Atlantic Niño induced sea surface salinity variability as observed from the satellite

The Atlantic Niño exerts great impact on surrounding weather and climate anomalies, leading to anomalous wind, temperature, precipitation, etc. However, the impact of Atlantic Niño on ocean salinity in the tropical Atlantic has not been well disclosed. The present study examines the Atlantic Niño induced sea surface salinity (SSS) distribution in both boreal summer and winter seasons by using the satellite data as well as various reanalysis and objective analysis data sets. It concludes that the summer Atlantic Niño leads to large fresh SSS anomalies in the eastern equatorial basin, while the winter Atlantic Niño leads to a meridional dipole structure of SSS anomalies. The former is mostly controlled by the dynamical processes of ocean, while the latter is largely controlled by the atmospheric processes. Accordingly, two SSS indices are developed to describe the relationships in the two seasons. The present study advances our understanding of the Atlantic Niño and its associated SSS variability and reveals the possible deficiencies of current reanalysis and objective analysis data sets in the tropical Atlantic Ocean.


Introduction
Ocean salinity is an important variable, as it affects key physical processes (Maes et al 2005, Reul et  Climate modes such as El Niño-Southern Oscillation and Indian Ocean dipole can exert great impact on the distribution of salinity in their respective ocean basins (Hasson et al 2013, Qu et al 2014, Du and Zhang 2015, Zhi et al 2020).In the tropical Atlantic Ocean, a leading mode at seasonal to interannual time scale is characterized by warm sea surface temperature anomalies (SSTA) in the eastern equatorial ocean basin, referred to as the Atlantic Niño, which shares many features with its Pacific counterpart (Keenlyside and Latif 2007).While there are considerable analysis and diagnosis on the Atlantic Niño induced SSTA and atmospheric anomalies (Chang et al 2006, Richter et al 2013, Losada and Rodríguez-Fonseca 2016, Lübbecke et al 2018, Liu et al 2023), the Atlantic Niño induced sea surface salinity anomalies (SSSA) had been underexplored until recently.Awo et al (2018) examined the sea surface salinity (SSS) signatures of the tropical Atlantic zonal mode, namely the Atlantic Niño.However, they only looked into the SSS variability in boreal summer and ignored that in the winter, when the Atlantic Niño can also have a peak (Okumura andXie 2006, Prigent et al 2020).It has been found that the winter Atlantic Niño appears more frequently and exhibits great climatic influences over recent years (Vallès-Casanova et al 2020, Richter et al 2022, Song et al 2023).In addition, Awo et al (2018) mostly relied on the objective analysis and model data for their study, which exists potential concerns due to the quality and accuracy of salinity data in the tropical Atlantic as will be discussed later.
Therefore, it is interesting to examine the Atlantic Niño induced SSS variability in the tropical Atlantic Ocean for both boreal summer and winter seasons.
To achieve the goal, an accurate and high quality data set of salinity is required.The historical lack of salinity measurements has long been a challenge for the study of ocean salinity, particularly in the tropical Atlantic Ocean (Da-Allada et al 2013, Schlundt et al 2014, Sena Martins et al 2015, Bourlès et al 2019, Yu 2020), where in situ observations are very sparse compared with observations in the tropical Pacific and India Ocean (Foltz et al 2019, Stammer et al 2021).While reanalysis and objective analysis data sets in the tropical Pacific and Indian Oceans are of good quality and often used as the proxy of observation, the accuracy and quality of these data sets in the tropical Atlantic are questioning.This study mainly relies on the satellite data to conduct the analysis.For reference and comparison, the SSS from several reanalysis and objective analysis data sets are also evaluated, highlighting biases and caveats of these data sets in the tropical Atlantic Ocean, which can serve as an alert of caution.
The rest of this paper is as follows: section 2 introduces the data sets and methods used in this study; sections 3 present the results, with emphases placed on developing two SSS indices to represent the SSS variability under the influence of Atlantic Niño, and on discovering the physical mechanisms regarding how the Atlantic Niño affects the tropical Atlantic SSS variability in summer and winter seasons; conclusions and discussions are presented in the last section.

Data and methods
The present study uses the eighth version of Level 3 soil moisture and ocean salinity (SMOS) SSS (CEC-Locean L3 Debiased v8; Boutin et al 2023) from the French Centre Aval de Traitement des Donnees SMOS, with a spatial and temporal resolution of 0.25 • × 0.25 • and nine days, respectively (SMOS SSS hereafter).Besides, salinity data from multiple data products are also used for comparison, including the objective analysis of EN4.2.1 (Good et al 2013) from the Met Office Hadley Centre, and the reanalysis of GLORYS (https://doi.org/10.48670/moi-00021)from E.U.Copernicus Marine Environment Monitoring Service (CMEMS).The EN4 and GLORYS have a horizontal resolution of 1 • and 1/12 • , respectively.Vertically, the salinity on the top layer of each product is taken as the SSS.The top layer is 5 m and 0.5 m for the EN4 and GLORYS, respectively.In this study, salinity is reported in practical salinity units (psu).The 850 hPa wind data are from the JRA-3Q reanalysis (Kosaka et al 2024), precipitation from the newest Global Precipitation Climatology Project 3.2 (CGCP; Huffman et al 2023), evaporation form the 1 • grided Objectively Analyzed air-sea Fluxes (OAFlux; Yu and Weller 2007), ocean surface currents from the CMEMS L4 global total velocity field at 0 m (Rio et al 2014; https://doi.org/10.48670/moi-00049),and river discharges from the Observation Service SO HYBAM (https://hybam.obs-mip.fr/data/).All data span from January 2010 to December 2022 (13 years) unless otherwise noted.All data are interpolated to a uniform grid of 1 • × 1 • using bilinear interpolation except that the SMOS is interpolated using the local area averaging, as bilinear interpolation can lead to some missing values near offshore.The data interpolation has little influence on the conclusions.
The Atlantic Niño is represented by the ATL3 index defined as the area-averaged SSTA (3 • S-3 • N and 20 • W-0 • ; Zebiak 1993), which is calculated with the HadISST (Rayner et al 2003) and shown in figure SI1.The EN4 and ERSST (Huang et al 2020) are also used to calculate the ATL3 time series for comparison and verification (figure SI1).It is defined that an Atlantic Niño/Niña event occurs when the normalized ATL3 index exceeds one standard deviation range in all three data sets.According to this criterion, there are three positive events (2010, 2016, and 2021) and four negative events (2011, 2012, 2014, and 2015) in JJA, and two positive events (2016 and 2019) and three negative events (2011, 2013, and 2014) in NDJ, from 2010 to 2022.Based on this definition, composite analysis is performed to explore the SSS distribution caused by the Atlantic Niño in both summer (JJA) and winter (NDJ).

Sea surface salinity mean state and variability in multiple data sets
To firstly have a general look at the difference between the satellite-derived SSS and the SSS from various sources, figure 1 shows the seasonal mean SSS and its variability in multiple data sets.The SMOS SSS displays its highest values at the northern and southern extremes of the studied region, reaching up to ∼37.5 psu, due to strong evaporation.It also displays the lowest values at about 5 degrees north of the equatorial band, varying around 34.5-35.5 psu, due to intense precipitation under the Intertropical Convergence Zone (ITCZ), consistent with Tchilibou et al (2015).In addition, the low SSS appears in the eastern equatorial and northwestern tropical regions, corresponding to where the high SSS variability locates (figure 1(g)), due to strong freshwater discharges from the Amazon, Niger, Congo, and Orinoco rivers, as shown by black dots of A, B, C, and D, respectively, in figure 1.These features exhibit apparent seasonal variations (figures (g) and (j)).The northwestern tropical SSS variability is stronger in JJA than in DJF, while the eastern equatorial SSS variability is stronger in DJF than in JJA.It is consistent with Tzortzi et al (2013) that the SSS variability in the tropical Atlantic is dominated by the seasonal displacement of ITCZ and outflows from large rivers.
Figures 1(b) and (h) reveal that the EN4 overestimates the mean SSS and underestimates its variability in the Amazon outflow region in JJA, while figures 1(c) and (i) show that the GLORYS is more in line with the SMOS.In addition, the EN4 underestimates the SSS variability near the river mouth of Congo in DJF (figure 1(k)).The EN4 is an objective analysis data set, which relies highly on the number and quality of in situ observations, while the GLORYS also relies on the performance of model.Figure SI2 shows that the EN4 has very few observations and large uncertainty in these regions, which could be responsible for its poor representation of the SSS in the regions.On the contrary, satellite can provide higher temporal and spatial coverage than in situ observations.The model-based GLORYS is more in line with the SMOS SSS, probably due to its 1/12

Sea surface salinity variability induced by the Atlantic Niño
A typical Atlantic Niño event begins with a relaxation of the equatorial trade winds in boreal summer, followed by a deepening of the thermocline and SST warming in the eastern basin, which in turn strengthens the anomalous winds, forming the so called Bjerknes feedback.The dominant mechanism responsible for the formation of Atlantic Niño in boreal winter is not conclusive, but evidences suggest that it may be associated with signals beyond A composite map of the SSSA difference between positive and negative Atlantic Niño events in both summer and winter seasons is shown in figure 2. In JJA, the SMOS exhibits an apparent fresh anomaly in the eastern equatorial region, resembling the canonical SSTA pattern induced by the Atlantic Niño but with an opposite sign (figure 2(a)).In contrast, the fresh anomaly is very weak in the EN4 (figure 2(b)) and the GLORYS (figure 2(c)).In NDJ, there is not only a fresh anomaly in the eastern basin, but also a salty anomaly in the northern tropical Atlantic Ocean (5 • N-15 • N and 20 • W-40 • W), forming a meridional SSSA dipole structure, which can be well captured by all the three data sets (figures 2(d)-(f)).It should be noted that the SSS anomalies in NDJ are stronger than those in JJA, indicating that the winter Atlantic Niño has greater impact on the tropical Atlantic SSS variability, since the climatological standard deviations of SSSA are comparable in the two seasons in these regions (figure 1).In addition, there are some large positive anomalies in the northwestern tropical Atlantic, which could be due to the river outflows of Amazon and Orinoco and is not included in the following analysis.
To further investigate how the Atlantic Niño affects the SSS distribution, two indices are developed to represent the SSS variability under the influence of Atlantic Niño in boreal summer and winter, respectively, considering their different patterns of SSSA as shown in figure 2 (black boxes).The SSS summer index (SSI) is defined as the area-averaged SSSA over the domain of 7 • S-3 • N and 10 • W-10 • E, whereas the SSS winter index (SWI) is defined as the difference between the area-averaged SSSA over the southeastern domain (10 • S-5 • N and 0-10 • E) and that over the northwestern domain (5 • N-15 • N and 20 • W-40 • W; the former minus the latter).Table 1 describes the correlations between the ATL3 and defined SSS indices.As can be seen, the SMOS exhibits a strong relationship between the ATL 3 and SSS indices for the two seasons, with a correlation value of −0.66 for JJA and −0.85 for NDJ, respectively.Such high

How the Atlantic Niño affects the SSS variability
This subsection explores the possible physical mechanisms responsible for the SSS distribution caused by the Atlantic Niño in boreal summer and winter seasons.Figure 3 displays the composite map of the difference between positive and negative events for different variables including SST, 850 hPa wind, E − P, and surface currents.As can be seen, during a summer Atlantic Niño, anomalous westerlies occur in the western equatorial basin (figure 3  Figure 4 displays the time series of E − P, U current, and V current anomalies, and their correspondence with the defined SSS indices in different seasons.Noted that the time series of E − P, U current, and V current anomalies in JJA and DJF are calculated in the same way as the SSI and SWI, respectively.The SSI exhibits a low correlation with the E − P and V current anomalies but a high correlation of −0.72 with the U current anomalies, conforming the above conclusions that the eastward surface current anomalies play an important role in the SSS variability during the summer Atlantic Nino.In contrast, the winter Atlantic Niño is more subject to the influence of E − P, as the correlation between the SWI and E − P anomalies is 0.58, much higher than the correlations between the SWI and surface current anomalies.Additionally, the magnitude of E − P anomalies is also much greater in DJF than in JJA.It is also noticed that the V current anomalies are stronger in DJF, consistent with the result of figure 3, although the correlation is weak.These results suggest that the oceanic processes play the more important role for the SSS distribution during the summer Atlantic Niño, while the atmospheric processes are more important during the winter Atlantic Niño. The river discharges may be subject to the influence of Atlantic Niño and play a role in the SSS variability (Tzortzi et al 2013, Grodsky andCarton 2018).Table 2 shows the correlations between different .And yet, whether the river outflow of Amazon can contribute to the summer SSS variability in the eastern equatorial basin is not sure and needs further investigation in the future.In addition, the river discharge of Congo shows a correlation of −0.50 (above the 90% significance level) with the SSI, indicating its potential influence on the summer SSS variability in the eastern equatorial basin.However, this influence seems to be unrelated to the summer Atlantic Niño, as the correlation between the JJA ATL3 and river discharge of Congo is not significant.In contrast, the winter Atlantic Niño has relatively low correlations with all three river discharges, and the SWI Y Chen shows low correlations with the river discharges as well, suggesting the small influence of river discharges on SSS variability during the winter Atlantic Niño.

Summary and discussion
In short, this study investigated the Atlantic Niño induced SSS variability in both boreal summer and winter seasons and explored possible mechanisms that control the distribution of SSS.Meanwhile, this study unveiled that the satellite data can well reflect the SSS variability induced by the Atlantic Niño.
Based on the satellite data, two useful SSS indices were defined to represent the SSS variability under the influence of Atlantic Niño.It concluded that the summer Atlantic Niño leads to large fresh SSS anomalies in the eastern equatorial basin, while the winter Atlantic Niño leads to a meridional dipole structure of SSS anomalies.The former is mostly controlled by the dynamical processes of ocean, particularly the Bjerknes feedback, while the latter is largely controlled by the atmospheric processes, particularly the seasonal displacement of ITCZ.Specifically, during the summer Atlantic Nino, there are large negative E − P anomalies in the central and western tropical Atlantic basin, resulting in great amount of fresh water, which can be transported to the east by the strong anomalies of eastward equatorial surface current, leading to the decreased SSS in the eastern equatorial basin; during the winter Atlantic Nino, the southward migration of ITCZ can give rise to increased convection activities and decreased E − P at the equator, along with the decreased convection activities and increased E − P in the northern tropical Atlantic, which ultimately results in the corresponding dipolar structure of SSS anomalies.
To fully understand and quantify the contribution of each process, a detailed salinity budget analysis is needed in the future.In addition, some studies pointed out that the Atlantic Niño exhibits multidecadal variations (Rodríguez-fonseca et al 2009, Losada and Rodríguez-Fonseca 2016, Martín-Rey et al 2018).Thus, it is worthy to further study whether the relationship between the Atlantic Nino and its associated SSS also shows a multidecadal variation.Nonetheless, this study can advance our understanding of the Atlantic Niño and the SSS variability in the tropical Atlantic.
al 2014, Zhang et al 2023), and therefore provides useful information regarding the hydrological cycle (Durack et al 2012, Yu et al 2021), biological productivity (Fine et al 2017, Brandt et al 2023), climate change (Du et al 2019, Li et al 2020), etc. Understanding the distribution and variation of ocean salinity is of significant scientific and societal interests (Foltz et al 2019, Vinogradova et al 2019, Reul et al 2020, Stammer et al 2021, Martin et al 2022, Kim et al 2023).

Figure 1 .
Figure 1.Seasonal mean SSS (a)-(f) and standard deviation of SSSA (g)-(l) in the tropical Atlantic Ocean for SMOS, EN4, and GLORYS.Black dots of A, B, C, and D in each panel indicate the river months of the Amazon, Niger, Congo, and Orinoco, respectively.

Figure 2 .
Figure 2. Composite of SSSA difference between the Atlantic Niño and Niña events in boreal summer (JJA; (a)-(c)) and winter (NDJ; (d)-(f)) for SMOS, EN4, and GLORYS.Black boxes indicate areas where great SSS anomalies exist, which are defined to calculate the Atlantic Niño SSS indices.Dotted areas indicate the differences are significant above the 95% confidence level compared to the climatology according to a two-tail Student's t test.

Figure 3 .
Figure 3. Composite of difference between the Atlantic Niño and Niña events in boreal summer (JJA; (a)-(b)) and winter (NDJ; (c)-(d)) for SST ( • C), 850 hPa wind (m s −1 ), E−P (m yr −1 ), and surface current anomalies (cm s −1 ).Black boxes indicate areas where great SSS anomalies exist, which are defined to calculate the Atlantic Niño SSS indices.Dotted areas indicate anomalies are significant above the 95% confidence level compared to the climatology according to a two-tail Student's t test.
(a)), and large SSTA in the eastern equatorial basin (figure 3(a)), indicative of a clear Bjerknes feedback configuration, which could lead to the deepened thermocline, depressed upwellings, and decreased salinity in eastern equatorial basin.At the same time, there are significant negative E − P anomalies in the western tropical Atlantic basin (figure 3(b)), leading to large amount of fresh water.The fresh water could be carried to the eastern equatorial basin by the strong eastward surface current anomalies (figure 3(b)), intensifying the negative SSSA in the eastern equatorial basin (figure 2(a)).During a winter Atlantic Niño, the SSTA exhibits a dipolar structure with cold anomalies in the northern tropical Atlantic and warm anomalies in the equatorial and southern tropical basin (figure 3(c)), suggesting an anomalous southward migration of ITCZ.The southward migration of ITCZ could lead to the suppressed E − P at the equator and the enhanced E − P in the northern tropical Atlantic (figure 3(d)).

Figure 4 .
Figure 4. Time series of SSS summer index (SSI; psu), E − P (m yr −1 ), U current (cm s −1 ), and V current anomalies (cm s −1 ; (a)-(c)).(d)-(f) are same as (a)-(c), but for DJF.Values at the top right of each panel indicates the correlation between the two time series in the panel.Correlations greater than 0.55 are significant above the 95% confidence level, according to a two-tail Student's t test.

Table 1 .
Correlations between the ATL3 and defined SSS indices for different data sets.The correlation higher than 0.55 is above the 95% significance level according to a two-tail Student's t test, indicated by the asterisk.

Table 2 .
Correlations between different indices and river discharges in the corresponding season.Noted that the Amazon and Orinoco data span from 2010 to 2019.The correlation above the 95% significance level is indicated by the asterisk, according to a two-tail Student's t test.