Search

Unstructured mesh models can resolve the model domain with a variable and very fine mesh resolution. Nevertheless, tuning the model setup is still required (for example because of parametrized sub-grid processes). Adjoint models are commonly used to calculate sensitivities of ocean models and optimize their parameters so that better agreement is achieved between model simulations and observations. One major obstacle in developing an adjoint model is the need to update the reverse code after each modification of the forward code, which is not always straightforward. Automatic differentiation is a tool to generate the adjoint model code without user input. So far this method has mainly been used for structured mesh ocean models. We present here an unstructured mesh, adjoint, tidal model using this technique, and discuss the sensitivities of the misfit between simulated and observed elevations with respect to open boundary values, the bottom friction coefficient and the bottom topography. The forward model simulates tides on the European Continental Shelf and we show that the tidal model dynamics in the adjoint simulations can be used to define regions, where parameters or mesh has to be optimized. We analyze the dependence of the sensitivities on the wave type and mesh resolution to specify whether the model misfit originates from physical or numerical model deficiencies. In the sensitivity patterns, it is possible to identify islands not resolved in the mesh. We propose to refine the mesh prior to the parameter optimization. [ABSTRACT FROM AUTHOR]

A new multi-trajectory highly vectorizable Gear chemical-transport model is discussed and tested against measurements from 25 sites in Europe for a two-week summer period. The simulations are compared with measurement data and with results of a Quasi-Steady-State Approximation (QSSA) based Lagrangian model using the EMEP mechanism.The model is developed from the Lagrangian EMEP model. The Regional Atmospheric Chemistry Mechanism (RACM) is used as the model’s chemical mechanism. To solve the stiff chemical rate equations, a sparse-matrix highly vectorizable Gear algorithm is used. The meteorological data used to run the model are obtained from the HIgh Resolution Limited Area Model (HIRLAM).The vectorized Gear based model improves the accuracy of the numerical solution for the chemistry compared with the QSSA based model. The differences between the two models are explained by alternative approaches of the chemical modules employed in the model: a Gear algorithm versus a QSSA solver, photolysis rate parameters in RACM are calculated from a radiation transfer code versus parameterized photolysis rate parameters used in the Lagrangian EMEP model, the RACM versus the EMEP atmospheric chemical mechanisms and two methods of aggregating the emissions of Volatile Organic Compounds (VOC). Photolysis rate parameters calculated from radiation codes (which are more accurate than simple parameterizations) and the Gear algorithm (which is a benchmark solver compared with the QSSA solver) are recommended for atmospheric chemistry modeling because of the high sensitivity of ozone concentrations to the chemical reaction scheme and to the photolysis rates (Stockwell and Goliff, 2004). [ABSTRACT FROM AUTHOR]