Modelling Tool

Science and Engineering at The University of Edinburgh

School of GeoSciences

Personal Home Pages

Section Contents

The Interface Modelling Tool (now available to all CASIX members)

A 1-d bio-geochemical ocean turbulence model for air-sea carbon flux is now available to members of CASIX. Using this model you can simulate the ocean carbon cycle and resulting CO2 fluxes at any time and space resolution. The model is ideal for studying short time scale processes at the air-sea interface. For example, diurnal variations in carbon flux governed by diurnal variation of light and turbulent mixing through the water column. Details of the model are given below.

Model Physics

The model physics are calculated using a general ocean turbulence model (GOTM). This deals with the air-sea interactions, boundary layer turbulence, internal wave breaking and shear instability. Forcing from the atmosphere eg surface wind stress, net heat flux, solar radiation penetration, sea surface elevation, surface slope are calculated within GOTM using observation data and bulk formulas. The equations for calculating turbulence contain more unknowns than equations and so they are generally solved using turbulence closure schemes which are simply parameterisations for the unknown variables. GOTM includes many such turbulent closure schemes for example the k-epsilon (Rodi 1987), k-omega (Umlaf et al. 2003), Kantha and Clayson (1994) and Mellor-Yamada (Mello & Yamada 1982) models which can be easily selected.

Model Bio-chemistry

The bio-chemistry routines which are fully coupled into the GOTM model are those developed by the UK Met Office for the HadOCC (Hadley Centre Ocean Carbon Cycle) model. The ocean biology is represented by 4 variables - nutrient, phytoplankton, zooplankton and detritus (known as a NPZD model), while the carbon chemistry is represented by dissolved inorganic carbon (DIC) and alkalinity. These 6 variables interact with each other and are also moved through the water column according to currents and turbulent dispersion calculated within the physical model. Additional routines on light attenuation have been added to the modelling tool so that it is possible to predict photosynthesis using a daily averaged and spectrally averaged light attenuation model (Anderson, 1993), an instantaneous spectrally averaged model (Anderson, 1993) or an instantaneous model using 61 wavebands of light (Morel 1988; 2001). To calculate the CO2 flux at the interface there are currently 3 different gas transfer velocity parameterisations to choose from (those of: Liss and Merlivat, 1986; Wanninkhof, 1992; Wanninkhof and McGillis, 1999).

Model Validation

Validating a 1-d (vertical) model is not easy because the validation data comes from a 3-d world where the governing processes act horizontally as well as vertically. Thus in regions where there is strong horizontal advection our 1-d model can not hope to reproduce the patterns in biological activity. The ocean turbulence model itself has been used by a large number people and there are many published papers verifying its ouptut so further validation of GOTM's modelled sea water temperatures was not necessary. However, substantial changes had to be made to the HadOCC code to allow coupling to GOTM and to allow calculations on any time and space step. To verify these were made correctly the model was rigorously tested against the 1-d testbed model for data assimilation developed for CASIX by John Hemmings. By doing this the bio-chemistry and gas transfer subroutines were thoroughly verified but it was not possible to test the turbulent mixing of these variables by GOTM. To do this, data from a study in the North Sea by Helmut Thomas et al (2004) were used. More info on North Sea validation.

Using the Interface Modelling Tool

The model is written in Fortran 90 with a makefile which requires a recent version of gnumake (free software) and a f90 or f95 compiler and netcdf libraries. The model only needs to be compiled once after which each model run can be specified using a comprehensive set of options contained in the (uncompiled) input files. The time taken for a model run will depend upon your processor power, the number of layers in your vertical grid and the length of your chosen time step. Very long time steps are not possible because of numerical instabilities - about 30 minutes is the max. However, the model is generally quick to run - it takes about 10-30 minutes to run a simulation lasting 1 year.

Inputs and Outputs It is possible to run the model without any input data - for example for controlled theoretical runs. In this case the initial conditions and the driving meteorological fluxes are specified by constant values or calculated according to latitude, longitude and time of day in the case of clear-sky solar radiation. If real data are available these can easily be read into the model (once correctly formatted) with no restrictions on the time step of the data - the model linearly interpolates the data to find the forcing values at each model time step. Typical forcing data are wind speed (U and V) at 10m, air pressure, air temperature, humidity and solar radiation or cloud cover. All model variables are available for output - for example CO2 flux, chloropyll, temperature profiles, oceanic pCO2, mixed layer depth etc. These are produced in a NetCDF file format which can be read by Matlab or IDL for example.

Data Assimilation If sea water temperature or salinity profiles are available then these may be assimilated into the model but this must be done carefully to avoid instabilties. Since the model is only one dimensional there are no influxes of nutrient due to advection - this means that nutrient must be assimilated otherwise it is rapidly used up. The assimilation is done by slowly relaxing towards Levitus' climatology below the mixed layer depth.

OBTAINING THE INTERFACE MODELLING TOOL (CASIX members only)

Email me at H.Kettle@ed.ac.uk for the latest version of the model and compilation instructions. I will also send you example input and data files which you can use to get going with the model, along with some matlab code for plotting the results if required.

References