Influence of Rossby waves on primary production from a coupled physical-biogeochemical model in the North Atlantic Ocean

Influence of Rossby waves on primary production from a coupled physical-biogeochemical model in the North Atlantic Ocean G. Charria, I. Dadou, P. Cipollini, M. Drévillon, and V. Garçon Laboratoire d’Etudes en Géophysique et Océanographie Spatiales, UMR5566/CNRS, Toulouse, France National Oceanography Centre, Southampton, UK CERFACS, MERCATOR-Océan, Toulouse, France Received: 30 October 2007 – Accepted: 13 November 2007 – Published: 29 November 2007 Correspondence to: G. Charria (gcha@noc.soton.ac.uk)


Introduction
The detection of westward propagating signals in surface chlorophyll concentrations 20 related to Rossby waves (e.g. Machu et al., 1999;Cipollini et al., 2001;Uz et al., 2001;Kawamiya and Oschlies, 2001) prompted the question of the underlying physical/biogeochemical interactions. Based on remotely sensed data and/or coupled physical/biogeochemical modelling, several studies investigated the coupled physical/biogeochemical mechanisms which might be involved (e.g. Charria et al., 2003; Printer-friendly Version Interactive Discussion

EGU
(1) the upwelling mechanism associated with nutrient injection Uz et al., 2001;Siegel, 2001), (2) the uplifting of a deep chlorophyll maximum towards the surface Kawamiya and Oschlies, 2001;Charria et al., 2003), and (3) the meridional advection of chlorophyll by geostrophic currents associated with baroclinic Rossby waves (Killworth et al., 2004). These three processes are described 5 using theoretical models and compared to the remotely sensed observations in Killworth et al. (2004). These authors pointed out the importance of the third process at the global scale. More recently, Charria et al. (2006a) showed a significant contribution of the vertical process of nutrient injection in the North Atlantic, north of 28 • N, using remotely sensed data and theoretical models from Killworth et al. (2004). However, 10 in the two latter studies, several assumptions (for example: a constant relaxation time for biology and a constant Chl N ratio) were made. We propose here to investigate the relative contribution of these mechanisms using a more realistic 3-D coupled physical/biogeochemical modelling approach.
Furthermore, the influence of Rossby waves on biogeochemical processes has to be 15 estimated for a better understanding of the carbon cycle through the biological pump mechanism. Kawamiya and Oschlies (2001) have found a ∼30% increase in primary production with the passage of a Rossby wave in the Indian ocean around 10 • S using a coupled physical/biogeochemical model. Near the Hawaiian Ocean Time Series (HOT) station ALOHA in the Pacific Ocean, a combination of in situ high-frequency measure-20 ments (shipboard and moored sampling) and remotely sensed data (altimetry) during a long period (1997)(1998)(1999) allowed Sakamoto et al. (2004) to estimate primary production enhancement up to 25% on average due to the Rossby wave's passage. Indeed, Rossby waves can significantly contribute to supply nutrients to oligotrophic surface waters fuelling primary production. According to several authors (e.g. Oschlies and EGU upwell nutrients all along their propagation path through an ocean basin. This effect is then potentially comparable to basin-scale processes as induction process in the subpolar gyre  or Dissolved Organic Nitrogen and Phosphorus advection in the subtropical gyre . The present work aims at investigating Rossby wave influence on primary produc-5 tion using a 3-D coupled physical/biogeochemical realistic model in the North Atlantic Ocean. After a brief description of the coupled physical/biogeochemical model in Sect. 2, as well as the remotely sensed data used in the validation of simulated wave features in the model (Sect. 3), we will show the surface chlorophyll and physical signatures of Rossby waves using model results. In Sect. 4, the features of these modelled propagating signals are identified and compared to those estimated with remotely sensed data. In Sect. 5, the procedure to extract the studied regions is given followed by a results description in Sect. 6. The influence of Rossby waves on modelled primary production and the relative contribution of the underlying physical/biogeochemical processes in the oligotrophic gyre are finally discussed in the last section.

A coupled physical/biogeochemical model
The numerical experiments described here are performed using a 3-D coupled physical/biogeochemical model. The ocean physics is solved by the OPA 8.1 model (Madec and Imbard, 1996)

20
N represents dissolved inorganic nitrogen, P phytoplankton, Z zooplankton, D particulate organic matter and DON the dissolved organic nitrogen. Parameters are described in Table 1. Parameter values are deduced from Oschlies and Garçon (1999) and Huret et al. (2005), as well as from a preliminary sensitivity analysis (Charria, 2005). 25 The J (z, t, N) represents the phytoplankton growth function of light and nutrient limi-937 The nutrient limitation follows the Michaelis -Menten formulation: where K N stands for the half-saturation constant for nutrient uptake. The light limited growth rate is based on the analytical method from Evans and Parslow (1985). One of the advantages of this method is that a diurnal cycle from 10 daily solar fluxes is simulated analytically. In fact, the light limited growth rate J(z, t) is averaged over 24 h and over a vertical layer as: where I(z, t) is the local light intensity and α the initial slope of the photosynthesis-light (P-I) curve. Using these Eqs. (8 and 9), the light is maximum at noon at surface (it decreases exponentially with depth). Evans and Parslow (1985) then compute analytically the limitation term integrated over the mixed layer depth and during the day using the following equation:

EGU
where M is the mixed layer depth, k=0.04 m −1 the light attenuation coefficient, τ the time at noon and F (y, t) = (y 2 + t 2 ) 1/2 − t * l n t + (y 2 + t 2 ) 1/2 y (11) G(P) represents the zooplankton grazing, it is written following a Holling type III function: with g, the maximum grazing rate for the high P values and p the G(P ) slope for the weak P values. The advection scheme is MUSCL (Monotonic Upstream centred Scheme for Conservation Laws - Lévy et al., 2001). This scheme is monotonic, positive with an implicit diffusion and a weak dispersion. 10 The interannual simulation has been initiated from Reynaud et al. (1998)  the third year of coupled integration (i.e. 1998). Statistical analyses and comparisons with remotely sensed and in situ data showed that the model reproduces well the seasonal cycle of the ecosystem as well as the primary production in the North Atlantic biogeochemical provinces (Charria et al., 2006b). However a few biases were highlighted in the simulated fields. The Gulf Stream position is too far North as compared

Printer-friendly Version
Interactive Discussion EGU to observations. This is a well known bias in this kind of general circulation model (Barnier et al., 2006). Concerning the biogeochemical fields, the northern boundary of the oligotrophic gyre is located too far South. Indeed, the northern boundary is located around 28 • N in the observations and around 23 • N in the simulated fields. The elevated chlorophyll concentrations in the oligotrophic gyre are linked to high inorganic 5 nitrogen contents due to a fast remineralization loop above the nitracline associated with regenerated production.

Chlorophyll-a concentrations
Chlorophyll-a concentrations (in mgChl m −3 ) were obtained from the ocean colour sen- The products are on a regular grid of 9 by 9 km for the year 1998 (similar to the simulated year). The predicted error on the single 1-km SeaWiFS estimates of chlorophyll-15 a concentration is 35% ; the accuracy of 9 km gridded data is comparable or better. As our study focuses on the anomalies of chlorophyll-a concentrations, a monthly zonal average from raw data at each latitude and for each month is removed, so also removing part of the seasonal cycle. To apply the spectral analysis and filtering described below, gaps in the data, mainly due to the presence of clouds, 20 are filled with a linear interpolation. EGU obtained from the combined processing of the Topex/Poseidon (T/P) and ERS-1/2 data. The two data sets were combined using an improved space/time objective analysis method taking into account long-wavelength-errors (noise correlated on large scales) with a 1-2 cm mean error (Le Traon et al., 1998). SLA are relative to a seven year average (1993)(1994)(1995)(1996)(1997)(1998)(1999) and were mapped every 7 days for the year 1998 with a spatial 5 resolution of 1/3 • . To have the same temporal resolution as for the surface chlorophyll-a concentrations, SLA data were averaged with a monthly time step.

Rossby waves in the coupled model
Rossby wave features in the coupled experiment results (Sea Level Anomalies -SLA -and surface chlorophyll concentrations) are first described and compared to those 10 deduced from remotely sensed data (Charria et al., 2006a). The SLA are analysed to detect the westward propagating signals. SLA are estimated from the temperature and salinity fields and from the barotropic stream function computed with the pressure compensation relation (Mellor and Wang, 1996).
We will consider four main parameters (wavelength, amplitude, phase speed and 15 phase relationship between SLA and surface chlorophyll concentrations) previously used in analysis of Rossby waves features. The wavelengths of westward propagating signals in SLA are estimated using a 1-D continuous wavelet analysis applied for each model latitude between 10 • N and 40 • N. Maxima in the local wavelet power spectra are associated with wavelengths 20 mainly located between 400 and 1000 km with a few occasions over 1000 km. They are representative of the first baroclinic mode of Rossby waves according to the linear and extended theories (Killworth and Blundell, 2003). These values are similar to those estimated from remotely sensed data with Fourier-based spectral analyses (Cipollini et al., 1997;Hill et al., 2000;Killworth et al., 2004), least-square based analysis combined Introduction

EGU
The modelled SLA are filtered reconstructing the signal for the wavelengths located between 400 and 1200 km after a 1-D continuous spatial wavelet analysis (Torrence and Compo, 1998). To estimate the Rossby wave amplitude, the westward propagating signals are extracted (Fig. 2). The simulated waves present amplitudes slightly weaker (reaching 5 cm) than those observed (7-10 cm) in remotely sensed data (Killworth et 5 al., 2004;Charria et al., 2006a) over the North Atlantic Ocean ( Table 2). The lower amplitudes in the simulations are due to the model spatial resolution. Higher resolution would be necessary to simulate the whole Rossby waves energy.
Propagation velocities (or phase speeds) associated with Rossby waves are calculated from the filtered time/longitude diagrams with a 2-D Radon transform method 10 . The modelled velocity values are between 3 and 4 cm s −1 (Table 2). They fall in the same range as those estimated in previous studies (Table 2). However, a narrower interval is observed for the present study partly due to the time period over which the phase speeds are estimated.
Similar analyses, based on the longitude/time diagrams and using the same wavelet 15 filter, are performed on the simulated surface chlorophyll concentration fields. Wavelengths between 400 and 1500 km are observed in agreement with those deduced from the SeaWiFS data for the same year. These wavelengths are similar to the SLA south of 35 • N and are slightly higher north of this latitude. Concerning the amplitudes associated with westward propagations, they are similar in the filtered modelled concen-20 trations and in the filtered remotely sensed concentrations estimated by the SeaWiFS sensor. Note that chlorophyll anomalies are used, not absolute values; areas of high absolute chlorophyll concentrations may not necessarily show large anomalies. Now we will examine the relationships between SLA and surface chlorophyll concentrations. Following Killworth et al.'s (2004) approach, the coupled processes in- 25 volved can be estimated from the relationships between SLA and surface chlorophyll concentrations. As shown by Charria et al. (2006a), the phase relationships between Rossby waves (signal in SLA) and their chlorophyll signature can be highly variable. We compare the phase relationships using a cross-wavelet analysis between filtered EGU SLA and filtered surface chlorophyll concentrations. From 23 • N to 35 • N, the modelled phase relationships (phase(Chlorophyll) -phase(SLA)) are mainly between π/2 and π (Fig. 3a). By contrast, between 16 • N and 23 • N in the simulated fields (Fig. 3a) and between 10 • N and 28 • N in the remotely sensed data (Fig. 3b), values of phase relationship are between −π/2 and 0. According to the theoretical approach from Killworth 5 et al. (2004), we can deduce that between 16 • N and 23 • N in the simulations (between 10 • N and 28 • N in the data), the meridional chlorophyll advection should be the dominant process as confirmed by Charria et al. (2006a). However, the phase relationships deduced from the modelled fields north of 23 • N do not allow to detect a clear dominant process when the meridional chlorophyll gradients are positive due to the same phase 10 relationship for vertical and horizontal processes (Killworth et al., 2004). After the analysis of modelled Rossby waves and chlorophyll signature features and the comparison with previous studies, we can conclude that Rossby waves are well represented in the model despite a southwards shift of the boundary between two different phase regimes due to modelled biogeochemical conditions.

Region selection with significant Rossby wave signatures
To quantify the influence of Rossby waves on primary production, our analysis is focused on the three parts of the wave chlorophyll signature: the positive chlorophyll anomaly linked to the wave's crest (hereafter CA+), the negative anomaly linked to the wave's trough (CA−) and the lack of chlorophyll anomaly (CA0). The accurate identifi-20 cation of these wave patterns in particular regions needs the following pre-processing on the simulated fields.
Before extracting any wave, the previously spatially filtered fields (see section 4) at each time step and each latitude are used to detect clear wave propagations in SLA and surface chlorophyll concentrations. A cross-spectrum analysis between SLA and the 25 decimal logarithm of surface chlorophyll concentrations (similar to the analysis applied on remotely sensed data in Killworth et al., 2004) is then performed on longitude/time 943 EGU windows. Windows are 60 grid points wide (around 20 degrees in longitude). The cross-spectrum energy is integrated over Rossby wave period and wavelength (3-24 months/400-1200 km). Figure 4 shows the amplitude of the integrated cross-spectrum peak. Larger values are observed in the north-west part of the domain where Rossby waves have larger amplitudes. Then, several maxima can be extracted for different lat- scheme doesn't resolve any horizontal diffusion).

Chlorophyll wave crest and primary production
First, the terms for the chlorophyll signature of the wave crest (CA+) are analysed. The influence on primary production is obtained for each region (Fig. 6)  the meridional advection processes (dissolved inorganic nitrogen and phytoplankton). Indeed, the phytoplankton meridional advection and the inorganic nitrogen meridional advection are reaching respectively more than +140% and +78% of the CA0 condition, respectively (Fig. 7a). At the opposite, a strong decrease in primary production (−34.3%) can be noticed in the southernmost latitude (39.5 N-34.3 W) (Fig. 7b). This 20 decreasing production is mainly explained by a strong diminution of the vertical inorganic nitrogen advection and the zonal phytoplankton advection. The CA+ is then explained by a simple meridional chlorophyll advection (Fig. 7b) EGU illustrate the complex variability in the Rossby wave influence on primary production. Furthermore, these two close latitudes are corresponding to very different time periods (winter at 40 N-41.3 W and spring at 39.5 N-34.3 W) associated with two different biogeochemical regimes (bloom in spring). In spring, the primary production associated with the CA+ will then decrease due to the limiting effect of the vertical inorganic ni-5 trogen advection. The combination of these different factors linked to biogeochemical background and Rossby wave features can easily explain such a difference between these two situations. Further south, 8 extracted regions between 29 and 39 N are considered together (38.9 N-42 W/37.1 N-32.7 W/34.7 N-37 W/34.7 N-40 W/32.2 N-31.3 W/31.9 N-10 41.7 W/29.9 N-32.7 W/29.6 N-47.3 W). In this band, primary production is slightly increasing between CA0 and CA+ (+0.2% at 38.9 N-42 W; +12.3% at 37.1 N-32.7 W; +7.2% at 34.7 N-37 W; +2.7% at 32.2 N-31.3 W; +3.1% at 31.9 N-41.7 W and +9.4% at 29.6 N-47.3 W) except for the 34.7 N-40 W and 29.9 N-32.7 W regions where the primary production is decreasing (−8.3% and −10.6% respectively). For these dif-15 ferent cases, the primary production increase is mainly driven by the horizontal and vertical advection of phytoplankton and dissolved inorganic nitrogen. For example, in Fig. 8a (37.1 N-32.7 W case), the CA+ is associated with a strong enhancement of the meridional phytoplankton advection. By contrast, in Fig. 8b, we observe that the primary production diminution is due to decreases in phytoplankton and inorganic ni-20 trogen vertical advection as well as horizontal phytoplankton and inorganic nitrogen advection.
Between 25 N and 26 N, two wave signatures are followed for two very close latitudes (25.8 N-38.7 W and 25.5 N-44.7 W) and with a common initial time but for a longer time period in the case of the easternmost longitude (from 26 March to 10 June at . Surprisingly, the effects of the wave passage on the primary production are completely different. For the easternmost area, we observe a strong decrease in primary production associated with CA+ (−19.9%). At the other location, we found a +20.3% augmentation. The source of the increase is  20.6 N and 18.7 N. First,at 20.6 N, primary production enhancements are observed. There are estimated at +3.4% over a slightly longer time period (26 July to 18 December) and +12.7% over a shorter time period (18 August to 18 December). The vertical diffusion of phyto-5 plankton allows to explain this production increase. The difference between these two values is associated with the integration time period and highlight a negative effect on primary production in the CA+ between mid-July and mid-August 1998 for this latitude. At 18.7 N, no significant primary production trend is observed in the western part (+1.6%) and a −13.9% decrease is obtained in the eastern part. This last decrease is 10 associated with a diminution of the meridional and vertical advection of phytoplankton and dissolved inorganic nitrogen during the wave passage.
Finally in the southern part of the domain (9 N-15 N), the analysed wave signatures are initially tracked east (14.8 N-34.7 W and 10 N-37.3 W) and over (14.8 N-41 W and 11.3 N-42.7 W) the mid-Atlantic ridge. The westernmost waves are observed earlier during the year than the easternmost propagations. In every case, an increase in primary production associated with CA+ is noticed: +15.8% at 14.8 N-34.7 W, +8.4% at 14.8 N-41 W, +18.2% at 11.3 N-42.7 W and +7.3% at 10 N-37.3 W. It appears that wave signatures from the eastern part of the basin observed during autumn have a weaker impact on primary production than those observed in spring and summer. This 20 is in agreement with a more stratified ocean in summer. Indeed, when we detail the processes involved, the vertical input of dissolved inorganic nitrogen explains the production increase (Fig. 9a). By contrast, during the less productive season, the primary production increase associated with the wave signature is due to meridional phytoplankton advection and vertical phytoplankton diffusion (Fig. 9b). There is no significant 25 inorganic nitrogen advective input. 947 EGU 6.2 Chlorophyll wave trough and primary production After analyzing the processes explaining the surface chlorophyll positive anomaly and the impact on the primary production, the opposite situation during the chlorophyll negative anomaly (CA−) is explored. It generally appears in the southern part of the domain that the CA− is associated with a negative effect on primary production (Fig. 6). 5 This first result confirms that the vertical advection of dissolved organic nitrogen is never the only process involved during the wave's passage. In fact, this mechanism inducing a inorganic nitrogen upwelling has an effect on primary production. This effect needs then to be offset by different mechanisms as horizontal advection to explain the decrease in primary production measured as well as the negative surface chlorophyll 10 anomaly.
Except in a few regions, the primary production trend (increase or decrease) in the CA− is then opposite to the situation in the CA+. This opposition is also observed when we consider each mechanism separately.
More surprisingly, we can remark that when the primary production is decreasing 15 over the CA+, the situation in the CA− is opposite and we generally observe an increase in primary production associated with the CA−. This result confirms the important role of meridional advective mechanisms as suggested by Killworth et al. (2004) and Charria et al. (2006a). In fact, in the case of an horizontal advection, the chlorophyll is similar to a passive tracer and only reacts to the physical forcing without significant 20 changes in the biogeochemical processes. These purely physical processes are theoretically symmetrical between the crest and the trough of a Rossby wave. This bipolar system reflects a symmetric effect of the Rossby waves on primary production which suggests that the net effect of Rossby waves on primary production would be very small.

Discussion
In the present work, the different biogeochemical-physical mechanisms explaining the surface chlorophyll signature of Rossby waves are investigated through the exploration of the effect of these waves on primary production. The use of a realistic coupled physical/biogeochemical model in the North Atlantic allows to detail separately the positive 5 and negative surface chlorophyll anomalies associated with these waves. The biogeochemical advective and diffusive fluxes analysed for 20 areas highlight the fact that the increase of phytoplankton biomass in the CA+ is due to a complex combination of vertical and meridional processes depending on the considered latitude. This surface chlorophyll positive anomaly is associated with various patterns of 10 primary production depending upon season and region (latitude and longitude). Previous studies with remotely sensed data and theoretical modelling from Killworth et al. (2004) and Charria et al. (2006a) showed that the Rossby wave signature in chlorophyll concentration can be explained by the meridional chlorophyll advection mostly south of 28 N. When we compare these results to the present work (Fig. 10), the role of the 15 horizontal advection is also noticed but several regions highlight also a strong effect of the dissolved inorganic nitrogen vertical advection. Based on the 20 present regions analysed, the different relative contributions north and south of 28 N (as described in Charria et al., 2006a) are not fully reproduced. This can be partly explained by the limited time period and the specific location of the selected regions. Furthermore, several 20 assumptions were made to compare remotely sensed data and theoretical modelling in Killworth et al. (2004) and Charria et al. (2006a). Indeed, a simple term was used to represent the source and sinks biogeochemical terms in the theoretical model and several statistical assumptions were made to fit the model outputs with observations. However, the present results of the realistic simulations confirm the weak effect of the 25 vertical advection of phytoplankton, apparent in the whole domain (Fig. 10).
The second part of this work addresses the local impact of Rossby waves on primary production following the time period and location. When we quantify the local enhance-949 EGU ment of primary production due to the wave passage, values between −34.3% (latitude 39.5 N-34.3 W) and +20.3% (latitude 25.5 N-44.7 W)  (latitude 37.1 N-32.7 W) are obtained. These values represent the difference between CA+ and CA0 primary productions (divided by the CA0 primary production for the per-5 centages). It appears that the increases are stronger when the dissolved inorganic nitrogen vertical advection is involved (i.e. between 9 N and 15 N). Furthermore, this last process seems to be systematically involved in spring and summer when the increases of primary production are exceeding 5%. By contrast, decreases in primary production associated with the CA+ are generally due to a decrease of the vertical 10 advection and diffusion of dissolved inorganic nitrogen. For the other situations, the primary production decreases associated with CA+ can be explained by a contribution of the meridional advection of phytoplankton. The different dominant processes explaining primary production decrease and the CA+ signature are probably the result of a strong stratification in summer which will induce a de-correlation of the processes in 15 the mixed layer and below. Kawamiya and Oschlies (2001), with the same kind of coupled physical/biogeochemical model applied in the Indian Ocean, showed that Rossby waves can induce a ∼30% increase of primary production associated with the uplifting of the deep chlorophyll maximum during the wave passage. This value obtained with a different approach in a different basin is comparable to our estimations. In the Pacific, 20 Sakamoto et al. (2004), based on local measurements at the ALOHA site and on an estimated relationship between primary production and Sea Surface Height anomalies estimated near the HOT station ALOHA, showed similar increases in primary production reaching 25%. They were linked to nitrate injection in the euphotic layer during two well identified passages of Rossby waves. Using the same relationship, these au- 25 thors conclude a limited role of Rossby waves inducing a mean primary production enhancement lower than 5-10 % over the 1997-1999 period. These values are also comparable to our estimations over the selected transects in the North Atlantic Ocean. Then, in our study similar increases are obtained using a different approach in a 950 EGU different ocean but decreases are also simulated. Consequently, our results suggest a net weak effect of Rossby waves on primary production. These estimations need however to be considered carefully. These results using a realistic ocean general circulation model in the North Atlantic are obtained on given latitudes mainly located in low production provinces for a given time period. Furthermore, 5 the method used to identify wave propagation in surface chlorophyll concentrations is based on several assumptions necessary to perform this first detailed analysis of the coupled physical/biogeochemical processes involved during the Rossby wave propagation. In particular, a recent study from Chelton et al. (2007) suggests that most westward propagating features observed in remotely sensed sea surface height are 10 the signature of non linear propagating eddies. In the North Atlantic, more than 25% of the variance is explained by eddies instead of Rossby waves west of 53 W and north of 32 N from the western to eastern coasts. These observations are based on the Okubo-Weiss parameter which is a measure of the relative importance of deformation and rotation (see Fig. 3 in Chelton et al., 2007). In our studied domain, 14 out of 20 15 regions are located south of 32 N and east of 53 W which is a region where non linear eddies are not predominant. North of 32 N, our approach does not allow to certify that Rossby wave signature is analysed instead of eddy signature. However, Sweeney et al. (2003) and Mouriño-Carballido and McGillicuddy (2006) showed that in the case of eddies, the primary production associated with the negative chlorophyll anomaly will EGU tions. The analyses based on 20 regions showed the important contribution of horizontal advection and of vertical inorganic dissolved nitrogen advection. The main mechanism involved differs according to the biogeochemical conditions (due to latitude and time period). Furthermore, Rossby waves have a non negligible influence on primary production which can be quantified using coupled 3-D modelling. Indeed, 5 positive surface chlorophyll anomalies are generally associated with an increase in primary production. This approach also allowed to explore the influence of wave trough on primary production and a negative effect was observed. The net impact of Rossby wave on primary production, based on the 20 regions analysed, seems to be weak due to the symmetrical shape of the wave effects (between crest and trough). How-10 ever, as suggested by Sakamoto et al. (2004), inputs of dissolved inorganic nitrogen can induce shifts in the phytoplankton community structure and consequently increase the exported production. This approach using different particular cases needs to be extended to the whole basin in order to quantify the net basin scale effect of Rossby waves on primary production. To perform this estimation, we need to investigate what 15 happens when the Rossby wave effect is removed. This calls for a more systematic study with coupled physical/biogeochemical models where Rossby waves can be switched on and off.