilation that WAP-1D-VAR provides within: 1) a forward model simulated by physical forcings and initial (initial conditions and model parameter guesses) and boundary conditions; 2) a cost function to evaluate misfits between the forward model results and the assimilated observations; 3) a tangent linear adjoint version of the forward model to compute the gradient of the cost function with respect to model parameters; and 4) an optimization procedure (M1QN3 3.1) to determine the direction and the optimal step size by which the model parameters should be modified to reduce the cost function based on the cost function gradient from 3). These four components are iterated sequentially to determine a set of the adjusted model parameters until present criteria are satisfied (e.g., low gradients), which then serves as an optimal numerical solution (equations with the optimized model parameters) for the final model outputs.

The model development and validation for version v1.0 of the model for Palmer Station, Antarctica, is presented in: Kim, H. H., Luo, Y.-W., Ducklow, H. W., Schofield, O. M., Steinberg, D. K., Doney, S. C (2021). WAP-1D-VAR v1.0: Development and Evaluation of a One-Dimensional Variational Data Assimilation Model for the Marine Ecosystem Along the West Antarctic Peninsula, Geoscientific Model Development, Under Review. The Kim et al. manuscript describes in more detail the variables and model equations used in the forward model, physical forcing data sets used for Palmer Station, and the resulting optimized model parameters.