Glider observations of the Northwestern Iberian Margin during an exceptional summer upwelling season


 <p>We present results from a 2 month deployment of an ocean glider over the Northwestern Iberian Margin. Glider observations during the exceptionally strong 2010 summer upwelling season resolved the evolution of physical and biogeochemical variables during two upwelling events. Upwelling brought low oxygen Eastern North Atlantic Central Water from 190 m depth onto the shelf up to a depth of 50 m. During the two observed periods of upwelling,<br>equatorward transport over the shelf increased from 0.13 (&#177; 0.04) Sv to 0.18 (&#177; 0.08) Sv and a poleward jet developed over the shelf-break. The persistent upwelling favourable winds maintained equatorward flow on the outer shelf for two months with no reversals during relaxation periods, a phenomenon not previously observed. During upwelling, near surface chlorophyll a concentration increased by more than 6 mg m<sup>-3</sup> . Dissolved oxygen concentration in the near surface increased by more than 40 &#956;mol kg<sup>-1</sup> , 6 days after the chlorophyll a maximum.</p><p><br>This was the first and, to date, only deployment of a glider over the North West Iberian Margin. A single glider was able to occupy a cross shelf section for two months, without the need for a costly ship based campaign. This presentation highlights some of the challenges of using gliders to study shelf break regions.</p>



Introduction
Eastern Boundary Upwelling Systems are some of the oceans' most productive areas, covering less than 1% of the ocean but accounting for up to 20% of the global wild fish take (Pauly & Christensen, 1995). The Northwestern Iberian Margin (NWIM) forms the northernmost extent of the Canary Current Upwelling System, an Eastern Boundary Upwelling System of the North Atlantic. The NWIM hosts a seasonally varying multicore flow that exhibits strong variability (Teles-Machado et al., 2015). With the northward movement of the Azores High and the intensification of the Icelandic Low in summer, episodic southward winds blow along the Iberian coastline (Nolasco et al., 2013;Peliz et al., 2002). These southward winds drive warm surface waters offshore by Ekman transport, inducing the upwelling of cooler, nutrient-rich water from as deep as 200 m and enhancing local primary production (Barton, 2001). During upwelling periods, typically 7-10 days Huthnance et al., 2002), the coastal sea level lowers, the thermocline rises, and an equatorward coastal jet develops over the shelf (Arístegui et al., 2009).
The NWIM extends 350 km along the west coast of the Iberian peninsula, from Cape Mondego to Cape Finisterre ( Figure 1). The NWIM consists of a shelf 50-60 km wide that slopes gently to the shelf break between the 200 and 300 m isobaths before dropping to 2,000 m over a distance of 15 km. The glider deployment area, in the neighborhood of Cape Silleiro, is known to feature intensified upwelling in summer (Huthnance et al., 2002;Relvas et al., 2007). The shelf and slope region host seasonally varying meridional flows detailed by Herrera et al. (2008) and Ferreira . Offshore, the North Atlantic subtropical gyre transports water equatorward in the broad, slow Portugal Current (Arístegui et al., 2009). The most variable current over the slope is the Iberian Poleward Current (IPC). The IPC transports water concentrations of zooplankton and pelagic fish, enhancing its biological and economic importance (Rossi et al., 2013). During upwelling events, substantial cross-shelf exchange can take place (Brink, 1998). These events of enhanced primary productivity and offshore transport are the focus of this study.
The 2010 summer upwelling season was unusually strong. Winds originated from the direction 0±45°(i.e., within 45°of north) for 82% of the deployment. The mean wind speed was 8.2 m s −1 , and the mean Upwelling Index (UI) was 950 (±40) m 3 km −1 s −1 . UI is an estimate of offshore Ekman transport. A positive value of UI is indicative of upwelling favorable conditions. UI for the region for each year over the same yearday range averaged 550 (±190) m 3 km −1 s −1 (Puertos del Estado, 2019). In 2010, UI was two standard deviations above the mean (Figure 2). Similarly strong upwelling conditions occurred during 1981, 2002, and 2016. These unusually strong conditions resulted in a summer dominated by upwelling.
To observe this variability during the upwelling season at high spatial and temporal resolution, an autonomous ocean glider was deployed at the NWIM during summer 2010. The deployment is described in section 2.1. Data processing and gridding are presented in section 2.2. In section 3, we present the results; in section 4, we discuss the results and make recommendations for future observational campaigns on the NWIM. In section 5, we summarize the key results.

Data Collection
From 1 June to 5 August 2010, Seaglider SG510 Orca occupied an onshore-offshore section at 42.1°N across the shelf and slope from 9.1 to 9.7°W ( Figure 1). Each passage through the section is referred to as a transect. Seagliders are small, buoyancy-powered vehicles that profile to 1,000 m with a sawtooth dive pattern (Eriksen et al., 2001). The glider profiled over bathymetry of 150 to 2,000 m. Individual dive cycle duration varied from 30 min on the shelf to 4 hr over the deep slope. During a dive the glider travels between 500 m and 4 km horizontally. Each dive cycle yielded two profiles, one when the glider was descending and one when it was ascending. The glider recorded measurements every 5 s above 200 m and every 10 s below 200 m. The glider has a typical vertical speed of 0.1 m s −1 , resulting in vertical sampling resolutions of approximately 0.5 and 1.0 m, respectively. The glider travelled zonally at 0.1-0.3 m s −1 relative to the ground.
The glider was equipped with a Paine Corporation pressure sensor, an unpumped Seabird CT sail, a WETLabs ECO Puck triplet sensor, and an Aanderaa 4330F oxygen optode. Transects covered on average 45 km and took 2-6 days. Transect time increased towards the end of the deployment due to biofouling that increased drag on the glider (Figure 1d). The glider completed the section 17 times. Some transects were truncated, but all were greater than 36 km (Figure 1c). Due to strong equatorward currents, the glider deviated meridionally from its intended zonal track with a standard deviation of 2.8 km (Figure 1c). Considering these deviations to be small, we have projected all samples onto a zonal section. We compared the temperature-salinity characteristics of all transects (not shown). Transects all sample the same water masses, even those with large meridional deviations. Transects 2 and 6 have been chosen as typical examples of non-upwelling and upwelling conditions, respectively. Transect 2 took place after a period of relaxation favorable conditions, whereas transect 6 was conducted at the peak of the first upwelling event.
The shelf break is defined as the 300 m isobath shown in Figure 1. Throughout the text, "shelf" refers to waters east of the shelf break, "slope" refers to waters west of the shelf break. Yeardays (YD) are used throughout, with 1 January 2010 as yearday 0. The first day of this deployment was 1 June 2010, yearday 151.

Data Processing and Gridding
The hydrodynamic flight model for the glider was tuned following the methods of Frajka-Williams et al. (2011). Dive average currents (DACs) were calculated from the difference between the glider's flight path calculated from GPS fixes at the beginning and end of each dive and the glider's flight path from the flight model. The flight model regression is very sensitive to drag coefficients, which varied greatly over the glider deployment. Parasitic drag increased by over 200% due to biofouling. To accommodate this, the glider flight model was calculated using batches of 30 dives, allowing the friction coefficients to vary over the 1,050 dives analyzed. The DACs were inspected for directional biases that can arise from a poorly calibrated compass, but no substantial differences were found. Thermal lag of the CT cell was corrected following Garau et al. (2011). These corrections were implemented with the UEA Seaglider Toolbox (Queste, 2014).
To remove tidal currents from the DAC time series, dives were separated into two subsets, onshore and offshore of the shelf break, following the method of Sheehan et al. (2018) who separated DACs into three spatial bins for tidal analysis. These two datasets, each comprising approximately 1 month of DAC observations, were treated as discontinuous time series, and harmonic analysis was used to extract the M 2 and S 2 tidal constituents. The combined M 2 +S 2 tidal current had a maximum amplitude of 0.5 cm s −1 over the slope and 2.0 cm s −1 on the shelf. The tidal constituents were validated against the TPXO tide model (Egbert & Erofeeva, 2002). The choice of two domains was made as the M 2 tidal component in the region varies substantially between shelf and slope (Quaresma & Pichon, 2013). Each bin also satisfies the Rayleigh criterion for distinguishing between the M 2 and S 2 tides with time series of greater than 14.8 days (Sheehan et al., 2018). For the purposes of this paper, DAC is assumed to be an approximate barotropic current where the glider sampled the full water column and an approximate vertical average current in the upper 1,000 m, where the bathymetry exceeded the glider's profiling depth. The M 2 and S 2 tidal constituents were subtracted from the DACs before using the DACs to reference geostrophic shear. DACs are typically considered accurate to within 1 cm s −1 (Eriksen et al., 2001;Merckelbach et al., 2008). Acknowledging that this detiding will not remove all tidal constituents from the DACs, we have incorporated a 2 cm s −1 uncertainty in our calculations of geostrophic currents. This uncertainty in geostrophic velocity is used for uncertainty estimates in alongshore transports.
The WETlabs ECO Puck measures fluorescence as a proxy for chlorophyll a concentration (henceforth chlorophyll). The ECO Puck excites chlorophyll by emitting at 470 nm and records florescence at 695 nm. Chlorophyll fluoresces at a range of wavelengths centered on 682 nm (Maxwell & Johnson, 2000). The chlorophyll florescence data are calculated using a linear equation y = m(x−c), where y is chlorophyll concentration (mg m −3 ) and x is the sensor output (counts). We used the manufacturer supplied gradient m = 0.121 and a dark counts level c = 48, 8% lower than the manufacturer supplied value, such that the sensor registered 0 chlorophyll at depths greater than 150 m. An in-water calibration was carried out with co-located CTD casts on 1 June (YD 151), 29 June (YD 179), and 29 July (YD 210) (Brown, 2013). Chlorophyll values were corrected for the effects of non-photochemical quenching following the methodology of Thomalla et al. (2018). As the principal interest of this study is the cross shelf and temporal variability of chlorophyll, we are not aiming for an approximation of chlorophyll concentration better than a factor of two.
The Aanderaa optode is a low-power foil-type sensor as described by Alkire et al. (2012). Dissolved oxygen concentration was calculated using manufacturer calibration constants. The oxygen concentration was then corrected for temporal drift by applying a linear correction in time such that oxygen concentrations at 850-950 m depth remained constant in time. Winkler bottle samples were used to calibrate the ship CTD O 2 sensor on 29 July (YD 210), 15 September (YD 257), and 29 September (YD 271). This calibration was applied to CTD casts on 1 June (YD 151), 29 June (YD 179), and 29 July (YD 210) (Brown, 2013).
Temperature and salinity data for each transect were interpolated with an Objective Analysis Barnes function (Barnes, 1994) onto a grid with spacing 1 km horizontal by 1 m vertical, using a horizontal smoothing distance of 8 km and vertical smoothing of 8 m. This horizontal distance was chosen as it is the first internal Rossby radius of deformation over the shelf slope at the middle of the section. These gridded values were then used to calculate the potential density, absolute salinity, and conservative temperature using the Gibbs Seawater toolbox (IOC, SCOR, & IAPSO, 2010). We found the geostrophic velocity field calculated from these interpolated data to be largely insensitive to smoothing distances from 0.5 to 15 km. Dissolved oxygen concentration and chlorophyll concentration were gridded using the same methodology. Geostrophic currents were calculated from thermal wind, using the detided glider DACs as a reference velocity. The geostrophic approximation is commonly used with glider datasets in upwelling regions (Pietri et al., 2013;Todd et al., 2011), with estimated uncertainties of 1-2 cm s −1 . Geostrophic currents calculated with this method compare well with ADCP data (Pietri et al., 2013). Bottom velocities were nearest neighbor extrapolated to fill gaps between glider sampling and bathymetry over the shelf and slope, with no extrapolation past the maximum measurement depth (1,000 m). A Monte Carlo method was used to estimate uncertainty in the alongshore transports by applying random Gaussian noise with a standard deviation of 2 cm s −1 to the DACs, the largest source of error in estimation of geostrophic currents from glider data.
UI for the Rías Baixas is calculated by the Puertos del Estado at 6 hr intervals using the FNMOC model (Puertos del Estado, 2019). Satellite sea surface temperature (SST) from CMEMS Atlantic European North West Shelf Seas-Reprocessed SST Analysis-ODYSSEA from AVHRR Pathfinder v5.3, a daily satellite product with a spatial resolution of 0.04°. Chl a satellite data from MODIS (Hu et al., 2012), daily satellite product with a 0.0104°spatial resolution. Bathymetry from EMODnet is used in this study (EMODnet Bathymetry Consortium, 2018).
We use units of conservative temperature and absolute salinity following IOC, SCOR, and IAPSO (2010). All densities are potential density anomalies σ θ = potential density − 1,000 with units of kg m −3 . Oxygen supersaturation, Δ(O 2 ) is calculated as where c(O 2 ) is the measured O 2 concentration and c eq (O 2 ) is the O 2 concentration at an absolute pressure of 101,325 Pa, calculated with potential temperature and salinity (Garcia & Gordon, 1992. A positive value represents oxygen supersaturation, and a negative one represents oxygen undersaturation.

Initial Conditions
Prior to upwelling (Figure 1d, transects 1 and 2), conditions across the section were typical of relaxation. Isopycnals were near horizontal, with a plume of warm >18°C, low salinity <35.9 g kg −1 , low density σ θ < 26.0 kg m −3 water occupying the upper 20 m over the shelf and slope (Figures 3c, 4a, and 4b).
Vertical chlorophyll and Δ(O 2 ) distributions were similar across the section, with a subsurface chlorophyll

First Upwelling Event
The first upwelling event began on 14 June (YD 164, Figure 6). This occurred during transects 4-7 of the deployment (Figures 4 and 5). The onset of upwelling was first apparent in the increase in the equatorward current over the outer shelf from 3 to 8 cm s −1 during transect 4 (14 June, Figure 6a). The buoyant plume of water was advected 30 km offshore in 4 days (Figures 4a and 4c). The front between the warm water of the buoyant plume and cooler upwelled water moved offshore at approximately 0.1 m s −1 , consistent with previous observations of frontal advection during upwelling (Rossi et al., 2013). This current speed is comparable to that of the glider and prevented it from reaching its eastern waypoint on transect 4 ( Figure 6). Because of this, we have no glider observations for the shelf more than 14 km inshore of the shelf break for 10 days. This period coincided with the wind speed peak of the first upwelling event. Satellite data show surface cooling and elevated chlorophyll a concentrations during this time period (Figures 6c and 7c). After the offshore advection of the buoyant plume, near-surface waters over the shelf became cooler and more saline (Figures 4g and 4i). Temperature near the surface decreased by as much as 3.0°C at the eastern end of the section (Figure 6a). An across slope temperature gradient of 0.1°C km −1 in the upper 20 m was observed (Figure 3b) typical of a front between warm surface and cool upwelled waters (Ferreira Cordeiro, 2018). Over the shelf, the σ θ ¼ 27:1 kgm − 3 isopycnal shoaled from 180 m to shallower than 100 m (Figures 4b and 4h). A core of cool, saline water with temperature-salinity characteristics between those of ENACW st and ENACW sp was upwelled onto the shelf during transects 6 and 7 (17-23 June, Figures 4g  and 4i). The presence of this water on the shelf suggests upwelling of waters from depths of greater than 190 m, as has been observed previously (Huthnance et al., 2002). The change from near-horizontal isopycnals pre-upwelling to isopycnal slopes of 4 m km −1 across the shelf break is pronounced (Figures 3c and 3d). The σ θ ¼ 27:0 kgm − 3 isopycnal shoaled by 20 m over the shelf, similar to that observed during summer 2009 by Ferreira . During the upwelling event, isopycnals outcropped over the shelf break ( Figure 3f).
Prior to the first upwelling event, average chlorophyll concentrations were similar on the shelf and over the slope, though concentrations over the slope exhibited more variability (Figures 7 and 8a). Chlorophyll concentrations increased after the development of full upwelling, coincident with the decrease in near-surface temperature (Figure 7). Higher chlorophyll concentrations were observed over the shelf than the slope for the entirety of the upper 100 m during transect 6 ( Figure 8c). The subsurface chlorophyll maximum over

Journal of Geophysical Research: Oceans
the shelf shoaled to 12 m and near-surface concentrations surpassed 6.0 mg m −3 over the inner shelf more than 10 km inshore of the shelf break during transects 6 and 7 (Figure 7). Δ(O 2 ) followed the same pattern as chlorophyll but peaked during transects 8 and 9, 6 days later (Figures 7 and 9). Δ(O 2 ) increased most in the near surface over the shelf to greater than 28%. During transect 8, the greatest supersaturation was observed during the deployment (Figures 8f and 9). Chlorophyll and Δ(O 2 ) over the slope increased only slightly during the same time period (Figures 8c-8f).
After peaking during transect 7, maximum chlorophyll concentration over the shelf decreased to 2.3 mg m −3 , and the subsurface chlorophyll maximum descended to 43 m during transects 8 and 9. Δ(O 2 ) over the shelf also decreased, reaching a minimum during transects 10 and 11, 8 days later than the minimum of chlorophyll in the near surface (Figures 7 and 9).
A brief period of strong equatorward wind around YD 188 (8 July, transects 10 and 11) increased equatorward current speed on the shelf (Figure 6a). However, this event was short lived and caused only a modest decrease in temperature on shelf at the eastern end of the transect (Figure 6a). Small increases in chlorophyll (0.5 mg m −3 ) and Δ(O 2 ) (10%) in the near surface were observed during the following transects 12 and 13 (Figures 7 and 9). The effects of this period of increased winds were mainly limited to the inner shelf, more than 10 km east of the shelf break (Figures 6,7, and 9).

Partial Relaxation and Second Upwelling Event
A relaxation of the southward winds during  July, transect 13) brought surface warming of 2.0°C over the slope and a decrease in the strength of equatorward flows (Figure 6). This relaxation was During the final three transects of the deployment (21 July to 8 August), a second upwelling event developed. This second event followed a similar pattern to the first with increased equatorward currents over the shelf and upwelling of cold, dense water decreasing near-surface temperature by 2.0°C ( Figure 6). During the final transect, chlorophyll concentrations over the shelf increased to similar levels as observed during the first upwelling event (Figure 7). The highest chlorophyll concentration in the near surface were observed on the outer shelf at 5 km east of the shelf break (Figure 7). Assuming a similar lag between chlorophyll and (Δ(O 2 ) in the near surface as observed during the first upwelling event, it is likely that Δ(O 2 ) increased past the end of the deployment.

Geostrophic Currents and Transports
As expected for the NWIM, along slope flows dominated, as shown by the DACs (Figure 6). During the deployment, the wind was primarily perpendicular to the glider transect, so Ekman flow contributed little to along slope velocities. After detiding and gridding (section 2.2), we assume the velocity structure we observe to be dominated by geostrophic flow. The DACs include ageostrophic contributions from wind stress.

Journal of Geophysical Research: Oceans
The alongshore flow averaged horizontally and vertically over the entire section, and over its shelf and slope subsections, was equatorward in every transect, even though poleward jets were present (Figure 6a). Average equatorward transport was 0.17 (±0.07) Sv over the shelf and 0.83 (±0.6) Sv over the slope. Averaged over the 17 transects (not shown) surface intensification of southward flow over the shelf is apparent, particularly at the near-shore end of the section, with a maximum flow speed of 15 cm s −1 . A minimum in southward flow speed of 2 cm s −1 was observed near the sea floor at the shelf break. Over the slope, equatorward flow is strongest at 50-150 m, the typical depth of ENACW st . Equatorward flow weakens with depth, reaching a minimum flow speed below 700 m, at the depth of MEDW (Figures 3e and 3f).
During the 2 months of observation there was substantial variability in the strength of the equatorward transport. During upwelling, current speed increased over the shelf, and the flow became more surface intensified (Figures 4d, 4f, and 4h). Transport on the shelf increased from 0.13 (±0.04) Sv to 0.18 (±0.08) Sv during the two upwelling events. This flow is strongest at the near-shore end of the section (Figure 3f). The flow can be reasonably expected to extend further inshore, as has been observed in previous upwelling seasons on the NWIM Rossi et al., 2013), and therefore, our transport is likely an underestimate.
Over the slope, a broad equatorward flow dominated in the upper 500 m (Figure 3e). This flow was observed to weaken at depth, with a sporadic poleward flow below 500 m (Figure 3f). No relation was found between the UI and meridional transport over the slope. Opposing jet pairs were observed in the near surface throughout the deployment (Figures 4e and 4f). These moved offshore (westward) during the first upwelling event at 2 cm s −1 , similar to the upwelling event observed by Rossi et al. (2013).

Discussion
During June and July of 2010, the shelf and slope near Cape Silleiro experienced summer upwelling similar in character to that of previous years, but stronger (Relvas et al., 2007). In contrast to previous years, which featured cycles of upwelling and relaxation Rossi et al., 2013), upwelling conditions dominated the observational period. A cross-shore temperature gradient between cool upwelled water over the shelf and warmer surface waters offshore was present during the majority of the deployment ( Figure 6). Isopycnal outcropping was frequently observed over the shelf and upper slope (Figure 4).
During the first upwelling event, mean temperature of the upper 20 m of the water column over the shelf decreased by 2.5°C in less than 8 days ( Figure 6). Latent heat loss to the atmosphere averaged 120 W m −2 Previous studies have established upwelling as the mechanism by which cool, nutrient-rich water reaches the near surface over the shelf Relvas et al., 2007). Additionally, satellite SST data show that cool water features centered on the capes, including Cape Silleiro, expand and contract zonally but do not migrate meridionally (not shown). This is strongly indicative of upwelling, not advection. Surface cooling would not explain the observed changes in salinity, chlorophyll, Δ(O 2 ), and near-surface currents, which can only be the result of upwelling.
Upwelling raised the subsurface chlorophyll maximum and increased chlorophyll concentration throughout the mixed layer over the shelf and shelf break (Figure 8). The upwelled ENACW is relatively low in oxygen, but the high-nutrient concentration promotes phytoplankton growth (Rossi et al., 2013). We take elevated chlorophyll ( Figure 7) and optical backscatter (not shown) to be indicators of elevated primary productivity and biomass, while acknowledging that processes such as photoacclimation or changes in pigment packaging and ecosystem composition can influence chlorophyll without necessarily increasing primary productivity and biomass (Cetinić et al., 2015). Increased primary productivity would account for the observed increase of Δ(O 2 ) ( Figure 8). Δ(O 2 ) peaked after chlorophyll (and optical backscatter at 650 nm, not shown) over the shelf, with a delay of approximately 6 days. Oxygen in the mixed layer will also be affected by air-sea gas exchange, which will cause a further lag in the peak response of c(O 2 ) with respect to the chlorophyll concentration. This measurable time delay is therefore an important result. Chlorophyll concentration provides a convoluted signal of productivity and biomass, whereas oxygen concentration is an integrated signal of production, giving cumulative net community production. One would therefore expect the integrated signal of oxygen to reach its maximum after the peak in chlorophyll, as is shown in our glider observations.
During crossings 12 and 13, the local maxima of chlorophyll and Δ(O 2 ) coincided (Figures 7 and 9). This could indicate that another mechanism affects the concentration of oxygen in near-surface waters. This could be a physical effect such as bubble injection or a different ecosystem response to that which contributed to the delay between maxima in chlorophyll and Δ(O 2 ) observed after the first upwelling event. The absence of observed lag could also be a result of the relatively long transect sampling interval; the time between between crossings 12 and 14 over the inner shelf was 10 days. As is apparent in Figure 7c, chlorophyll over the inner shelf can increase and decrease in as little as 3 days.
During the first upwelling event, the chlorophyll maximum and near-surface temperature minimum were observed at the eastern end of the section. During the second upwelling event, the chlorophyll maximum and near-surface temperature minimum were observed 5 km east of the shelf break by the glider. The observations of minimum near-surface temperature and maximum chlorophyll near the shelf break during the second upwelling event may be due to observational limitations. During the second upwelling event the glider was travelling slowly due to biofouling. The glider reached the shelf break more than a day after its final sampling of the inner shelf. Satellite SST and chlorophyll data support this interpretation (Figures 6c and 7c). The bloom initiating with the second upwelling event spread further offshore than the first bloom (Figures 7a and 7c). Westward DACs over the outer shelf and shelf break may have contributed to this offshore spreading of chlorophyll (Figure 6a).
In deeper water (>50 m) over the shelf, Δ(O 2 ) decreased during and after upwelling events (Figure 8). A potential cause is the advection of low-oxygen ENACW sp onto the shelf. The upwelling of ENACW sp is visible in the temperature and density transects (Figure 3) in the shoaling of the σ θ = 27.1 kg m −3 isopycnal over the inner shelf. Biological activity also contributes to low-oxygen values in deeper water (Rossi et al., 2013). During upwelling, nutrients depleted by near-surface phytoplankton are replenished at depth by microbial remineralisation, consuming oxygen (Álvarez-Salgado et al., 1997;Rossi et al., 2013). Our observations of decreased Δ(O 2 ) below 50 m over the shelf agree with observations of near-bottom low-oxygen layers by Rossi et al. (2013).
The persistence of equatorward flow over the shelf throughout the deployment is atypical for the NWIM. Prior studies of the summer upwelling season have observed a reversion to poleward flow over the shelf 10.1029/2019JC015804

Journal of Geophysical Research: Oceans
during relaxation of equatorward winds Peliz et al., 2002;Rossi et al., 2013). The absence of poleward flow over the shelf in our observations may be due to the time taken for the glider to return to the shelf. After the relaxation of the southward winds during , the glider was not present again over the shelf until YD 199 (19 July, Figure 6). Ferreira Cordeiro et al. (2018)  Offshore of the shelfbreak, a strong equatorward flow persisted throughout the deployment. We do not observe the poleward flow of the Iberian Poleward Current seen in models (Teles-Machado et al., 2015) and observations Torres & Barton, 2007). This could be because the glider does not sample far enough offshore, turning around at 9.7°W over bathymetry of 2,000 m, midway down the slope (Figure 1). An observational campaign in June and July of 2009 only observed the poleward flow west of 9.8°W . Earlier observational studies have shown a similar pattern of poleward flows in the upper 200 m west of 9.8°W during the summer months (Torres & Barton, 2007). The observed pattern of equatorward flow dominance over the shelf and upper slope would be expected during upwelling conditions, with the upwelling jet keeping the IPC offshore as has been suggested by Nolasco et al. (2013).
The slow speed of the glider resulted in considerable time lapse between transects (6 days on average). Due to this, the glider did not observe some events apparent in the satellite chlorophyll data such as the increases in near-surface chlorophyll concentration over the shelf YDs 188-190 and 205-208. The time gap between observations of the shelf limited our ability to constrain the timing of some events, such as the lag between the chlorophyll and Δ(O 2 ) maxima (Figures 7a and 9a). The strong currents over the shelf also prevented the glider from reaching its eastern waypoint during the development of the first upwelling event. The glider's short section limited our observations of alongshore currents over the deep slope and of upwelling features inshore of the 160 m isobath. Future glider deployments in the region will need to consider the trade-off between section length and the frequency of observations at either end of the section. Alternatively, multiple gliders could be deployed concurrently.

Summary
An autonomous ocean glider was used to observe the 2010 summer upwelling season over the NWIM. Upwelling of cold ENACW from below 190 m contributed to an increase of near-surface chlorophyll concentrations from less than 1 mg m −3 to greater than 7 mg m −3 . The increase in primary production contributed to a near-surface increase of Δ(O 2 ) of 16%, 6 days after the chlorophyll maximum. Decreasing Δ(O 2 ) was observed near the sea floor over the shelf during upwelling.
The 2010 summer upwelling season featured atypically strong upwelling favorable winds. Persistent net equatorward flow was observed on the shelf throughout the 2 month deployment, a phenomenon not previously observed. Equatorward flow increased and became more surface intensified during upwelling, and a sporadic, weak poleward jet was observed over the shelf break.
This was the first, and to date only, deployment of a glider to observe summer upwelling over the NWIM. This study highlights some of the challenges of using gliders to study shelf break regions, particularly when the length of time between observations over the shelf is longer than the time period of current reversals on 10.1029/2019JC015804 Journal of Geophysical Research: Oceans the shelf. Despite these difficulties, a single glider was able to occupy a cross shelf section for 2 months, without the need for a costly ship-based campaign.