Extended Abstract

We present a new coupled, one dimensional biological/physical model applied to the subtropical region near Bermuda. The physical component of the model, which is driven by smooth climatological forcing, successfully reproduces the long-term seasonal cycles of upper ocean temperature, salinity and boundary layer depth from Hydrostation S. The nitrogen based biological model, which includes the effects of photoadaptation, phytoplankton aggregation, and particle remineralization in the aphotic zone, shows significant skill in capturing the major features of the annual chlorophyll field (e.g. spring bloom, deep chlorophyll maximum) and depth integrated chlorophyll and primary production as exhibited by the U.S. JGOFS Bermuda Atlantic Time Series (BATS) data. The introduction of variable phytoplankton chlorophyll to nitrogen ratios is found to be important for simulating the subsurface chlorophyll maximum, and the model solutions show a realistic deep nitracline in the summer and a low annual average f-ratio of 0.21 compared to previous modeling work. The performance of the model solutions are weakest during the late summer, when the model can not supply enough nutrients to support the high production observed in the stratified near surface waters. The coupled model has large winter production values, leading to a substantial export of organic material from the euphotic zone via downward turbulent mixing. The model predicts a total export production from the euphotic zone of 0.24 mol N m y, approximately equally partitioned between particle sinking and suspended matter detrainment. The bulk of the export production is remineralized in the shallow aphotic zone, and only a small fraction is transported below the depth of the maximum winter mixed layer and thus contributes to ``biological pump''.

The biological production and export of organic matter from the surface ocean is a key component of ocean biogeochemical cycling and the global carbon system (e.g. Broecker and Peng, 1982; Najjar, 1992). Physical processes play an important role in marine ecosystem dynamics (Mann and Lazier, 1991) and can modify or limit biological production through the nutrient supply and mean irradiance field (e.g. McClain et al., 1990; Mitchell, et al., 1991). The most relevant physical phenomena for understanding the patterns and timing of biological productivity, therefore, are often the seasonal variation in the planetary boundary layer (PBL) depth (e.g. Evans and Parslow, 1985) and the turbulent fluxes through the seasonal thermocline. A new one-dimensional, coupled biological--physical model has been designed at NCAR as part of a longer term global biogeochemical modeling project. The effort is guided by two criteria: the essential physical processes of the ocean PBL must be simulated correctly as a prerequisite if one wishes to accurately model the behavior of an upper ocean ecosystem, and the coupled model should capture the essential characteristics of the marine ecosystem while remaining simple enough, both conceptually and computationally, to allow for incorporation into global-scale ocean circulation models. The overall goal of this paper is to introduce the numerical model and verify its skill for the oligotrophic ocean using the U.S. JGOFS Bermuda Atlantic Time Series (BATS) data.

The initial focus on the oligotrophic Sargasso Sea region is governed by the great extent of oligotrophic conditions in the global ocean, the strong, physically driven seasonal signals at Bermuda, and the availability of the BATS record and other historical data for model validation (WHOI and BBSR, 1988; Knap et al., 1991; 1992; 1993; 1994; Lohrenz et al., 1992; Michaels et al., 1994). Biological simulations can be judged by many different measures depending on the questions one wants to ask. The constraints of a one-dimensional approach, the BATS data set, and our long-term biogeochemical goals impose a particular emphasis on our modeling efforts, stressing the vertical profiles of nutrients and chlorophyll and the export production of fixed organic matter. Additional parameters such as primary production from the model are also examined because of their utility in validating coupled ocean biological models. Using the new coupled model, we explore the interaction of upper ocean physics and biology at Bermuda on seasonal timescales, addressing in particular the factors affecting chlorophyll and nutrient depth structure, vertical nutrient and particle fluxes, and aphotic zone remineralization. The model also serves as a framework for interpreting the JGOFS Bermuda time series.

In our paper an overview of the current understanding of biogeochemical cycling at Bermuda is presented. The formulation of the coupled numerical model is discussed, and the physical model solutions, which are driven by seasonal climatological forcing are described. We present the coupled biological solutions for Bermuda and judge their general validity, where possible, against the JGOFS BATS data. The dynamics of the model nitrogen cycle and general relevance are then discussed. Future modeling work will include more realistic, transient physical forcing and/or additional biogeochemical interactions.

The coupled, 1-D biological--physical model forced with smooth, climatological surface fluxes replicates the basic features of the seasonal cycle for Bermuda including the mixed layer depth, seasonal thermocline, spring bloom and deep chlorophyll maximum. These successes occur despite the much more realistic, low nutrient levels throughout the seasonal thermocline and corresponding low f-ratios (0.2) compared to previous modeling efforts for Bermuda (e.g. Fasham et al., 1990). The model is less able to simulate the seasonal variation in primary production, particularly the high surface values during the summer, and has too much zooplankton biomass with a larger annual cycle than the observations. As mentioned by Fasham et al. (1990) the comparison of model zooplankton values with observations can be problematic, and only a limited data set exists for comparison at Bermuda. In contrast, the high summer primary production values appear to be a robust feature of the observations, and the mismatch between the model and data requires some consideration.

The low surface summer production in the model is caused by nutrient limitation. Following the spring stratification, particle export continues to lower the total nitrogen levels of the surface, which can not be made up by an upward turbulent flux because of the deep nitracline. In fact, the high surface production in the BATS observations occurs despite nutrient concentrations being below detection limits to a depth of about 80 m during the summer (Michaels et al., 1994). One possibility is that the community structure at Bermuda shifts during the stratified period to a very tightly coupled ecosystem with little or no nitrogen export from the near surface layer (e.g. Jenkins and Goldman, 1985; Goldman, 1987). Production would then be maintained by a constant but low total nitrogen level. Alternatively or perhaps in combination, the summer production could be supported by other sources of new nitrogen such as nitrogen fixation. More speculatively, evidence exists for a possible decoupling of carbon and nitrogen in the oligotrophic surface ocean (Sambrotto et al., 1993). The Bermuda site in particular shows a large draw down of dissolved inorganic carbon over the summer that is difficult to reconcile with the available nitrate supply using traditional C/N Redfield ratios (Toggweiler 1993; Michaels et al., 1994). The C based production rate may not, therefore, be a good indicator for dissolved nitrogen uptake during this time, and may also explain some of the difference of oxygen and nitrogen based new production numbers. These uncertainties clearly highlight the need for coupled modeling studies including carbon, nitrogen and oxygen.

The spring bloom in the BATS observations and coupled model results actually occur during the deep winter convective period and is associated, at least in the model, with a large downward export of fixed PON by turbulent mixing. As shown in Figure 10, phytoplankton biomass increases during the day when the boundary layer shoals to about 25--50 m under the influence of the diurnal physical cycle (e.g. Large et al., 1994; Doney et al., 1995). The winter surface irradiance is sufficient to initiate the bloom at Bermuda as soon as nutrients are entrained into the euphotic zone and does not require the water column to restratify. In fact, even minor restratification, such as those due to the diurnal cycle, can trigger bloom like events (e.g. Stramska and Dickey, 1994). The mid-day turbulent boundary layer depth in Figure 10 is much shallower than the mixed layer depth diagnosed from the nearly homogeneous model profiles of temperature and salinity, and the resulting surface trapping of the daily phytoplankton production leads to an enhancement of the total daily production relative to solutions that do not resolve the diurnal physical cycle. In the evening, the boundary layer deepens back to its maximum depth near 235 m, and the daily accumulated PON in the euphotic zone is exported as a downward turbulent pulse across 140 m (Figure 10) (Stramska and Dickey, 1994).

Over the annual cycle, the downward turbulent mixing and subsequent detrainment of PON contributes about half, 112 mmol N m y, of the model export production. Fasham et al.'s (1990) box model results suggests a similarly important role for detrainment of PON, and Altabet (1989) estimates from observations that 40% of the export production (+130 mmol N m y) occurs via this mechanism. Recent data on the seasonal cycle of DOC at Bermuda show a large winter production and subsequent export of 990--1210 mmol C m y to the aphotic zone by deep mixing (Carlson et al., 1994). It is difficult to calculate a corresponding nitrogen flux without more information on the C/N ratio of the DOM. Clearly, however, both the model and the DOC results support larger total export production via an additional pathway in regions like Bermuda with deep winter convection, i.e. downward physical transport of organic material by mixing rather than gravitational particle sinking.

The coupled model predicts that much of the export production from the euphotic zone at Bermuda is remineralized in the shallow aphotic zone and is thus available to be reentrained during the next winter's deep mixing and to drive the subsequent spring bloom. This result is at least qualitatively consistent with the depth profile of oxygen utilization determined by Jenkins and Goldman (1985). For global models directed towards the ``biological pump'' and the sequestration of carbon from the atmosphere, it is the flux of fixed organic matter that escapes the depth of the maximum winter mixing that is important (e.g. Najjar et al., 1992; Sarmiento et al., 1993). In areas of weak seasonality, this ``winter--mixed--layer export'' will differ only slightly from the more traditional export production, defined as the flux of material out of the euphotic zone. The winter--mixed--layer export will be much smaller, however, for regions with deep winter mixing such as that near Bermuda and or the subpolar North Atlantic where much of the downward flux is regenerated within the seasonal thermocline. The winter--mixed--layer export is almost entirely due to sinking particles, and the partitioning of the export production between sinking particles and suspended particle detrainment may thus impact the size of the biological pump without changing the total flux out of the euphotic zone.

The low summer production in the coupled model could be indicative of limitations with either the model physics or biology. The one dimensional model can not, by its nature, simulate all of the effects of ocean transport. The Ekman convergence and downwelling of nutrient poor surface water creates a nitrogen sink in the euphotic zone equivalent to about 10% of the upward turbulent nutrient flux in the annual mean (Figure 9), with even larger relative effects during the summer when the turbulent fluxes are minimal. Upper ocean advection can be divided into the surface geostrophic flow and the divergent and non-divergent wind-driven Ekman circulation (e.g. Paduan and DeSzoeke, 1986). The parameterization of downwelling in the 1-D model accounts only for the divergent Ekman circulation, neglecting temporal changes from the advection of horizontal property gradients past Bermuda. Bermuda is on the southern edge of the Gulf Stream recirculation region, with southwesterly surface velocities of approximately 5 cm s (McWilliams, 1983). Eriksen (1987) found displacements of order 10 cm s in the seasonal thermocline of the LOTUS mooring data. Jenkins (1982) noted the presence of positive oxygen and salinity perturbations in the Hydrostation S data from March through July deeper in the water column below the upper part of the subtropical mode water ( 26.4, 200 m), which he ascribes to the arrival of renewal events from ventilation in the northeast corner of the recirculation region with apparent propagation rates of about 5 cm s.

The large-scale horizontal gradients of nutrients and other biological parameters necessary for converting these velocities to temporal changes are not well described, and thus it is difficult to estimate the effects of large-scale advection on the nitrogen budget at Bermuda. Nor can the model capture the nutrient transport arising from submesoscale processes, mesoscale eddies and horizontal advection and mixing along isopycnal surfaces (e.g. Jenkins, 1988). Although most likely incorrect because of the large advective nitrate input, the only available nutrient fluxes available for comparison come from the three dimensional model of Fasham et al. (1993): convection +343 mmol N m y; horizontal advection +398 mmol N m y; vertical advection -161 mmol N m y; and turbulent mixing +13 mmol N m y. The physical nutrient transports into and out of the euphotic zone in the 1-D model are significantly smaller: convection and turbulent mixing +235 mmol N m y and vertical advection -23 mmol N m y (Figure 8).

The biological model also neglects the possibly important effects of size class and ecosystem structure. Total primary production and grazing rates are often dominated by the smallest size classes, pico- and nano-phytoplankton (e.g. Malone et al., 1993) and microzooplankton (Frost, 1987; Roman et al., 1993). The vertical particle flux, however, may originate largely from larger size classes in the food web (e.g. Michaels and Silver, 1988). Roman et al. (1993) observed comparable integrated zooplankton biomass levels in August and March/April at Bermuda but a much larger proportion of large zooplankton during the bloom period. Changes in the phytoplankton size class distribution also accompany the seasonal succession in community structure following the bloom (e.g. Siegel et al., 1990; Michaels et al., 1994). Jenkins and Goldman (1985) and Goldman (1987) present a two layer hypothesis for oligotrophic systems suggesting vertical spatial variability in size structure during the summer stratified period. The surface layer would be dominated by the small size classes of phytoplankton with tight coupling and high nutrient regeneration; most of the new production would come from a different community at the base of the euphotic zone, perhaps due to episodic events. Recent work by Hurtt and Armstrong (1995) suggests a novel approach for parameterizing size effects based on a continuous size distribution function and allometric relationships. Differentiation of size structure may be required to capture seasonal and geographic variations in export production (Goldman, 1987; Toggweiler et al., 1987), and the role of size effects will be explored in future work.