Table 2 Best models (lowest AIC) explaining seasonal variation in three response variables: log10 cell abundance, \(\sqrt[2]{{\alpha {\text{ species richness}}}}\) and \(\sqrt[2]{{\alpha {\text{ ENS}}}}\). Best models were determined by AIC (ΔAIC > 2). For each response variable, all possible combinations of ln DIN, \(\sqrt[2]{{{\text{silicate}}}}\), and the interaction between ln DIN and mean station-level salinity were used. Mean station-level salinity was included in all models. This resulted in six models in total. When no clear model emerged, all equivalent models are reported.
Model | Marginal r2 | Conditional r2 | AIC | AIC weight |
|---|---|---|---|---|
log10 cell abundance | ||||
(int.) + ln DIN + salinity + \(\sqrt[2]{{{\text{silicate}}}}\) + ln DIN:salinity | 0.33 | 0.37 | 6131 | > 0.99 |
\(\sqrt[{\mathbf{2}}]{{{\mathbf{\alpha }}\,{\mathbf{species }}\,{\mathbf{richness }}}}\) | ||||
(int.) + ln DIN + salinity + \(\sqrt[2]{{{\text{silicate}}}}\) | 0.022 | 0.18 | 7523 | 0.38 |
(int.) + ln DIN + salinity | 0.021 | 0.18 | 7524 | 0.30 |
(int.) + ln DIN + salinity + ln DIN:salinity | 0.021 | 0.18 | 7525 | 0.16 |
(int.) + ln DIN + salinity + \(\sqrt[2]{{{\text{silicate}}}}\) + ln DIN:salinity | 0.022 | 0.18 | 7525 | 0.15 |
\(\sqrt[{\mathbf{2}}]{{{{\varvec{\upalpha}}}\,{\mathbf{ ENS }}}}\) | ||||
(int.) + ln DIN + salinity + \(\sqrt[2]{{{\text{silicate}}}}\) + ln DIN:salinity | 0.013 | 0.10 | 6286 | 0.54 |
(int.) + ln DIN + salinity + ln DIN:salinity | 0.012 | 0.093 | 6286 | 0.44 |
