ENSO-induced co-variability of Salinity, Plankton Biomass and Coastal Currents in the Northern Gulf of Mexico

The northern Gulf of Mexico (GoM) is a region strongly influenced by river discharges of freshwater and nutrients, which promote a highly productive coastal ecosystem that host commercially valuable marine species. A variety of climate and weather processes could potentially influence the river discharges into the northern GoM. However, their impacts on the coastal ecosystem remain poorly described. By using a regional ocean-biogeochemical model, complemented with satellite and in situ observations, here we show that El Niño - Southern Oscillation (ENSO) is a main driver of the interannual variability in salinity and plankton biomass during winter and spring. Composite analysis of salinity and plankton biomass anomalies shows a strong asymmetry between El Niño and La Niña impacts, with much larger amplitude and broader areas affected during El Niño conditions. Further analysis of the model simulation reveals significant coastal circulation anomalies driven by changes in salinity and winds. The coastal circulation anomalies in turn largely determine the spatial extent and distribution of the ENSO-induced plankton biomass variability. These findings highlight that ENSO-induced changes in salinity, plankton biomass, and coastal circulation across the northern GoM are closely interlinked and may significantly impact the abundance and distribution of fish and invertebrates.


Results
Main patterns of salinity and plankton biomass. Figure 1a,b shows the leading EOFs of a surface salinity anomaly (SSA) and a surface phytoplankton anomaly (SPA) (hereinafter anomaly implies data with the climatological annual cycle removed) derived from our ocean-biogeochemical model. These two leading modes are eminently coastal patterns with the largest variability occurring over the Louisiana-Texas inner shelf (Fig. 1a,b). The temporal variation in the EOF mode for these two variables, represented by the Principal Components (PCs), are significantly correlated, making clear the link between salinity and phytoplankton variability over the shelf (Fig. 1c). Both PCs also closely match the variability of the integrated river discharge anomaly from the main northern GoM rivers (Fig. 1c), indicating that the leading driver of interannual variability for salinity and phytoplankton biomass is river runoff. Accordingly, the greatest SSAs and SPAs occur under extreme river discharge conditions during severe drought years (e.g., 1981, 1988, 2000, and 2006) and wet years (e.g., 1979, 1983, and 1991). Positive discharge anomalies, concomitant with negative SSA and positive SPA, prevailed during the 1980s and 1990s relative to the climatology for 1979-2014, indicating an interdecadal modulation of the river runoff signal. The temporal coupling between river discharge and phytoplankton biomass is also observed in the PCs of satellite chlorophyll anomaly derived from the SeaWiFS and MODIS sensors (satellite data description in Methods), which closely resembles the model-derived patterns (Fig. 1d). Similar patterns to those in the model SPA are also found in the model surface zooplankton anomaly (SZA, Supplementary Fig. S1).
ENSO impacts on the northern GoM. The influence of ENSO on precipitation patterns over the southeastern continental United States is usually phase-locked to the seasonal cycle, such that the strongest anomalies occur during winter (positive during El Niño and negative during La Niña) e.g. [35][36][37][38] . As a consequence, El Niño's impact on river discharge has a marked seasonality (Fig. 2a), with the largest positive anomalies occurring in late fall and winter and declining values occurring in spring. The sign of the river discharge anomalies reverses during La Niña (Fig. 2b), although La Niña anomalies for the Mississippi-Atchafalaya rivers are non-significant. Since interannual changes in salinity and plankton biomass along the coastal areas of the northern GoM are mainly driven by river discharge (Fig. 1), it is logical to hypothesize that an ENSO signal for salinity and plankton biomass can be expected during winter. To evaluate this hypothesis, the correlation coefficients between the N34 and PC series of the model SSA, SPA, and SZA were estimated for each calendar month. We presented the correlation at zero-lag, but similar results are derived when N34 leads the PC series by 1-4 months (not shown). Consistent with the ENSO signal in river discharge, the correlation patterns between the N34 and PC1 series show a strong seasonal modulation (Fig. 2c), with the maximum correlation in February (r = −0.62, 0.48, and 0.58 for SSA, SPA, and SZA, respectively) and statistically significant values occurring only during December-May. The derived patterns are supported by observational data, which also show a significant correlation between the N34 and PC series for SeaWiFS and MODIS chlorophyll (r = 0.78 and 0.55 for the January-March averaged time series of SeaWiFs and MODIS, respectively), as well as between the N34 and the in situ zooplankton dry weight series from Dauphin Island (r = 0.83 for March; see in situ zooplankton data in Methods) (Fig. 2d). To visualize the spatial variability of salinity and coastal circulation due to ENSO, we derived El Niño and La Niña composites of SSA and surface velocity for winter (December-February) and spring (March-May). During El Niño winters (Fig. 3a), the SSA displays significant negative values across most of the northern GoM. The largest anomaly magnitude (about 2 psu) is located along the inner shelf (onshore of the 25-m isobath) off Mississippi, Louisiana, and Texas (87°-96°W). Concurrent with this pattern in salinity, anticlockwise circulation anomalies are observed along the outer shelf (offshore of the 25-m isobath), as well as along the Texas inner shelf. This implies a strengthening of the prevailing westward flow during El Niño on the Louisiana-Texas shelf (the average climatological circulation is shown in Supplementary Fig. S2). The negative winter SSA condition persists throughout spring, but the magnitude of the anomalies decreases significantly nearshore (Fig. 3c). An offshore spread of the salinity anomalies is evident, linked to predominantly southeastward current anomalies. On the other hand, the derived La Niña SSA composite is non-significant across most of the northern GoM shelf (Fig. 3b,d), reflecting the asymmetry between El Niño and La Niña discharge patterns. An examination of the PC1 of the SSA reveals that the weaker La Niña signal is partly explained by the two weak La Niña events in 1984-85 and 1998-99, as fresher conditions prevailed during these events ( Supplementary Fig. S3). Still, La Niña composites display the opposite pattern to El Niño composites during winter, but with about half of the El Niño anomaly magnitude. The circulation anomalies linked to La Niña winters are mainly clockwise and located in the northwestern GoM. The saltier pattern breaks in spring, as negative SSAs associated with the Mississippi-Atchafalaya plumes spreads over the Louisiana-Texas shelf (the mean La Niña discharge anomalies for the Mississippi-Atchafalaya rivers are positive during March-May; Fig. 2b). However, positive SSAs are observed nearshore across most of the northern GoM, with the largest values located northeast of the Mississippi delta (~89°W), in the northeastern GoM (83°-85°W), and near the U.S.-Mexico border (~26°N, ~97°W).
We also examined spatiotemporal patterns in plankton anomalies induced by ENSO. Circulation patterns significantly influence the distribution of SPA and SZA during El Niño, generating distinct winter and spring  Fig. 4a,c). On the other hand, southeasterly current anomalies during El Niño springs lead to an increased offshore export of plankton biomass, especially in the north-central GoM (Fig. 4b,d). Because zooplankton growth responds to phytoplankton growth, the largest accumulation rates of zooplankton biomass occur downstream of the phytoplankton biomass maximum, producing the greatest zooplankton anomalies westward from the phytoplankton maximum in winter, and southward in spring. La Niña composites for the SPA show mostly non-significant anomalies across the northern shelf ( Supplementary Fig. S4). Consistent with the pattern in salinity, the SPA and SZA during La Niña winters are predominantly negative. This low biomass pattern largely vanishes during La Niña spring, as positive SPAs and SZAs appear over the north-central GoM.

Drivers of ENSO circulation anomalies.
On a seasonal time scale, the predominant downwelling favorable winds during winter compress the Mississippi and other river plumes against the coast, inducing a sharp salinity gradient that drives westward flow along the northern GoM 29 . This gradient can be seen in the simulated climatological pattern of salinity and alongshore flow (Fig. 5a) from a vertical section across the Louisiana-Texas shelf (section A, location depicted in Supplementary Fig. S5). There, salinity displays almost vertically-oriented isohalines, ranging from ~28 psu nearshore to >36 psu over the outer shelf (bottom depth >150 m), and the maximum alongshore currents (~10 cm s −1 at the surface) occur in response to the strongest salinity gradient.
Since the winter alongshore-flow in the northern GoM shelf is, to a great degree, in geostrophic balance 7 , we can hypothesize that the decrease in nearshore salinity and, consequently, the increase in the cross-shore density gradient, drives the westward current increase during El Niño (Figs 3a and 5b,c). To evaluate this hypothesis, we derived geostrophic currents from the thermal wind relationship (see equation (1) in Methods) using the model density field (Fig. 5d). The comparison revealed a similar structure and amplitude of the anomalies for the modeled current and the current derived from the thermal wind balance, with maximum values (~4 cm s −1 )  Fig. S6). Across the northwestern shelf (southern Texas and northern Mexico coasts), the winter alongshore-flow variability associated with changes in salinity is reinforced by winds. Northerly winds anomalies during El Nino   Fig. 6a) induce onshore Ekman transport, which increases the zonal gradient of sea surface height, triggering an anomalous southward barotropic flow. On the other hand, southerly wind anomalies during La Nina (Fig. 6b) induce offshore Ekman transport and trigger an anomalous northward barotropic flow. The wind influence on circulation can be seen in the velocity patterns of a cross-shore section off southern Texas (section B, location shown in Supplementary Fig. S5), where the thermal wind approximation captures main features in the model flow anomaly but underestimates the anomaly's magnitude, especially during La Niña ( Supplementary Fig. S7). During El Niño spring, the alongshore-current anomalies over the Louisiana-Texas shelf depart from the thermal wind-derived flow anomalies (not shown), and wind-driven barotropic dynamics become more prominent. This is explained by the strengthening of El Niño wind anomalies, which progress from northerly during winter to northwesterly (i.e., upwelling favorable) during spring (Fig. 6a,c), inducing an anomalous southeastward flow into the north-central shelf during spring (Fig. 3c).

ENSO impacts on the deep GoM.
Additional ENSO-related anomalies in plankton biomass can be expected in the surface layers of the deep GoM (bottom depth >500 m), where river inputs are not dominant. Changes in plankton production in the deep GoM are mainly linked to mixing and stratification changes, the latter mostly driven by temperature 12,34 . The link between thermal stratification and phytoplankton biomass is evident in the northern deep GoM series of SSTs, the vertical mixing of nitrate, and surface phytoplankton (Supplementary Fig. S8a; northern deep GoM series are extracted from the deep ocean region north of 25°N), which show positive phytoplankton anomalies associated with cold and increased vertical mixing periods. It is well know that El Niño increases the frequency of cold fronts, determining the northwesterly anomalies shown in Fig. 6, promoting increased vertical mixing and negative temperature anomalies during late winter and early spring 39 and, consequently, impacting plankton biomass. Indeed, we found significant correlations between N34 and the model derived time series of the vertical mixing of nitrate, SSTs, phytoplankton, and zooplankton anomalies (N34 leading by 3 months) during spring ( Supplementary Fig. S8b). This result is consistent with the expected ENSO modulation of plankton biomass due to changes in vertical mixing, as suggested by Melo-Gonzalez et al. 33 . This ocean signal reinforces the positive phytoplankton anomalies during El Niño, especially over the outer shelf.

Summary and Discussion
Using the outputs of a regional high-resolution ocean-biogeochemical model, we determined that the leading mode of salinity and plankton biomass variability in the northern GoM is associated with river discharge variability. The variability in the PC1 time series compares well with the patterns derived from satellite chlorophyll, as well as in situ zooplankton biomass observations. We found significant correlations between the EOF modes of surface salinity and plankton biomasses and the Nino3.4 time series. The correlations are largest during winter and early spring, reflecting the seasonal phase locking of ENSO signal. Further composite analysis revealed an asymmetry between El Niño and La Niña impacts. The El Niño-induced anomalies can be more than two times larger than the La Niña-induced anomalies.
Our study reports ENSO-induced anomalies in the coastal circulation over the northern GoM, which has not been address in previous studies. ENSO disturbances in the cross-shore salinity gradient modulate the intensity of the alongshore current in the Louisiana-Texas shelf during winter via thermal wind relationship. ENSO-induce wind anomalies during winter reinforce the alongshore-current anomalies over southern Texas and the northeastern Mexican coast. During El Niño springs, the wind impact on alongshore circulation anomalies is more prominent, and the alongshore-current anomalies over the Louisiana-Texas shelf deviate from the thermal wind relation approximation. These coastal circulation anomalies during El Niño explain the largest plankton anomalies west of 89°W during winter and off the north-central shelf during spring. We also found that ENSO wind anomalies impact the seasonal patterns of mixing and stratification in the deep GoM, and thus modulate plankton biomass during late winter and early spring, consistent with the hypothesis of Melo-Gonzalez et al. 33 .
The above-described anomalies in salinity and plankton biomass could have significant impact on the reproductive success and biological condition of upper trophic levels, including commercially important species. Indeed, an improved Gulf menhaden condition (measured as fish oil content) is associated with El Niño years, presumably due to increased prey biomass 17 . Additionally, ENSO disturbances in river discharge and coastal circulation patterns influence the dispersal and recruitment of Gulf menhaden, as previous studies have indicated low recruitment levels associated with increased Mississippi-Atchafalaya river discharge 15,40 . Salinity anomalies may also have a direct impact on fish growth and condition, such as for red snapper larvae that have experienced declining conditions during low salinity periods 16 . Although the link between ENSO and upper trophic level variability has been suggested for several species of fish and invertebrates, the ENSO-related patterns of salinity, plankton biomass, and circulation-three variables hypothesized as driving mechanisms of recruitment and condition variability-have been scarcely described. In this context, our model results provide a framework to better comprehend ENSO-related variability in the northern GoM ecosystem and advance understanding of the larger-scale climate variability mode as a driver of ecosystem and marine population changes.
Finally, ENSO-induced anomalies in river discharge, phytoplankton biomass, and winds could potentially influence hypoxia development over the Louisiana-Texas shelf 41,42 . However, estimations of midsummer hypoxia size during 1985-2011 6 do not support an evident link between ENSO conditions and hypoxia (not shown). This could be explained by the difference in seasonality between ENSO and hypoxia. More specifically, the strongest ENSO anomalies in river discharge, salinity and plankton biomass occur during winter and early spring, while conditions for the development of bottom hypoxia appear to occur mainly during late spring and early summer 41

Methods
Ocean-Biogeochemical Model. To describe the ENSO-related variability in salinity, plankton biomass, ocean currents, and surface winds we estimated mean composite for El Niño and La Niña conditions. The definition of the El Niño/La Niña periods was based on the N34 time series, with warm (cold) ENSO conditions linked to N34 values > 0.5 °C (<−0.5 °C). We only considered El Niño and La Niña events that persisted until late spring (May) because ENSO events that persist throughout the spring have a more significant effect on the atmospheric circulation anomalies that influence U.S. rainfall e.g. 35,[53][54][55] . This criterion was met for six El Niño events (1982-83, 1986-87, 1991-92, 1997-98, 2004-05, and 2009-10) and six La Niña events (1984-85, 1988-89, 1998-99, 1999-00, 2007-08, and 2010-11). The statistical significance of these ENSO events was assessed with Monte Carlo experiments 52 . For each variable, 1,000 independent realizations of the composite were generated by randomly selecting 6 years (six is the number of El Niño/La Niña events) from 1979-2014. An El Niño (or La Niña) composite value was significant at the 90% level when it fell outside the interval defined by the percentiles of 5% and 95% from the randomly generated composite distribution.
Thermal wind balance. To analyze ENSO-related alongshore flow variability, mean values of the geostrophic current were derived from the model density field using the thermal wind equation: where u g (z) is the geostrophic current at depth z, g is the gravitational acceleration (9.8 m s −2 ), ρ o is a reference density (1,025 kg m −3 ), f is the Coriolis parameter, δρ δy is the cross-shore density gradient, and H is the bottom depth. Following Zhang et al. 7 , it was assumed that horizontal geostrophic velocity vanishes at the bottom.

Data Availability
The ocean-biogeochemical model outputs used in this study are in the Network Common Data Form (NetCDF) format on the NOAA-AOML server, available upon request to the corresponding author.