Journal of Plankton Research Vol.18 no.l pp.63-85,1996 Modeling phytoplankton growth rates Jeffrey D.Haney and George A Jackson1 Department of Oceanography, Texas A&M University, College Station, TX 77843, USA 'To whom correspondence should be addressed Abstract. Mathematical models of planktonic ecosystems use a variety of different formulations to relate phytoplankton growth rates to environmental conditions. Does the formulation influence the model result? We have modified the model of Fasham, Ducklow and McKelvie (/. Mar. Res., 34,591639, 1990) to test how its results would respond to changes in algal growth rate formulations. The original model uses a Monod relationship between nutrient concentration and relative growth rate, and a multiplicative rule to combine light and nutrient effects. Use of a Droop formulation for algal growth rate or a threshold (Blackman's law) mechanism to combine light and nutrient limitation produced significant changes in simulation results. One important effect was to increase zooplankton population and, as a result, the regenerated production. While there are aesthetic reasons to prefer these alternate formulations, a more accurate formulation will require more laboratory work on algal physiology. Such laboratory work should be encouraged as an adjunct to modeling work. Introduction Models of planktonic processes become ever more important as oceanographers work to understand how the different parts of planktonic ecosystems function together. Such models are being used to study global carbon balances (e.g. Najjar etal., 1992; Fasham et al., 1993; Sarmiento et al., 1993), distribution and movement of copepods (e.g. Hofmann, 1988; Hofmann and Ambler, 1988), and deep chlorophyll layer dynamics (Varela et al., 1992). Each new model seems to use different formulations to describe interactions between biological variables. The large differences in model formulations and the paucity of discussion about the implications of the formulations are surprising. Steele and Henderson (1992) have studied model formulations in a simple nitrogen-phytoplankton-zooplankton (NPZ) system that focused on zooplankton interactions, primarily grazing. They examined how grazing of zooplankton on phytoplankton affects nutrient, phytoplankton and zooplankton populations. They noted that the results are very sensitive to the choice of zooplankton grazing formulation and of parameters used in it, and need to be studied further. The most commonly used relationship between phytoplankton growth concentration originates with the works of Caperon (1967) and of Dugdale (1967), which used Michaelis-Menten kinetics to describe nitrogen uptake rate by algal cells as a function of ambient nutrient concentration. This hyperbolic relationship has been used to justify a Monod relationship between specific growth rate and nutrient concentration (e.g. Dugdale, 1967). In an alternative approach, Droop (1973) has argued that growth actually responds to the size of an internal nitrogen pool (cell quota) rather than directly to the external nutrient concentration. The cell quota is, in turn, determined by nutrient uptake rates (given by Michaelis-Menten kinetics) and cell division. Burmaster (1979) has shown that these two formulations © Oxford University Press 63 J.D.Haney and G.AJactson are equivalent under steady-state conditions. Both formulations have been used in recent phytoplankton growth models. Nitrogen is the nutrient most often considered to limit phytoplankton growth rate and, hence, used as model currency. Its use to control phytoplankton nutrient uptake/growth in models is complicated by the fact that seawater has at least two important chemical forms of nitrogen, nitrate and ammonia, which act differently. Nitrate is usually considered to be added to the euphotic zone by physical movement of subeuphotic zone water up into the light; ammonia is usually considered to be released by animals within the euphotic zone (e.g. Dugdale and Goering, 1967). The two forms are referred to as new and regenerated nitrogen. Ammonia is frequently assumed to be taken up preferentially by algal cells, both by being more rapidly taken up at equal concentrations of nitrate and ammonia, and by suppressing nitrate uptake. Computer simulations use a variety of formulations and constants to express the suppression of nitrate uptake. Walsh (1975) used a multiplicative factor to suppress nitrate uptake in the presence of ammonia. Wroblewski (1977) used the data of Walsh and Dugdale (1972) to develop a different inhibition expression. Light limitation of algal growth has been an extensively studied physiological process (e.g. Platt etal., 1977). Although irradiance and resulting photosynthetic rates vary through the water column, photosynthesis in models is usually averaged through the mixed layer. The problem of joining expressions for light- and nutrient-limited algal growth rates has resulted in quite different solutions. The difficulty arises when describing algal growth rate under conditions where either light irradiance or nutrient concentration could limit it. Traditional approaches start with the calculation of specific growth rates relative to maximum rates for light limitation and for nutrient limitation. One approach then uses the smaller of the fractional growth rates as the appropriate one for the model. This is Blackman's 'law of the minimum' (threshold) approach. A second approach multiplies the two fractional growth rates to generate a combined fractional rate. In either case, there is a single combined value for the relative specific growth rate. The different choices that have been made to express the relationships between algal growth rates and environmental conditions may or may not have consequences for simulation results. A problem with interpreting published modeling studies is that it is difficult to know the extent to which the environmental nature of a given situation determines the results and the extent to which the details of the model implementations do. Fasham et al. (1990) published a model (hereafter referred to as FDM) that they developed to describe the euphotic zone ecosystem in the Sargasso Sea, near Bermuda. The model is a fairly simple description of a planktonic ecosystem in which temporal changes in the thermocline depth affect the ecosystem. It is a welldocumented and well-thought-out model. The authors compared its results with field observations. This model forms the basis for recent global models of carbon cycling (e.g. Fasham et al., 1993; Sarmiento et al., 1993). We have used the FDM model modified with different algal growth formulations to test their effects on simulation results. Our goal has been to determine 64 Modeling pbytoplankton growth rates which formulations have minimal effect on the results and which have profound effects. Model description Base model The FDM model follows the evolution of average mixed-layer concentrations of seven nitrogen compartments, including phytoplankton (P), zooplankton (Z), bacteria (B), nitrate (Na), ammonia (Nr), dissolved organic nitrogen (jVd) and detritus (D). Unless otherwise noted, we have used the same expressions used in FDM for the non-phytoplankton compartments and will not discuss them further. The readers should consult their paper for further details. Changes for the vertically averaged concentrations of phytoplankton and nitrate in the mixed layer in FDM are given by: where crn is the phytoplankton specific growth rate for nitrate uptake, G, is the rate of phytoplankton loss to grazing, u,, is the phytoplankton specific mortality rate, No is the subsurface nitrate concentration, •>, is the fraction of phytoplankton growth excreted as DON, M is the mixed layer depth, h* is the rate of increase of M when the mixed layer is getting deeper, 0 when it is not, and m is a mixing velocity. Equation parameters and values are listed in Table I. Much of the uncertainty that we will address concerns the model formulation used to calculate the algal specific growth rate a. Nutrient uptake Almost all formulations of nutrient uptake rate by cells as a function of nutrient concentration N use a hyperbolic relationship of the form N/(K + N), where K is the half-saturation constant. Ammonia uptake inhibition of nitrate uptake has been expressed by Wroblewski (1977) as an exponential reduction factor e"**r multiplying the nitrate uptake term. He used a value ofty= 1.5 (J.M"1. Hofmann and Ambler (1988) used the factor with t|i = 5.59 JJLM-'. Walsh (1975) developed a different inhibition factor of nitrate uptake that is essentially a two-term Taylor series approximation to the other, (1 - tyN,) - e*"; with a much smaller v|/ (= 0.02 u,M"'). FDM used Michaelis-Menten kinetics to describe phytoplankton uptake of nitrate and ammonia. Nitrate and ammonia uptake rates were summed to calculate the total nutrient uptake rate. Ammonia inhibition of nitrate uptake was expressed using Wroblewski's (1977) inhibition factor: 65 J.D.Hanej and G.A Jackson Table I. Notation. The first value shown is the standard FDM value. Note that u.M is equivalent to mmol Symbol Meaning Units Values B C (C/N) U.M-N mol mol/mol (iM-N d a y ' 21 P Bactenal concentration Carbon content of algal cell Maximum C/N composition of algae Phytoplankton loss to zooplankton grazing Rate of mixed-layer depth increase Relative light-controlled specific growth rate Half-saturation concentration for nitrate uptake and growth Half-saturation concentration for ammonia uptake and growth Half-saturation concentration for nitrate uptake in Droop formulation Half-saturation concentration for ammonia uptake in Droop formulation Minimum cell quota Eddy diffusion mixing rate Mixed-layer depth Dissolved organic N concentration Nitrate (new N) concentration Subsurface nitrate concentration Ammonia (regenerated N) concentration Phytoplankton concentration Q. + Q. Relative nitrate uptake rate Relative ammonia uptake rate Nutrient content of cell (cell quota) Zooplankton concentration Fraction of phytoplankton growth excreted as DON Specific growth rate Maximum specific growth rate Droop model specific growth rate Droop model maximum specific growth rate Phytoplankton specific mortality rate Cell concentration Nutrient uptake rate P. Maximum nutrient uptake rate a Phytoplankton average daily specific growth rate Phytoplankton average daily specific growth rate for nitrate uptake Nitrate inhibition constant c. hJ Kn K KD Ko m M jVd N No Nr P Q Qo Q, Qat z7. M-I a. N. a-r^ • e~*Nr m day' U.M-N 0.5 JJLM-N 0.5,0.05 M.M-N 2.625 u,M-N U.M-N cell' m day-' m u.M-N M.M-N M.M-N ^M-N H.M-N u.mol-N cell"' U.M-N 2.625 0.24 x 10^ 0.1 2,5 0.05 day-' day' 2.9 day-' day-' 3.58 day ' 0.045,0.54 cell 1' u.mol-N cell ' day' jimol-N cell 1 day' 3.6 x 10-* day-' day' HM-' 1.5,5.59,0.02 (3) (4) (5) 66 Modeling phytoplankton growth rates FDM used the same values for half-saturation constants for nitrate and ammonia uptake (0.5 u.M). Ammonia uptake half-saturation constants can be much lower than those for nitrate (e.g. McCarthy and Goldman, 1979). We have tested the effect of using different expressions for the Qn and Q, in the model calculations. We have tested both the effect of a different ammonia inhibition factor by substituting the factor (1 - \\>Nr) for e-*"' in equation (3) and the effect of using different \\i values found in the literature. We also investigated another formulation of ammonia inhibition by using the Michaelis-Menten inhibition kinetics to combine both the nitrate and ammonia substrates in a form used to express competitive inhibition of a single enzyme (Neame and Richards, 1972; O'Neill etal., 1989): where N'0 = NJKB and N', = NJ/Kr. This was the formulation recommended by Cullen etal. (1993). Nutrient-growth model FDM used a Monod formulation for the algal specific growth as a function of nutrient uptake: M- = VJQ (8) where \i.m is the maximum algal specific growth rate. Droop (1973) suggested that algal growth rate is related to the internal nutrient cell content, known as the cell quota Q^: m where Ko is the minimum cell quota and (i.nU3 is the maximum specific growth rate for the Droop formulation, different than |xm. The cell quota, in turn, depends on the cell division rate and on the nutrient uptake rate per cell p: where pm is the maximum uptake rate per cell and Q is, as before, the fraction of maximum nutrient uptake rate (not the cell quota). The concentration of nitrogen in phytoplankton is the product of algal cell number concentration <}> and cell quota: 67 J.D.Haney and G.A Jackson P = ^Qat (11) Changes in the phytoplankton nitrogen concentration are calculated using (12) Changes in nitrate and ammonia concentrations must be similarly altered. Equations (11) and (12) can be used to calculate changes in Q^. dP dt dt ^ dt ^ dt dt n dt v oQ (14) This use of the cell-quota approach requires additional parameters: KQ, pm, and half-saturation constants for nutrient uptake, KDjl and KDj, that can differ from those for growth. For the comparison of the Droop and FDM models, we required that Kr = Kn and KDj = KDjl ** KD. We derived the Droop parameters by setting the Droop and Monod formulations equal at steady state (e.g. Burmaster, 1979): where C is the carbon content of the cell and (CIN)^ is the maximum value of algal C:N composition. We set (C/N)m = 21 after Tett and Droop (1988). We calculated C (mol) as 1.67 x lO^cm)" 8 (Mullin et ai, 1966), assuming r = 5 x l ( H c m ( 5 jim). The maximum specific growth rate ( f i ^ ) in the Droop equation is the specific growth rate at an infinite cell quota. The parameter ^ in the Monod equation is the maximum growth rate at infinite external nutrient concentration and is related toj 68 Modeling phytoplankton growth rates PmM-n, (17) Pm ~ Light limitation Evans and Parslow (1985) derived an analytic solution for algal specific growth rate averaged over the mixed layer for a simplified light profile as a function of daylength and peak surface irradiance. It was based on the relationship between photosynthesis and irradiance developed by Smith (1936), and recommended by Jassby and Platt (1976). Daytime irradiance is described by triangular function and does not explicitly consider cloudiness (Evans and Parslow, 1985). FDM used the Evans and Parslow result. The algal specific growth rate under light limitation can be expressed as: V = »mJ(t,lMP) (18) where J is the relative growth rate as a function of peak daily surface irradiance /, day of the year t, mixed-layer depth M and phytoplankton concentration P. Joint nutrient-light limitation FDM expressed the algal specific growth rate in the presence of both light and nutrient control as the product of the two relative rates: a=\K*JQ (19) An alternative multiplicative formulation to express joint light-nutrient limitation using the Droop equation is: (20) The interaction of light and nutrient has also been represented as the smaller of the two relative growth rates (e.g. Walsh, 1975; Tett etal., 1986; Varelae/a/., 1992): <r = ,xmmin(/,Q) (21) This is known as the 'threshold' hypothesis or as the 'law of the minimum' after Blackman (1905). Mixed-layer model Deepening of the mixed layer in FDM decreases phytoplankton concentrations by diluting mixed-layer phytoplankton with phytoplankton-free water, but it does not decrease the total, vertically integrated phytoplankton population. Shoaling in mixed-layer depth has the opposite effect of decreasing the total, integrated phytoplankton population, but not directly affecting the concentrations. 69 J.D.Haney and G.AJackson Mathematically, this is calculated using h* = max(—— ,0) (Evans and Parslow, d 1985). ' Unlike the phytoplankton, the total zooplankton population, summed through the mixed layer, does not change with the mixed-layer depth. However, zooplankton concentration does increase or decrease as animals are spread out or concentrated by the deepening or shoaling of the mixed layer. FDM calculated the mixed-layer depth, M, as a function of time by fitting straight-line segments to seasonal mixed-layer depths measured over a 3 year period off Bermuda. We tried two approximations to the mixed-layer depth as a function of time: their use of a series of straight-line approximations between data points, and a smooth spline fit to the data points (Figure 1A). Sub-mixed layer water is the source for nitrate when the mixed layer deepens. FDM used a value for its concentration No = 2 JJLM. We also examined the case where JV0 = 5 \LM. Other interactions The FDM model includes interactions with dissolved organic nitrogen (DON), detritus, bacteria and zooplankton. Because our interest was in the effect of the phytoplankton model formulation, we have not modified the FDM formulations for these other compartments. FDM examined the effect of two different settling rates for detritus. We have standardized on one: 10 m day-1. In conjunction with this settling rate, FDM used a phytoplankton mortality rate ji, = 0.045 day 1 . FDM included DON loss as a factor in the phytoplankton population dynamic. They set it at a small constant fraction of primary production -y, = 0.05. Numerical solution The model was solved numerically by coding in FORTRAN, using an AdamsBashford differential equation subroutine (DDEABM from the Slatec subroutine library). Each variation was run for three simulated years, with only the results from the last year presented here. The first two simulated years were used to initialize the model. Thefinalyear's results for simulations showed no appreciable difference with either 2 or 4 years of initialization. Results The base model Phytoplankton in the basic FDM model has two concentration peaks during an annual cycle (Figure 2A). The spring bloom, driven by nutrients entrained in submixed layer water, accounts for thefirstpeak (t ~ 100 days) with a phytoplankton concentration of 0.42 u,M. This is followed by a rapid decline in phytoplankton as summer shoaling of the mixed layer cuts nitrate input and increases zooplankton grazing pressure. The second phytoplankton peak (0.48 n-M) is associated with the start of autumn mixed-layer deepening (/ ~ 280 days). Phytoplankton growth is low during the winter because of low winter light input, 70 Modeling phytoplankton growth rates 0 50 100 150 200 250 300 .150 Time (day of year) Fig. 1. Representation of mixed-layer depth and its effects. (A) Mixed-layer depth through time using straight-line segments and a cubic spline curve to interpolate between observations. (B) Phytoplankton, zooplankton and nitrate concentrations (P, Z, JV.) for different interpolation schemes between measured mixed-layer depths. Straight line segments are used in FDM. Cases are differentiated by line thickness, as indicated on the figure. The peak nitrate concentration, 0.42 u,M, is also associated with the winter mixed-layer deepening, although earlier (at —100 days) than that of the phytoplankton. Nitrate and phytoplankton concentrations show bumps at 230 and 260 days that are associated with how mixed-layer depth is calculated (see below). The peak ammonia concentration is much smaller (0.05 u,M), occurring during the winter mixed-layer deepening. The net rate of phytoplankton change is the sum of all the processes that add and subtract from the phytoplankton pool (Figure 2B). Growth rate (uP) shows strong pulses that are associated with changes in the rate of mixed-layer deepening h+ that are partially mirrored in the rate of loss to mixed-layer deepening and mixing. By far the largest phytoplankton loss is to simply mortality u,,P. Dead phytoplankton contribute to the detritus which could, in turn, settle from the euphotic zone or be consumed by zooplankton. Loss to zooplankton is always small, having a primary peak when the zooplankton is concentrated by mixed-layer shoaling and a secondary peak at the time of maximum phytoplankton concentration at the end 71 J.D.Haney and G.AJackson 0.5 50 100 150 200 250 300 350 Time (day of year) Fig. 2. Simulation results for the standard multiplicative-Monod (FDM) formulation. Time is in days from 1 January. (A) Concentrations of phytoplankton, zooplankton, nitrate and ammonia (P, Z, N, Nt) through an annual cycle. (B) Terms composing — through time. Shown are the net rate of change d( (—), algal growth corrected for DON leakage [(1 - yt)<jP], algal death (ji,/1), losses to mixed-layer At changes and mixing (MLD), and zooplankton grazing loss (C,). Note that some of the terms are subtracted rather than added to calculate —. Algal death is by far the largest loss term, nearly equal df to the growth term. Losses to zooplankton grazing are small at all times. of summer. The relatively slow algal specific growth rate can be inferred from the fact that phytoplankton specific growth rate, which has a maximum of \Ln = 2.9 day 1 , actually has a value almost equal to the phytoplankton specific mortality rate u., = 0.045 day-1. Mixed-layer model Using straight-line segments to approximate the mixed layer (Figure 1A) introduces transients that using a smoothed curve does not. There are shght changes in the timing and size of phytoplankton concentrations as a result (Figure IB). 72 Modeling phytoplankton growth rates Ammonia inhibition of nitrate uptake Ammonia inhibition of nitrate uptake is increased by larger i|; values, resulting in higher nitrate concentrations, and delayed but higher peak phytoplankton concentrations in the spring bloom (Figure 3). The maximum spring nitrate concentration is 0.58 u,M for »»| = 5.59 u,M~' and 0.38 n-M for i|i = 0.02 n-M'1, compared to 0.44 fiM for the FDM \\i = 1.5 JJLM"1 case. The larger \\i results in a 30% increase in maximum spring nitrate concentration relative to the FDM case. The associated peak spring phytoplankton concentrations are 0.41 and 0.46 JAM compared to 0.42 ^M for FDM. There is virtually no difference in either nitrate or phytoplankton concentrations during summer (t = 180-270 days), when there is essentially no nitrate or ammonia in the water. The Michaelis-Menten inhibition formulation of ammonia inhibition gives results very similar to those for the FDM model (not shown). The slightly higher nitrate and ammonia concentrations during the winter help to create a slightly higher phytoplankton maximum during the spring bloom. Half-saturation constants for ammonia uptake A smaller half-saturation constant for ammonia uptake [K, = 0.05 u,M; equation (4)] has little effect on simulation results (Figure 4). Ammonia concentrations are lower and uptake greater than with the standard K, = 0.5 u.M. The resulting phytoplankton and nutrient concentrations are similar to those for the case of lower ammonia inhibition of nitrate uptake (Figure 3), a greater spring phytoplankton concentration occurring earlier. The second half of the year shows essentially no differences between the two formulations. Nutrient-growth model Nutrient and phytoplankton concentrations are very similar during summer and early fall (/ = 180-300 days), but very different during winter and spring for the Droop and Monod (FDM) growth formulations (Figure 5A). Maximum winter nitrate concentrations, 0.43 JJLM for FDM, are almost 10 times greater than the 0.05 H,M maximum for the Droop case. The cell quota for the Droop model ranged from 2.5 x 10-7 to 8 x 10-7 u,mol-N cell1 (not shown). The integrated net primary production (NPP) is higher in the Droop model during almost all of the year (Figure 5B). The higher phytoplankton concentrations for the Droop model in the winter contribute to an earlier spring bloom peak, with the maximum primary production being greater than for FDM. Both formulations show strong stepping in NPP and new net primary production (NNPP) during the latter half of the year that is associated with the sudden changes in the rate of mixed-layer depth deepening (/i+). Light-nutrient growth model There are major differences, some subtle, some striking, between the results from using the multiplicative light-nutrient interactions (FDM) and those using the 73 J.D.Haney and G.A Jackson 1 1 0.5 » / */ \ 0.4 / / p \ \ — Z - Nn Nr V \ 1)1=15 ¥=5 59 ^.02 ^T A J \ '' c o 0.3 / .—^_X ^b 0.2 4> o c o U -1 j 0.1 v, >.•/*- + /.• "-•••••v--- 0 H 0.05 1 H dP/dt (1-TlXJP 1——i—i—i— ¥=u - - 0.04 B MLD GiP 0.03 / > / \ ! u 0.02 te of chan — yfc5 59 i(»002 i ^ .•-•-. •v>... 0.01 0 - ^ O.01 0 50 100 150 200 250 300 350 Time (day of year) Fig. 3. Effect of different ammonia inhibition formulations. (A, B) as in Figure 2. Shown are the results for the standard FDM (if = 1.5 M-M"')- and for higher and lower values of I\I of 5.59 and 0.02 fj.M"'. Differences are greatest at the beginning of the year when the mixed layer is deepening. By / = 210 days, there is little difference in the results for the different values of I|I. threshold formulation. For the threshold formulation, the spring phytoplankton bloom occurs earlier during the time of maximum mixed-layer thickness and lasts longer, but has a slightly smaller maximum concentration (Figure 6A). Phytoplankton concentrations fall rapidly when the mixed layer shoals. Nitrate and ammonia concentrations are small all year, ~0.01 u-M, compared to the winter ammonia and nitrate concentrations of ~0.05 and 0.4 jtM in the multiplicative formulation. Zooplankton concentrations during the winter are ~0.1 jtM, approximately 10 times those for the multiplicative model. The maximum zooplankton concentration is ~0.18 u.M, about three times that of the multiplicative model. Values of relative light and nutrient phytoplankton specific growth rates / and Q help explain the difference (Figure 6D). In FDM, phytoplankton growth is light limited during the winter, with Q ~ 0.5 and / ~ 0.05. Cell division is nutrient limited in the shallow mixed layer of summer, with Q ~ 0.5 and / ~ 0.05. The product QJ determines the actual specific growth rate and ranges from —0.02 to 0.025. In contrast, the threshold formulation has Q almost always less than J, although the 74 Modeling phytoplankton growth rates p — — 0.5 Kr=*05 N, 0.4 \ I \ \ 1 ;, ^ 0.2 I / \ rm X / / \ o c / / V \ o O ; V ' I \ 0.3 1) \ / "</ ^ f o A KH=0J N.-- X \X r 0.1 .V, 0 1— /-^ \ 1— 1 0.05 H B dPAh' 0 Ml?* *" 0.04 -a - Nfl-D G|P 0.03 '• 3. \^\ 1 <u Vy ed . . . - • • . 1 .- - * JC U - ;_- " >• 0.01 1 . • mi- • • - . J • X - -0.01 50 • - f\A' &'''' Rat . 100 150 200 250 300 350 Time (day of year) Fig. 4 Effect of half-saturation constant for ammonia uptake (K,). (A, B) as in Figure 2. The standard FDM uses K, - 0.5 JAM; the variant uses K, = 0.05 M-M. The lower K, results in a higher winter growth rate [(1 - y,)uP] and grazing rate (G,P). The overall rate of change (—) shows little difference d( except for a slightly earlier spring peak. two have close to the same values during the winter. The values of J are similar for the two formulations. The difference is in the value of Q, which ranges only from 0.01 to 0.04 for the threshold case and is always less than Q for the multiplicative case. As a result of the lower nitrate concentrations, the threshold case is nutrient limited essentially all year. Despite the lower nutrient concentrations, the phytoplankton division rate is usually greater for the threshold case. The effect of continual nutrient limitation in the threshold case can be seen in the high growth rates aP during the winter mixed-layer deepening (Figure 6B). Furthermore, sharp changes in the mixed-layer depth representation cause an even greater saw-toothed pattern in aP. Phytoplankton dying, JJL,/), is still a major term removing phytoplankton, although zooplankton grazing is nearly comparable during the winter and greater during the spring zooplankton pulse. Net primary production is significantly greater during the winter for the threshold case (Figure 6C). Because NPP is the product of phytoplankton growth per unit volume uP and the mixed-layer depth (A/), low rates of concentration during 75 J.D.Haney and G.A Jackson ' 50 100 150 200 250 300 350 50 100 150 200 250 300 o 350 Tune (day of year) Fig. 5. Effect of using Droop and Monod (FDM) nutrient-growth models. (A) Effects on nutrient pools (P, Z, JVW N,). (B) Effects on terms of phytoplankton rate of change and net change for different lightnutrient interactions through time for threshold formulation. (C) Effects on rates of vertically integrated primary productivity measures, including net primary production (NPP), net new primary production (NNPP) and net regenerated primary production (NRPP). (D) Effect on relative nutrient (Q), light (/) and joint [JQ for FDM, min(<r, (?) for threshold] through time. The standard Monod case is not shown in (C), but is in Figure 2B. the winter are balanced by the large values of M. Most of the difference between the threshold and the multiplicative cases is in the greater importance of production resulting from regenerated production (NRPP) in the threshold case. Subsurface nitrate concentration Changing the subsurface nitrate concentration, No, has a great effect on the FDM formulation. Using No = 5 jiM rather than 2 u.M resulted in peak nitrate concentrations during the spring bloom of 2.2 |i,M, ~5 times that of the standard case (Figure 7A). Peak ammonia and zooplankton concentrations are also much greater, ~20 and 30 times higher. Maximum phytoplankton concentrations are about the same, although there are pronounced oscillations after the peaks. All compartments show oscillations after the spring bloom peak. These are intensified for higher values of No (not shown). Phytoplankton population regulation changes from predominantly natural mortality to predominantly zooplankton grazing with the increase in Ng, although natural mortality and zooplankton grazing are comparable during the pre-spring bloom winter deepening of the mixed layer. The oscillations in phytoplankton concentrations, as well as those of other compartment, are the most dramatic effect of higher N^ Excepting these large oscillations, it is surprising how small the changes in phytoplankton concentrations in 76 Modeling phytoplankton growth rates 0.05 dPAh - Thresh U-YIXJP • 05 0.04 MLD G|P 0.03 OJ 0.02 §0.2 00 0.01 "o C o 3. -0.01 ~ 6 0J 04 a -5 3 § 03 > fi 2 02 g •a 0 1, 0 01 50 100 150 200 250 300 350 50 100 150 200 250 300 350 0 Time (day of year) Fig. 6. Effect of formulation of nutrient-light interactions through time (A-C) as in Figure 5. Shown are the results for multiplicative (FDM) and threshold formulations of the Monod growth model. (D) Equivalent relative growth rates for the two formulations. time are. There is a slight increase in pre-spring bloom concentrations with larger No, but remarkably little else. A decrease in No to 1 ^.M does dramatically decrease P to about half the values when No = 2JJLM. The results for the threshold formulation with the higher value of No are similar for the multiplicative formulation (Figure 7B). Winter nutrient concentrations are slightly lower, 1.9 versus 2.2 |xM, and peak zooplankton concentrations during the spring are slightly higher, 2.25 versus 1.6 (xM. The Droop formulation shows similar results to those of the threshold formulation at this higher value of No, as it did for the case of No = 2 u,M. However, the Droop formulation shows little oscillation in phytoplankton concentration. Discussion We wanted to know whether the formulation of phytoplankton growth rate relationships affects simulation results. The answer is clearly yes. For this particular system, the biggest difference is during the deep mixing of early winter. There are two different types of model response. The first has phytoplankton nutrient uptake and growth rates that are too low to effectively deplete nutrient concentrations during the winter. The lower nutrient usage formulations include those with higher ammonia suppression of nitrate uptake and the multiplicative growth and light interaction formulation. The second response has higher uptake and growth rates that result in a depletion of nutrients, larger phytoplankton concentrations and a smaller decrease in winter zooplankton populations. The extra winter plant growth and higher phytoplankton concentrations resulting from more 77 J.D.Haney and G.AJackson 2 0 50 100 150 200 250 300 350 Time (day of year) Fig. 7. Effect of higher sub-mixed layer nitrate concentration, No = 5 p.M. The FDM standard value of No = 2 JJ.M. (A) Effect on concentrations of phytoplankton, zooplankton, nitrate and ammonia (P, Z, N* ^r) through time for the standard FDM model. (B) Effect on phytoplankton and zooplankton concentrations for higher sub-mixed layer nitrate concentration for the FDM, threshold and Droop formulations. efficient nutrient usage keep the zooplankton population from having as large a winter decrease and allow it to consume more of the plant production during the spring. The threshold growth model and the Droop formulation are examples of models with this effect. That the ammonia suppression of nitrate uptake results in low winter uptake is not surprising, but it might be surprising that the multiplicative formulation has a similar effect. To see why, consider a case where light and nutrient concentration would each limit the algal specific growth rate to 0.25 of maximum. With the multiplicative approach, the effective algal specific growth rate would be 0.0625, a quarter of the rate of the threshold approach. As noted earlier, the relative specific growth rate for the non-limiting factor in the multiplicative case is always less than one and frequently less than 0.5 (Figure 6D). As a result, the algal specific growth rate is always significantly less than the rate of the lower, controlling process, i.e. the multiplicative formulation yields specific growth rates for phytoplankton significantly lower than those determined by the more limiting factor. 78 Modeling pbytoptankton growth rates Despite the formulation of growth rate as a multiplicative function of light and internal cell quota, the Droop formulation has an effect similar to that of the threshold formulation. Adding the threshold to the Droop formulation results in little additional change (not shown). The decoupling of nutrient uptake and growth [equations (9-14)] allows the cell quota to become large enough during the winter deepening of the mixed layer so that the relative specific growth rate for nutrients becomes almost one, with a similar effect to that of the threshold formulation. Probably more important is the fact that the maximum specific rate of nutrient uptake is greater than that for growth when the cell quota is small. This decoupling of nutrient and growth rates is in the spirit of the ideas advanced by Goldman and McCarthy (1978), Morel (1987) and McCarthy (1981). Of all their parameters, FDM found that their model results are most sensitive to the choice of subsurface mixed-layer nitrate concentration. We find that this extreme sensitivity results in noticeable transients in the phytoplankton/nutrient/ concentrations when the depth of the thermocline is approximated using straightline segments to interpolate between observations. Similar transients can be seen in the FDM results. When we use a smooth spline fit to approximate the mixedlayer depth, the oscillations disappear. Platt et al. (1989), Goldman (1993) and Goldman et al. (1992) have argued that storms can be important periodic nutrient sources, particularly for large diatoms in the Sargasso Sea. Such pulsing of nutrient input would also have a large effect on this model and suggests that it might be important to include in future models. A comparison of the various annual productivity values highlights similarities and differences for the different formulations (Table II). The annual new production is essentially unchanged in all of the simulations with No = 2 \iM because essentially all of the nitrate that arrives in the surface mixed layer is taken up by the phytoplankton. There is a small, 5%, increase for the threshold Monod formulation. The annual regenerated productions of the threshold and Droop formulations are 2.6 and 2.3 times that of the standard case. The resulting /-ratios are 0.58, 0.60 and 0.77. Thus, the enhanced zooplankton population associated with the higher winter growth of the threshold and Droop formulations gives rise to greater nitrogen recycling and greater annual total production. FDM were interested in understanding the seasonal cycle in the plankton off Bermuda. They used historical data both to 'tune' their model and to test it. They observed the effects of varying algal mortality rate u.,, subsurface nitrate concentration N& and detrital sinking rate V on the annual total net and new primary production. They ultimately chose tofixNo = 2 u.M and to choose the value JJL, that best fit the productivity data for a given value of V. It could be argued that a better comparison of the original FDM model with any of the variants that we have discussed would involve the same parameter selection procedure used by FDM. This would involve the selection of a specific death rate that gave similar values for the annual total net and new production rates. Contour plots of the values of annual total production and/-ratios show that this technique can be used for the threshold and Droop formulations only over a more limited range of/-ratios (Figure 8). Such a procedure yields a value for p., = 0.054 day-', 18% higher than the previous value. However, the range of/-ratios that can befitis 79 J.D.Haney and G.A Jackson Table II. Integrated annual phytoplankton production. New production is production resulting from nitrate uptake; regenerated production is that resulting from ammonia uptake; total production is the sum of new and regenerated production, / i s the ratio of new to total production Case Production (mol-N nr 2 year 1 ) / Total % change / % change Total New Regenerated Multiplicative Monod Standard Smoothed mix layer <\i = 5.59 M.M1 I|I = 0.02 M-M-' K, = 0.05 No = 5 (JIM 0.52 0.52 0.49 0.53 0-57 1.61 0.40 0.40 0.39 0.40 0.40 0.80 0.12 0.12 0.10 0.13 0.17 0.81 0.77 0.77 0.80 0.76 0.71 0.49 _ _ 0 -5.8 1.9 9.6 209.6 0 3.9 -1.3 -7.8 -36.4 Multiplicative Droop Standard /V 0 =5(iM 0.70 1.89 0.41 0.85 0.28 1.04 0.60 0.45 34.6 263.5 -22.1 ^U.6 Threshold Monod Standard u., = 0.054 day-' iV 0 =5nM 0.72 0.57 1.77 0.42 0.42 0.85 0.31 0.15 0.92 0.58 0.73 0.48 38.5 9.6 240.4 -24.7 -5.2 -37.7 smaller for the Droop and threshold formulations. For these, the maximum value of/is 0.78, compared to 0.88 for the FDM case, over the region of interest. Note that the fact that essentially all of the nitrate is consumed fixed the annual NNPP. Choosing parameters to match NPP then also determines NRPP and the/-ratio. FDM noted that the results suggested that algal death was the dominant loss for the phytoplankton growth, although they also noted that the inclusion of sizebased algal-grazer interactions would alter the model results. It is disconcerting that the predominant loss of phytoplankton is to solitary death, whose rate depends on a parameter used to tune the model. The importance of this term is partially the result of the low specific growth rates calculated using the multiplicative growth formulation and the low zooplankton populations. While Walsh (1983) has argued that such death should be important in the sea, there is little experimental evidence that this is so. However, p., should be a measurable quantity. Using a threshold growth formulation with the same specific death rate (0.045 day 1 ) increases the average specific algal growth rate and makes the zooplankton grazing a comparable or, occasionally, a greater loss term for the algae. Fasham (1994) has derived a new expression for algal death which decreases its role as the major loss of phytoplankton. It is intriguing that model zooplankton dynamics are so sensitive to changes in algal growth formulations, that relatively small changes in phytoplankton concentrations, —30%, can cause such large changes in winter zooplankton concentrations. Although the zooplankton directly respond only to phytoplankton concentration, they are able to effectively respond to large changes in phytoplankton growth rates and productivity that show only as small changes in phytoplankton concentration. 80 Modeling phytoplankton growth rates /-ratio Annual total production 1.4 1 6 1.8 2.0 2.2 1.2 1.4 1.6 1.8 2.0 2.2 Sub-mixed layer nitrate concentration (uM) Fig. 8. Values of annual total production (A, C, E) and/-ratios (B, D, F) for standard Monod-multiplicative (A, B), threshold-Monod (C, D) and Droop-multiplicative (E, F) formulations as functions of /Voand(i.,. In a similar study of the zooplankton mortality term formulation, Steele and Henderson (1992) have shown that the zooplankton/phytoplankton dynamics are influenced by the form of the zooplankton mortality term. Changes in the description of zooplankton grazing rates can lead to pronounced predator-prey oscillations in a simple NPZ model. FDM also noted that their results were very sensitive to the choice of grazing model. Of the terms and constants that we examined, predator-prey oscillations were most sensitive to the subsurface nitrate concentration (e.g. Figure 7). While an algal preference for the reduced nitrogen form of ammonia rather than nitrate has been justified on energetic grounds, Dortch (1990) has argued that the experimental evidence that this actually occurs is weak. On the contrary, Levasseur et al. (1993) have observed that nitrate can be taken up by diatoms preferentially to ammonia under the low-light conditions where energy should be limiting. Usage of the ammonia inhibition factor can result in total nitrogen uptake rates being lower in the presence of the additional nutrient source of ammonia than the uptake of nitrate alone without ammonia present. That adding a nutrient source to a nutrient-limited situation can decrease total uptake does not seem reasonable. A 81 J.D.Haney and G.A Jactsod similar preference for ammonia uptake can be obtained by using a smaller value of Kt. For these reasons, we suggest that ammonia inhibition be dropped until such time as there is better experimental evidence to describe it. The Droop formulation for nutrient-limited growth has several nice properties that the Monod one does not. Its decoupling of nutrient uptake and growth rate does agree with experimental observations. Furthermore, the adaptation of algal nutrient content allows a modicum of organism response to changing environmental conditions. This additional flexibility comes at the cost of an additional variable and several additional parameters needed to describe the system. Despite these nice properties, it is not clear the Droop formulation is optimal. Morrison et al. (1987) examined the effect of algal growth formulation on the ability of a water quality model to predict phytoplankton concentrations in a Canadian lake and reservoir system. For each formulation, they used an optimization procedure to calculate coefficients that provided the best fit during a 3 year period at one of three stations. They compared simulation results to observations for the remaining data set,findingthat storage models such as that of Droop did a poor job of predicting concentrations. Their models differed significantly from those of FDM because they imposed grazing losses on the phytoplankton using field measurements of zooplankton populations rather than let zooplankton populations respond to phytoplankton concentrations. However, they do indicate that deciding the best phytoplankton growth formulation will require the comparison of observations and predictions at multiple sites. The threshold formulation for incorporating both nutrient and light limitation of algal growth appears to be better than the multiplicative one because the latter has the algae growing at rates that are substantially less than that expected for the single most limiting factor. In either case, there is a more subtle problem in combining photosynthesis and nutrient uptake which should be manifest in the composition of the cells (Eppley, 1981; Cullen et al., 1993). If a cell is taking nutrients up at a rate equivalent to 0.25 day"1, but photosynthesizing at 0.5 day"1, its carbon content should continue to increase relative to that of nitrogen, easily to unrealistic levels. In fact, the photosynthesis rate also needs to fall to be in balance with the nutrient uptake. This lack of balance between photosynthetic and nutrient uptake rates is another sign that the present formulations for algal growth have been patched together rather than developed as a unified idea. O'Neill et al. (1989) examined the ability of different expressions for nutrient limitation by two factors to fit experimental observations. There are methodological problems with using the same formulation to predict rate of cell division for algae grown in chemostats and with accumulated harvest yield in terrestrial crops. Nevertheless, the comparisons are informative. O'Neill et al. found that a Blackman formulation using two Monod relationships did notfitthe data well when both elements might be limiting, although it worked well when only one element was limiting. Expressing growth as a product of two Monod relationships, in what they called a 'Baule model', provided one of the betterfitsto the data, but the constants derived from fitting all the data were different than those derived by using data with only one limiting nutrient. The implication for oceanic models is that the con82 Modeling phytoplankton growth rates stants which would make a multiplicative expression work are not the ones derived from typical laboratory measurements. Rather, the constants need to be developed from experiments examining interactions between limiting factors. This paper has avoided making comparisons to experimental observations to focus on the differences that choices in model formulation can make. Furthermore, the results of Menzel and Ryther (1960) that FDM used to calibrate their model have been superseded by newer, presumably more accurate, measurements made as part of the Bermuda time series (Michaels et al., 1994; Siegel et al., 1995). The newer measurements, made using trace metal-clean techniques, provide estimates of annual primary production twice as high as that of Menzel and Ryther (1960) (Siegel etal., 1995). Were FDM to have used this data for their model, they would have made different parameter choices and presumably have done a better job fitting the new data, just as we would have also. It would not, however, have changed the fact that choices in model formulation affect the results in ways that need to be resolved. When model predictions are compared against field measurements, it is important to remember that there are a range of measurements whose values must be consistent with model predictions. J.Cullen (personal communication) has noted that C:Chl data provide a powerful check, both directly, and also because the assimilation index (g-C g-Chl"1 day 1 ) divided by the C:Chl ratio (g-C g-Chl-1) should equal the algal specific growth rate (day 1 ). With the specific growth rate calculated in the model and the assimilation number measured in the field, a value of C:Chl that is consistent with them can be calculated and compared to the range of values found in culture studies (e.g. Geider, 1993). Such values need to be consistent. The proper formulation of algal growth rate as a function of combined nutrientlight limitation is an experimental problem. There has been extensive work on the role of nitrogen concentration on growth rates (e.g. Carpenter and Capone, 1983), and extensive work on the interaction of light and photosynthetic rate (e.g. Platt et al., 1977), but remarkably little progress in uniting the two. Doing so will require describing the interaction of cellular carbon and nitrogen concentrations and light absorbance, as well as understanding the specific growth rate for a given combination of irradiance and nutrient concentration (Cullen etal., 1993). Models such as that of Laws and Chalup (1990) may help describe some of the nutrient-light interactions, but are too complicated to be used directly in large ecosystem models. They may be useful, though, to test the errors associated with computationally simpler growth models, such as the threshold formulation. Although being able to construct the relevant interactions will not be easy, it will be important for the development of models describing planktonic ecosystems without the present arbitrariness and uncertainty. As computers become bigger and faster, as society focuses on global change, it is natural and right that oceanographers try to use our understanding of oceanic ecosystems by building global biological models. The grandeur of such efforts should not blind us to the continuing need for experimental and theoretical work to refine the relationships upon which the large models are built. 83 J.D.Haney and G.AJackson Acknowledgements We were helped by conversations with P.Eldridge, A.Murray and P.Harrison, J.Cullen and B.Zakardjian provided useful and helpful comments on the manuscript. This work was supported by Office of Naval Research Contract N00014 87-K0005 and by the DOE Computational Science Graduate Fellowship Program. References Blackman.F.F. (1905) Optima and limiting factors. Ann. Bol., 19,281-295. Burmaster.D.E. (1979) The continuous culture of phytoplankton: mathematical equivalence among three unsteady state models. Am. Nat., 113, 123-134. CaperonJ. (1967) Population growth in micro-organisms limited by food supply. Ecology, 48,715-722. Carpenter.EJ. and Capone.D.G. (1983) Nitrogen in the Marine Environment. Academic Press, New York. CullenJ.J., Geider.RJ., IshizakaJ., Kiefer.D.A., MarraJ., Sakshaug.E. and RavenJ.A. (1993) Toward a general description of phytoplankton growth for biogeochemical models. In Evans,G.T. and Fasham,M J.R. (eds), Towards a Model of Ocean Biogeochemical Processes. Springer-Verlag, pp. 153-176. Dortch.Q. (1990) The interaction between ammonium and nitrate uptake in phytoplankton. Mar. Ecol. Prog. Ser., 61,183-201. Droop.M.R. (1973) Some thoughts on nutrient limitation in algae. J. PhycoL, 9, 264-272. Dugdale.R.C. (1967) Nutrient limitation in the sea: Dynamics, identification, and significance. Limnol. Oceanogr, 12, 685-695. Dugdale,R.C. and GoeringJ J. (1967) Uptake of new and regenerated forms of nitrogen in primary production. Limnol. Oceanogr., 12,196-206. Eppley,R.W. (1981) Relations between nutrient assimilation and growth in phytoplankton with a brief review of estimates of growth rate in the ocean. In Platt.T. (ed.), Physiological Bases of Phytoplankton Ecology. Can. Bull. Fish. Aquat. 5c/., 201, 251-263. Evans,G.T. and Parslow J.S. (1985) A model of annual plankton cycles. BioL Oceanogr., 3, 327-347. Fasham,M.J.R. (1994) Modelling the marine biota. In Heimann,M. (ed.), The Global Carbon Cycle. Springer-Verlag, Berlin. Fasham,MJ.R., Ducklow.H.W. and McKelvie.S.M. (1990) A nitrogen-based model of plankton dynamics in the oceanic mixed layer. J. Mar. Res., 34,591-639. Fasham.M.J.R., SarmientoJ.L., Slater.R.D., Ducklow.H.W. and Williams.R. (1993) Ecosystem behavior at Bermuda Station 'India': a general circulation model and observational analysis. Global Biogeochem. Cycles, 7,379^*15. Geider.RJ. (1993) Quantitative phytoplankton physiology: implications for primary production and phytoplankton growth. ICES Mar. ScL Symp., 197,52-62. Goldman J.C. (1993) Potential role of large oceanic diatoms in new primary production. Deep-Sea Res., 40,159-168. GoldmanJ.C, Hansell.D.A. and Dennett,M.R. (1992) Chemical characterization of three large oceanic diatoms: potential impact on water column chemistry. Mar. Ecol. Prog. Ser., 88, 257-270. GoldmanJ.C. and McCarthyJJ. (1978) Steady state growth and ammonium uptake of a fast-growing marine diatom. Limnol Oceanogr., 23, 695-703. Hofmann,E.E. (1988) Plankton dynamics on the outer southeastern U.S. continental shelf. Part III: A coupled physical-biological model./ Mar. Res., 46,919-946. Hofmann,E.E. and Ambler J.W. (1988) Plankton dynamics on the outer southeastern U.S. continental shelf. Part II: A time-dependent biological model. / Mar. Res., 46,883-917. JassbyvA. and Platt.T. (1976) Mathematical formulation of the relationship between photosynthesis and light for phytoplankton. Limnol Oceanogr., 21,540-547. Laws,E.A. and Chalup.M.S. (1990) A microalgal growth model. Limnol Oceanogr., 35,597-608. Levasseur,M., Thompson.P.A. and Harrison,PJ. (1993) Physiological acclimation of marine phytoplankton to different nitrogen sources. J. Phycol., 29,587-595. McCarthyJJ. (1981) The kinetics of nutrient utilization. In Platt,T. (ed.), Physiological Bases of Phytoplankton Ecology. Can. Bull Fish. Aquat. Sci., 210, 211-233. McCarthyJJ. and GoldmanJ.C. (1979) Nitrogenous nutrition of marine phytoplankton in nutrientdepleted waters. Science, 203,670-672. 84 Modeling pbytoplankton growth rates Menzel.D.W. and Ryther J.H. (1960) The annual cycle of primary production in the Sargasso Sea off Bermuda. Deep-Sea Res., 6,351-367. Michaels.A.F., Knap.A.H., Dow,R.L_, Gundersen,K., Johnson.RJ., SorensenJ., CloseA-, Knauer.G.A., Lohrenz,S.E., Asper.V.A., Ruel,M. and Bidigare,R. (1994) Seasonal patterns of ocean biogeochemistry at the U.S. JGOFS Bermuda Atlantic Time-Series Study site. Deep-Sea Res. 1,41, 1013-1038. Morel^F.M.M. (1987) Kinetics of nutrient uptake and growth in phytoplankton. J. Phycol., 23,137-150. Morrison,K.A., Th6rien,N. and Marcons3- (1987) Comparison of six models for nutrient limitations on phytoplankton growth. Can. J. Fish. Aquat Sci., 44,1278-1288. Mullin,M.M., Sloan,P.R. and Eppley.R.W, (1966) Relationship between carbon content, cell volume, and area in phytoplankton. LimnoL Oceanogr., 11, 307-311. Najjar.R., SarmientoJ.L. and ToggweilerJ.R. (1992) Downward transport and fate of organic matter in the ocean: simulations with a general circulation model. Global Biogeochem. Cycles, 6,45-76. Neame JC.D. and Richards,T.G. (1972) Elementary Kinetics of Membrane Carrier Transport. Blackwell Scientific, London, 120 pp. O'NeiU,R.V., De Angelis,D.L., Pastor JJ., Jackson,BJ. and Post,W.M. (1989) Multiple nutrient limitations in ecological models. EcoL Modelling, 46,147-163. Platt,T., Denman JC.L. and JassbyA-D. (1977) Modeling the production of phytoplankton. In Goldberg JE.D., McCaveJ.N., O'Brien J J. and Steele J.H. (eds), The Sea. Vol. 6. Marine Modeling. Wiley, pp. 807-856. Platt,T., Harrison,W.G., Lewis,M.R., Li,W.K.W., Sathyendranath,S., Smith.R.E. and Vezina,A.F. (1989) Biological production of the oceans: the case for a consensus. Mar. Ecol. Prog. Ser., 52,77-88. SarmientoJ.L., Slater,R.D., Fasham.MJ.R., DucklowJ^.W., ToggweilerJ.R. and Evans.G.T. (1993) A seasonal three-dimensional model of nitrogen cycling in the North Atlantic euphotic zone. Global Biogeochem. Cycles, 7,417^450. SiegelJD.A., MichaelsAF., SorensenJ.C, O'BrienJvI.C. and Hammer,M.A. (1995) Seasonal variability of light availability and utilization in the Sargasso Sea. /. Geophys. Res., 100, 8695-8713. Smith,E.L. (1936) Photosynthesis in relation to light and carbon dioxide. Proc. NatlAcad. So. USA, 22, 504-511. SteeleJ.H. and Henderson.E.W. (1992) The role of predation in plankton models. J. Plankton Res., 14, 157-172. Tett,P., Edwards.A. and JonesJC. (1986) A model for the growth of shelf-sea phytoplankton in summer. Estuarine Coastal Shelf Sci., 23,641-672. Tett,P. and Droop.M.R. (1988) Cell quota models and planktonic primary production. In WimpennyJ.W.T. (ed.), Handbook of Laboratory Model Systems for Microbial Ecosystems. CRC Press, Vol. 2, pp. Ml-232. Varela,R.A., Cruzado,A., Tintorey. and Ladona,E.G. (1992) Modelling the deep-chlorophyll maximum: A coupled physical-biological approach. J. Mar. Res., 50,441-463. WalshJ.J. (1975) A spatial simulation model of the Peru upwelling ecosystem. Deep-Sea Res., 22, 201-216. Walsh J J . (1983) Death in the sea: enigmatic phytoplankton losses. Prog. Oceanogr., 12, 1-86. Walsh J J . and Dugdale,R.C. (1972) Nutrient submodels and simulation models of phytoplankton production in the sea. In Allen.H.E. and KramerJ.R. (eds), Nutrients in Natural Waters. Wiley, New York, pp. 171-191. Wroblewski J.S. (1977) A model of phytoplankton plume formation during variable Oregon upwelling. J. Mar. Res., 35, 357-394. Received on March 8, 1995; accepted on September 22, 1995 85
© Copyright 2026 Paperzz