After 16 August, the wind turned to the S and SW (data not shown), thus causing the upwelling to relax. Several cold upwelling filaments developed along the southern coast between longitudes 23 and 27°E, and a few of them transformed into eddies (Figure 4). The filaments were persistent at three locations: north of Hiiumaa, and off Pakri and Tallinn (Figure 4b). In situ Chl a concentrations
along the transect varied in a wide range from 1.57 to 15.54 mg m− 3 during the period of field measurements ( Figure 6). Low Chl a values were observed during the first half of July in the upwelling region along the northern coast. From 25 July, when upwelling along the northern RG7204 concentration coast was in the relaxation phase, the Chl a concentrations increased off the northern coast, and decreased off the southern coast. The highest (15.5 mg m− 3) and lowest (1.6 mg m− 3) Chl a concentrations were observed on 8 August
off the northern and southern coasts respectively. The Chl a concentrations calculated with the FUB processor from the MERIS data was correlated with in situ Chl a for two time windows: 24 h and 2 h intervals (before or after) from the satellite overpass. A scatterplot of selected data pairs is shown in Figure 7. Androgen Receptor antagonist A total of 7 data pairs fulfilled the 2 h criterion: 3 samples (TH9, TH11 and TH13) from 11 July and 4 samples (TH11, TH13, TH15 and TH17) from 25 July ( Table 1). For the 2 h window the FUB processor underestimated Chl a compared with in situ Chl a ( Figure 7); the average underestimation was 25% (1.3 mg m− 3), which is of the same magnitude as in previous studies in the Baltic Sea ( Kratzer et al. 2008). The correlation (r2) for data points within the 2 h window was 0.67 and for the 24 h window was also relatively high at 0.56. The linear regression for the 2 h window with 95% confidence limits was Chl a = 0.48(± 0.39) × X + 3.6(± 1.8), where X is the FUB processor
output. The standard deviation of the residuals (i.e. standard error of the estimation – SEE) was 0.51. For the 24 h window the slope and y-intercept of the linear regression were 0.44 (± 0.097) BCKDHA and 2.9 (± 0.60) respectively. The standard deviation of the residuals for the 24 h window was 1.43. In addition to the FUB processor we also evaluated the case 2 regional water processor (C2RW) for Chl a (data not shown). The correlation for the FUB processor (0.67) was much higher compared to the C2RW processor (0.17). Also, the Chl a overestimation of C2RW by 52% is poorer compared with the underestimation (25%) by the FUB processor. On the basis of the above analysis, the FUB algorithm was used to calculate Chl a from MERIS data in the Gulf of Finland. The equation obtained with linear regression for the 2 h window was applied to calibrate MERIS Chl a data.