Generalized additive models for biomass simulation of submerged macrophytes in a shallow lake

Submerged macrophytes are widely distributed primary producer that play important roles in maintaining healthy aquatic ecosystems. Generally, the relationships between macrophytes and environmental factors are complicated, so nonlinear nonparametric models with relatively flexible structures are optimal for macrophyte habitat simulation. In this study, generalized additive model (GAM) was used to evaluate the response of the submerged macrophytes biomass to water environmental factors in the Baiyangdian Lake. Forward stepwise method was used to implement model optimization. Likelihood ratio test was used to determine whether adding a variable enhances the model performance. Four individual variables (water depth, transparency, total nitrogen, and total phosphorus) and two interaction terms (water depth × transparency and water depth × total phosphorus) were included in the optimal GAM. The optimal model explained 70.5% of the biomass variation with a relatively low residual deviance value (22.40). There was a significant correlation between the measured and predicted data (R² = 0.716, p = 0.0004). The response lines generated by the model indicated that macrophyte biomass had a positive correlation with transparency but negative correlations with total nitrogen and nitrite nitrogen in water. The response patterns of macrophyte biomass to water depth and total phosphorus were unimodal. The biomass reached the maximum value when the water depth was about 2.1 m and the total phosphorus concentration was 0.07 mg/L. Water depth and transparency, which affect light availability, are critical physical variables affecting the conditions associated with the submerged macrophytes, and excess nitrite and phosphorus limiting macrophyte biomass.