Robert R. Sterling

A validation of the resulting SSS products is performed using in situ TSG salinity data from four different cruise campaigns carried out during 2012. We used 4 Thermosalinograph (TSG) data sets available during the period of the study by the following ships: Explorer (C6TN4), TARA (FVNM), Ronald H. Brown (WTEC) and the TMM Sinaloa (ZCDJ6). The TSG data are smoothed with a running Hanning filter of half width of ∼12.5 km (approximately half width of the quarter degree grid resolution) to have a spatial resolution to those of L3 and L4 SSS products. Due to the poor quality of SMOS SSS data near to the coast (the land-sea contamination is intense in this area), TSG data with distance from the coast smaller than 350 km were not considered.

Figure6.7a shows the salinity values measured by TSG during the year 2012. The histograms of the differences between the different satellite products and TSG salinity data are shown in Figure 6.7, panels b-d. SMOS L3 (Figure 6.7b) presents the largest discrepancies from in situ observations, both in terms of the standard deviation and the inter-quartile range (iqr). In contrast, L4 data fusion products (Figure 6.7c-d) present a reduction of both the central value of the difference (mean or median) and the error when compared to the L3 product (Figure

6.7b). Of all the fused products, the L4-ADT (Figure 6.7c) is the closest to the in situ data (bias of -0.19 and error of 0.35), while the L4-SST (Figure 6.7d) presents worse scores (bias of -0.21 and error of 0.40).

Figure 6.7: TSG in situ salinity (top left). SSS difference histogram with in situ data L3 SSS (top right), L4-ADT (bottom left) and L4-SST (bottom right).

Figure 6.8 shows a TSG transect from the Explorer vessel in the period August 13-17. In Figure6.8a, the transect (thick solid line) is drawn on top of the L4-ADT map from August 15.

The SSS map suggests that the Explorer TSG transect crosses the GS as well as two structures with apparent CCR signature (those around 56^◦W and around 50^◦W). Notice that west of the GS (65^◦W,40^◦N), the in situ SSS is higher than 35.5 indicating the presence of a WCR. The various salinity products show this salty anomaly and the location of the saline front associated with the GS (61-62^◦W,40^◦N).

In Figure 6.8b, the SSS data along the transect is shown for all the salinity products. The local salinity minimum around 56^◦W, as indicated by the TSG data, is completely missing in the L3 SSS product. The L4-SST product displays a slightly shifted (at 55^◦W) and a secondary minimum, while the L4-ADT displays a well-defined salinity minimum at the location of the CCR, i.e., in line with the TSG data. Although the amplitude of the minimum salinity is

Figure 6.8: L4-ADT for the 15th of August with TSG positions of Explorer from 13-17 August 2012 in black dots (a).

Comparison of in situ data along the transect in (a) with L3 SSS and L4 SSS maps of 15th of August (b).

weaker than that observed by in situ data (almost 1.5), the salinity difference between the inner-core and its western limit is as large as 1. On the other hand, the TSG salinity minimum located at 50^◦W is completely missing in L4-SST. Both SMOS L3 and the L4-ADT detect a negative salinity anomaly, although slightly displaced towards the west. Again, the L4-ADT product recovers the low salinity anomaly, showing a salinity difference between the CCR outer and inner parts (∼1) similar to the difference of the in situ data, although the absolute value is fresher.

In the Brazil Current area there not available TSG data for the year 2012, therefore Argo profiles are used as an independent source of validation for the SSS maps. Figure 6.9a shows the upper-most salinity value and positions of Argo for the year 2012. The histograms of the differences between the different satellite products and the upper-most salinity measured by Argo floats data are shown in Figure 6.9b-d. The L3 (Figure6.9b) present the largest discrepancies from in situ observations, both in terms of the standard deviation and of the inter-quartile range (iqr). In contrast, L4 data fusion products (Figure6.9c-d) present a reduction of both the central value of the difference (mean or median) and the error when compared with the L3 products (Figure6.9b-c). The L4-SST (Figure6.9d) is the closest to in situ data (bias of 0.10 and error of 0.43), while the L4-ADT (Figure 6.9c) presents slightly worse scores (bias of 0.13 and error

of 0.43).

An example 9-day SSS map of L4-ADT starting the 21st of January with the available Argo measurements (black dots) in the area the same days are shown in Figure 6.10a. An Argo value was found inside a WCR the 23st of January at (46.9W,40.2S) (red dot). The salinity and temperature measured along this Argo profile with depth is shown in Figure 6.10c. The upper-most salinity for the Argo profiler was 35.6 at 6 meters depth. The matchup of satellite products closer to the Argo measurement is the L4-ADT (35.18), while L4-SST and L3 SSS give an estimate of 34.99 and 34.98 respectively. Regarding the temperature of this WCR, the Argo upper-most value is 22.42^◦C , while the AVHRR SST estimate is 21.53^◦C . The difference between satellite and in situ temperature value (-0.89^◦C ), corresponds to differences in salinity of -0.62, -0.61 and -0.42 for L3, L4-SST and L4-ADT respectively. In this case L4-ADT product achieves the best score against Argo measurement. Notice that the effect of stratification is not negligible in this case, although the observed differences, especially in temperature, are more likely due to representativity errors (as the time and space scales of SSS products are quite different from those of an Argo measurement).

Another Argo float is found inside a CCR ring the 25th of February; the corresponding 9-day map of L4-ADT starting the 23rd of February with the position of the Argo (red dot) is shown in Figure 6.10b. The temperature and salinity profile for this Argo in the first 1000 meters is shown Figure 6.10d. The upper-most salinity for the Argo profiler was 34.3 at 4.7 meters depth. The matchup of satellite products closer to the Argo measurement is in this case the L4-SST (34.09), followed by L4-ADT (33.86) and finally L3 SSS (33.49) give the worse estimate.

Regarding the temperature inside the CCR, the Argo upper-most value is 19.98^◦C , while the AVHRR SST estimate is 18.84^◦C . The difference between satellite and in situ temperature value (-1.14^◦C ), corresponds to differences in salinity of -0.81, -0.21 and -0.44 for L3, SST and L4-ADT respectively. In this case L4-SST product achieves the best score as compared to Argo measurement. Again, stratification and scale representativity are probable causes explaining the observed differences.

Figure 6.9: Available Argo positions and upper-most salinity value for the year 2012 (a). SSS difference histogram with Argo data for L3 SSS (b), L4-ADT (c) and L4-SST (d).

Figure 6.10: L4-ADT with available Argo positions overimposed for 9-day periods starting on 21st of January (a) and starting on 23rd of February (b) of 2012. Argo profile values for salinity and temperature for the first 1000 meters with col located surface values of AVHRR SST, L3 SSS, L4-ADT and L4-SST at (50.3W,39.1S) for the 23rd of January (c) and at (46.9W,40.2S) for the 25th of February (d) of 2012.