Journal of Geophysical Research: Oceans

Continued CO2 outgassing in an upwelling area off northern Chile during the development phase of El Niño 1997–1998 (July 1997)

Authors


Abstract

[1] Carbonate system parameters were measured in the upper 200 m of the water column during July 1997 in an upwelling area off northern Chile (22.6°–24°S), and the CO2 fluxes were estimated. At this time (during the onset of El Niño 1997–1998), the water column that feeds the coastal upwelling was less dense, warmer, and saltier than in non-El Niño winters. Nevertheless, the major vertical gradients in pH, total inorganic carbon (CT), carbon dioxide fugacity (fCO2), and apparent oxygen utilization (AOU) remained confined to the upper 100 m of the water column, so that the active upwelling forced by southerly winds caused the upwelling of CO2-rich water leading a CO2 flux from the ocean to the atmosphere. However, these fluxes were found to be highly variable. Grid surveys 2 weeks apart show a change in CO2 flux from +3.9 mol C m−2 yr−1 to +0.4 mol C m−2 yr−1: the change is thought to be associated with a pulsed upwelling forcing in combination with an active biological uptake of CO2. This high short-term variability of CO2 fluxes makes it difficult to assess the interannual variability of CO2 outgassing in this area based on low-frequency direct CO2 observations. The fact that the oxycline, whose location usually coincides with the carboncline, also remained within the upper 100 m during the remarkably warm 1972 and 1983 El Niño winters seems to imply that the CO2 outgassing during those warm periods can be as strong as we report for 1997 under similar upwelling favorable winds.

1. Introduction

[2] Equatorial upwelling is a major marine source of CO2 to the atmosphere [e.g., Keeling and Revelle, 1985; Tans et al., 1990]. The cessation of upwelling in the eastern equatorial Pacific during ENSO (El Niño/Southern Oscillation) warm events reduces drastically the flux of CO2 out of the ocean [Feely et al., 1987; Wong et al., 1993; Feely et al., 1999; Chavez et al., 1999], producing negative anomalies in the trends of atmospheric CO2 content [Feely et al., 1999]. Coastal upwelling areas, which are mostly located along Eastern Boundary Currents (EBC) transport cold and CO2-rich water to the surface [Copin-Montégut and Raimbault, 1994; Torres et al., 1999, 2002]. The resulting sea to air CO2 fluxes and their variability, including the ENSO cycle, have been poorly documented and understood. The recent indication that the anticipated greenhouse warming due to increased atmospheric CO2 could lead to changes in the frequency of ENSO [Timmerman et al., 1999], and in the local forcing of coastal upwelling [Bakun, 1990], underlines the need to understand the relation between ENSO, upwelling and CO2 fluxes.

[3] The EBC off the west coast of South America, which is one of the most productive coastal upwelling areas on Earth [Summerhayes et al., 1995], is also the most directly affected by the ENSO cycles. In this region the alongshore, equatorward winds which normally promote upwelling of cold, CO2-and nutrient-rich water do not always decrease during the ENSO warm events [Enfield, 1981; Huyer et al., 1987, 1991]. However, the physical and chemical characteristics of the upwelled water (e.g., temperature, salinity, nutrients) can be substantially altered during the El Niño (warm) phase [e.g., Bernal et al., 1982, Barber and Chavez, 1983], suggesting an alteration in the distribution of CO2 and of CO2 fluxes. In fact many processes involved in CO2 ventilation (upwelling, turbulence, circulation, respiration, primary production, sensible and latent heat fluxes, etc.) are expected to be sensitive to ENSO, so that direct assessment of the distribution and variability of the CO2 parameters during those cycles is required. This is especially true within the largely unexplored coastal upwelling area off northern Chile.

[4] The Mejillones peninsula at 23°S is one the most important irregularities in the essentially straight and meridionally oriented coastline of northern Chile (Figure 1), where the topography and the characteristic southwesterly winds promote year-around upwelling [Figueroa and Moffat, 2000; Pizarro et al., 1994]. The nearshore equatorward alongshore winds, derived from 6 years of satellite scatterometer observations (ERS-1 and 2) off Mejillones, are on average weaker than those at other latitudes along the Chile-Peru coast (e.g., at 15°S and 30°S) [Shaffer et al., 1999]. However, during the austral winter energetic southerly coastal wind events are interrupted by short relaxation periods [Pizarro et al., 1994] leading to pulsed upwelling particularly during the developing stage of the ENSO warm phase [Rutllant et al., 1998]. The upwelling focus off the Mejillones Peninsula was first identified by Gunther [1936], and confirmed later as one of the most persistent upwelling foci in northern Chile on the basis of satellite sea surface temperature (SST) imagery for the period 1987–1992 [Barbieri et al., 1995]. During the austral winter, the colder waters can be identified at approximately 10–20 km from the coast, although well-structured surface cold filaments are not as common as in the austral summer [Barbieri et al., 1995]. The deflection of the isopycnals and isotherms suggests that upwelled waters come from the upper 100 m of the water column [Wyrtki, 1963; Morales et al., 1996]. These waters are normally low in oxygen and high in nutrients [Marin et al., 1993; Morales et al., 1996], and are therefore able to support the relatively high phytoplankton biomass (pigment concentration) and primary production near the coast [Forsberg and Joseph, 1963; Marin et al., 1993; Thomas et al., 1994; Morales et al., 1996; González et al., 1998; Thomas, 1999; Daneri et al., 2000] and within the bays [Marin et al., 1993]. However, little is known about the interception of the upwelled CO2 by biological sequestering [see Wilkerson et al., 1987]. Though satellite SST observations show higher SSTs during the 1987 and 1992 El Niños, they also suggest that the number of upwelling episodes off the Mejillones peninsula increases during this phase of the ENSO cycle [Barbieri et al., 1995]. However, the impact of the warm periods on CO2 fluxes has not been evaluated in coastal upwelling areas off Chile. In this work we estimate these fluxes for July 1997, during the development phase of the strongest El Niño of the last century [McPhaden, 1999]. We also present the hydrography and carbonate system parameters for the upper 200 m of the water column, and discuss the key hydrographic, meteorological and biological factors involved in the ventilation of CO2.

Figure 1.

Study area and oceanographic stations; numbers identify regular stations. The locations of the Antofagasta tidal gauge and the Caleta Constitucion meteorological station are indicated by white arrows. Solid lines depict the 500 m and 1000 m isobaths.

2. Methods

2.1. Oceanographic Stations

[5] Discrete seawater samples and CTD data were collected onboard the R.V. Abate Molina off northern Chile (22.6°–24°S), during the 1997 austral winter. We report here on two grid surveys carried out between 1–4 and 18–21 July, respectively. The initial survey (grid 1) comprised 31 stations arranged as five east-west transects covering an area of approximately 21000 km2 (Figure 1). During the second survey (grid 2) we repeated the three northernmost transects of grid 1 (stations 1–19, Figure 1). At each station, Neil Brown Mark III CTD-oxygen-rosette casts were made to varying depths; results are reported here only for the upper 200 m of the water column. Water samples from 11 or 12 depths were taken for pH, alkalinity and several other measurements as described by Rutllant et al. [1998] and González et al. [1998]. The CTD oxygen probe was calibrated using a semiautomatic Winkler method [Williams and Jenkinson, 1982]. In addition, a continuous pumping system whose inlet was located at about 2 m depth was used for semicontinuous measurements of fCO2 in surface seawater.

2.2. CO2 Measurements and CO2 Flux Calculations

[6] The estimation of fCO2 in surface seawater (ca. 2 m depth) was based on the equilibration of a carrier gas (air) with seawater flowing continuously through an equilibrator (described by Cooper et al. [1998]), and its subsequent determination of the mole fraction of CO2 (XCO2) in the carrier gas by infrared absorption using a Licor 6262 instrument mounted in a automated measurement system [Cooper et al., 1998], fCO2 was calculated from XCO2 and in situ sea temperature following Dickson and Goyet [1994]. Intermittent estimation of atmospheric fCO2 every 7 min was used to calculate ΔfCO2 (fCO2 in the seawater − fCO2 in the atmosphere). Carbon flux (fCO2) was calculated as k * ΔfCO2, where the gas transfer coefficient k was calculated following Wanninkhof and McGillis [1999] on the basis of the wind speed at 10 m above sea level.

2.3. Coastal and Onboard Wind Speed

[7] The wind speed and direction were sampled every 15 s and averaged over 30 min periods at a Campbell automatic meteorological station located at Caleta Constitucion (Figure 1). The strong diurnal cycle of the coastal winds results in low wind speeds between midnight and noon. Since this effect quickly dampens offshore [Beardsley et al., 1987], the wind forcing of the upwelling at Caleta Constitucion was assessed using the mean of the relatively constant afternoon winds (between 1400 and 2000 local time) on an alongshore axis (009°–189°). The turbulent transfer coefficients used to estimate the CO2 flux were calculated from onboard wind measurement with a similar Campbell meteorological station sited above the bridge about 12 m above sea level. Corrections to the apparent wind speed measured while under way were made automatically using data from the ship's GPS system.

2.4. Measurement of pH and Alkalinity; Calculation of fCO2 and CT

[8] Seawater samples were collected in 5 L Niskin bottles mounted on the CTD rosette. Samples for pH measurement were taken in 50 mL syringes using a Tygon tube and then transferred to a closed 25 mL cell thermostated at 25 ± 0.1°C. pH was measured with a Metrohm 713 pH meter (input resistance > 1013Ω, 0.1 mV sensitivity and nominal resolution 0.001 pH units) using a glass combined double junction Ag/AgCl electrode (Metrohm model 6.0219.100) calibrated with 8.089 tris and 6.786 2-aminopyridine buffer [Dickson and Goyet, 1994] at 25 ± 0.1°C; pH values are therefore reported on the total hydrogen ion scale [Dickson and Goyet, 1994]. The overall uncertainty in the measured pH values was estimated by Torres et al. [1999] as 0.006 pH for surface waters (pH near 8) and less than 0.009 pH for very acid waters (pH ≈ 7.2). Samples for total alkalinity (AT) were collected at five stations at different depths in the 0–200 m depth range. 250 mL samples were drawn into high-density polypropylene bottles, fixed with 20 μL of saturated mercuric chloride solution, and stored in darkness for three weeks before analysis using an automated potentiometric titration method [Haraldsson et al., 1997]. The AT values from the 0–200 m depth range were used to determine a linear regression between total alkalinity and salinity in order to predict the alkalinity for the remaining stations in the same depth range (Figure 2). In order to assess the error in the other (calculated) carbonate system parameters, we use the average of the absolute difference between measured and estimated AT values (6 μmol kg−1) as the uncertainty of each extrapolated surface AT value. This gives uncertainties of 9 μmol kg−1 and 9 μatm for CT and fCO2, respectively [Torres et al., 1999]. Calculation of CT and fCO2 from pH and AT followed the methods given by Dickson and Goyet [1994], which use dissociation constants measured in artificial seawater. Although recent studies indicate that there are consistent differences in the dissociation constants of carbonic acid in natural and artificial seawater [Wanninkhof et al., 1999; Lee et al., 2000], these differences are too small to have a significant effect on the results reported here.

Figure 2.

Relationship between alkalinity and salinity in the upper 200 m of the water column. The error bars correspond to one standard deviation estimated from duplicate or triplicate measurements. The fitted linear regression through the origin has an absolute mean residual of 5.5 μmol kg−1: a regression including an intercept gave only slightly lower residuals (absolute mean 5.0 μmol kg−1) and was not used.

2.5. Historical Hydrographic Data, Time Series Records, and Satellite Data

[9] Data for temperature, salinity and dissolved oxygen collected in previous expeditions were supplied by SHOA (Servicio Hidrográfico y Oceanográfico de la Armada, Chile). Sea level and temperature data from the tide gauge at Antofagasta (23.6°S) were obtained by ftp from TOGA Sea Level Center (University of Hawai'i, http://uhslc.soest.hawaii.edu) and from SHOA, respectively. The daily coastal sea surface temperature consists of a single daily measurement of temperature: the validation of this data set is given by Fonseca [1985]. Sea surface temperature anomalies in the El Niño 1 + 2 area (0°–10°S, 90°–80°W; see Figure 1) were obtained by ftp from the NOAA Climate Prediction Center (http://www.cpc.ncep.noaa.gov). Values for the standardized Southern Oscillation Index (SOI) were obtained by ftp from IRI/LDEO Climate Data Library (http://ingrid.ldgo.columbia.edu).

[10] Pathfinder sea surface temperature data derived from the advanced very high resolution radiometer (AVHRR) on board the NOAA Polar Orbiters were obtained from the Jet Propulsion Laboratory (JPL) by ftp (http://podaac.jpl.nasa.gov/sst/). A description of the methods and of the validation of this data set is given by Smith et al. [1996].

[11] Wind stress data derived from space-borne scatterometers (NSCAT-ADEOS and AMI-ERS) are distributed (on CD) from Département d'Océanographie Spatiale, IFREMER, France (http://www.ifremer.fr/droos) as a WOCE contribution. The validation of these measurements is described by Graber et al. [1996] and Atlas et al. [1999].

3. Results and Discussion

3.1. Warm and Cold Periods at a Coastal Upwelling Area Off Northern Chile (23.5°S)

[12] Observationally, the most striking features of the ENSO cycle are the positive SST anomalies (El Niño warm phase) and the negative SST anomalies (La Niña cold phase) in the central and eastern equatorial Pacific. The last three warmest Julys (austral winters 1972, 1983, and 1997) observed around 23.6°S (Antofagasta; Figure 3a) and off the west coast of South America (i.e., El Niño 1 + 2 region; Figure 3a) have been associated with three very strong El Niño events. The austral winters of 1972 and 1997 correspond to the development stages of El Niños 1972–1973 and 1997–1998, respectively, while the austral winter of 1983 marked the end of El Niño 1982–1983 (Figure 3b). During these three events, the anomalies in SST and sea level (SL) remained positive for several months, however the major changes in SST occur after the corresponding SL changes (Figure 3b).

Figure 3.

(a) July means of the SST anomaly in the El Niño 1 + 2 area (Figure 1) and SST at the Antofagasta station, from 1968 to 1998. (b) Daily SST and SL anomalies at Antofagasta from 1 March to 31 October for the years 1972, 1983, and 1997. The anomaly for a given day was calculated as the difference between the measured value and the average for that date of a 51 years (1947–1998). The resulting daily anomalies were filtered using a 121-Lanczos cosine filter. (c) Vertical sections measured at the times indicated by solid circles in Figure 3b. The station positions are shown in Figure 1; for 1997, data from stations 20–26 were used.

[13] Across-shelf transects carried out during these warm periods (Figure 3c) show that the pycnocline and thermocline weaken toward the coast especially during the development stages of the El Niño (i.e., 1972 and 1997). The isopycnals and isotherms slope upward at depths less than 50–70 m indicating upwelling, at depths greater than 50–70 m the isopycnals and isotherms slope downward. In spite of the divergence of isopycnals and isotherms toward the coast, the main portion of the oxycline (AOU < 200 μmol kg−1) remained within the upper 100 m (Figure 3c), allowing water undersaturated in oxygen to well up to the surface close to the coast.

[14] The hydrographic characteristics of the El Niño warm phase during austral winter contrast strongly with the La Niña cold period (e.g., July 1989). During July 1989 (La Niña) the entire upper 100 m of the water column was 0.6–0.8 kg m−3 denser than in July 1997 (Figure 4a) despite a decrease in salinity (Figure 4b), indicating that the change in water temperature (Figure 4c) is the cause of the density increase. During the July 1989 cold period the water column below 20 m depth had a greater oxygen deficit than in July 1997 (Figure 4d). The greatest differences in AOU (80–100 μmol kg−1) were found at 45–80 m depth at the stations closest to the coast, an area which remained poorly oxygenated even during the warm period (100–160 μmol kg−1 AOU) as well as in other El Niño austral winters (Figure 3c). Consequently, even during warm periods upwelling pulses can maintain surface water undersaturated in oxygen, but the levels of O2 undersaturation within the upwelling centre can be lower than during colder periods, assuming the same wind driven upwelling regime. Since the oxycline depth virtually coincides with the maximum vertical gradient in pH, fCO2 and CT [Copin-Montégut and Raimbault, 1994] (also see section 3.5.) it is highly probable that during El Niño periods (as 1972, 1983, and 1997) the main gradients in pH, fCO2 and CT are also confined to the upper 100 m as we will document for July 1997.

Figure 4.

Vertical sections of (a) σt, (b) salinity, (c) temperature, and (d) AOU on 16 July 1989 and 2–3 July 1997 (stations 14–17, see Figure 1) off the Mejillones Peninsula at 23.3°S.

3.2. Wind Patterns and Hydrographic Anomalies at 23°–24°S During the Onset of El Niño 1997–1998

[15] The Southern Oscillation Index (SOI), dropped precipitously to negative values beginning in March 1997, when the SST anomalies were weakly positive (≈1°C) in the study area and in the El Niño 1 + 2 region (Figure 1 and Table 1). A positive SST anomaly was observed at the Antofagasta station beginning in late April (Table 1) and continuing during May, when the monthly average of the alongshore wind stress was especially weak and the SL was particularly high (Table 1).

Table 1. Meteorological and Hydrographic Anomalies During Early 1997
 Feb.MarchAprilMayJuneJuly
  • a

    22.3°–24.3°S; 70.8°W–71.8, based on 8 years of satellite data.

  • b

    Based on 50 years of measurements.

  • c

    Alongshore wind stress anomalies between 23°–24°S and 71°–72°S, based on 6 years of satellite data from ERS-1 and 2.

SOI1.61.1−0.9−1.8−2−1
SSTA (El Niño 1 + 2 area, °C)011344
SSTA (study area, °C)a011233
SSTA (Antofagasta, °C)b000233
SLA (Antofagasta, cm)b371217109
WSA (study area, mPa)c810−4−241713

[16] However in June 1997, while the SST anomalies reached a maximum of about 3°C in the study area (Table 1), the upwelling favorable winds were particularly strong (Table 1). Both the local and large-scale wind data suggest that a prolonged and relatively strong upwelling favorable wind event occurred between 17 June and 3 July (Figure 5), although during early June short episodes of winds reversals were observed (Figure 5).

Figure 5.

Alongshore wind at Caleta Constitucion. (time 1400–2000, solid line). The alongshore wind between 22.5°–24°S and 71°–73°S is derived from NSCAT scatterometer measurements (dashed line).

[17] The SST, SL and offshore wind anomalies remained positive throughout July 1997 (Table 1). Coastal wind strength was also above normal [Rutllant et al., 1998]. The grid 1 survey (1–4 July) was carried out in rough sea conditions following 13 days of upwelling favorable winds generally exceeding 6 m s−1 (Figure 5). During the survey the afternoon winds remained strongly upwelling favorable with the exception of the last day when the alongshore winds reversed. The second survey (grid 2) was carried out immediately after a period of relatively weak but upwelling favorable winds which continued throughout the survey (Figure 5).

3.3. Transects at 23.0°S in July 1997: Subsurface Distribution of CO2

[18] Off Punta Angamos (23°S, Figure 1) the uplift of isopycnals and isotherms from about 50 m depth (T1, Figure 6) suggests an active upwelling during early July, although the upwelled water was 3°–4°C warmer than in La Niña years (Figure 4). Below about 70 m depth the isotherms and isopycnals sloped steeply downward toward the coast.

Figure 6.

Vertical sections for the 23°S transect (stations 8–13) of the two grid surveys (Figure 1). Transects T1 and T2 were measured during grids 1 and 2, respectively.

[19] The salinity was high and constant near the coast (34.8–35.0 in the upper 200 m). Another striking feature of the salinity sections is the minimum centered at 70–80 m depth in the oceanic stations. This “Shallow Salinity Minimum” [Reid, 1973], indicates the intrusion of subantarctic water into the study area. The main vertical gradients in pH (measured at 25°C), fCO2, CT and AOU were located between the base of the mixed layer and about 100 m depth (Figure 6). Within this depth range the fCO2 isolines shoaled toward the coast, suggesting upwelling of CO2 supersaturated water.

[20] During grid 2 (T2), the 13°–14°C isotherms are shallower than in T1, although the upper isotherms (i.e., 19°–17.5°C) deepen in the same period. The observed subsurface cooling close to the coast (>50 m depth, stations 11–13, Figure 6) was associated with advection of subantartic water (note the reduction in salinity) which in turn was associated with a moderate reduction in fCO2 at 50–100 m depth range (Figure 6). Certainly Figure 7 shows that the less salty water found within the 25.5 to 26.0 σt range (oceanic, Figure 7a) was characterized by relatively low fCO2 and AOU values (Figures 7b–7d) compared to the higher-salinity waters (coastal). Continuous records of dissolved oxygen and salinity allowed us to obtain the relationship between AOU and salinity at two constant σt ranges 26.0 ± 0.01 (n = 30) and 25.5 ± 0.01 (n = 31). Both relationships (Figures 8a and 8e) show that the AOU is proportional to salinity at a given density. The horizontal property distributions on both isopycnals (Figure 8) show that while the 26.0 isopycnal deepens sharply toward the coast (Figure 8b), the 25.5 isopycnal remains at about 60 m depth (Figure 8f), however in both cases salinity and AOU increase toward the coast (Figures 8c–8d and 8g–8h, respectively). It is clear that the oxycline and carboncline does not necessarily rise together with the thermocline or pycnocline since the water mass characteristics can vary substantially. For example, the rise in the CO2/density and AOU/density ratios toward the coast (Figures 7 and 8) results in the fCO2 isolines sloping upward toward the coast even as the isopycnals deepen (Figure 6).

Figure 7.

(a) T-S diagrams where isopycnals are depicted by solid lines and (b) AOU-density, (c) fCO2-density, and (d) CT-density relationships. Coastal (<70.8°W) and oceanic (>71.6°W) observations are depicted by crosses and circles, respectively.

Figure 8.

AOU-salinity relationship and the horizontal distribution of properties at two density surfaces: (top) σt = 26.0 and (bottom) σt = 25.5. The horizontal distributions of depth, salinity, and AOU at each isopycnal are shown in the contour diagrams.

3.4. Surface Hydrography and Air-Sea CO2 Fluxes

[21] The synoptic satellite observations of SST (9 × 9km resolution) showed especially warm surface waters during early June 1997 (e.g., SST > 20°C on 9 June, Figure 9a) when the equatorward winds were weak or even reversed (poleward winds; Figure 5). In contrast, the persistent upwelling favorable winds observed during late June (Figure 5) were accompanied by particularly cold surface water around Punta Angamos (SST < 18.5°C during 25–29 June, Figure 9a). In July 1997, the in situ SST showed that the relatively cold surface water was located very close to the coast and in the south west edge of the study area (Figures 9b and 9C1), agreeing reasonably well with the SST distribution suggested by remote sensing (Figures 9c and 9g), although those measurements were not exactly simultaneous with the sea truth data. Surface warm water tend to be more salty (Figures 9d and 9h).

Figure 9.

(a) Satellite SST within an 81 km2 area (1 pixel) centered on 22.98°S, 70.44°W near Punta Angamos. The map shows satellite SST data for 9 × 9 km pixels on six selected dates. Surface fields of (b)–(c) temperature, (d) salinity, and (e) ΔfCO2 during grid 1. Surface fields of (f)–(g) temperature, (h) salinity, and (i) ΔfCO2 during grid 2. In situ SST is derived from continuous measurements; satellite SST is derived from 9 × 9 km resolution data collected on 6 July (Figure 9c) and 18 July (Figure 9g); salinity is derived from CTD profiles (3–5 m depth range); and fCO2 is derived from semicontinuous underway measurements of surface water fCO2 and atmospheric fCO2.

[22] During grid 1, the colder coastal water (70.8°–70.6°W; SST < 18.5°C) was more strongly CO2 supersaturated (ΔfCO2 = 280 ± 45 μatm; n = 13) than the warmer surface water (SST > 18.5°C) observed within the same area (ΔfCO2 = 133 ± 44 μatm; n = 313; Figures 9b–9e). Certainly the cooling near the coast was associated with a strong ΔfCO2 increase and a reduction in pH (correlation coefficients for ΔfCO2 and pH25 with temperature were −0.8 and 0.9, respectively, for longitudes westward of 70.8°W). A different situation was observed further offshore where the surface fCO2 in colder and less salty oceanic water (>71.8°W; temperature < 18.5°C) was closer to equilibrium with the atmosphere (ΔfCO2 = 21 ± 7 μatm; n = 47) than in the warmer oceanic water (ΔfCO2 = 33 ± 13 μatm, SST > 18.5°C, n = 237). Moreover a conspicuous surface ΔfCO2 maximum close to station 18 (Figures 9b–9e and Figure 1) occurred in moderately warm and salty water.

[23] During grid 2 the colder coastal water (70.8°–70.6°W, SST < 18.5°C) remains strongly CO2 supersaturated (ΔfCO2 = 193 ± 33 μatm; n = 101) compared with the warmer surface waters (ΔfCO2 = 94 ± 47 μatm; n = 91; Figure 9i). However, the surface ΔfCO2 in both warm and cold surface waters has decreased by more than 30% compared to grid 1. At the oceanic stations the fCO2 decreases further and large areas become CO2 undersaturated (Figure 9f–9i; see also Figure 6).

[24] In general, a reduction in ΔfCO2 from 81 μatm to 5 μatm (about 93%) was observed between grid 1 and grid 2 in the area sampled in both grids (70.6°–71.9°W, 22.6°–23.4°S, ∼12800 km2), while the mean transfer velocity k (calculated from onboard wind speed records, Figure 5) showed little difference between the two periods (0.044 and 0.036 mol C yr−1 m−2 μatm−1 for grids 1 and 2, respectively). Thus the area average CO2 flux drops from 0.55 Tg C yr−1 (mean weight flux = +3.9 mol C m−2 yr−1; Figure 10) to 0.02 Tg C yr−1 (mean weight flux = +0.4 mol C m−2 yr−11; Figure 10) for grids 1 and 2, respectively.

Figure 10.

CO2 fluxes during grids 1 and 2 (mol C m−2 yr−1); positive values depict a flux toward the atmosphere.

3.5. Relationship Between CO2 System Parameters and AOU

[25] CT, pH25 and fCO2 show similar distributions to AOU, and are better correlated with AOU (r = 0.89, −0.92, 0.90, respectively, n = 617) than with density (r = 0.77, −0.78, 0.57, respectively, n = 617), or other hydrographic variables in the upper 200 m of the water column. The relationship with CT arises because aerobic respiration consumes O2 and produces CO2, increasing both AOU and CT. The equation ΔCT = 0.83 ΔAOU + 0.5 ΔAT has been derived from previous studies [Brewer, 1978; Goyet and Brewer, 1993]. In this study the variation in AT is very small, especially near the coast (Figure 2) so that most of the CT variability should be explained by the variability in AOU, as is confirmed by the correlation coefficients given above. In addition, the equally good correlations of pH25 and fCO2 with AOU also indicate that alkalinity variations are not significant. In order to make an accurate assessment of AOU variability we have used AOU values derived directly from Winkler titrations in order to avoid errors associated with the oxygen electrode mounted on the CTD. The linear relationship between CT (μmol kg−1) and Winkler-derived AOU (μmol kg−1) has a mean slope of 0.89 (r = 0.99, n = 12) in the upper 200 m, close to the slope quoted above. Although this relationship can be eroded toward the surface because of exchanges of CO2 and oxygen between the ocean and the atmosphere, within the oxycline or below it (at least down to 200 m depth) the AOU variations (presented in this work) provide a relative good indicator of the CT variations (carboncline) as observed in others coastal upwelling areas [Simpson, 1986; Copin-Montégut and Raimbault, 1994].

3.6. CO2 Outgassing Continues During July 1997

[26] The first SOI minimum of the 1997–1998 El Niño occurred in May–June 1997, at a time when the local equatorward alongshore winds (which force the upwelling) shifted from weak to strong (Table 1). However, in spite of those particularly strong upwelling favorable winds (e.g., at the end of June and during July 1997; Figure 5) the SST anomalies remained strongly positive (+3°C). We conclude that the upwelling at that time was supplied by anomalously warm water, and note that the source of upwelled water (within the upper 100 m of the water column) was both warmer and saltier than in colder (La Niña) periods.

[27] The main hydrographic anomalies (high coastal SST and salinity) indicate a southward displacement of the water properties along the coast, as has been reported for previous strong El Niño events (e.g., El Niño 1972) [Bernal et al., 1982]. These hydrographic anomalies (onset of El Niño 1972 and 1997) are explained by enhanced poleward transport near the coast, consistent with ocean currents measured on previous El Niño onsets (1976 and 1982–1983) [Smith, 1983; Huyer et al., 1991], which suggest the intensification of the poleward current on the shelf off Peru when the sea level and temperature rise. A recent report of geostrophic surface currents derived from sea surface height observations (TOPEX/Poseidon and ERS-2 altimeters) shows a strong poleward flow during May–June 1997 off northern Chile [Strub and Corinne, 2000]. Thus it seems that during the study period (i.e., the onset of El Niño 1997–1998) upwelling continues off the Mejillones Peninsula in response to relatively strong equatorward winds (Table 1), but the upwelled water was salty and warm, having apparently been advected south along the coast. This water has a similar T-S shape when compared to waters found at lower latitudes (as ca. 15°S) during non-El Niño austral winter period [e.g., Copin-Montégut and Raimbault, 1994].

[28] This particularly salty and warm coastal water was characterized by a high CO2/density ratio (Figure 7), with the result that the carboncline and the oxycline remained within the upper 100 m, although this entire layer was less dense than during the cold phase (e.g., Figure 4). Since the carboncline was not particularly deep, increases in the upwelling favorable winds were able to induce coastal upwelling of already CO2 supersaturated water. ΔfCO2 ranged generally between 100 and 200 μatm near the coast (Figures 9b–9i), resulting in a strong coastal CO2 flux to the atmosphere exceeding 10 mol C m−2 yr−1 in both grid surveys (Figure 10). This figure does, however conceal strong spatial and temporal variability which is discussed below. Although no CO2 measurements have been reported for this area in a non-El Niño austral winter, the strong oxygen depletion observed in the upper 100 m of the water column during a typical cold La Niña austral winter (e.g., 1989, Figure 4d) suggests that during cold periods the recently upwelled water can be less oxygenated and thus more strongly CO2 supersaturated. However, the strong short-term variability of the CO2 fluxes [e.g., Torres et al., 1999], and the scarcity of CO2 data in the region prevent any estimation of anomalies in the local CO2 fluxes.

3.7. Variability of Surface fCO2 and CO2 Flux During July 1997

[29] The high intramonth variability in the CO2 flux revealed by the difference between grids 1 and 2 can be explained by changes in the local wind field (Figure 5), which affect both (1) the upwelling and (2) the ventilation of CO2 through changes in the transfer velocity k.

[30] 1. Changes in the upwelling of CO2 rich waters are as follows. The persistence and strength of the upwelling wind forcing in late June suggest that the strongest upwelling occurred on about 24–26 June (Figure 5), at which time the SST in the coastal water reached a minimum (<18.5°C over an area of 400–500 km2; Figures 8a–8d). Our measurements from grid 1 suggest that this relatively cold coastal water was invariably highly CO2 supersaturated, so the strong CO2 supersaturation observed in most of the study area during grid 1 was most probably associated with a prolonged and strong upwelling pulse in late June. Conversely, on grid 2 the alongshore wind before the survey was less constant and in general less intense, decreasing to a minimum of 2 m s−1 just before grid 2 (Figure 5).

[31] 2. Changes in CO2 ventilation (transfer velocity k) are as follows. We can identify periods of very low wind stress (i.e., 4, 9–10, and 17 July; Figure 5) and very high wind stress (i.e., 2–3, 11–12, and 18–19 July; Figure 5). These rapid wind speed fluctuations can cause order of magnitude variations in the transfer velocity k, which increases approximately as the cube of the wind speed [Wanninkhof and McGillis, 1999]. Since variations in wind speed can cause such large changes in CO2 flux, reliance on longer term averages of wind speed and k will lead to erroneous results [Wanninkhof, 1992]. In the specific case of grids 1 and 2, the mean k value was similar, although the alongshore wind which forces the upwelling (calculated from the coastal records) was weaker in grid 2 than in grid 1 (particularly if the last day of each grid is ignored because the response of the upper ocean lags the meteorological forcing by about 1 day) [Brink et al., 1980; Smith, 1981]. So the difference in the CO2 flux in grids 1 and 2 was mainly driven by changes in ΔfCO2, which are in turn related to variations in the upwelling intensity as discussed above.

[32] In addition, we have shown that over a period of 18 days a significant change in the depth of the 13°–16°C isotherms occurred (T1 to T2; Figure 6). However, this upward movement of the “lower thermocline” (which could be remotely forced) [Shaffer et al., 1997] is accompanied by a moderate decrease in fCO2 and CT in the 50–100 m depth range, which is the source of upwelling water. This is in contrast to the expectation that cold waters are CO2-rich and oxygen-poor. In this case the changes appear to be due to the introduction of less saline subantarctic water (see salinity variations in Figure 6) which has somewhat lower fCO2 and CT (Figure 7). Thus we suggest that movements of the pycnocline do not necessarily involve a corresponding movement of the carboncline, so that the interpretation of density changes or their indicators (e.g., sea level) in the context of CO2 (and also oxygen) distributions should take into account the variability of the water mass characteristics.

3.8. Biological CO2 Uptake

[33] Close to the coast the fertilization effect associated with the CO2-and nutrient-rich upwelled water will tend to enhance primary production (PP) and thus decrease fCO2. During grid 1, PP was estimated to be approximately 2 gC m−2 d−1 on the basis of measurements of 14 profiles [González et al., 1998]. This integrated PP is not sufficient to stop CO2 outgassing from the surface water to the atmosphere, although the surface gross primary production exceeds the community respiration [Daneri et al., 2000]. Consequently, the outgassing of CO2 during this period can be ascribed to the active supply of CO2-rich waters to the surface mixed layer. During grid 2, the occurrence of weak upwelling favorable winds (Figure 5), together with the tendency of the upper isopycnals to slope downward toward the coast (Figure 6), suggest a weaker supply of CO2-rich water due upwelling, which in association with the biological CO2-uptake could explain the observed reduction in the surface fCO2 from grid 1 to grid 2. At oceanic stations where some areas were strongly CO2 undersaturated (e.g., Figures 9f–9i), the primary production was higher and more variable that in pre-El Niño conditions (0.8 ± 0.3 gC m−2 d−1 and 1.6 ± 1.2 gC m−2 d−1 for January and July 1997, respectively) [González et al., 1998].

[34] At the coastal Stations, the PP decreased moderately from January (pre-El Niño, 3.2 ± 1.8 gC m−2 d−1) to July 1997 (El Niño, 2.0 ± 1.4 gC m−2 d−1) González et al., 1998], while respiration of the microplanktonic community decreased strongly within the same interval (from 2.3 to 0.7 gC m−2 d−1) [González et al., 1998]. Therefore the particularly high PP/respiration ratio in July 1997 [Eissler and Quiñones, 1999] provides conditions for effective uptake of the upwelled CO2. The development of CO2 undersaturated areas between grids 1 and 2 (Figure 8) cannot be explained by temperature changes (no sharp cooling observed), air-sea exchange (leads to equilibration, not undersaturation), or advection (no obvious supply of undersaturated water). Biological CO2 uptake may, however, provide an explanation for this rapid change. We can estimate the maximum effect of PP as follows: gross PP is 2 gC m−2 d−1, which is equivalent to 4.2 μmol C L−1 d−1 in a mixed layer 40 m deep; this amount of PP without any respiration taking place would reduce the surface water fCO2 by 120 μatm in 18 days (the approximate time period between the surveys). If we make allowance for respiration (as 1/3 of the PP) [see Eissler and Quiñones, 1999] we can conclude that the observed fCO2 reduction of the order of 50 μatm is consistent with the expected biological uptake, even though advection and other exchange processes will tend to reduce the impact on the surface waters measured in grid 2.

4. Conclusions

[35] 1. During the onset of El Niño 1997–1998 (July 1997) the carboncline remained within the upper 100 m even when this entire stratum was warmer allowing the pulses of strong southerly wind to promote the upwelling of anomalously warm but CO2 supersaturated water, producing a CO2 flux from the ocean to the atmosphere. The high short-term variability in the CO2 flux can be ascribed to the combined effect of the local wind fluctuations and to uptake of the upwelled CO2 by the phytoplanktonic community.

[36] 2. The oxygenation of the water column during the El Niño period, revealed by the absence of very low oxygen concentrations in the upper 100 m (Figures 3 and 4), suggests that fCO2 (positively correlated with AOU in the upper 200 m) can be higher within the recently upwelled water during La Niña cold phase. However, the anomalously intense alongshore winds observed during the study period (July 1997, El Niño warm phase) enhanced the CO2 outgassing by a increment in wind-driven mixing and upwelling forcing, thus the CO2 fluxes reported here are not necessarily weaker compared with non-El Niño periods.

Acknowledgments

[37] We acknowledge extensive cooperation from the participants of the FONDECYT 5960002 Program directed by Humberto González, especially Dante Figueroa (CTD data) and Renato Quiñones (oxygen data). We thank Oscar Pizarro and Marcus Sobarzo for fruitful discussions and for help in handling satellite data. We thank the Department of Inorganic Chemistry, Laboratory of Pelagic Ecology and EULA-chemistry Laboratory at the University of Concepción for provision of laboratory facilities. We thank Servicio Hidrográfico y Oceanográfico de la Armada de Chile (SHOA), Département d'Océanographie Spatiale IFREMER France, University of Hawai'i Sea Level Center, and the Jet Propulsion Laboratory for the supply of hydrographic (local or remote sensing) data. This work was mainly funded by FONDECYT 5960020 (DICLIMA Experiment) and by the concurrent 5960002 Program. Additional financial support was provided by Göteborg University, the University of Chile (PRODAC) and the FONDAP Humboldt Program.

Ancillary