Generalized Additive and Fuzzy Models in Environmental Flow Assessment of Yangtze River in Jiangsu Province

With the ever-increasing hydrological changes of rivers and the resulting environmental degradation, the environmental flow assessment discipline has been established to assess the quantity and quality of water required for ecosystem conservation and resource protection. Habitat rating methodology (HRM) is a widely use model to quantify the environmental flow of the river, which provided a means to differentiate habitat changes resulting from physical, chemical and/or biological factors or a combination of such factors. However, the conventional HRM cannot deal with the nonlinear effect in aquatic ecosystem, or the uncertain ecological thresholds. To solve these problems, the generalized additive model (GAM) and fuzzy variable method (FVM) are introduced into the environmental flow research, and an improved habitat rating methodology (I-HRM). GAM can be applied to treat high-dimensional data and it is compatible with real-time applications, which has a number of advantages such as flexibility in shapes of the relationships as well as the distribution of the output variable. FVM can reasonably identify the relative membership degree and relative membership function between the sample index and standard interval of each level index and properly determine the assessment level of each sample by varying the model and its parameters. I-HRM applies GAM represent the nonlinear relationships among hydrological, hydrodynamic and environmental factors; and then uses FVM to quantify the membership that the habitat belongs to each pattern. At last, IHRM is applied into the environmental flow assessment of Yangtze River in Jiangsu Province. The result shows that, (i) the environmental flow demand of Yangtze River in Jiangsu Province is 7430 m³/s, and the current runoff volume can satisfy this requirement; (ii) compared with the conventional HRM, I-HRM has better capacities in dealing with the nonlinear relationships in the aquatic ecosystem, and the uncertainties in ecological thresholds. Therefore, we think that the improvement of HRM is feasible and I-HRM should be applied to more rivers and basins.