Advertisement

A Comparison of Satellite-Derived Sea Surface Salinity and Salt Fluxes in the Southern Ocean

  • Brady S. Ferster
  • Bulusu Subrahmanyam
Original Paper

Abstract

Sea surface salinity (SSS) derived from the multi-satellite missions, NASA’s Aquarius/SAC–D and Soil Moisture Active and Passive (SMAP), and ESA’s Soil Moisture and Ocean Salinity (SMOS) are compared and used to estimate horizontal advective salt fluxes in the Southern Ocean (SO). In comparison with an Argo product, all three satellites estimate similar SSS in the Southern Hemisphere mid-latitudes (30° S–45° S) with low variability among the products. At high latitudes, there are temporal patterns of bias (relative to Argo) in Aquarius during Austral summer and in SMOS during Austral winter. Differences in the satellite products and Argo exist along coastal boundaries, low temperatures, and strong currents. Satellite-derived salinity indicates low temporal–mean standard deviations with Aquarius (0.215) and moderate standard deviations with SMOS (0.294) and SMAP (0.325) against Argo in the SO. Differences in satellite-derived zonal and meridional SSS gradients are large; standard deviation values are 2.52 and 1.49 × 10−6 psu m−1, respectively, and similarly located within the sub-tropical salinity maxima, Antarctic Circumpolar Current, and coastal zones. Differences in the horizontal advective fluxes are on average small, but large variability greater than 275 mm month−1 indicates errors of similar magnitude to the estimated Argo flux. Based on these results, the use of satellite-derived salinity may prove to be a useful resource for observing salinity and horizontal salt fluxes, outside the inaccuracies associated with the high latitudes and coastal currents between the various remotely sensed products, and could significantly influence the results depending on the product.

Keywords

Aquarius Salt flux Sea surface salinity SMAP SMOS Southern Ocean 

1 Introduction

Salinity is an essential variable used to quantify density and the state of the ocean. Monitoring the Southern Ocean (SO) salinity is important due to its strong influence on density stratification, ocean circulation, water cycle, and biological productivity [1, 2, 3, 4, 5, 6, 7]. The SO is heavily influenced by the Southern Hemisphere Westerlies and the Antarctic Circumpolar Current (ACC), driving exchange between the Atlantic, Pacific, and Indian Ocean basins. As a result, the SO is dominated by zonal advective (geostrophic transport) processes and diffusive (Ekman transport) properties in the meridional component. Strong eddies interact within the SO on the spatial scales of tens-to-thousands of meters, driving regional changes in surface mixed layer and deep water formation [8, 9, 10].

The introduction of the Argo float program significantly increased and improved the sampling of the SO [11]. There are currently over 3800 active Argo floats throughout the global ocean, although still lacking extensive spatial coverage in the SO compared to other regions, typically covering a 3° × 3° area. This limitation of spatial resolution alone acts as a basis for the need to use satellite-derived salinity in monitoring the SO. Current projects, such as the Southern Ocean Carbon and Climate Observations and Modeling (SOCCOM) (https://soccom.princeton.edu/), have goals of increasing the number of floats within the SO, to account for the limited spatial and temporal coverage. Argo floats (and most in situ profilers) generally quantify surface measurements at a depth of 5–10 m (bulk surface measurement) compared to the surface 1–2 cm (skin) of satellite-derived estimates. Moon and Song [12] and Song et al. [13] discuss differences between bulk and skin measurements that exist in regions of high stratification and describe the skin layer as having higher seasonal variability than bulk measurements. Despite measuring at different depths, differences have been shown to be marginal outside the tropics [14, 15, 16, 17].

Satellite-derived salinity estimations in the SO are heavily influenced by the strong winds and low sea surface temperatures of the high-latitude Southern Hemisphere [18, 19, 20, 21]. Despite the potential inaccuracies of satellites and differences within the satellite salinity products, the increased temporal and spatial scales and measuring the top few centimeters of the ocean surface have made satellite-derived salinity a useful quantity for air–sea interaction studies. In recent years, satellite-derived salinity has been incorporated with the balance of evaporation (E) and precipitation (P), documenting the strong relationship between the ocean water cycle and near-surface salinity [22, 23]. Johnson et al. [22] compared the salinity advection to atmospheric freshwater forcing using in situ and climatology products, but only for the tropical Pacific region. Both Johnson et al. [22] and Yu [23] stressed the importance of surface salinity estimates for ocean processes, hydrological forcing, and model simulations.

Satellite-derived sea surface salinity (SSS) data have recently been used to improve seasonal climate predictions and ocean state estimates [24, 25, 26, 27, 28]. As SSS begins to be incorporated into such forecasts, reanalyses, or state estimates, the accuracy and precision of salinity have become more significant. Satellite-derived salinity plays an important role in the spatial and temporal scales in monitoring salinity and salt fluxes in the global circulation and developing the need to compare satellite-derived products in the SO. In the recent state estimate of ECCO version 4 run 3, the use of Aquarius has been incorporated between 2011 and 2013, although limiting the use of high-latitude data [24]. Furthermore, Vinogradova et al. [28] compared satellite-derived salinity to both in situ synthesized and model data globally. However, more recent versions of the satellite products have been released since their comparison.

The SO is known to have limited observations in an eddy-dominated system; therefore, the objective of this paper is to statistically compare the satellite-derived salinity with an in situ Argo product. Through comparing SSS, the goal of this analysis is to emphasize the significance of finite differences in SSS as well as the influence on horizontal advective flux estimations with the most current versions of satellite-derived SSS. Advective fluxes are important to balance the hydrological cycle [22] and are directly related to the salinity gradient and surface current velocity. We hypothesize that satellite-derived level 3 (L3) monthly salinity can confidently be used at similar spatial and temporal resolutions to that of in situ data collected for the SO, with expected accuracy below the 0.2-psu standard set by the NASA missions, although originally set for the low latitudes. We further anticipate finding small differences in the satellite-derived products and SSS gradients. Potential implications of this analysis are to enable effective utilization of satellite-derived SSS within model and reanalysis-based products to obtain a more accurate representation of the ocean state.

To test the hypothesis, a comparison is made between the respective satellite-derived SSS measurements, as well as between the satellite-derived SSS and Argo product. The purpose of the analysis is to test for low mean differences and standard deviations (SDs) in relation to in situ-based products between May 2010 and December 2016. By investigating the reliability and worth of L3 satellite salinity products, we examine differences in the salinity gradient. Salinity gradients are then used with the surface currents to estimate the horizontal advection of salt [22].

2 Materials and Methods

2.1 Satellite and Observational Data

In this study, SSS from the European Space Agency (ESA) Soil Moisture and Ocean Salinity (SMOS) mission is used. SMOS has previously been found to exhibit bias: significant freshening in coastal areas, over-estimation of salinity in high southern latitudes and near the sea ice due to low temperatures, and in the presence of strong winds [18, 20, 21]. In this study, the operational version 2.0 L3 unbiased binned product from the Barcelona Expert Centre (BEC) [29] at monthly 1° × 1° resolution is used. This product is computed from the SMOS ESA version 6.22 SSS and empirically debiased using climatology to correct for land–sea contamination [29]. This SMOS product should allow for reduced errors commonly found within the high latitudes and areas with strong surface currents (i.e., Malvinas Current region). SMOS data from May 2010 through December 2016 is used in this study.

Aquarius/SAC–D (Satélite de Aplicaciones Científicas–D), a joint NASA and the Argentinean Space Agency (CONAE) mission, provided nearly 4 years of continuous SSS data before ending on June 7, 2015. The current version of the Aquarius dataset, version 5.0, has a global root–mean-square error under 0.2 psu [30] (i.e., a mission objective), but limited peer-reviewed papers have been published on version 5.0. The previous version of the Aquarius dataset is known to be positively biased with respect to in situ products in high latitudes and negatively biased in low latitudes [19, 31]. In this study, we use Aquarius v5.0 L3 monthly 1° × 1° resolution binned product from September 2011 to May 2015 (duration of data collection for Aquarius). The dataset was retrieved from NASA’s Jet Propulsion Laboratory, Physical Oceanography Distributed Active Archive Center (PO.DAAC).

NASA’s Soil Moisture Active Passive (SMAP) is an Earth Satellite mission that primarily measures and maps soil moisture. The SMAP mission provides continuity in SSS estimates as the data is generated using the same core algorithm as Aquarius and also distributed in a similar manner. To correct for surface roughness, other satellites on similar paths provide the surface wind speeds from the Remote System Sensing (RSS) WindSat and the Special Sensor Microwave Imager/Sounder (SSMIS) F17 [32, 33, 34]. SMAP provides uninterrupted data continuity between the Aquarius–SMAP missions, improved spatial resolution (resolve mesoscale eddies, fronts, subduction, and upwelling zones) near coastal zones [32, 35]. SMAP gridded resolution results in a 0.25° × 0.25° product, but further increases noise within the data. Monthly averaged products are available through JPL/NASA PODACC. Version 2.0 has been improved from the beta-version by significantly lower high latitude (zonal) and seasonal biases along the southern hemisphere continents [32]. For this study, the monthly averaged L3 product was binned into 1° × 1° resolution to compare with that of Argo, SMOS, and Aquarius from May 2015 through December 2016.

Temporal and spatial scale variations are essential in the study of the ocean. The SO can have large SSS gradients along the ACC, seen in Fig. 1. The small-scale variability (i.e., eddies, coastal processes) can be accounted for through SMAP in the original 0.25° × 0.25° gridded resolution format. Previous work has shown SMOS to have very similar estimations to Aquarius [18]. However, SMOS regionally derives overall higher SSS in cold waters due to differences in dielectric constant models and dependence on latitude and temperature compared to Aquarius [18].
Fig. 1

Monthly mean sea surface salinity during January 2016 from SMAP level 3 product at 0.25° resolution, encompassing latitudes south of 30° S

An Argo data product is obtained from the Asia-Pacific Data Research Center (APDRC) of the International Pacific Research Centre (IPRC). The specific APDRC Argo data product used in this study is the gridded monthly mean product on standard levels from January 2010 until December 2016. The Argo product is interpolated and smoothed in a 1° × 1° grid as far as 62.5° S in monthly format, with limited horizontal resolution and lower sampling frequency than that of satellite-derived products. This particular product from the APDRC is created through binned quality controlled salinity measurements, interpolated into a monthly grid (0 to 5 m depth). This version of APDRC Argo product is hereon referred to as Argo.

2.2 Mean Difference and Variability

In order to statistically analyze the remotely sensed L3 data, spatial and temporal comparisons are made using the defined SO region south of 30° S. The regions of interest are the entire SO and the Southern Hemisphere mid-latitude region within the SO (30° S–45° S), which allows for our estimate of the SO with limited bias from the low sea surface temperatures and strong winds. Comparisons are made between the monthly products by estimating the differences and the standard deviation (SD) of the differences. This would allow for bias (error) to be determined and the variability in the differences of the products. In terms of variability, we used the 0.2-psu standard set by the Aquarius mission, although the mission goal is for the low latitudes. To compare the satellites with in situ data products, SMOS is compared from May 2010 to December 2016, Aquarius from September 2011 until May 2015, and SMAP from April 2015 until December 2016. To inter-compare satellite-derived products, Aquarius and SMOS are compared for the duration of the Aquarius product, September 2011 until May 2015, and SMAP and SMOS are compared between April 2015 and December 2016.

Seasonal comparisons are the months taken from [36]. Therefore, in this study, austral summertime months are defined as January, February, and March, while austral winter months are defined as July, August, and September. Seasonal differences in salinity and horizontal salt fluxes represent a near maxima and minima for sea ice extent and yearly salinity variations [36]. Although a seasonal comparison is made, there are still very limited in situ observations available within the high latitudes of the SO, stressing the importance of satellite-derived estimations.

2.3 Salinity Gradients and Estimation of Horizontal Salt Fluxes

The salinity products used within the analysis do not estimate the ocean surface current velocities. In estimating the horizontal advective salt flux, the differences between the results would be the finite differences in salinity. Therefore, we will make a comparison of the salinity gradients and the horizontal salt advection. Estimations of the zonal and meridional gradients of salinity are computed individually using Aquarius, SMOS, and SMAP. Ocean Surface Current Analysis Real-time (OSCAR) [37] currents are used to infer the zonal and meridional surface currents (m s−1) between 2010 and 2017.

The surface gradients are calculated similar to Johnson et al. [22]. The salinity gradient (gradSSS) is decomposed into the zonal (x), meridional (y), and vertical (z) coordinates and a, b, and c are unit vectors in the zonal, meridional, and vertical directions, respectively. The salinity gradient is in units of × 10−6 psu m−1.
$$ \mathrm{gradSSS}=\nabla \mathrm{SSS}=\frac{\partial \mathrm{SSS}}{\partial x}\mathrm{a}+\frac{\partial \mathrm{SSS}}{\partial y}\mathrm{b}+\frac{\partial \mathrm{SSS}}{\partial z}\mathrm{c} $$
(1)
To compare the importance of SSS, the salinity gradients are applied to the steady-state balance between the atmospheric forcing (E – P); advection of salt is described similar to Johnson et al. [22].
$$ \mathrm{E}-\mathrm{P}\cong U\cdotp \nabla \mathrm{SSS}=\frac{H_0}{S_0}\ \left[U\frac{\partial \mathrm{SSS}}{\partial x}+V\frac{\partial \mathrm{SSS}}{\partial y}+W\frac{\partial \mathrm{SSS}}{\partial z}\right] $$
(2)
For simplicity, H 0 is the surface layer depth of 32.5 m (depth of valid satellite-derived surface currents) and S 0 is the mean salinity of 35 psu, the same values as in Johnson et al. [22]. U is the velocity vector. Velocities and coordinates are defined with U and x as the zonal, V and y the meridional, and W and z the vertical components. In this analysis, we focus only on the horizontal salt advection (A h , i.e., the zonal and meridional components), as the vertical component is relatively small [22].
$$ {A}_{\mathrm{h}}=\frac{H_0}{S_0}\ \left[U\frac{\partial \mathrm{SSS}}{\partial x}+V\frac{\partial \mathrm{SSS}}{\partial y}\right] $$
(3)

The units used for advection are presented in millimeters per month.

3 Results

3.1 Comparison of Satellite-Derived Salinity Products with Argo

Spatial averages of mean differences and SDs can be found in Table 1 and Figs. 2 and 3. The spatiotemporal mean difference with respect to Argo in the entire SO for Aquarius is − 0.025 and 0.001 psu for SMOS, with mid-latitude discrepancies of − 0.052 psu (Aquarius) and − 0.056 psu (SMOS), respectively. SMOS has positive bias with Argo in the high latitude and negative bias in the mid-latitudes that negate in the spatial mean. SMAP differs from the other satellites by having a near ubiquitous negative bias (Fig. 2c, − 0.150 psu) in the SO. The negative bias of SMAP is between − 0.10 and − 0.25 psu for the SO (Fig. 3a). This consistent bias is important to note as SMAP has similar variability by latitude as SMOS and Aquarius (Fig. 3b). Between all three satellites, the largest SDs occur near the Drake Passage and Malvinas current. The lowest spatiotemporal mean SD (30° S–62.5° S) is from Aquarius (0.215 psu). Spatially averaged SDs in SMOS are 0.294 and 0.325 psu in SMAP.
Table 1

The mean temporal difference and standard deviation (psu) between satellite-derived salinity and Argo. The values noted are taken from Fig. 2. In comparison with the Argo product, global values refer to the area between 62.5° N and 62.5° S

Satellite

Mean difference (psu)

Standard deviation (psu)

62.5° N–62.5° S

30° S–62.5° S

30° S–45° S

62.5° N–62.5° S

30° S–62.5° S

30° S–45° S

Aquarius

− 0.040

− 0.025

− 0.052

0.195

0.215

0.134

SMOS

− 0.010

0.001

− 0.056

0.265

0.294

0.148

SMAP

− 0.062

− 0.150

− 0.157

0.226

0.325

0.176

Fig. 2

Mean differences (ac) and standard deviations (SDs) of SSS differences (df) between satellite-derived salinity and Argo (psu). Aquarius (a, d) is compared from September 2011 to May 2015, SMOS (b, e) May 2010 through December 2016, and SMAP (c, f) April 2015 through December 2016. Corresponding mean values are in Table 1 and units for salinity are in psu

Fig. 3

Zonal averages of mean difference (a) and standard deviation (SD) (b) between satellite-derived salinity and Argo surface salinity (psu) in 1° resolution. Differences between (red) Aquarius and Argo are from September 2011 to May 2015, (blue) SMOS and Argo are May 2010 through December 2016, and (magenta) SMAP and Argo are from April 2015 through December 2016. All data points are for corresponding spatial locations within the domain of Fig. 1

The lowest variability (Figs. 2d–f and 3b) occurs in all three satellites within the Southern Hemisphere mid-latitude region (30° S–45° S), with much of the variability below 0.2 psu, the low-latitude mission goal of Aquarius. Within the high latitudes, the variability increases along the ACC, sea ice, and Drake Passage region. These regions are typically under-sampled with Argo floats for 1° resolution and have sharp salinity gradients.

The zonally averaged mean difference values shown in Fig. 3a display the low value of discrepancies for Aquarius and SMOS between 30° S and 45° S, with values on average below 0.1 psu. Poleward of 45° S, the magnitude of mean difference within Aquarius and SMAP increases in magnitude, whereas SMOS increases in magnitude to a lesser extent due to offset of spatial positive/negative biases. The zonally averaged SD values further show the low variance within the Southern Hemisphere mid-latitude region. For much of the SO, Aquarius has the lowest zonally averaged variation, below 0.2 psu to approximately 52° S.

To compare the satellites in a temporal-scale comparison, the latitudinal distributions of monthly mean differences and SD (Fig. 4) are analyzed for the SO. Fluctuating variability indicates seasonal biases in SMOS and Aquarius. There is a consistent austral summer bias in Aquarius and an austral winter bias in SMOS. There is potentially an austral winter bias in SMAP, but a longer temporal scale would be required. Figure 4 further shows the low variability within the mid-latitude region of the SO, with values below 0.2 psu in both Aquarius and SMOS south of 50° S. Although the values are averaged by latitude, the low variability in Aquarius and SMOS is indicated by much of the SO with low standard deviation, but all three satellites still show large variability to the Argo product in the high latitudes. These biases in the satellite products are likely caused by a combination of the surface roughness, low surface temperatures, and salinity derivation methods [18, 20, 21, 30, 31, 35] and are further discussed in Section 4.1 of Section 4.
Fig. 4

Comparison of salinity derived from Aquarius (a, d), SMOS (b, e), and SMAP (c, f) to the Argo product from 2010 through 2016. Each mean difference satellite minus Argo (ac) and standard deviations (SDs) (df) are averaged over latitude. Aquarius is compared from September 2011 to May 2015, SMOS during May 2010 through December 2016, and SMAP during April 2015 through December 2016. Units are in psu

3.2 Comparison Between Satellite-Derived Salinity Products

To visually compare the satellite-derived salinity against another, mean differences between monthly Aquarius, SMOS, and SMAP for all corresponding data points from September 2011 until December 2016 have been plotted to compare the satellites spatially (Fig. 5) and temporally (Fig. 6). Statistical values of the difference and variability are found in Table 2. Differences between Aquarius and SMOS (Figs. 5a, b and 6a, b) exhibit strong seasonal bias, as there is increased variability in Aquarius during Austral Summer and in SMOS during Austral Winter. Despite the seasonal biases, the mean differences are similar to the comparisons with the in situ data. However, the variability between Aquarius and SMOS is greater in the SO (0.284) than compared to Argo. Similar variability exists between SMAP and SMOS, having strong SO of 0.342. The differences between SMAP and SMOS are increasing in magnitude and variability dramatically increased towards the end of 2016. The differences between SMAP and SMOS are not analyzed beyond December 2016, but further analysis should investigate the disparity between the two products. The largest mean difference and variability between the satellite-derived salinity exist in the in the high latitude, similar to the Argo product.
Fig. 5

Mean sea surface salinity (SSS) (psu) differences (ac) and standard deviation (SD) (d) between Aquarius minus SMOS from September 2011 to May 2015 and similarly SMAP minus SMOS (eh) from April 2015 to December 2016. a, e The austral summer, b, f the austral winter, and c, g the temporal comparisons between the satellites

Fig. 6

Comparison of Aquarius minus SMOS salinity (a, c) and SMAP minus SMOS salinity (b, d) between 2011 and 2016. Each mean difference between these satellite products (a, b) and standard deviations (SD) (c, d) are averaged by latitude. Aquarius minus SMOS is compared from September 2011 to May 2015 and SMAP minus SMOS compared from April 2015 to December 2016. Units are in psu

Table 2

The mean temporal difference and standard deviation (psu) between satellite-derived salinity products. The values noted are taken from Fig. 5

Satellite

Mean difference (psu)

Standard deviation (psu)

Global

30° S–62.5° S

30° S–45° S

Global

30° S–62.5° S

30° S–45° S

Aquarius – SMOS

− 0.083

− 0.078

− 0.010

0.362

0.284

0.180

SMAP – SMOS

− 0.096

− 0.216

− 0.115

0.388

0.342

0.236

Although the satellites derive similar regional mean differences to in situ observations, there is increased variability between the satellite-derived SSS. Comparing the values in Figs. 5 and 6 to the Argo product comparisons in Figs. 2 and 4, differences in Aquarius minus SMOS and SMAP minus SMOS are much larger than differences to the Argo product. Differences and SDs from Aquarius minus SMOS and SMAP minus SMOS are larger than compared to Argo, with mean differences larger in magnitude and SDs well above the 0.2-psu bar.

3.3 Comparison of Salinity Gradients and Horizontal Fluxes

Differences between Aquarius and SMOS (Fig. 7c) zonal gradients are similarly co-located within regions of high SSS variability. The mean (SD) zonal gradient difference between Aquarius and SMOS is 0.123 × 10−6 psu m−1 (2.52 × 10−6 psu m−1) in the SO. Differences in SMAP and SMOS (Fig. 7g) zonal gradients are on average larger than those of Aquarius and SMOS, with a greater area of high variability (Fig. 7h). The mean (SD) zonal gradient difference between SMAP and SMOS for the SO is 0.006 × 10−6 psu m−1 (3.00 × 10−6 psu m−1). In both comparisons, large differences are located along coastal boundaries and sea ice extent, regions known to produce errors in satellite-derived SSS. Differences between the satellites are further driven by the area of swath coverage and the amount of smoothing within the L3 products. The gradients reveal the increased smoothing in SMOS (Fig. 7b, f) compared to Aquarius (Fig. 7a) and SMAP (Fig. 7e). Similar results are found within the meridional gradients, although there is less variability between the differences in meridional gradients (Fig. 8) than zonal. The mean (SD) meridional gradient difference between Aquarius and SMOS is − 0.182 × 10−6 psu m−1 (1.49 × 10−6 psu m−1) and between SMAP and SMOS is 0.037 × 10−6 psu m−1 (1.76 × 10−6 psu m−1) in the SO.
Fig. 7

Temporal mean zonal surface salinity gradient (× 10−6 psu m−1) for Aquarius (a), SMOS (b, f), and SMAP (e). Differences in the mean zonal salinity gradient between Aquarius and SMOS (c) and SMAP and SMOS (g) and the standard deviations (SDs) between Aquarius and SMOS (d) and SMAP and SMOS (h). The duplication of SMOS shows the average zonal flux taken during Aquarius time period (b) and that of SMAP (f)

Fig. 8

Temporal mean meridional surface salinity gradient (× 10−6 psu m−1) for Aquarius (a), SMOS (b, f), and SMAP (e). Differences in the mean meridional salinity gradient between Aquarius and SMOS (c) and SMAP and SMOS (g) and the standard deviations (SD) between Aquarius and SMOS (d) and SMAP and SMOS (h). The duplication of SMOS shows the average meridional flux taken during Aquarius time period (b) and that of SMAP (f)

To quantitatively show how satellite-derived SSS plays a role in the discrepancies in horizontal advective salt flux estimates, we compared with horizontal advective salt fluxes estimated from Argo (Fig. 9 and Table 3). The comparison is made for all four products between 2010 and 2016. Since the same OSCAR currents are used in each calculation, the resulting difference in advection is due to the discrepancies in zonal and meridional salinity gradients. The monthly averaged horizontal advection is estimated through Argo (Fig. 9a). The scale of advection (mm month−1) is the same as Johnson et al. [22], a typical scale to compare salinity and the atmospheric freshwater forcing. In the SO, the largest advection occurs along strong coastal currents and the ACC. The Malvinas and Agulhas regions, two regions of large advection, are also noted to have large variability between in situ and satellite-derived salinity (Figs. 2 and 5). In comparison of the horizontal advective fluxes, the largest differences occur within the ACC, Agulhas, and Malvinas regions, all regions of high variability. Moreover, the differences are as large as the mean fluxes. The mean SO (SD) difference between Argo and Aquarius is 9.64 mm month−1 (346.2 mm month−1), 1.61 mm month−1 (276.6 mm month−1) with SMOS, and 0.74 mm month−1 (417.5 mm month−1) with SMAP.
Fig. 9

Mean difference in horizontal surface salt advection (mm month−1). Argo monthly mean horizontal surface salt advection (a) between 2010 and 2016 (mm month−1). Differences between Aquarius (b), SMOS (c), and SMAP (d) minus Argo horizontal surface salt advection. Aquarius is compared from September 2011 to May 2015, SMOS is from May 2010 through December 2016, and SMAP is compared April 2015 through December 2016. Corresponding values are in Table 3

Table 3

The mean temporal difference and standard deviation (psu) between satellite-derived and Argo horizontal advective fluxes. The values noted are taken from Fig. 9. In comparison with the Argo product, global values refer to the area between 62.5° N and 62.5° S

Satellite

Mean difference (mm month−1)

Standard deviation (mm month−1)

62.5° N–62.5° S

30° S–62.5° S

30° S–45° S

62.5° N–62.5° S

30° S–62.5° S

30° S–45° S

Aquarius

4.85

9.64

11.34

451.7

346.2

304.8

SMOS

1.85

1.61

0.50

275.7

276.6

253.0

SMAP

3.60

0.74

− 0.46

416.0

417.5

330.9

4 Discussion

4.1 Satellite-Derived Salinity

Aquarius and SMOS have positive biases between 45° S and 60° S, and SMAP has a negative bias in the SO, likely resulting from the influence of low sea surface temperatures and strong Westerlies. The L-band radiometer at low sea surface temperatures has previously been noted to have low sensitivity in the high latitudes [38]. Previous SMOS products discussed SDs between 0.4 and 0.5 psu for the different basins of the Southern Ocean [39], values much larger than those of the current unbiased SMOS product [29]. Our analysis indicates that the monthly SO SD differences is below previous estimates, but regionally has mean differences and standard deviations greater than 0.5 psu. Previous analyses found SDs of Aquarius and SMAP to Argo to be 0.25 psu or less globally, but as high as 0.5 psu in the higher latitudes [30, 31, 35]. Similar to their findings, these calculations show a time–mean SO (global) SD of 0.215 (0.195) psu for Aquarius V5 and 0.325 (0.226) psu for SMAP V2, but regionally has mean differences and SDs larger than 0.5 psu for both Aquarius and SMAP. Similar to Kao et al. [30] and Lee [38], the greatest discrepancies and variability compared with Argo and the Aquarius data product (Fig. 2) are within the Malvinas, Agulhas retroflection region, and the southern extent of sampling: all regions with sharp surface salinity gradients or are heavily under-sampled. This is a common signature found within all three L3 products compared with the Argo product, and not just Aquarius.

Both SMOS and Aquarius experience strong interannual variability. The increased austral summer bias in Aquarius and austral winter SMOS are likely related to the low sea surface temperature, a result from the accuracy of brightness temperature measurements in the high latitudes. Both Aquarius and SMOS seasonal biases are evident within high latitudes. The largest differences and variability occur at the high latitudes and the ACC region. There is no clear seasonal pattern observed in SMAP due to the short temporal period, but could potentially have a similar high latitude seasonal bias of SMOS (Fig. 4).

The resulting inter-comparison of satellite-derived SSS indicates disagreement between the satellites along the sea ice extent, the continents, and regions of the ACC. The differences in satellite-derived salinity support the strong variability in regions of strong winds, low sea surface temperatures, and coastal regimes, similar to what is indicated in the comparison with the Argo product. Differences between the satellites could result from bias in algorithms. For example, SMOS has previously been described to have freshening in coastal regions due to contamination from land surfaces, over-estimation in high latitudes due to the methods in dielectric constant, and errors associated with surface roughness [18, 20, 21] and Aquarius to be positively biased in high latitudes [30, 31, 35].

The differences between respective satellite-derived salinity are greater than the differences to Argo, with SDs between Aquarius and SMOS nearly twice as much as compared with Argo. This increased mean difference and larger variability in the entire SO indicate the high degree of variability within the surface centimeters, but mainly the large variability in the remotely sensed products. The comparisons of satellite-derived salinity indicate large variability between products in the high latitudes and the ACC region of the Southern Hemisphere, despite newer releases of the satellite-derived products better accounting for high-latitude biases. Not only is there large variability in the high latitudes when compared with Argo, but larger variability exists between the satellites. It is important to consider the vast differences in the derived products based on the ability to alter models, reanalyses, or estimated hydrological balances.

4.2 Salinity Gradients and Horizontal Fluxes

The variability between SSS gradients has the ability to significantly influence the results of SO analyses on the hydrologic cycle [22, 23]. The zonal SSS gradients plotted in Fig. 7 further show the similarities and differences between the satellites. Large gradients greater than 1.0 × 10−6 psu m−1 exist near coastal boundaries and within strong current regions such as the ACC, Agulhas, and Malvinas. Gradients are relatively small within the open ocean for each of the satellite-derived product. In previous analyses, there are larger salinity gradients and variability in the SO meridional gradient component than the zonal [23]. The meridional gradients are greatest along the ACC and near the sub-tropical maxima regions, greater than 2.0 × 10−6 psu m−1. Here, the meridional gradients (Fig. 8) are nearly two times greater than the zonal gradients (Fig. 7) in satellite-derived products, but differences between the zonal and meridional gradients in satellite products are large. The mean difference in the SO zonal gradients between Aquarius and SMOS is 0.123 and 0.006 psu m−1 between SMAP and SMOS; however, the SDs are 2.52 × 10−6 and 3.00 × 10−6 psu m−1, respectively. The mean difference in the SO meridional gradients between Aquarius and SMOS is − 0.182 × 10−6 and 0.037 × 10−6 psu m−1 between SMAP and SMOS; however, the SDs are 1.49 × 10−6 and 1.76 × 10−6 psu m−1, respectively. The zonal SSS gradient is shown to exhibit relatively large variability in gradient differences compared to the meridional gradient in each comparison of Aquarius, SMOS, and SMAP. Therefore, estimations of SSS and the associated gradients are of utmost importance, as differences in the satellite products are on the same magnitude as the gradients.

Horizontal advection estimated through Argo and differences with the satellites indicate the largest discrepancies are in the high latitude and coastal currents, both areas described to have large differences between salinity products. Outside of the very high latitudes and coastal regions, the main driving force between oceanic and atmospheric forcing is the balance in salinity advection and net precipitation [23]. The monthly SD for the Argo product horizontal advective flux is 87.2 mm month−1 in the SO. The mean difference in the satellite-derived products with respect to Argo for the SO are all less than 2 mm month−1 for SMOS and SMAP, but larger than 9.6 mm month−1 with Aquarius. SMOS further has relatively low standard deviations of the monthly differences, 276.6 mm month−1, but has larger variability in Aquarius and SMAP. SMOS indicates the lowest differences and variability to that of Argo derived fluxes, but could further be a result of smoothing within the products. The SMOS product is smoothed more than the other satellite-derived products, which could result in more similar gradients to that of the smoothed in situ product.

From the comparison in Fig. 9, the variability in SSS alone is enough to derive differences on the same order of magnitude as the horizontal advection, without including the vertical flux component or runoff from land. As a result, the spatial resolution, smoothing, and satellite-derived SSS algorithm all can strongly influence hydrologic balance calculations, especially south of 45° S. The SO is known to have limited temporal resolution with in situ observations, especially in austral winter months. The need to reduce the differences in salinity and horizontal salt fluxes between satellite products alone would help our understanding of the hydrologic cycle.

5 Conclusions

To conclude, this paper emphasizes the use of satellite-derived salinity data to better understand the SO and its interacting air–sea processes. With modern remote sensing techniques, the ability to spatially and temporally monitor SSS has been greatly enhanced compared to in situ techniques. Using Aquarius, SMOS, and SMAP salinity measurements, we find seasonal patterns of salinity discrepancies. In order to support the hypothesis, statistical analyses prove all three satellites to have low mean difference and variability in the Southern Hemisphere mid-latitudes, below a 0.2-psu limit. Although in all three satellites, there are increasing differences and variability south of 50° S. This analysis found the temporal variability between the APDRC Argo product and Aquarius v5 to be 0.215 (0.195) in the SO (globally), 0.294 (0.265) with SMOS, and 0.325 (0.226) with SMAP. In comparison with an Argo product in the Southern Hemisphere mid-latitudes, all three satellites averaged standard deviations below 0.2 psu.

Spatial distributions show the strong high latitude positive bias of Aquarius and SMOS and the negative Aquarius bias along sea ice. SMOS is found to have a low mean difference within the SO, but further spatial comparison reveals the negating of negative/positive biases. Comparatively, SMAP has a ubiquitous negative bias to the Argo product for the SO, except for the Drake Passage. Additionally, the largest differences and variability is found in the Drake Passage region with all three satellites. A temporal analysis further showed the seasonal high latitude bias, increasing austral summer variability in Aquarius and austral winter variability in SMOS. In comparison of the satellite-derived salinity, the differences and variability are greater than that of the variability with in situ observations.

In the estimations of horizontal advective fluxes, the zonal SSS gradient indicates that differences in satellite-derived products are the same order of magnitude as the zonal gradients. The zonal (meridional) gradient SD between Aquarius and SMOS is estimated to be 2.52 psu m−1 (1.49 psu m−1) and between SMAP and SMOS is 3.00 psu m−1 (1.76 psu m−1). The estimation of meridional gradient indicates differences in satellite-derived techniques differ slightly, being an order of magnitude less than the meridional gradient and lower variability. The horizontal advection is estimated in all Argo, Aquarius, SMOS, and SMAP products for a single month. The results indicate advective fluxes estimated in all products strongly vary in the SO, particularly poleward of the ACC and along coastal regions. The mean differences in SO horizontal advective fluxes are on average small with SMOS and SMAP, less than 2.0 mm month−1, but only SMOS has variability below 300.0 mm month−1 in the SO.

These results are important to consider as differences in salinity are influenced by sampling depths, under-sampling of Argo in high latitudes, satellite footprint, data smoothing, and grid size. Moreover, the use of L-band-derived techniques is known to be strongly biased in low temperatures, but yet the high latitudes are important to monitor to better understand global cycles and climate variability. Using various satellite products could provide a useful tool to analyze and monitor the hydrological cycle, but differences in the products and smoothing could significantly skew the results. Although satellite-derived SSS has improved throughout the SO, the large variability between remotely sensed techniques indicates the importance to further improve spatial and temporal scales that can lead to ground-breaking advances in global climate, circulation, and hydrological cycles.

Notes

Acknowledgments

Brady S. Ferster is supported by the NASA/South Carolina Space Grant Graduate Fellowship. Aquarius version 5.0 L3 (ftp://podaac-ftp.jpl.nasa.gov/allData/aquarius/L3/mapped/V5/monthly/SCI/) and OSCAR ( https://doi.org/10.1175/1520-0485(2002)032<2938:DMAAOT>2.0.CO;2) datasets are obtained from the NASA’s JPL Physical Oceanography Distributed Active Archive Center (PO.DAAC). The SMOS unbiased binned data used for this study is the L3 Operational version 2.0 provided by the ESA obtained from the SMOS Barcelona Expert Center Data distribution and visualization services ( https://doi.org/10.1016/j.rse.2016.02.038). The SMAP data are produced by Remote Sensing Systems, Santa Rosa, CA, and version 2.0 level 3 is obtained from NASA JPL PO.DAAC (ftp://podaac-ftp.jpl.nasa.gov/allData/smap/L3/RSS/V2/monthly/SCI/). Argo data (Argo DOI:  https://doi.org/10.17882/42182) is obtained from the Asia-Pacific Data Research Center (APDRC) of the International Pacific Research Centre (IPRC), the 1° gridded on standard levels product. The authors would like to thank the anonymous reviewers and the editor, whose comments significantly contributed to the improvement of this paper.

References

  1. 1.
    Gille ST (2002) Warming of the Southern Ocean since the 1950’s. Science 295:1275–1277.  https://doi.org/10.1126/science.1065863. CrossRefGoogle Scholar
  2. 2.
    Dong S, Sprintall J, Gille ST, Talley L (2008) Southern Ocean mixed–layer depth from Argo float profiles. J Geophys Res 113:C06013.  https://doi.org/10.1029/2006JC004051. Google Scholar
  3. 3.
    Liu J, Curry JA (2010) Accelerated warming of the Southern Ocean and its impacts on the hydrological cycle and sea ice. PNAS 107:14987–14992.  https://doi.org/10.1073/pnas.1003336107 CrossRefGoogle Scholar
  4. 4.
    McNeil BI, Matear RJ (2008) Southern Ocean acidification: a tipping point at 450–ppm atmospheric CO2. PNAS 105:18860–18864.  https://doi.org/10.1073/pnas.0806318105 CrossRefGoogle Scholar
  5. 5.
    Montes-Hugo M, Doney SC, Ducklow HW, Fraser W, Martinson D, Stammerjohn SE, Schofield O (2009) Recent changes in phytoplankton communities associated with rapid regional climate change along the western Antarctic Peninsula. Science 323:1470–1473.  https://doi.org/10.1126/science.1164533 CrossRefGoogle Scholar
  6. 6.
    Sarmiento JL, Hughes TMC, Stouffer RJ, Manabe S (1998) Simulated response of the ocean carbon cycle to anthropogenic climate warming. Nature 393:245–249.  https://doi.org/10.1038/30455 CrossRefGoogle Scholar
  7. 7.
    Toggweiler JR, Samuels BL (1995) Effect of sea ice on the salinity of Antarctic bottom waters. J Phys Oceanogr 25:1980–1997.  https://doi.org/10.1175/15200485(1995)025<1980:EOSIOT>2.0.CO,2 CrossRefGoogle Scholar
  8. 8.
    Meyer A, Sloyan BM, Polzin KL, Phillips HE, Bindoff NL (2015) Mixing variability in the Southern Ocean. J Phys Oceanogr 45(4):966–987.  https://doi.org/10.1175/JPO-D-14-0110.1 CrossRefGoogle Scholar
  9. 9.
    Rintoul, S. R., Naveira Garabato, A. C. (2013) Dynamics of the Southern Ocean circulation, ocean circulation and climate, Gerold Siedler, Stephen M. Griffies, John Gould and John A. Church, Elsevier Ltd, 103:471–492Google Scholar
  10. 10.
    Sheen KL, Naveira Garabato AC, Brearley JA, Meredith MP, Polzin KL, Smeed DA, Forryan A, King BA, Sallée JB, Laurent LS, Thurnherr AM (2014) Eddy–induced variability in Southern Ocean abyssal mixing on climatic timescales. Nat Geosci 7(8):577–582.  https://doi.org/10.1038/ngeo2200. CrossRefGoogle Scholar
  11. 11.
    Lyman JM, Johnson GC (2008) Estimating annual global upper–ocean heat content anomalies despite irregular in-situ ocean sampling. J Clim 21(21):5629–5641.  https://doi.org/10.1175/2008JCLI2259.1 CrossRefGoogle Scholar
  12. 12.
    Moon J-H, Song YT (2014) Seasonal salinity stratifications in the near–surface layer from Aquarius, Argo, and an ocean model: focusing on the tropical Atlantic/Indian oceans. J Geophys Res Oceans 119:6066–6077.  https://doi.org/10.1002/2014JC009969 CrossRefGoogle Scholar
  13. 13.
    Song YT, Lee T, Moon J–H, Qu T, Yueh S (2015) Modeling skin–layer salinity with an extended surface–salinity layer. J Geophys Res Oceans 120:1079–1095.  https://doi.org/10.1002/2014JC010346 CrossRefGoogle Scholar
  14. 14.
    Boutin J et al (2016) Satellite and in-situ salinity: understanding near–surface stratification and subfootprint variability. Bull Amer Meteor Soc 97:1391–1407.  https://doi.org/10.1175/BAMS-D-15-00032.1. CrossRefGoogle Scholar
  15. 15.
    Boutin J, Martin N, Reberdin G, Yin X, Gaillard F (2013) Sea surface freshening inferred from SMOS and ARGO salinity: impact of rain. Ocean Sci 9:183–192.  https://doi.org/10.5194/os-9-183-2013 CrossRefGoogle Scholar
  16. 16.
    Drucker R, Riser SC (2014) Validation of Aquarius sea surface salinity with Argo: analysis of error due to depth of measurement and vertical salinity stratification. J Geophys Res Oceans 119:4626–4637.  https://doi.org/10.1002/2014JC010045 CrossRefGoogle Scholar
  17. 17.
    Henocq C, Boutin J, Petitcolin F, Reverdin G, Arnault S, Lattes P (2010) Vertical variability of near–surface salinity in the tropics: consequences for L–band radiometer calibration and validation. J Atmos Ocean Technol 27:192–209.  https://doi.org/10.1175/2009jtecho670.1. CrossRefGoogle Scholar
  18. 18.
    Dinnat EP, Boutin J, Yin X, Le Vine DM (2014) Inter–comparison of SMOS and Aquarius sea surface salinity: effects of the dielectric constant and vicarious calibration. IEEE Microrad 55–60:14528317.  https://doi.org/10.1109/MicroRad.2014.6878907 Google Scholar
  19. 19.
    Le Vine DM, Dinnat EP, Meissner T, Yueh SH, Wentz FJ, Torrusio SE, Lagerloef G (2015) Status of Aquarius/SAC–D and Aquarius salinity retrievals. IEEE J Sel Topics Appl Earth Observ Remote Sens 8(12):5401–5415.  https://doi.org/10.1109/JSTARS.2015.2427159 CrossRefGoogle Scholar
  20. 20.
    Reul N, Tenerelli J, Boutin J, Chapron B, Paul F, Brion E, Gaillard F, Archer O (2012) Overview of the first SMOS sea surface salinity products. Part I: quality assessment for the second half of 2010. IEEE Trans Geosci Remote Sens 50(5):1636–1647.  https://doi.org/10.1109/LGRS.2012.2207943. CrossRefGoogle Scholar
  21. 21.
    Yin X, Boutin J, Martin N, Spurgeon P, Vergely J–L, Gaillard F (2014) Errors in SMOS sea surface salinity and their dependency on priori wind speed. Remote Sens Environ 146:159–171.  https://doi.org/10.1016/j.rse.2013.09.008 CrossRefGoogle Scholar
  22. 22.
    Johnson ES, Lagerloef GS, Gunn JT, Bonjean F (2002) Surface salinity advection in the tropical oceans compared with atmospheric freshwater forcing: a trial balance. J Geophys Res Oceans 107(C12).  https://doi.org/10.1029/2001JC001122
  23. 23.
    Yu L (2011) A global relationship between the ocean water cycle and near-surface salinity. J Geophys Res Oceans 116(C10).  https://doi.org/10.1029/2010JC006937
  24. 24.
    Fukumori, I., Wang, O., Fenty, I., Forget, G., Heimbach, P., Ponte, R. M. (2017) ECCO version 4 release 3. http://hdl.handle.net/1721.1/110380
  25. 25.
    Hackert E, Busalacchi AJ, Ballabrera-Poy J (2014) Impact of Aquarius sea surface salinity observations on coupled forecasts for the tropical Indo–Pacific Ocean. J Geophys Res Oceans 119:4045–4067.  https://doi.org/10.1002/2013JC009697 CrossRefGoogle Scholar
  26. 26.
    Köhl A, Martins MS, Stammer D (2014) Impact of assimilating surface salinity from SMOS on ocean circulation estimates. J Geophys Res Oceans 119:5449–5464.  https://doi.org/10.1002/2014JC010040 CrossRefGoogle Scholar
  27. 27.
    Toyoda T, Fujii Y, Kuragano T, Matthews JP, Abe H, Ebuchi N, Usui N, Ogawa K, Kamachi M (2015) Improvements to a global ocean data assimilation system through the incorporation of Aquarius surface salinity data. Q J R Meteorol Soc 141(692):2750–2759.  https://doi.org/10.1002/qj.2561 CrossRefGoogle Scholar
  28. 28.
    Vinogradova NT, Ponte RM, Fukumori I, Wang O (2014) Estimating satellite salinity errors for assimilation of Aquarius SMOS data into climate models. J Geophys Res Oceans 119:4732–4744.  https://doi.org/10.1002/2014JC009906 CrossRefGoogle Scholar
  29. 29.
    Olmedo E, Martínez J, Umbert M, Hoareau N, Portabella M, Ballabrera J, Turiel A (2016) Improving time and space resolution of SMOS salinity maps using multifractal fusion. Remote Sens Environ 180:246–263.  https://doi.org/10.1016/j.rse.2016.02.038 CrossRefGoogle Scholar
  30. 30.
    Kao, H.-Y., Lagerloef, G., Lee, T., Melnichenko, O., Hacker, P. (2017) Aquarius salinity validation analysis, data version 5.0. Aquarius Science Team Technology Document, http://podaac.jpl.nasa.gov/aquarius.
  31. 31.
    Lagerloef, G., Kao, H.-Y., Meissner, T., Vazquez, J. (2015) Aquarius salinity validation analysis, data version 4.0. Aquarius Science Team Technology Document, http://podaac.jpl.nasa.gov/aquarius.
  32. 32.
    Meissner, T., Wentz, F. J. (2016) Remote Sensing Systems SMAP ocean surface salinities [level 2C, level 3 running 8–day, level 3 monthly], Version 2.0 validated release. Remote Sensing Systems, Santa Rosa, CA., USA.  https://doi.org/10.5067/SMP20-3SMCS
  33. 33.
    Wentz, F., Le Vine, D.M. (2012) Aquarius salinity retrieval algorithm, ATBD. RSS Technical Report 082912, ftp://podaac-ftp.jpl.nasa.gov/allData/aquarius/docs.
  34. 34.
    Wentz, F., Ricciardulli, L., Gentemann, C., Meissner, T., Hilburn, K. A., Scott, J. (2013) Remote Sensing Systems Coriolis WindSat [Daily] Environmental Suite on 0.25 deg grid, Version 7.0.1, [WSPD_LF]. RSS Technical Report 091316 19, Remote Sensing Systems, Santa Rosa, CA, USA. www.remss.com/missions/windsat.
  35. 35.
    Meissner, T., Wentz, F., Scott, J. (2015) Remote Sensing Systems SMAP level 3 ocean surface salinities [running 8–day, monthly] on 0.25 deg grid, version 1.0 (BETA). Technical report, Remote Sensing Systems, Santa Rosa, CA., USA. (http://www.remss.com/missions/smap)
  36. 36.
    Ren L, Speer K, Chassignet EP (2011) The mixed layer salinity budget and sea ice in the Southern Ocean. J Geophys Res 116:C08031.  https://doi.org/10.1029/2010JC006634. Google Scholar
  37. 37.
    Bonjean F, Lagerloef GSE (2002) Diagnostic model and analysis of the surface currents in the tropical Pacific Ocean. J Phys Oceanogr 32:2938–2954.  https://doi.org/10.1175/1520-0485(2002)032<2938:DMAAOT>2.0.CO;2
  38. 38.
    Lee T (2016) Consistency of Aquarius sea surface salinity with Argo products on various spatial and temporal scales. Geophys Res Lett 43(8):3857–3864.  https://doi.org/10.1002/2016GL068822 CrossRefGoogle Scholar
  39. 39.
    Boutin J, Martin N, Yin X, Font J, Reul N, Spurgeon P (2012) First assessment of SMOS data over open ocean: part II—sea surface salinity. IEEE Trans Geosci Remote Sens 50(5):1662–1675.  https://doi.org/10.1109/TGRS.2012.2184546 CrossRefGoogle Scholar

Copyright information

© Springer International Publishing 2018

Authors and Affiliations

  1. 1.School of the Earth, Ocean and EnvironmentUniversity of South CarolinaColumbiaUSA

Personalised recommendations