Marine Ecosystem Dynamics Modeling Laboratory

Flexible Biological Module

Various ecosystem models have been implemented in FVCOM, including NPZ, NPZD, NPZDB, and water quality models. To make FVCOM more flexible for various needs of ecosystem studies, we have built a generalized Fortran 95 biological module into FVCOM to allow users to select either a pre-built biological model (such as NPZ, NPZD, etc) or construct their own biological model using the pre-defined pool of biological variables and parameterization functions. The generalized biological module includes seven groups: (1) nutrients, (2) autotrophy, (3) heterotrophy, (4) detritus, (5) DOM, (6) bacteria, and (7) auxiliary. A biological model can be constructed using “function pointers” to select both model structures and parameterization functions. This model can be run simultaneously together with FVCOM with parallelization (so-called “online” mode) or driven separately by FVCOM output (“offline” mode).  Named the Flexible Biological Module (FBM), this new module could be driven by the physical model in FVCOM or by other popular ocean models. This module acts like a platform that allows us to examine the relative importance of different physical and biological processes under well-calibrated physical fields. Validation tests are presently being conducted to use the generalized biological module platform to re-build the state-of-the-art NPZ (Franks and Chen, 1996, 2001) and multi-species NPZD models (Ji, 2003; Ji et al., 2005a-c) for use in GB/GoM applications.

FBM can be run for 1-D (vertical) and 3D cases. For the 1-D case, a special setup was built into FVCOM to allow the model to be run with tidal forcing. This module provides users an easy and fast way to check their own built biological model with pre-defined biological pool in FBM. A brief description of this module is given below.

1. Flow Chart of FBM

The structure of the Flexible Biological Module was developed by dividing lower trophic food web processes into 7 state variable groups: 1) nutrients [N(i), i=1, nn], 2) phytoplankton [P(i) , i=1,np], 3) zooplankton [Z(i), i=1, nz], 4) detritus [D(i), i=1,nd], 5) dissolved organic matter [DOM(i), i=1,nm], 6) bacteria [B(i), i=1, nb], and 7) auxiliary state variables [Y(i), i=1, ny]. The flow chart of the transformation among these variables is shown in Fig. 8.1. We named this system “Flexible Biological Module (FBM)” to emphasize two points. First, this module provides a platform that allows users to build their own parallelized biological model from a discrete set of functions that is independent of the physical model. This module can be run simultaneously with linkage to unstructured-grid (e.g.: FVCOM) and structured-grid ocean models through the connection to the physical model dependent 3-D advection and diffusion variables or it can be run separately by itself in 1-D applications. The second reason we chose the descriptor “Flexible” is that we realized that the range of existing biological models is too vast and complex to try to encompass in a generalized way.

Schematic of Flexible Biological Module (FBM). This is an example of our original module. This module can be easily modified to add more components. A DO module, for example, was recently added into FBM for the study of hypoxia.

In the FBM code, the biological module is an independent 1-D system that is self-maintained and upgraded without linking to a physical model.  It is easy to extend the FBM to a 3-D case by linking to the advection and diffusion modules of any physical model.  It can be also converted to a Lagrangrian-based biological model by linking it with the 3-D Lagrangian particle-tracking module.

The biological module in FVCOM includes point source input from rivers, nudging at lateral boundaries, air-sea interaction at the surface and benthic flux at the bottom. Because these physical processes are already documented in the FVCOM manual and code, we will focus our description of the FBM here on the internal biological processes.

2. Examples of FBM Application to Gulf of Maine

We used FBM to build a NPZ model for the Gulf of Maine with the same structures as Franks and Chen (1996). Numerical experiments were made to examine the influences of model geometrical fitting and turbulence parameterization on the temporal and spatial distributions of simulated phytoplankton in the Gulf of Maine. The assessment of the role of geometrical fitting was made by running a state-of-the-art Nutrient-Phytoplankton-Zooplankton (NPZ) model with physical fields provided from FVCOM (unstructured-grid, finite-volume coastal ocean model) and ECOM-si (structured-grid, finite-difference coastal ocean model), respectively. The impact of turbulence parameterization was studied by running a coupled NPZ-FVCOM system with various vertical turbulence modules implemented in the General Ocean Turbulence Model (GOTM). Comparisons were focused on three large tidal dissipation regions: Georges Bank (characterized by strong tidal rectification over steep bottom topography and tidal mixing fronts), Bay of Fundy (featuring large semidiurnal tidal oscillations due to the gulf-scale resonance) and Nantucket Shoals (a tidal energy flux convergence zone). For the same given tidal forcing and initial physical and biological conditions, the ability of a model to accommodate irregular coastal geometry and steep bottom topography is critical to determine the robustness of the simulated spatial and temporal structure of N and P. For the same given external forcing in FVCOM, turbulence parameterizations have less impact on N and P in mixed regions than in stratified regions. In mixed regions, both qe and qql models reproduced the observed vertical mixing intensity. Since biological variables remained vertically mixed in these regions, their structures were little affected by turbulence closure schemes. In stratified regions, q-e models predicted stronger mixing than qql models, which produced more nutrient fluxes over the slope and thus influenced the growth and distribution of P around the tidal mixing front. A direct comparison between observed and model-predicted turbulence dissipation rates suggested that q-e models with a mixing cutoff at Richardson number of 1.0 predicted more realistic mixing intensity than qql models in stratified regions on Georges Bank. A brief of description of the model results are shown here. For details, please go to our recent manuscript that is in press on Deep Sea Research II-GLOBEC/GB special issue:

Tian, R. and C. Chen, 2006. Influence of model geometrical fitting and turbulence parameterization on phytoplankton simulation in the Gulf of Maine. Deep Sea Research II, in press.

Why does FVCOM predict different results from POM/ECOM-si in the Gulf of Maine, particularly on Georges Bank, Bay of Fundy and Nantucket Shoal?

The reasons are:

FVCOM predicts more realistic residual flow on Georges Bank.

On Georges Bank, the distribution of the residual current predicted by FVCOM and ECOM-si differed significantly. On the northern flank, for example, FVCOM predicted a residual current that followed mainly the local isobath, whereas ECOM-si showed a stronger on-bank flow across isobaths. As pointed out by Chen et al. (2006b), the orthogonal curvilinear grid used in ECOM-si was not built along the local isobath on GB and, as a result, the 3D slope on the northern flank was treated as step-like bottom topography, which can cause a significant on-bank residual current. As the horizontal resolution increased, the on-bank residual flow predicted by ECOM-si was significantly reduced, and the ECOM-si numerical simulation tended to converge to the FVCOM solution. This finding is also applicable to an idealized stratified case shown in the present study. When the horizontal resolution doubled, the ECOM-si-predicted residual current on the northern flank of GB turned significantly in the along-isobath direction.

Click image on the right to view the full size images published on Tian&Chen’s paper.

FVCOM provides a better resolving of residual eddy circulation in the Bay of Fundy.

Chen et al. (2006d) examined the impact of Grand Manan Island on the formation of anticyclonic and cyclonic eddies around the island and a cyclonic eddy in the inner bay of the BF. They suggested that in order to resolve realistically the subtidal circulation in the BF, a model must be capable of resolving the complex coastline of Grand Manan Island and the strong tidal currents in the inner bay.

With adequate matching of the coastline of the BF and Grand Manan Island, FVCOM predicted a pair of anticylonic and cyclonic eddies in the southern area of Grand Manan Island. A relatively strong cyclonic residual eddy along the 60-m isobath was also resolved. The strong tidal currents predicted by FVCOM resulted in energetic tidal mixing in the inner bay, forming a cold-core area between the 60- and the 100-m isobaths, given the linear distribution of temperature in the initial conditions.

In contrast, the low-resolution grid of ECOM-si could not capture the complex geometry of Grand Manan Island and the coast of the BF. Consequently, ECOM-si resolved only one anticyclonic eddy in the southern area of Grand Manan Island. No cyclonic residual eddy was generated along the 60-m isobath. Because this model significantly underestimated the amplitude of tidal currents and vertical mixing, surface waters remained warmer than that in the FVCOM simulation. With horizontal resolution doubled, the cyclonic residual eddy appeared along the 60-m isobath. With improved simulation of tidal amplitudes and currents, the “high-resolution” ECOM-si produced much stronger vertical mixing, which caused the surface water temperature to drop significantly, converging toward the FVCOM results.

Click image on the right to view the full size of images published on Tian&Chen’s paper.

FVCOM showed more complex residual current patterns on Nantucket Shoals

In terms of tidal dynamics, Nantucket Shoals (NS) is a transition region between the southern New England shelf to the southwest and the resonant Gulf of Maine to the northeast (Chen et al. 2006a). The tidal-induced residual currents and mixing in that region are controlled by complex physical processes of tidal wave interaction and tidal flushing around Nantucket Island. With detailed representation of the coastal geometry of the island, FVCOM predicted multiple residual eddies around Nantucket Island and a relatively strong clockwise circulation on NS. As a result, a warm core patch formed inside the clockwise circulation area south of Nantucket Island. As the coastal geometry of Nantucket Island was not adequately represented by the low-resolution grid, ECOM-si did not resolve the tidal-flushing-induced eddies around the island. Although this model did show a clockwise residual circulation on NS similar to FVCOM, the velocity was significantly weaker. The warm water predicted by ECOM-si covered a broader area from Nantucket Sound to NS. By increasing the horizontal resolution, the simulation result of ECOM-si was improved, particularly in the NS area offshore from Nantucket Island. A distinct clockwise residual eddy appeared on NS, which tended to form a warm-core patch like that in the FVCOM solution. Since even the higher resolution structured grid resolved poorly the complex coastline of Nantucket Island, ECOM-si did not capture the tidal-flushing-induced residual eddies around the island.

Click image on the right to view the full size of images published on Tian&Chen’s paper.

Selected NPZ comparison results of FVCOM and ECOM-si
Georges Bank Cross-bank section Bay of Fundy

Which turbluence closure schedule is suitable for the biological study in the Gulf of Maine?

The GOTM is a community turbulence module developed by Burchard et al. (1999) and continuously upgraded by a team effort led by Burchard ( This module contains two types of turbulence models: (1) qql equations and (2) qe equations. Both turbulence model groups include the original code with a Richardson number cut off at 0.2 and a modified version with a Richardson number cut off between 0.2 and 1.0. Modules tested here are:

BB: Burchard-Baummert (1995); CK: Kantha-Clayson (1994); CA: Canuto et al. (2001); CB: Canuto et al. (2001) and MY: Mellor-Yamada (1982)

For stratified water, CA (with a Richardson number 1.0) is better!

For vertically well-mixed water, all turbulence modules works well.

Click image on the right to view the full size of the comparison figure on both northern and southern flanks.

Vertical Eddy Viscosity




Posted on January 17, 2014