A Coupled Epipelagic-Meso/Bathypelagic Particle Flux Model for the Bermuda Atlantic Time-series Station (BATS)/Oceanic Flux Program (OFP) Site

David M. Glover and Maureen H. Conte
Dept. of Marine Chemistry and Geochemistry
Woods Hole Oceanographic Institution
Woods Hole, MA 02543


The overarching goal of this project is to mechanistically connect euphotic zone processes with meso- and bathypelagic zone processes. It is our long term goal to accomplish this by means of a prognostic model that can be used to further our understanding of unparalleled time-series of deep-water sediment traps (21+ years) at the Oceanic Flux Program (OFP), euphotic zone measurements (10+ years) at the Bermuda Atlantic Time-series Site (BATS). In order to realize this goal we will derive a meso/bathypelagic ecosystem structure and use it to model the flux of biogeochemically active constituents (carbon, nitrogen and silica) through the water column. In this follow-up to the initial ecosystem kernel build, we present the mesopelagic ecosystem in a pseudo one-dimensional, nitrogen only framework. The equations and initial parameters are presented along with insight as to how this kernel fits into the over all scheme.


Of considerable scientific interest is the role remineralization plays in the global carbon cycle. It is the ``biological pump'' that fixes carbon in the upper water column and exports it for long time periods to the deep ocean. Yet the biological pump can only respond to changing climate indirectly (Denman et al., 1996) via modification of remineralization processes. From a global carbon cycle point-of-view, it is the processes that govern remineralization in the mid- to deep-ocean waters that provide the feedback to the biogeochemical carbon cycle. And yet, our understanding of these processes is very limited. This modeling study will allow mechanistic, prognostic experimentation of these remineralization processes through formulation of the ecosystem structure. In this manner we will be able to elucidate the importance of (e.g. bacterial mediated decomposition vs. zooplankton repackaging) processes, for example, on the over all remineralization of organic matter at this site and, by extension, the globe.

A number of deep-sea ecosystem models have been proposed in recent years (Dadou et al., 2001; Boehm and Grant, 2001; Armstrong et al., 2002; Jackson and Burd, 2002) in an one-dimensional context. Armstrong et al. (2002) challenge the Martin et al. (1987) concept that to a large extent remineralization is a function of depth only. Dadou et al. (2001) found that remineralization of DON and zooplankton excretion were the dominate processes in remineralization in the Northeastern Tropical Atlantic. Boehm and Grant (2001) found steady state solutions that do not require bacterially mediated remineralization to explain the exponential decay of POC flux in the mesopelagic zone. Jackson and Burd (2002) also produced steady state solutions that predict an exponential decay of organic matter flux with depth, but perturbations of these steady state solutions can create oscillations in the populations and could explain the observed large swings in deep flux. Here we create an ecosystem model that tries to explicitly model all processes believed to be important in the regulation of organic matter flux during the long fall from the euphotic to the sediment at the bottom of the sea.

The Model

Figure 1

The model under construction is actually a combination of three models. An epipelagic zone ecosystem model (Moore et al., 2002) will be coupled with an advanced physical model of the upper ocean mixing regime (Large et al., 1994) forced by buoyancy flux, wind stress, and surface irradiance. This combination will, in turn, be used to drive the meso/bathypelagic portion of the model. The overall linkage between the epipelagic and mesopelagic models is provided by the ecosystem model (Moore et al., 2002) schematically shown in Fig. 1. Through the flux of large detritus on a one-dimensional (to start with) grid productivity in the euphotic zone will be linked to our ecosystem model. Figure 2 shows a comparison of the large (sinking) detritus from the epipelagic model to the organic nitrogen flux intercepted by the PIT traps deployed at BATS, the agreement is reasonably good. Figure 2

Figure 3 shows a close up of the meso- bathypelagic portion of the model and it is this kernel that is the main topic of this poster. This model consists of an active feeding habit zooplankton, a passive feeding habit zooplankton, large detritus (that sinks), small detritus (non-sinking), and a nutrient pool. The active feeder (A), passive feeder (P), large detritus (Dl), small detritus (Ds), and nutrients (at this time N-based only) are modeled using equations 1-5 shown below. The parameters used in this simulation are given in Table 1 along with units, values, and reference sources. As the detritus, the primary source of food, moves through the water column it will be fed upon by the active/passive pair and will also undergo bacterially mediated remineralization into nutrients. The large detrital pool at depth will gain material from the formation of fecal pellets from the passive and active feeding zooplankton. Sloppy feeding habits of the active feeder will contribute to the small detrital pool. Zooplankton mortality (both classes) will also contribute directly to the large detrital pool. Aggregation and disaggregation (Boehm and Grant, 1998; Jackson and Burd, 2002) transform detrital particles from one pool to the other and back again. The model will yield a particulate flux between levels, at the trap levels this flux can be directly compared to the data collected at the OFP traps. The nutrients at each depth will gain from detrital remineralization and zooplankton excretion. Figure 3

Figure 4 shows, in a schematic manner, the general top-level lay out of the model and the manner in which an one-dimensional framework is approximated. The model integration starts with the large detritus output from the epipelagic model of Moore et al. (2002) stored in an off-line manner. This flux of large detritus can be modified to move the starting depth of the model deeper into the water column. In these simulations this multiplicative factor has been set to one, essentially the base of the euphotic zone. The large, sinking, detritus is fed as an input to prtflx_v0p7_L1, the code containing the differential equations below. An estimate of the rates of change of the state variables is then passed to the integrator (Runge-Kutta 45) and a time step is taken. The process repeats itself, integrating forward in time, until a previously set stop time (usually 10 years) is reached. At each time step the value of each state variable is piped off to a display so that the trace of the evolution of the model can be watched as the model progresses.

Figure 4

One state variable is additionally piped off (the large detritus) and passed to a transport delay module. This module simulates a delay that corresponds to the amount of time it would take the particle to sink to the next level at a settling velocity of 20 m/d. After passing through the transport delay the large detritus is passed to a carbon-copy of the previously described model with the important difference that it takes as input the sinking large detritus from the layer above. In this manner, transport through the water column is approximated for the upper 600 m. Of course, this is only a crude approximation as the processes being modeled are continuous throughout the water column and the transport delay merely delays the arrival at the next layer of the model. A more complete and continuous model is being constructed that will model these processes over the entire 4000 m water column near the OFP/BATS sites.





Results and Discussion

In Fig. 5 we show an one-year trace of the five state variables from the mesopelagic ecosystem model just below the euphotic zone (nominally 200 m) and in Fig. 6 we show similar results from from a nominal depth of 500 m. These two figures are year ten of a ten year model run, repeatedly forced with one year's worth of large detritus flux from Moore et al's (2002) model. While this is only a one-dimensional simulation, some of the features seen agree qualitatively with data published in Deevey and Brooks (1971) and Conte et al. (2001) and some do not. In particular, we see the same sort of timing in the populations of passive and active with the active feeders preceding, slightly, the passive feeders. The large detrital pool follows a similar seasonal cycle as do the OFP sediment traps, including the double peak. However, the near instantaneous flux from the surface to depth (see Conte et al. poster OS52B-0221) is not captured by this model. Figures 7 and 8 show a full 10 year simulation and the diminishment of the N flux in the various pools in the model also follows the OFP trap data.

Figure 5 200m Figure 6 500m

Nevertheless, in order to address some of the more interesting issues brought to light by the analysis of the OFP time-series (increased homogenization of particle composition and the increasing C/N ratios with depth) the model will have to address the stoichiometry of the consumer---food relationship. We will extend the approach of Geider et al. (1998) to the meso/bathypelagic zone by assuming that zooplankton also have cell quotas for nitrogen, carbon, and phosphorus. In this manner, the composition of the falling detrital material will change via the repackaging of the two zooplankton classes, the scavengers directly and the predators indirectly.

Figure 7 200m Figure 8 500m

Additionally there are other issues that need to be addressed and these issues point the way for the future work with this model. The following features are planned to be added:

 zooplankton that migrate vertically (Steinberg et al., 2000);
 ecosystem still needs explicit nitrification;
 sloppy predation COULD be split between Dl and Ds;
 bacteria are not explicitly modeled, will be added as a state variable;
 only two size classes for detritus;
 particle sinking velocity is constant with depth, needs to be variable.


© 2002, 2003 -- David M. Glover, WHOI --