δ15NPOM ~ glm[f(class) + f(season) + f(layer) + Lon + Lat + T + S] (4). All the parameters were statistically significant (ANOVA, p < 0.001), and the r2 value of the model was found to be 0.451. The longitude had a positive impact on δ15NPOM, i.e., δ15NPOM increased eastward (Fig. 9e). The δ15NPOM was negatively affected by the latitude, temperature, and salinity (Fig. 9d, f, g). The lsmean values indicated that the δ15NPOM of classes I (lsmean ± SE: –1.6 ± 0.41‰) and II (–3.1 ± 0.22‰) were significantly lower and higher than those of the other three classes, respectively (pair-wise test with Tukey’s adjacent, p < 0.01). Moreover, classes III (2.1 ± 0.34‰) and IV (2.0 ± 0.29‰) were not significantly different (p = 0.9735). The seasonal variations exhibited significant differences between spring (1.3 ± 0.25‰) and summer (1.9 ± 0.18‰, pair-wise test with Tukey’s adjacent, p < 0.01), while the lowest lsmean values were recorded in winter (0.9 ± 0.43‰). No significant differences of lsmean δ15NPOM were found between spring and summer (p > 0.0897). The layer indicated that δ15NPOM at a depth of 100 m (2.3 ± 0.39‰) was significantly higher (pair-wise test with Tukey’s adjacent, p < 0.01) than that at the surface (0.9 ± 0.15‰) and subsurface (1.0 ± 0.21‰).