Construction of aδ15NBase isoscape
We obtained δ 15NBulk values for six copepod species (N. cristatus : n = 130; N. plumchrus : n = 56; N. flemingeri : n = 28; O. venusta : n = 39; P. aculeatus : n = 12; P. parvus : n = 66), including our own samples (n = 356) and data from the literature (Pomerleau et al . 2014) (n = 4). Based on δ 15NAAs, the mean TPs of each copepod species were 2.7 (range: 2.5–2.9, n = 6) forN. cristatus , 2.6 (2.4–2.9, n = 5) for N. plumchrus , 2.7 (2.4–2.8, n = 4) for N. flemingeri , 2.2 (2.1–2.5, n = 3) for O. venusta , 2.3 (1.9–2.6, n= 5) for P. parvus , and 1.9 (1.9–2.0, n = 3) for P. aculeatus (Table S3). Theδ 15NBase andδ 15NBulk were highly correlated in each species (r 2 = 0.905 for N. cristatus , 0.827 for N. plumchrus , 0.993 for N. flemingeri , 0.974 for O. venusta , 0.871 for P. parvus , and 0.818 for P. aculeatus ; Fig. S2). Therefore, theδ 15NBase values were calculated using the regression equations withδ 15NBulk data (Eq. 4). The slope and intercept of the species (a and b in Eq. 4) were 1.455 and –6.128 for N. cristatus , 1.229 and –7.010 forN. plumchrus , 1.866 and –11.496 for N. flemingeri , 2.400 and –11.028 for O. venusta , 0.977 and –5.242 for P. parvus , and 1.461 and –6.590 for P. aculeatus , respectively (Fig. S2).
The estimated isoscape showed a large spatial variation inδ 15NBase in our study area (see Fig. 1). Specifically, theδ 15NBase was different among the Okhotsk Sea and coastal area of Japan (2.5–5.0‰), the central NP (–5.0‰ to 0.0‰) and the Bering Sea Shelf (5.0–8.0‰). Linear regression analysis showed that spatial gradients inδ 15NBase were largely explained by δ 15NesPhy (explained deviance: 45.3%, Fig. S4). Therefore, estimatedδ 15NesPhy complementedδ 15NBase values in areas of chum salmon habitat where zooplankton samples were not available, such as the Okhotsk Sea and Gulf of Alaska (Fig. 1). The root-square-mean error in δ 15NBase values between actual measurements (copepod samples) and values predicted by DIVA gridding software was 1.42‰.