ENSO-driven CO2 efflux variability and the role of the upwelling region on the carbon exchange in the Northern Humboldt Current System

Journal Pre-proof ENSO-driven CO2 efflux variability and the role of the upwelling region on the carbon exchange in the Northern Humboldt Current System

Rodrigo Mogollón PII:

S0924-7963(19)30377-X

DOI:

https://doi.org/10.1016/j.jmarsys.2019.103240

Reference:

MARSYS 103240

To appear in:

Journal of Marine Systems

Received date:

11 February 2019

Revised date:

5 September 2019

Accepted date:

20 September 2019

Please cite this article as: R. Mogollón, ENSO-driven CO2 efflux variability and the role of the upwelling region on the carbon exchange in the Northern Humboldt Current System, Journal of Marine Systems(2019), https://doi.org/10.1016/j.jmarsys.2019.103240

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier.

Journal Pre-proof

ENSO-driven CO2 efflux variability and the role of the upwelling region on the carbon exchange in the Northern Humboldt Current System Rodrigo Mogoll´on∗

ro

of

Laborat´ orio de Dinˆ amica e Modelagem Oceˆ anica (DinaMO), Universidade Federal do Rio Grande - FURG, RS, Brazil

Abstract

Jo

ur

na

lP

re

-p

The role played by El Ni˜ no-Southern Oscillation (ENSO) on modulating the oceanic carbon outflux in the Northern Humboldt Current System (NHCS) is assessed with the use of outputs from a coupled physical-biogeochemical hindcast which permits to evaluate the interannual variability of the surface seawater partial pressure of carbon dioxide, hereinafter pCOsw 2 , and reconstruct associated air-sea CO 2 fluxes from 1998 through 2015 at monthly timescales. In addition, surface dissolved inorganic carbon (DIC) concentrations were inferred from a recent neural network methodology prescribed with the same outputs of the biogeochemistry hindcast. Results show a large no and La Ni˜ na spatiotemporal variability of the CO 2 exchange to nine El Ni˜ episodes throughout the period of analysis. Within the coastal and equatorial upwelling region, that variability results from combined ENSO-driven air ΔpCO2 (pCOsw 2 - pCO2 ) and wind speed variations, although the latter in a minor proportion during ENSO peaks. Specifically, it is found that the weak CO2 source behavior during an average El Ni˜ no episode is mainly caused by a weakening of ΔpCO2 which is partially compensated by more efficient CO2 exchange at the air-sea interface due to the strengthening of the upwelling-favorable winds. Conversely, the relatively strong CO 2 source behavior during an average La Ni˜ na episode results from more efficient upwelling which brings colder but CO 2 -rich waters to the surface, therefore, increasing pCOsw 2 and the associated CO 2 efflux. Moreover, it is estimated ∗

Corresponding author Email address: [email protected] (Rodrigo Mogoll´ on)

Preprint submitted to Journal of Marine Systems

October 21, 2019

Journal Pre-proof

that during an average La Ni˜ na episode, 3.3 million metric tons of carbon is additionally emitted contributing to the atmospheric CO 2 accumulation. In contrast, during an average El Ni˜ no episode, the total amount of carbon retained is about 24.7 million metric tons of carbon that normally would have been lost to the atmosphere as CO 2 , rendering the NHCS a key region for the global atmospheric carbon budget.

of

Keywords: El Ni˜ no-Southern Oscillation, Air-sea CO 2 fluxes, Humboldt Current System, CO2 -carbonate system, Interannual variability 1. Introduction

Jo

ur

na

lP

re

-p

ro

It is known that El Ni˜ no-Southern Oscillation (ENSO) induces strong effects on the upwelling intensity and associated nutrient-rich source waters, disturbing the Eastern Pacific marine ecosystem from a physical, biogeochemical and ecological perspective (Huyer et al. 1987; Chavez et al. 1999; Colas et al. 2008; Espinoza-Morriber´on et al. 2017). As such, Eastern Boundary Upwelling Systems (EBUS) are particularly impacted by basin-scale modes of climate variability at interannual and also decadal scales. Specifically, the proximity to the Equator and direct subsurface connections between the Humboldt EBUS and the Equatorial Current System (ECS) (Montes et al. 2011), renders the Northern Humboldt Current System (hereinafter NHCS) extremely sensitive to equatorial oceanic perturbations. The warm phase of ENSO, El Ni˜ no, is commonly associated with several negative environmental disturbances and adverse effects on nutrients supply and primary production rates (Barber and Kogelschatz 1990), rendering the Peruvian coast a marine ecosystem characterized by transient unfavorable conditions in terms ˜ of fishery productivity ( Niquen and Bouchon 2004). However, El Ni˜ no may be considered as a favorable event in terms of Earth’s climate. For instance, as evaporation requires large amounts of thermal energy, an El Ni˜ no episode tends to cool the ocean due to larger rates of evaporation that are induced in the Eastern Pacific (Trenberth et al. 2002). Moreover, Bakun and Weeks (2008) suggest that El Ni˜ no re-starts the Peruvian marine ecosystem from an initial highly productive state, by interrupting any longer-lived nonlinear adverse feedback that may negatively affect primary production. Additionally, the NHCS goes through an oxygenation phase during El Ni˜ no (Montes et al. 2014; Espinoza Morriber´on et al. 2018) modulating the spatiotemporal variability of the oxygen minimum zone. As a consequence, the 2

Journal Pre-proof

Jo

ur

na

lP

re

-p

ro

of

physical ventilation that suppress suboxic processes during El Ni˜ no peaks, would lead to a substantial reduction of coastal outgassing of nitrous oxide gas (N2 O) to the atmosphere (Mogoll´on and Calil 2017), which is a wellknown greenhouse gas with a warming potential 300 times larger than CO 2 . Together with a N2 O outflux decrease, a significant reduction of the average CO2 outgassing, i.e. an anomaly ocean CO 2 uptake response, during El Ni˜ no conditions is well documented in several studies (Inoue and Sugimura 1992; Feely et al. 1995; Rayner et al. 1999; Jones et al. 2001; McKinley et al. 2004; Doney et al. 2009; R¨odenbeck et al. 2014; Brady et al. 2019). In spite of the fact that the lack of consistent low-frequency pCOsw 2 measurements and the paucity of spatially comprehensive observations do not allow a robust characterization of the long-term variability in the coastal region of the NHCS, limited available observations indicate that the CO 2 source behavior is significantly influenced by the ENSO phenomenon. However, the average effect has not been well quantified because the majority of studies focus on the Equatorial/Tropical Pacific oceanic region. For instance, Feely et al. (1997) found lower values of ΔpCO2 in the central and eastern equatorial region during 1991-1994 ENSO period. Similarly, Feely et al. (1999) showed a reduction of the regional air-sea CO 2 outflux in the equatorial Pacific region during the 1991-1994 El Ni˜ no period. Chavez et al. (1999) also found that during the 1997-1998 El Ni˜ no, the equatorial Pacific Ocean retained 0.7 Pg C. R¨odenbeck et al. (2014) demonstrated a reduced CO 2 outgassing during El Ni˜ no events in the Tropical Pacific, and its magnitude depends on the relative contributions of the physical-biogeochemical caused responses. Studies undertaken in other EBUS, such as the California Current System (Chavez et al. 1999; Friederich et al. 2002), demonstrated a switching from sink to source during ENSO events. However, the Humboldt EBUS does not appear to shift to a CO 2 sink nature in response to ENSO-driven perturbations. In this study, the contribution of changes in wind speed, CO 2 content, and the cross-correlations between them to the interannual air-sea CO 2 flux anomalies are diagnosed. Moreover, the first attempt to estimate how much carbon, as CO2 , is retained or released throughout an average ENSO episode within the productive coastal and adjacent equatorial upwelling region is carried out. A better understanding of how the carbon cycle responds to ENSO may provide valuable insight into how the NHCS may respond to future climate change since it resembles an El Ni˜ no-like warming pattern (Meehl and Washington 1996).

3

Journal Pre-proof

This work is organized as follows: Section 2 describes the methodology, the products used, the computation of the air-sea CO 2 flux and the approach utilized to assess the sensitivity of the forcing factors behind its variability. The criteria used for ENSO events are presented as well. Section 3 shows the results. The main findings of the work are summarized in Section 4. Additionally, a series of appendices is provided in order to support the methods and results.

of

2. Methodology

lP

re

-p

ro

2.1. Domain and period of analysis The domain of study spans the region between 70–92 ◦ W and from 2◦ N to ◦ 22 S (see Fig.1). The horizontal resolution is 1/4 ◦ . The period of analysis is 18 years, from 1998 to 2015, at monthly timescales. The domain of analysis (coastal + equatorial), namely the upwelling region, is defined from the coast to the isoline of 0.2 mg m −3 of the annual mean observed surface chlorophyll derived from the European Space Agency Ocean Colour Climate Change Initiative (ESA OC-CCI Version 3.1) (http://www.esa-oceancolour-cci.org/). The upwelling region encompass both the adjacent ETP along the equatorial band, and the eutrophic zones (values higher than 0.2 mg m −3 ), thus, excluding the oligotrophic oceanic region.

Jo

ur

na

2.2. Global ocean biogeochemistry hindcast The product GLOBAL REANALYSIS BIO 001 029 is a global biogeochemical simulation produced at Mercator-Ocean (Toulouse, FR) provided by the Copernicus Marine Environment Monitoring Service (CMEMS). It uses the Nucleus for European Modelling of the Ocean (NEMO version-3.6stable) hydrodynamical model (Madec et al. 2015) and the Pelagic Interactions Scheme for Carbon and Ecosystem Studies (PISCES v2) biogeochemical model (Aumont et al. 2015). The outputs of the hindcast experiment provides global ocean 3D fields for the period 1993-2018 at 1/4 ◦ horizontal resolution, encompassing the 1998-2015 period of analysis. The biogeochemistry hindcast starts at rest in December 1991 (the first two years of simulations are considered as a spin-up phase) and it is initialized with World Ocean Atlas (WOA) 2013 for nitrate, phosphate, oxygen and silicate (Garcia et al. 2013, 2014). Similarly, GLODAP v2 climatology (Key et al. 2015) is used for Dissolved Inorganic Carbon (DIC) and Alkalinity (Alk), and model ouputs from a climatological run, for dissolved iron and dissolved organic carbon. 4

Journal Pre-proof

0

Ecuador Galápagos Islands

-5 Coastline 0.2 mg Chla m -3

of Upwelling region

ro

-10

Perú

-p

Latitude

2.0 mg Chla m -3

re

-15

lP

Oligotrophic region

-85

-80

-75

-70

Longitude

ur

-90

na

-20

Jo

Figure 1: Domain of study. The upwelling region is defined from the coast to the isoline of 0.2 mg Chla m−3 (dashed black line), comprising the coastal and equatorial eutrophic zone. Dot-dashed red line depicts the highly productive coastal area (> 2 mg Chla m−3 ).

Monthly atmospheric pCO2 (global values) was prescribed at the air-sea interface. The physical forcings come from the numerical simulation of a global ocean reanalysis, namely GLORYS2V4 (Garric et al. 2017). A comprehensive assessment of the quality/validation of the global biogeochemical hindcast may be found at the quality information document (Coralie Perruche and collaborators) published on the CMEMS dissemination server. See also Appendix A.

5

Journal Pre-proof

ur

na

lP

re

-p

ro

of

2.3. Inferred dissolved inorganic carbon Since DIC is unavailable online, the CArbonate system and Nutrients concentration from hYdrological properties and Oxygen using a Neural-network (CANYON-B) (Bittig et al. 2018) was implemented, which is a re-develop approach of Sauz`ede et al. (2017). CANYON-B is a Bayesian neural network mapping that accurately reproduces GLODAP v2 bottle data and the biogeochemical relations contained therein. It thus provides mappings from one set of inputs variables (temperature, salinity, oxygen, pressure, location, and time) to another set of variables such as the ones associated with the CO 2 carbonate system. Furthermore, CONTENT routines (Bittig et al. 2018) are then applied in order to refine the CO 2 -carbonate system variables to be consistent with the carbonate chemistry. The chose of CANYON-B/CONTENT also relies on the advantage of being a “dynamic climatology” since it uses a wider range of dynamic predictors, as previously described, which allows it to react more flexibly to different water mass characteristics and reflect the influence of surface driving mechanisms that may affect DIC concentrations, particularly under ENSO phenomena, that would not be assessed with the use of a classical climatological mapping approach. For a more detailed description of the method and a robust assessment of the validation globally, the reader is referred to (Bittig et al. 2018). CANYON-B/CONTENT is prescribed, for any grid point at the surface throughout the entire period of analysis, with outputs of sea surface temperature (SST), sea surface salinity (SSS) and sea surface oxygen from the biogeochemistry hindcast (section 2.2). An additional evaluation on an annual basis for both DIC and Alk is also provided in Appendix D.

Jo

2.4. Air-sea CO2 flux computation The CO2 flux at the air-sea interface (F CO2 ) is the product of three quantities, calculated as: air FCO2 = K0 kw ΔpCO2 = K0 kw (pCOsw 2 − pCO2 )

(1)

where atmospheric pCO2 , or pCOair 2 , was extracted from www.esrl.noaa.gov/gmd/ccgg/trends (see Appendix A). K0 is the CO2 solubility computed using the formulation of Weiss (1974) which depends on temperature and salinity. kw is the gas transfer velocity and is proportional to the square of the wind speed following the revisited formulation of Wanninkhof (2014), which is appropriated for regional estimates of CO 2 fluxes when using the Cross Calibrated Multi Platform (CCMP) V.2 (Wentz et al. 2015) wind product (monthly-means at 1 /4◦ 6

Journal Pre-proof

horizontal resolution). This new formulation provides both new estimates of the Schmidt number and an updated coefficient for the gas exchange-wind speed relationship. The convention is that positive values of F CO2 denote an outgassing of CO2 , i.e. from the ocean to the atmosphere, while negative values indicate an uptake by the surface ocean, i.e. from the atmosphere to the ocean.

-p

ro

of

2.5. Sensitivity of air-sea CO2 flux anomalies To diagnose the individual contributions of the forcing factors governing interannual physical-biogeochemical variability of F CO2 (Equation 1), a Reynold’s decomposition was perfomed following the same methodology of Doney et al. (2009). Two forcing factors are considered, i.e. (K 0 kw ) and (ΔpCO2 ), and the corresponding linear decomposition terms of the air-sea CO2 flux anomalies are: F0CO2 = (K0 kw )0 × ΔpCO2

(2)

lP

re

+ (K0 kw ) × ΔpCO2 0 h i 0 0 0 0 + (K0 kw ) × ΔpCO2 − (K0 kw ) × ΔpCO2

Jo

ur

na

(see Appendix B). The anomalies ( 0 ) are either spatial or temporal, relative to a domain or a monthly mean (overline bar), respectively. The first term in the right-hand side of Equation 2 reflects the contribution to the variability of the air-sea CO2 flux from gas transfer and solubility anomalies. However, (K0 kw )0 basically reflects variations in wind speed. This is because the temperature dependence of both K 0 and kw approximately cancel each other out (increased SST will reduce the solubility of CO 2 favoring the outgassing, but increased temperatures will decrease the transfer velocity and so reduce the fluxes). Hereinafter, this term is referred to as the gas transfer velocity term for short. The second term reflects the variability induced by ΔpCO2 anomalies, where ΔpCO02 is primarily driven by changes in pCOsw 2 in response to the variability of its drivers, i.e. SST, SSS, DIC, and Alk. Finally, the third term reflects the simultaneous cross-variations of both forcing factors, i.e. CO2 content and wind speed anomalies, corrected for its mean value. 2.6. ENSO criteria The Coastal El Ni˜ no Index (ICEN) (Takahashi et al. 2014) is used, which is a 3-month running mean of SST anomalies for the El Ni˜ no 1+2 region 7

Journal Pre-proof

lP

re

-p

ro

of

available at http://www.met.igp.gob.pe/variabclim/indices.html (last assessed July 15). Nine ENSO episodes were found, disregarding the events that last less than 3 months. The months while the ICEN index is larger than +1.7 and smaller than -1.4 are considered in order to better represent El Ni˜ no and La Ni˜ na events, respectively. By doing so, only strong/extraordinary El Ni˜ no and strong La Ni˜ na events are taking into consideration for the composites. The monthly anomalies, for any variable, were constructed by removing the monthly mean climatology computed for the entire period of analysis (i.e. mean January, mean February, etc.) at each grid point. This work considers the average El Ni˜ no episode as the temporal mean of the monthly composite of the warm phase of ENSO. Similarly, the average La Ni˜ na episode is the temporal mean of the monthly composite of the cold phase of ENSO, for any variable. El Ni˜ no episode typically last 9-12 months, which rarely persist more than a year at a time. La Ni˜ na event could even last longer than El Ni˜ no. This is in part because the normal conditions, or “ENSO neutral conditions”, in the NHCS resemble more with La Ni˜ na conditions than El Ni˜ no ones. Throughout the analysis period, the average El Ni˜ no episode lasts 10 months and the average La Ni˜ na episode lasts 14 months. Consequently, it is reasonable to consider an average ENSO phase duration of, approximately, 12 months.

na

3. Results and discussions

Jo

ur

3.1. Model evaluation In order to evaluate the simulated CO 2 -carbonate system, comparison between modeled and observed pCOsw 2 are used as a proxy. Fig.2 only shows the second half of the period of analysis, from 2007 to 2015, because observations (almost a quarter of the total valid points: 120) have been sparse and spatially isolated from 1998 to 2006. The database used is the Surface Ocean CO2 Atlas database, SOCAT V.6 (Bakker et al. 2016). For consistency, the observed fugacity of CO 2 from SOCAT was converted to pCO2 using the CO2SYS program (Van Heuven et al. 2011), and only the grid boxes containing observations were considered in order to perform the spatial means to plot the time series. A set of quantitative metrics comparing modeled pCOsw 2 with observational data was used. These statistical metrics are the correlation coefficient (r), the root mean square error (RMSE), the reliability index (RI), and the average error or bias (AE). The correlation coefficient measures the tendency of modeled and observed values to vary together. 8

Journal Pre-proof

a)

b) 500

-5

CO2 [ atm]

0 Observation

-10 -15 -20

450

-5

400

-10

350

-15

-70

-90

c)

ro

520

-80

Observation Model

-p

480

-70

of

-80

Model

-20

300

-90

440 400

re

CO 2 [ atm]

0

2008

2009

2010

2011

2012

2013

2014

2015

Years

na

320 2007

lP

360

Jo

ur

Figure 2: Annual mean spatial map of a) observed (SOCAT V6) and b) modeled pCOsw 2 . c) Time series of both observed (with their respective error bars of unit standard deviation) throughout the second half of the analysis period. These values and modeled pCOsw 2 are the spatial average into the whole domain of study considering only the grid pixels containing observations.

Ideally, this value will be close to one. RMSE and AE are measures of the size of the discrepancies between modeled and observed pCOsw 2 , and values near zero indicate a close match. RI quantifies the average factor by which modeled values differ from observations. A good prediction yields RI values close to one. For a more detailed explanation of these quantitative metrics, the reader is referred to Stow et al. (2009). Results (r = 0.80, RI = 1.01, RMSE = 13.8, and AE = -3.5) show that modeled pCOsw 2 are highly correlated with observations, exhibit low prediction errors, oscillates at the right amplitude, and underestimate observations by 3.5 μatm, respectively (see 9

Journal Pre-proof

of

Fig.2c). The model also captures the southeast–northwestward progression of high pCOsw 2 along the coastal and equatorial upwelling regions as shown in Fig.2a,b. An EOF analysis was performed (see Appendix E). Results show that the first mode of variability explains 63% of the total variance, and its respective principal component is highly correlated with ICEN (r = 0.75). Furthermore, a periodicity of 3.5 years was detected in the associated power spectrum, which is unequivocally related to the ENSO’s signal. Therefore, the mechanisms linking ENSO and the variability of the CO 2 efflux could be assessed. SST and chlorophyll are also evaluated during ENSO peaks in Appendix C.

Jo

ur

na

lP

re

-p

ro

3.2. ENSO composites Fig.3 shows the anomalies, with respect to the monthly climatological mean, of SST (panels a-c), wind velocity (panels d-f) and DIC (panels g-i) under an average El Ni˜ no and La Ni˜ na episode, as clearly indicated. Positive SST anomalies indicate warming and negative ones, a cooling (see Fig.3ac). This is well known due to advection of warm waters from the ECS toward the Peruvian coast during El Ni˜ no events. Less well known are the intensified alongshore southeasterly winds off Peru during an average El Ni˜ no episode, as seen in Fig.3d-f. This is paradoxical due to the expected weakening of the large-scale Pacific Trade winds during the warm phase of ENSO (Oort and Yienger 1996). One reasonable explanation is related to the Bakun’s upwelling intensification hypothesis (Bakun 1990), which stated that the increase in anthropogenic greenhouse gases, such as CO 2 or water vapor (ETP becomes moister during El Ni˜ no events), would lead to the increase of the land-sea thermal gradient along the Peruvian coast, thus, inducing a strengthening of the upwelling-favorable local winds (Bakun and Weeks 2008; Bakun et al. 2010). However, recent studies suggest that it is not likely to occur in the NHCS. For instance, Belmadani et al. (2014) found that a 40% increase of the land-sea thermal gradient represents a 10% alongshore wind decay, also associated with a poleward displacement of the South Pacific Anticyclone (SPA), in a climate change context. Chamorro et al. (2018) found that the increase in humidity did not support the Bakun’s strengthened land-sea thermal contrast. Instead, they suggest that the coastal wind intensification during El Ni˜ no 1997-1998 was mainly driven by the enhancement of the SST-driven alongshore pressure gradient. While the mechanisms behind the wind intensification are still open to debate, is clear that during an average El Ni˜ no episode the wind speed in the highly productive region of 10

Journal Pre-proof

of

the NHCS increases more than 1 ms −1 with an upwelling-favorable direction (Fig.3e), in agreement with Fig.1 in Chamorro et al. (2018), which also shows positive alongshore (equatorward) wind anomalies southern 5 ◦ S. In contrast, the wind anomalies during an average La Ni˜ na episode (Fig.3f) exhibit an opposite spatial pattern than during El Ni˜ no. This is evidenced by positive anomalies in the southern oceanic domain associated with the strengthening of the SPA gyre. Conversely, negative wind anomalies (decrease in the wind speed with an upwelling-unfavorable direction) in the central NHCS are found. Overall, the weakening of the winds during La Ni˜ na is about half the strengthening during El Ni˜ no in the coastal area.

24 22

-10

-10

-15

-15

b)

-5

-20

-20 -90

8 m/s

0

7

d)

-5 6

-15 -20

-2

-15 -20 -90

-80

-70

1

2 m/s

1 m/s 0

0.5

f)

0

-10

0

-15

-0.5

-20

-5 -10

-10

1950

-15

-15

-20

-20

Jo -90

-80

-80

-70

-90

-80

-70

40

0

2000

g)

-5

-1

-90

Anomaly [ mol kg ]

ur

-70

-5 -10

-20

-80

2050

1900

c)

0

-70

e)

-5

0

2

-4

-80

-15

-90

2100

0

La Niña - Climatology

4

0

-1

4

-90

-10

-10 5

3

DIC [ mol kg-1]

-70

na

Wind velocity [m s -1]

8

-80

lP

16

re

18

-1

20

Anomaly [°C]

0

a)

Anomaly [m s ]

0 -5

-p

SST [°C]

26

El Niño - Climatology

ro

Climatology 28

h)

-5

20

i)

-5 -10

0

-15

-20

-20 -40

-70

-90

-80

-70

-90

-80

-70

Figure 3: a) Annual mean simulated SST (climatology). b) SST anomaly during an average warm phase of ENSO (El Ni˜ no minus climatology). c) SST anomaly during an average cold phase of ENSO (La Ni˜ na minus climatology). Similarly, panels d-f and g-i are for wind velocity (CCMP V2) and inferred DIC (hindcast+CANYON-B/CONTENT), respectively.

11

Journal Pre-proof

Jo

ur

na

lP

re

-p

ro

of

Mogoll´on and Calil (2018) demonstrated that DIC and SST, associated with the physical circulation and CO 2 solubility mechanisms, are the most important drivers of pCOsw 2 in the NHCS. As a consequence, anomalies of DIC (Fig.3g-i) are highly related with changes in the solubility pump. This is partially in agreement with Archer et al. (1996), who found that biological changes are not responsible for changing pCOsw 2 values in the equatosw rial Pacific. Therefore, pCO2 changes could be explained by both, anomalies in upwelled/downwelled DIC together with the thermal solubility effect, rather than variations caused by biology, particulary during transient ENSO events. As the relationship between DIC and temperature is relatively linear (Sarmiento and Gruber 2006), a warming during El Ni˜ no would represent a decrease in surface DIC concentrations together with a decrease of CO2 solubility. However, the effect of warming in increase pCOsw 2 has been demonstrated to be up to three times smaller than the effect of DIC in deon and Calil crease pCOsw 2 , at least under climatological conditions (Mogoll´ 2018). This decrease of DIC during El Ni˜ no may also be related to remotelygenerated forcings, such as the Intraseasonal Equatorial Kelvin Waves and the subsequent poleward propagation of the downwelling-favorable coastaltrapped waves (CTW) which deepen the thermocline (Echevin et al. 2014), rendering the upwelled water warm and and CO 2 -depleted. A vertical displacement of the pycnocline and nutricline is also expected, modulating the vertical supply of nutrients and carbon above the euphotic layer during the passage of the CTW. In contrast, during La Ni˜ na, positive DIC anomalies are found in response to a more effective upwelling, also associated with an increase of CO2 solubility (cooling). Despite the relatively slight decrease of the upwelling-favorable winds during La Ni˜ na, upwelled waters carried by the Equatorial Undercurrent and the Peru-chile Undercurrent are known to be nutrient-repleted and the position and intensity of these major upwelling sources waters are much more favorable when compared to El Ni˜ no conditions (Colas et al. 2008; Montes et al. 2010, 2011; Espinoza-Morriber´on et al. 2017). Thus, large rates of vertical transport of inorganic carbon to the surface may explain positive DIC anomalies found under La Ni˜ na conditions. These considerations, while speculative, provide at least some degree of explanation of ENSO-driven surface DIC anomalies. Fig.4 shows the ENSO-driven anomalies of all the terms involved in the computation of the air-sea CO 2 fluxes. Larger anomalies are obtained during El Ni˜ no than La Ni˜ na phases when compared to the ENSO-neutral phase 12

-20

-20 -90

-80

-10 -20

0 j)

28

-10

-10

-20

-20 -90

-80

-80

-70

]

-1

[cm h

-70

k)

30

-10

0

-20 -90

-80

-70

-90

-80

0

2

i) -10

0

-2 -4

-20 -90

-80

-70

4

0

-70

2

l) -10

0 -2 -4

-20 -90

-80

-70

lP

K 0 [mol m

32

-90

-p

-1

0

34

-3

atm

36

-70

re

K

-20 -80

anomaly

h) -10

-90

0

4

g)

-20

-70

f)

-100

-70

0

-10

-80

of

-80

]

0

-90

ro

] -1

[cm h

-70

w

14 12 10 8 6 4 2

-2

anomaly [ atm]

e)

0

-80

-20 -90

100

d)

-20

-10

0

-70

0

-10

0

-2

-90

atm

CO 2 [ atm]

-70

0

100

]

CO 2 transfer velocity

-80

La Niña - Climatology c)

[mol m

2

-2

-90

solubility

y -1 ]

-10

0

-100

2

b)

anomaly

a) -10

2

-1

F CO [mol C m

2

200

CO

El Niño - Climatology 0

[mol C m

Climatology 0

-3

-2

4

anomaly

y -1 ]

Journal Pre-proof

na

Figure 4: a) Annual mean air-sea CO 2 flux (climatology). b) Air-sea CO 2 flux anomaly during an average warm phase of ENSO (El Ni˜ no minus climatology). c) Air-sea CO 2 flux anomaly during an average cold phase of ENSO (La Ni˜ na minus climatology). Similarly, panels d-f, g-i, and j-l are for ΔpCO2 , CO2 transfer velocity, and CO 2 solubility, respectively.

Jo

ur

(climatology), i.e. the effect of ENSO on the CO 2 efflux associated components becomes more severe during El Ni˜ no than La Ni˜ na events. Thus, there is exist an asymmetric response favoring the warm ENSO phase, probably because the La Ni˜ na phase resembles more with normal conditions. ΔpCO2 and CO2 solubility decrease by an amount of, approximately, 35 μatm and 1.5 mol m−3 atm−1 , respectively, during an average El Ni˜ no episode. In addition, CO2 transfer velocity increases by an amount of, approximately, 1 no phase in response to increased winds, as previously cm h−1 during El Ni˜ demonstrated. In contrast, during an average La Ni˜ na episode, anomalies are mirrored (as seen in Fig.4) since tend to oppose El Ni˜ no spatial patterns, albeit at a lower magnitude.

13

Journal Pre-proof

3.3. Forcing factors interannual variability

mol C m -2 year -1

3 2 1

0 -5 Upwelling

region

a)

0

0

0

-5

-5

-5

b)

c)

0

-10

-10

-10

-10

-1

-15

-15

-15

-15

-2

-20

-3

-20 -80

-70

-20 -90

-80

-70

-20 -90

-80

-70

-90

-80

-70

of

0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2

ICEN

15

14

20

13

20

20

12

11

20

20

10 20

09

08 20

06

07 20

05

20

04

lP

03

4 3 2 1 0 -1 -2

La Niña

years

na

20

02

20

01

20

00

20

99

El Niño

20

19

19

98

f)

20

re

-p

ro

e)

20

mol C m -2 year -1

19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15

-90

d)

Jo

ur

Figure 5: Sensitivity of the forcing factors of air-sea CO 2 flux (FCO2 ) anomalies. a) root-mean-square of FCO2 anomalies. Solid black line denotes the upwelling region. Contribution to the variability of F CO2 from b) gas transfer anomalies, c) ΔpCO2 anomalies, and from d) gas transfer velocity and ΔpCO2 anomalies cross-term. Dashed black line in panels b-d depicts the zero value. e) Temporal variability of the forcing factors computed within the upwelling region. f) ENSO index: ICEN.

Large values of the root-mean-square (rms) of air-sea CO 2 flux anomalies are found in the central domain within the upwelling region (onshoreward from the solid black line in Fig.5a), which is considered the major region of variability. Following the methodology described in Section 2.5, Fig.5b-d show the forcing factors driving the interannual variability of F CO2 . As seen, FCO2 variability (Fig.5a) is dominated by ΔpCO2 -driven anomalies (Fig.5c), gas transfer velocity-driven anomalies (Fig.5b), and their cross-correlations (Fig.5d), in this order of importance. The effect of the gas transfer velocity on driving air-sea CO2 flux variations results to be unusually high in the NHCS. 14

Journal Pre-proof

This is due to ENSO-induced large wind speed variability (Section 3.2), together with a large ΔpCO2 mean of, approximately, 70 μatm in the upwelling region (see Fig.4d). This is in agreement with a global study (Doney et al. 2009) which found maxima rms values of F 0CO2 in the Tropical Pacific, where a noticeable contribution from gas transfer variability also occurs.

Jo

ur

na

lP

re

-p

ro

of

Fig.5e shows the detrended monthly time series (3-months moving average) of the forcing factors for the upwelling region. It is noticeable that the interannual variability of the air-sea CO 2 flux in this region is dominated by both the contribution of the gas transfer velocity and ΔpCO2 anomalies (dash-dotted blue and solid-dotted red time series, respectively). The former is related to an increase and a decrease of the wind speed (+2 and -1 ms −1 ) during El Ni˜ no and La Ni˜ na phases, respectively (see Section 3.2). The latair ter is mainly related with changes in pCOsw 2 because pCO2 varies little at the interannual timescale. As seen in Fig.5e,f, variations in ΔpCO2 induce a negative contribution to the air-sea CO 2 fluxes anomalies during El Ni˜ no events which may be interpreted as a decrease of the CO 2 outgassing. This weaker CO2 source behavior could be explained due to the downwelling pattern that occurs during the warm phase of ENSO, leading to a substantial reduction of the carbon supply to the surface, thus decreasing pCOsw 2 and na, ΔpCO2 -driven diminishing the CO2 outgassing. Conversely, during La Ni˜ flux variability exhibits large positive amplitudes, which is interpreted as an increase of the CO2 outgassing. This response results from enhanced and more effective upwelling during La Ni˜ na which brings more DIC-rich waters sw to the surface, thus increasing pCO2 and favoring the CO2 outgassing. The contribution from the cross-terms is minor (solid purple line in Fig.5e). However, it becomes more significant during strong El Ni˜ no and La Ni˜ na peaks. Results obtained partially agree with Jones et al. (2001), which found that under El Ni˜ no conditions the majority of the reduction in CO 2 outgassing is due to a reduction in the ΔpCO2 , with a smaller contribution from the gas transfer velocity from a large-scale domain perspective. However, it is important to notice that a comparable contribution from both forcing factors during non-extreme ENSO conditions was found, although the magnitude of the air–sea CO2 fluxes seems to be slightly more sensitive to ΔpCO2 variations than gas transfer velocity ones. The difference between the contribution to the interannual air-sea CO 2 anomalies from the gas transfer velocity and ΔpCO2 anomalies, i.e. dash15

Journal Pre-proof

19

98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15

of

dotted blue line minus solid-dotted red line in Fig.5e, is correlated (0.71) with the ICEN (Fig.5f). This high correlation permits identify some general aspects associated with ENSO events within the upwelling region: i) during an average El Ni˜ no episode, gas transfer anomalies tend to increase the airsea CO2 flux, while ΔpCO2 anomalies tend to diminish it; ii) during an average La Ni˜ na episode, gas transfer anomalies tend to decrease the air-sea CO2 flux, while ΔpCO2 anomalies tend to increase it. These generalizations should be considered with caution because each El Ni˜ no and La Ni˜ na event has unique characteristics of timing and intensity.

120

ro

140 Interannual Climatology

a)

80

re

60

lP

20

3

La Niña

1 0

15

14

20

13

20

12

20

11

20

20

10 20

09

08

20

20

07

06

20

20

05

-2

04

03

2

-1

20

ur 02

20

20

20

20

01

00

El Niño

Jo

99 19

98

b)

19

4

na

0

ICEN

40

20

Tg C year

-1

-p

100

years

Figure 6: a) Interannual variability of the spatially-integrated air-sea CO 2 flux (FCO2 : gray bars) in [Tg C year−1 ]. The repetitive annual climatology is depicted as a solid black line with its annual mean value (dashed line). b) ENSO index: ICEN. The dashed lines denote the ICEN threshold values for both ENSO phases.

3.4. Integrated carbon flux The spatially integrated evasion flux of carbon in Tg C year −1 was calculated (see Fig.6). This computation was performed over the upwelling region, 16

Journal Pre-proof

re

-p

ro

of

an area of 2.86 ×106 km2 . Under climatological conditions, the integrated carbon flux was calculated in, approximately, 53.1 Tg C year −1 . This value is consistent with previous modeled and observed estimates (Friederich et al. 2008; Mogoll´on and Calil 2018). During an average El Ni˜ no and La Ni˜ na episode, the carbon evasion flux was estimated in, approximately, 28.4 and 56.4 Tg C year−1 , respectively. However, these fluxes may be larger depending on the temporal scale. For example, during the strongest La Ni˜ na-2007, more than 100 Tg C year −1 was obtained. To put in context, this value equals, in magnitude, the global coastal ocean uptake of anthropogenic carbon (Bourgeois et al. 2016). Considering a one-year period as the duration of an average ENSO episode (Section 2.6), the upwelling region contributes to the evasion flux of carbon with 3.3 Tg C more during La Ni˜ na when compared to neutral (climatological) conditions. In contrast, El Ni˜ no represents a period when the upwelling region emits to the atmosphere 24.7 Tg C less when compared to climatological conditions, i.e. almost half the amount emitted during an ENSO-neutral year. 4. Conclusions

Jo

ur

na

lP

In this work, the role played by the ENSO phenomenon on modulating the oceanic carbon outflux is assessed from 1998 through 2015 at monthly timescales. Furthermore, the sensitivity of the forcing factors behind the interannual variability of the air-sea CO 2 fluxes in the NHCS is also addressed. To this end, outputs from a global ocean biogeochemistry hindcast were used. In addition, a recent neural network-based methodology was implemented by means a combined statistical, modeled and observed approach in order to reconstruct surface DIC. It is found that the air-sea CO 2 flux variability in the upwelling region results from both ENSO-driven ΔpCO2 and wind speed variations, although the latter in a slight minor proportion. In general, during an average El Ni˜ no episode, gas transfer velocity anomalies tend to increase the air-sea CO 2 flux, while ΔpCO2 anomalies tend to diminish it. Conversely, during an average La Ni˜ na episode, variations in gas transfer velocity tend to decrease the air-sea CO 2 flux, while ΔpCO2 anomalies tend to increase it. The CO 2 source variability during ENSO phases could be explained mainly due to variations in the underlying mechanisms that directly impact surface DIC concentrations (consequently affecting pCOsw 2 ), as well as the efficiency of the CO 2 exchange across the air-sea interface. Specifically, the downwelling-like pattern induced during El Ni˜ no peaks renders the up17

Journal Pre-proof

welled water warm but DIC-depleted. This reduction in the upward carbon supply is compensated by more efficient gas exchange at the air-sea interface due to the strengthening of the winds. In contrast, the strong CO 2 source behavior during La Ni˜ na conditions results from enhanced effective upwelling which brings CO2 -repleted waters to the surface, thus increasing pCOsw 2 and the associated CO2 outflux.

Jo

ur

na

lP

re

-p

ro

of

In summary, based on the model assumptions and approach, during El Ni˜ no the total amount of carbon retained within the upwelling (eutrophic and mesotrophic) region in the NHCS is about 24.7 million tons of carbon that normally would have been lost to the atmosphere as CO 2 . During La Ni˜ na, 3.3 million tons of carbon is additionally released from the upwelling region contributing to the atmospheric CO 2 accumulation. Despite the similarities between different El Ni˜ no/La Ni˜ na episodes of different strengths throughout the analysis period, each ENSO event has their own characteristics, hence, generalizations here presented are mean to be to describe the average response of the NHCS to both ENSO phases. Another caveat is related to inferred surface DIC concentrations, because they are limited by the ability of the approach in simulating ENSO phenomena. In addition, the anthropogenically driven increase of atmospheric CO 2 was not separated from computed CO2 fluxes, and their associated anomalies still contain the combined natural and anthropogenic variability. Furthermore, complex positive and negative feedbacks between simulated ocean physics and biology may impact pCOsw 2 over multidecadal time scales, therefore, even though the results here presented have realistic sensitivity to ENSO, the CO 2 -carbonate variables may be biased because it is assumed that the ocean carbon cycle is in steady-state in face of changing climate and ocean chemistry. Although good agreement at large scale, too high values of pCOsw 2 in EBUS are informed in the global ocean biogeochemistry product used, and the associated simulation is not long enough for each variable to be fully balanced, at least at subsurface layers. In spite of these caveats and within the framework of the model assumptions, results obtained in this study are considered robust and may help to predict how the NHCS will react during ENSO events and probably in a climate change context. Further studies are still needed in order to understand the complete response of the carbonate system throughout an ENSO event in the Humboldt EBUS, as such, a sensitivity analysis of the drivers and the mechanisms behind the ENSO-driven spatiotemporal variability of pCOsw 2 would be highly desirable. This could be addressed in 18

Journal Pre-proof

future work. Acknowledgments

Jo

ur

na

lP

re

-p

ro

of

R. Mogoll´on acknowledges support from the Coordena¸ca˜o de Aperfei¸coamento de Pessoal de N´ıvel Superior (CAPES) scholarship. GLORYS2V4 and GLOBAL REANALYSIS BIO 001 029 data were provided by the Copernicus Marine Environment monitoring service (CMEMS). CCMP Version-2.0 vector wind analyses are produced by Remote Sensing Systems. Data are available at www.remss.com. The Surface Ocean CO2 Atlas (SOCAT) is an international effort, endorsed by the International Ocean Carbon Coordination Project (IOCCP), the Surface Ocean Lower Atmosphere Study (SOLAS) and the Integrated Marine Biogeochemistry and Ecosystem Research program (IMBER), to deliver a uniformly quality-controlled surface ocean CO2 database. The many researchers and funding agencies responsible for the collection of data and quality control are thanked for their contributions. No conflict of interest is declared.

19

Journal Pre-proof

Appendix A. Oceanic model

lP

re

-p

ro

of

The outputs from the Mercator-Ocean (Toulouse, FR) GLORYS2V4 global ocean reanalysis (Garric et al. 2017) (monthly-means at 1/4 ◦ horizontal resolution covering the 1993-2015 period) are used. GLORYS2V4 is forced by daily means ERA-Interim atmospheric variables (produced at ECMWF) (Dee et al. 2011) with data assimilation of temperature and salinity profiles as well as sea level anomalies, sea ice concentration, sea surface temperature, and mixed layer depth. The simulation was forced with temperature and salinity based on EN4 (Good et al. 2013) replacing the Levitus Climatology. For a more detailed description of the ocean model, the data assimilation method, the assimilated observations, and the validation, the reader is referred to http://marine.copernicus.eu/documents/QUID/CMEMS-GLO-QUID-001-025.pdf (last accessed 15 July 2019). Similarly, a comprehensive assessment of the quality of the global biogeochemical hindcast simulation may be found at http://cmems-resources.cls.fr/documents/QUID/CMEMS-GLO-QUID-001-029.pdf (last accessed 15 July 2019). The global average atmospheric pCO2 monthly values were extracted from www.esrl.noaa.gov/gmd/ccgg/trends/ (Ed Dlugokencky and Pieter Tans, NOAA/ESRL), from 1998 to 2015, as shown in Fig.A.7. 405

395

ur

385 380

Jo

CO 2 [ atm]

390

na

400

375 370 365

360 1998

2000

2002

2004

2006

2008

2010

2012

2014

Years

Figure A.7: Monthly time series of atmospheric pCO2

20

Journal Pre-proof

Appendix B. Reynolds decomposition Let y be a function defined as y = ab, composed by two factors, namely a and b. Those factors may be separated in both, a steady or ensemble average component, denoted as an overline bar, and in a component containing the deviations or fluctuations from the mean, denoted as ( 0 ), as follows: y = y + y 0 ; a = a + a0 ; b = b + b0 . Replacing, the following is obtained: y = a.b = (a + a0 ).(b + b0 ) = ab + ab0 + a0 b + a0 b0

(B.1)

ro

of

Reynolds-averaging each term of Equation B.1 and considering the corresponding laws: y = ab + ab0 + a0 b + a0 b0 = ab + a0 b0 (B.2) Replacing B.1 and B.2 in y 0 = y − y, it yields:

-p

y 0 = y − y = (ab + ab0 + a0 b + a0 b0 ) − (ab + a0 b0 )

re

Finally: y 0 = a0 b + ab0 + [a0 b0 − a0 b0 ] If y =FCO2 , a = (K0 kw ), and b = ΔpCO2 , then:

lP

F0CO2 = (K0 kw )0 × ΔpCO2

(B.3)

na

+ (K0 kw ) × ΔpCO2 0 i h + (K0 kw )0 × ΔpCO2 0 − (K0 kw )0 × ΔpCO2 0

ur

Appendix C. SST and chlorophyll

Jo

Two of the most anomalous months in the period of analysis were chosen in order to illustrate the comparison between modeled and observed SST and chlorophyll-a (Chla) during the largest ENSO peaks. These months are January-1998 and November-2007, for El Ni˜ no (ICEN=+3.4) and La Ni˜ na (ICEN=-1.92), respectively. Satellite-derived observed SST was extracted from the Advance Very High Resolution Radiometer (AVHRR) Pathfinder (Casey et al. 2010) and from the Moderate-resolution Imaging Spectroradiometer (MODIS) (http://modis.gsfc.nasa.gov/), for Jan-1998 and Nov2007, respectively, both at 1/12 ◦ horizontal resolution. Chla was obtained from ESA OC-CCI Version 3.1, which is a long-term multi-sensor satellite ocean-colour data at 1/12 ◦ horizontal resolution. Fig.B.8 and Fig.B.9 demonstrate that the large scale spatial pattern and main features during both 21

Journal Pre-proof

El Niño (Jan-1998)

La Niña (Nov-2007)

0

a)

-15

-20

-20 -85

-80

-75

C

-15

25

°

-10 20

-70

-90

of

-10

-90

b)

-5

-85

15

-80

-75

-70

ro

La Niña (Nov-2007)

-p

El Niño (Jan-1998)

10

0

c)

3 d)

-5

re

-5

-10

lP

-10

1

-3

0

-20 -90

-85

na

-15

-80

-75

0.3

mg m

SST (Model)

-5

Chlorophyll-a (Model)

30

0

-15

0.1 -20

0.03 -70

-90

-85

-80

-75

-70

Jo

ur

Figure B.8: Upper row: modeled SST (GLORYS2V4/PISCES) of a) January-1998 and b) November-2007, for El Ni˜ no (ICEN=+3.4) and La Ni˜ na (ICEN=-1.9) peak, respectively. Similarly, the bottom row is for modeled chlorophyll (GLORYS2V4/PISCES) in panels c and d.

ENSO peaks are very well captured by the model, despite that the model horizontal resolution is three times coarser than observations. During El Ni˜ no Jan-1998, anomalous warm waters are found in almost the entire domain. In contrast, during La Ni˜ na Nov-2007, the equatorial and coastal upwelling fronts are reproduced, evidenced by cold waters along the Equator and the Peruvian coast. The spatial distribution of Chla during both ENSO peaks is consistent. Low Chla concentrations are found at the nearshore strip during El Ni˜ no Jan-1998. The highly productive band (2 mg Chla m −3 22

Journal Pre-proof

El Niño (Jan-1998)

La Niña (Nov-2007)

a)

-15

-15

-20

-20 -80

-75

C

-10

25

°

-10

-85

b)

-5

20

-70

-90

of

-5

-90

-85

15

-80

-75

-70

La Niña (Nov-2007)

-p

El Niño (Jan-1998)

10

0

3 d)

re

-5

-10

lP

-10

1

-3

c)

-5

-20 -90

-85

na

-15

-80

-75

0.3

mg m

0

Chlorophyll-a (Observation)

30

0

ro

SST (Observation)

0

-15

0.1 -20

0.03 -70

-90

-85

-80

-75

-70

Jo

ur

Figure B.9: Upper row: observed SST of a) January-1998 (AVHRR) and b) November2007 (MODIS), for El Ni˜ no (ICEN=+3.4) and La Ni˜ na (ICEN=-1.9) peak, respectively. Similarly, the bottom row is for observed chlorophyll (ESA OC-CCI V3.1) in panels c and d.

isoline depicted in Fig.1) extended over a wider offshore area during La Ni˜ na Nov-2007, and the oligotrophic oceanic region at the southern domain remains with low Chla concentrations but an order of magnitude larger than during El Ni˜ no Jan-1998 peak.

23

Journal Pre-proof

Appendix D. Inferred DIC and Alk

re

-p

ro

of

Fig.D.10 shows the comparison between inferred and observed annual mean dissolved inorganic carbon (DIC) and total alkalinity (Alk). Observations were extracted from the Global Ocean Data Analysis Project (GLODAP v2.2016b at 1 ◦ ) (Olsen et al. 2016; Lauvset et al. 2016) and are shown as color maps (mapped climatology) merged with the original transects (noted the limited number of available observations). The latitudinal transition from south to north and from high to low concentrations of both DIC and Alk is very well reproduced. However, larger mapping errors (> 30μmol C l−1 ) associated to the GLODAP database could explain the overestimation of inferred surface DIC and Alk within the northern portion of the oceanic domain (0-10 ◦ S and 85-90◦ W). Overall, the large scale pattern is well captured from an annual scale perspective although the gridded GLODAP-derived DIC and Alk fields could be biased due to mapping errors arisen from the sparse and infrequent cruise data. Appendix E. ENSO Signal

Jo

ur

na

lP

Empirical Orthogonal Functions analysis (EOF) was performed in order to evaluate the ENSO signal. Firstly, modeled pCOsw 2 was detrended, the mean removed, and the seasonal cycle also removed. The first mode of variability (EOF1) of pCOsw 2 anomalies within the whole domain of analysis represents 63 % of the total variance, and its respective principal component (PC1) is correlated with the ENSO index: ICEN (r = 0.75) (see Figs.E.11a,c,d). A positive linear trend is obtained in almost the entire region of analysis, particularly within the central and southern domain (see Fig.E.11b). This trend was calculated using Least Squares and ranges from -1 to +3 μatm/year, in agreement with Takahashi et al. (2003), who estimated a trend of 1.5 μatm/year along the Equatorial Pacific. The power spectrum was plotted in Fig.E.11e specifying a sampling frequency of 12 samples per year, plotted on a log x axis, with x values in units of years rather than frequency. A periodicity of 3.5, and 1.7-1.4 (El Ni˜ no/La Ni˜ na successive events) years was detected, which is unequivocally related to the ENSO’s signal.

24

Journal Pre-proof

b) observed DIC

0

0

-5

-5

-10

-10

-15

-15

-20

-20

-75

-70

c) inferred Alk

lP

0 -5

na

-10

-20 -90

ur

-15

-85

-80

-75

-1

mol Kg

of ro -90

1950

1900

-85

-80

-75

d) observed Alk 2350

0

2300

-5

-1

-80

2000

2250

-10 2200

mol Kg

-85

-p

-90

Latitude

2100

2050

re

Latitude

a) inferred DIC

-15 2150

-20 -70

2100

-90

-80

-75

Longitude

Jo

Longitude

-85

Figure D.10: Annual mean inferred (a,c) and observed (b,d) DIC and Alk, respectively. Inferred fields comes from GLORYSv2/PISCES/CANYON-B/CONTENT methodology, while observed fields, from GLODAP v2 2016b. Black dots represent the stations.

25

26

-80

-60

-40

-20

8

9 19

-0.5

0

0.5

1

98 19

-90

20

00

00

20

02

20

02

20

-80

04 20

-70

ur

Jo

04

20

08

10

20

-80

b)

10

12 20

La Niña

20

12

20

14

20

-70

14 20

-1

0

1

2

3

0.2

-2

0

2

0

-80

c)

e)

9 76 5 4 3 2

1

periodicity (years)

of

4 0.1

Nyquist

-90

-20

-10

0

ro

-p

re

lP

20

08

20

Years

06

20

El Niño

d)

06

20

na

-90

-20

-10

-10

-20

0

0

a) atm/year

-70

0

500

1000

1500

Figure E.11: a) First spatial mode of variability: EOF1 map (scaled). b) pCO2 trend. c) Variance of pCO2 . d) First principal component: PC1 (scaled). ICEN index: positive values indicate El Ni˜ no events while negative ones, La Ni˜ na. e) Power spectrum.

EOF1

0

PC1

ICEN

2

( atm)

20

Journal Pre-proof

Journal Pre-proof

References Archer, D., Takahashi, T., Sutherland, S., Goddard, J., Chipman, D., Rodgers, K., Ogura, H., 1996. Daily, seasonal and interannual variability of sea-surface carbon and nutrient concentration in the equatorial pacific ocean. Deep Sea Research Part II: Topical Studies in Oceanography 43, 779–808.

of

´ e, C., Tagliabue, A., Bopp, L., Gehlen, M., 2015. PiscesAumont, O., Eth´ v2: an ocean biogeochemical model for carbon and ecosystem studies. Geoscientific Model Development Discussions 8.

-p

ro

Bakker, D.C., Pfeil, B., Landa, C.S., Metzl, N., O’brien, K.M., Olsen, A., Smith, K., Cosca, C., Harasawa, S., Jones, S.D., et al., 2016. A multidecade record of high-quality fco 2 data in version 3 of the surface ocean co 2 atlas (socat). Earth System Science Data 8, 383.

re

Bakun, A., 1990. Global climate change and intensification of coastal ocean upwelling. Science 247, 198–201.

lP

Bakun, A., Field, D.B., Redondo-Rodriguez, A., Weeks, S.J., 2010. Greenhouse gas, upwelling-favorable winds, and the future of coastal ocean upwelling ecosystems. Global Change Biology 16, 1213–1228.

ur

na

Bakun, A., Weeks, S.J., 2008. The marine ecosystem off peru: What are the secrets of its fishery productivity and what might its future hold? Progress in Oceanography 79, 290–299.

Jo

Barber, R., Kogelschatz, J., 1990. Nutrients and productivity during the 1982/83 el ni˜ no, in: Elsevier oceanography series. Elsevier. volume 52, pp. 21–53. Belmadani, A., Echevin, V., Codron, F., Takahashi, K., Junquas, C., 2014. What dynamics drive future wind scenarios for coastal upwelling off peru and chile? Climate dynamics 43, 1893–1914. Bittig, H.C., Steinhoff, T., Claustre, H., Fiedler, B., Williams, N.L., Sauz`ede, R., K¨ortzinger, A., Gattuso, J.P., 2018. An alternative to static climatologies: Robust estimation of open ocean co2 variables and nutrient concentrations from t, s, and o2 data using bayesian neural networks. Frontiers in Marine Science 5, Art–Nr. 27

Journal Pre-proof

Bourgeois, T., Orr, J.C., Resplandy, L., Terhaar, J., Eth´e, C., Gehlen, M., Bopp, L., 2016. Coastal-ocean uptake of anthropogenic carbon. Biogeosciences 13, 4167–4185. Brady, R.X., Lovenduski, N.S., Alexander, M.A., Jacox, M., Gruber, N., 2019. On the role of climate modes in modulating the airsea co2 fluxes in eastern boundary upwelling systems. Biogeosciences 16, 329–346. URL: https://www.biogeosciences.net/16/329/2019/, doi:10.5194/bg-16-329-2019.

ro

of

Casey, K.S., Brandon, T.B., Cornillon, P., Evans, R., 2010. The past, present, and future of the avhrr pathfinder sst program, in: Oceanography from Space. Springer, pp. 273–287.

re

-p

Chamorro, A., Echevin, V., Colas, F., Oerder, V., Tam, J., Quispe-Ccalluari, C., 2018. Mechanisms of the intensification of the upwelling-favorable winds during el ni˜ no 1997–1998 in the peruvian upwelling system. Climate Dynamics , 1–17.

lP

Chavez, F., Strutton, P., Friederich, G., Feely, R., Feldman, G., Foley, D., McPhaden, M., 1999. Biological and chemical response of the equatorial pacific ocean to the 1997-98 el ni˜ no. Science 286, 2126–2131.

ur

na

Colas, F., Capet, X., McWilliams, J., Shchepetkin, a., 2008. 1997–1998 El Ni˜ no off Peru: A numerical study. Progress in Oceanography 79, 138–155. URL: http://linkinghub.elsevier.com/retrieve/pii/S0079661108001687, doi:10.1016/j.pocean.2008.10.015.

Jo

Dee, D.P., Uppala, S.M., Simmons, A., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U., Balmaseda, M., Balsamo, G., Bauer, d.P., et al., 2011. The era-interim reanalysis: Configuration and performance of the data assimilation system. Quarterly Journal of the royal meteorological society 137, 553–597. Doney, S.C., Lima, I., Feely, R.A., Glover, D.M., Lindsay, K., Mahowald, N., Moore, J.K., Wanninkhof, R., 2009. Mechanisms governing interannual variability in upper-ocean inorganic carbon system and air–sea co 2 fluxes: Physical climate and atmospheric dust. Deep Sea Research Part II: Topical Studies in Oceanography 56, 640–655. 28

Journal Pre-proof

Echevin, V., Albert, A., L´evy, M., Graco, M., Aumont, O., Pi´etri, A., Garric, G., 2014. Intraseasonal variability of nearshore productivity in the Northern Humboldt Current System: The role of coastal trapped waves. Continental Shelf Research 73, 14–30. URL: http://linkinghub.elsevier.com/retrieve/pii/S0278434313003804, doi:10.1016/j.csr.2013.11.015.

of

Espinoza-Morriber´on, D., Echevin, V., Colas, F., Tam, J., Ledesma, J., V´asquez, L., Graco, M., 2017. Impacts of e l n i˜ no events on the p eruvian upwelling system productivity. Journal of Geophysical Research: Oceans 122, 5423–5444.

-p

ro

Espinoza Morriber´on, D., Echevin, V., Colas, F., Tam M´alaga, J.L., Guti´errez Aguilar, D.A., Graco, M.I., Ledesma Rivera, J., Quispe Ccalluari, C., 2018. Oxygen variability during enso in the tropical south eastern pacific. Frontiers in Marine Science 5, 526.

lP

re

Feely, R.A., Wanninkhof, R., Cosca, C.E., Murphy, P.P., Lamb, M.F., Steckley, M.D., 1995. Co2 distributions in the equatorial pacific during the 1991–1992 enso event. Deep Sea Research Part II: Topical Studies in Oceanography 42, 365–386.

ur

na

Feely, R.A., Wanninkhof, R., Goyet, C., Archer, D.E., Takahashi, T., 1997. Variability of c02 distributions and sea-air fluxes in the central and eastern equatorial pacific during the 199–1994 el nin o. Deep Sea Research Part II: Topical Studies in Oceanography 44, 1851–1867.

Jo

Feely, R.A., Wanninkhof, R., Takahashi, T., Tans, P., 1999. Influence of el ni˜ no on the equatorial pacific contribution to atmospheric co 2 accumulation. Nature 398, 597. Friederich, G., Walz, P., Burczynski, M., Chavez, F., 2002. Inorganic carbon in the central california upwelling system during the 1997–1999 el ni˜ no–la ni˜ na event. Progress in Oceanography 54, 185–203. Friederich, G.E., Ledesma, J., Ulloa, O., Chavez, F.P., 2008. Air–sea carbon dioxide fluxes in the coastal southeastern tropical pacific. Progress in Oceanography 79, 156–166.

29

Journal Pre-proof

Garcia, H., Locarnini, R., Boyer, T., Antonov, J., Baranova, O., Zweng, M., Reagan, J., Johnson, D., 2014. World ocean atlas 2013, volume 3: Dissolved oxygen, apparent oxygen utilization, and oxygen saturation. NOAA Atlas NESDIS 75, 27. Garcia, H.E., Locarnini, R.A., Boyer, T.P., Antonov, J.I., Baranova, O.K., Zweng, M.M., Reagan, J.R., Johnson, D.R., Mishonov, A.V., Levitus, S., 2013. World ocean atlas 2013. volume 4, dissolved inorganic nutrients (phosphate, nitrate, silicate) .

-p

ro

of

Garric, G., Parent, L., Greiner, E., Dr´evillon, M., Hamon, M., Lellouche, J.M., R´egnier, C., Desportes, C., Le Galloudec, O., Bricaud, C., et al., 2017. Performance and quality assessment of the global ocean eddypermitting physical reanalysis glorys2v4., in: EGU General Assembly Conference Abstracts, p. 18776.

lP

re

Good, S.A., Martin, M.J., Rayner, N.A., 2013. En4: Quality controlled ocean temperature and salinity profiles and monthly objective analyses with uncertainty estimates. Journal of Geophysical Research: Oceans 118, 6704–6716.

na

Huyer, A., Smith, R.L., Paluszkiewicz, T., 1987. Coastal upwelling off peru during normal and el ni˜ no times, 1981–1984. Journal of Geophysical Research: Oceans 92, 14297–14307.

ur

Inoue, H.Y., Sugimura, Y., 1992. Variations and distributions of co2 in and over the equatorial pacific during the period from the 1986/88 el ni˜ no event to the 1988/89 la ni˜ na event. Tellus B 44, 1–22.

Jo

Jones, C.D., Collins, M., Cox, P.M., Spall, S.A., 2001. The carbon cycle response to enso: A coupled climate–carbon cycle model study. Journal of Climate 14, 4113–4129. Key, R.M., Olsen, A., van Heuven, S., Lauvset, S.K., Velo, A., Lin, X., Schirnick, C., Kozyr, A., Tanhua, T., Hoppema, M., et al., 2015. Global ocean data analysis project, version 2 (glodapv2). Lauvset, S.K., Key, R.M., Perez, F.F., 2016. A new global interior ocean mapped climatology: the 1x1 glodap version 2. Earth System Science Data 8, 325. 30

Journal Pre-proof

Madec, G., et al., 2015. Nemo ocean engine. Scientific Notes of Climate Modelling Center Issue 27. doi:10.5281/zenodo.1464816. McKinley, G.A., Follows, M.J., Marshall, J., 2004. Mechanisms of air-sea co2 flux variability in the equatorial pacific and the north atlantic. Global Biogeochemical Cycles 18. Meehl, G.A., Washington, W.M., 1996. El ni˜ no-like climate change in a model with increased atmospheric co2 concentrations. Nature 382, 56.

ro

of

Mogoll´on, R., Calil, P.H., 2017. On the effects of enso on ocean biogeochemistry in the northern humboldt current system (nhcs): A modeling study. Journal of Marine Systems .

-p

Mogoll´on, R., Calil, P.H., 2018. Modelling the mechanisms and drivers of the spatiotemporal variability of pco2 and air–sea co2 fluxes in the northern humboldt current system. Ocean Modelling 132, 61–72.

na

lP

re

Montes, I., Colas, F., Capet, X., Schneider, W., 2010. On the pathways of the equatorial subsurface currents in the eastern equatorial Pacific and their contributions to the PeruChile Undercurrent. Journal of Geophysical Research 115, C09003. URL: http://doi.wiley.com/10.1029/2009JC005710, doi:10.1029/2009JC005710.

ur

Montes, I., Dewitte, B., Gutknecht, E., Paulmier, A., Dadou, I., Oschlies, A., Gar¸con, V., 2014. High-resolution modeling of the eastern tropical pacific oxygen minimum zone: Sensitivity to the tropical oceanic circulation. Journal of Geophysical Research: Oceans 119, 5515–5532.

Jo

Montes, I., Schneider, W., Colas, F., Blanke, B., Echevin, V., 2011. Subsurface connections in the eastern tropical Pacific during La Ni˜ na 1999–2001 and El Ni˜ no 2002–2003. Journal of Geophysical Research 116, C12022. URL: http://doi.wiley.com/10.1029/2011JC007624, doi:10.1029/2011JC007624. ˜ Niquen, M., Bouchon, M., 2004. Impact of el ni˜ no events on pelagic fisheries in peruvian waters. Deep sea research part II: topical studies in oceanography 51, 563–574.

31

Journal Pre-proof

Olsen, A., Key, R.M., van Heuven, S., Lauvset, S.K., Velo, A., Lin, X., Schirnick, C., Kozyr, A., Tanhua, T., Hoppema, M., et al., 2016. The global ocean data analysis project version 2 (glodapv2)–an internally consistent data product for the world ocean. Earth System Science Data 8, 297. Oort, A.H., Yienger, J.J., 1996. Observed interannual variability in the hadley circulation and its connection to enso. Journal of Climate 9, 2751– 2767.

ro

of

Rayner, P.J., Law, R.M., Dargaville, R., 1999. The relationship between tropical co2 fluxes and the el ni˜ no-southern oscillation. Geophysical Research Letters 26, 493–496.

-p

R¨odenbeck, C., Bakker, D.C., Metzl, N., Olsen, A., Sabine, C.L., Cassar, N., Reum, F., Keeling, R.F., Heimann, M., 2014. Interannual sea-air co 2 flux variability from an observation-driven ocean mixed-layer scheme. Biogeosciences Discussions 11, 3167–3207.

re

Sarmiento, J.L., Gruber, N., 2006. Ocean biogeochemical dynamics.

na

lP

Sauz`ede, R., Bittig, H.C., Claustre, H., Pasqueron de Fommervault, O., Gattuso, J.P., Legendre, L., Johnson, K.S., 2017. Estimates of water-column nutrient concentrations and carbonate system parameters in the global ocean: A novel approach based on neural networks. Frontiers in Marine Science 4, 128.

Jo

ur

Stow, C.A., Jolliff, J., McGillicuddy Jr, D.J., Doney, S.C., Allen, J.I., Friedrichs, M.A., Rose, K.A., Wallhead, P., 2009. Skill assessment for coupled biological/physical models of marine systems. Journal of Marine Systems 76, 4–15. Takahashi, K., Mosquera, K., Reupo, J., 2014. El ´ındice costero el ni˜ no (icen): historia y actualizaci´on. Bolet´ın T´ecnico del Instituto Geof´ısico del Per´ u 1, 8–9. Takahashi, T., Sutherland, S.C., Feely, R.A., Cosca, C.E., 2003. Decadal variation of the surface water pco2 in the western and central equatorial pacific. Science 302, 852–856.

32

Journal Pre-proof

Trenberth, K.E., Stepaniak, D.P., Caron, J.M., 2002. Interannual variations in the atmospheric heat budget. Journal of Geophysical Research: Atmospheres 107, AAC–4. Van Heuven, S., Pierrot, D., Rae, J., Lewis, E., Wallace, D., 2011. Matlab program developed for co2 system calculations. Oak Ridge, Tennessee: Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, US Department of Energy. pp. ORNL/CDIAC-105b .

ro

of

Wanninkhof, R., 2014. Relationship between wind speed and gas exchange over the ocean revisited. Limnology and Oceanography: Methods 12, 351– 362.

-p

Weiss, R., 1974. Carbon dioxide in water and seawater: the solubility of a non-ideal gas. Marine chemistry 2, 203–215.

Jo

ur

na

lP

re

Wentz, F., Scott, J., Hoffman, R., Leidner, M., Atlas, R., Ardizzone, J., 2015. Remote sensing systems cross-calibrated multi-platform (ccmp) 6-hourly ocean vector wind analysis product on 0.25 deg grid, version 2.0. Proc. Remote Sens. Syst .

33

Journal Pre-proof

Highlights • Weakened ocean-atmosphere pCO2 gradient decreases the CO 2 efflux during EN. • Enhanced ocean-atmosphere pCO2 gradient increases the CO 2 efflux during LN. • Variations in ΔpCO2 are partially compensated by changes in the wind speed.

ro

of

• 25 million tons of carbon is retained in the upwelling region under EN conditions.

Jo

ur

na

lP

re

-p

• 3 million tons of carbon is released in the upwelling region under LN conditions.

34

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6