TABLE 2.

Results of GLM testing hypothesis II^a

Independent variableb	Regression coefficient	SE	P
βMPTI	0.79	0.06	<0.01
Log (salinity)	−22.81	1.53	<0.01
Log (total inorganic nitrogen)	0.4	0.02	<0.01
Log (phosphate)	−0.63	0.03	<0.01
Log (PAR)	−0.05	0.02	<0.01
Log (chlorophyll a)	0.37	0.03	<0.01

^aHypothesis II is that based on cross-cruise comparison, the bacterial β diversity effect is stronger (more positive) when β diversity in increased more by deterministic and less by stochastic assembly processes (Fig. 1C). The model presented here is the results after backward stepwise selection to statistically control for potential confounding factors. Steps of the backward stepwise selection are listed in Table S3. The results indicate that, after accounting for environmental variables, the effect of bacterial β diversity on the summed bacterial community biomass increased with deterministic processes selecting for dissimilar species. This conclusion is qualitatively the same as that drawn from the univariate GLMM (Fig. 3).

^bBacterial β diversity effects on the summed bacterial community biomass was the dependent variable.