Advances of modeling on model requirements (based on assessment in model requirments, Task 5.6, deliverable 5.6.1., WP5 – Environmental Risk Assesment)
Jaromir Jakacki, Institute of Oceanology PAS
Mariya Golenko, Institute of Oceanology RAS
As the main tools in the modeling part of this project it is planned to use two models:
1) Community Earth System Model (CESM) recently developed at National Centre for Atmospheric Research (NCAR) and University Corporation for Atmospheric Research (UCAR). The Baltic Sea version of this model was created and developed at the Institute of Oceanology. The model was implemented for the following configuration:
a) advection is represented by central difference operator,
b) biharmonic horizontal mixing,
c) modified k-profile parameterization for vertical mixing. The modification has been done for vertical viscosity coefficient for turbulence profile similar to Mellor-Yamada which is mostly used for shelf seas like Baltic Sea,
d) the dependence on the bottom cell thickness was added to the bottom drag for better representation of the bottom friction,
e) at the model boundary, in the Kattegat sea level from Goteborg was assimilated – gradient of the difference between real and modeled sea level has been added to the barotropic equation,
f) Orlanski open boundary was also implemented at the model boundary
g) the model has horizontal resolution about ~2.3 km (1/48 degrees) and 66 vertical levels. Model domain, bathymetry and vertical levels are presented on the figure 1.
h) The model is forced by ECMWF ERA40 reanalysis and by data from model UM used as the European operational weather forecasting system.
2) Princeton Ocean Model
POM is a free surface, hydrostatic, sigma coordinate hydrodynamic model with an imbedded second and a half moment turbulence closure sub-model (Mellor, Yamada 1982). The modeling domain comprises of a wide area from the Arcona to the Gotland Basins (fig. 2) with a horizontal resolution of ~ 1.8 km along X and Y directions. The bottom topography is taken from source (Seifert, Kayser 1995). 36 sigma layers are specified. The vertical grid size is logarithmically refined towards the bottom in order to resolve BBL.
On the partly opened lateral boundaries the radiation condition (Blumberg, Mellor 1987, Androsov, Voltzinger 2005) as well as the data of other models comprising the considering area are used. Previously this configuration of POM was adapted for the Central and South-East Baltic and verified with detailed CTD and ADCP field data (Golenko et al. 2009, 2012).
Real atmospheric forcing is prescribed on the surface. Two components of the wind stress, air temperature at 2m height above the sea, air pressure, humidity, precipitation and total cloudiness are interpolated from the HIRLAM grid (10km resolution in both horizontal directions) into the POM grid. All these variables are used to calculate the heat flux onto the surface. The initial temperature (T) and salinity (S) stratifications can be interpolated from other models (for ex. HIROMB: 1nm resolution in horizontal directions and 5m in vertical) as well as set horizontally homogeneous corresponding to mean profiles observed in the considering region in the particular season.
Figure 1. Model domain, bathymetry (color scale is in levels) and vertical resolution plot (inside the picture).
Figure 2. The bottom topography of the Central and South-East Baltic. The presented area is the modeling domain on the base of POM. Red spots denote the points where the lagrangian particles were released from in the BBL. Green spots denote dumpsites of the chemical munitions and supposed to be considered in the project.
As a first step it is planned to have the models at the same state. It means the models will be started from the same initial conditions; the same atmospheric fields will force both of them. Then the results from both models will be compared and it would be focused on the bottom currents. The main goal of this important part of the project is to combine measurements and modeling. Data from moored stations and AUV’s will be compared with both model results and then the models will be tuned up for having the most similar results. The last step is to apply assimilation where it will be possible. The main planned work is presented on the diagram (on the next page).