Influences of Human-induced Habitat Modifications on Basin-wide Fish Species Richness in the Danshuei River Watershed of Taiwan

The intensive exploitation of water resources has seriously degraded riverine environments and threatened inhabitant biota. In this study, we aim to assess the influences of human-induced habitat modifications on basin-wide fish species richness based on multi-year heterogeneous datasets collected from the Danshuei River Watershed of Taiwan. We aggregated long-term datasets (2003-2012) of fish composition, river network structures, dam locations and water quality parameters including water temperature, pH, conductivity, turbidity, dissolved oxygen and total phosphorus, at 45 sampling sites across the Danshuei River Watershed. We first used a multiple linear regression model to relate river network structures, water quality parameters, land-use changes and dam locations with fish species richness. Then we performed an unsupervised learning and clustering method, the self-organizing map (SOM), to nonlinearly interrelate the complex hydro-chemo-ecosystems. Following that, we compared the major forcing factors detected by different models to evaluate the anthropogenic influences on fish species richness. Our results showed that although based on the same datasets, the forcing factors identified by different methods may not be consistent, and therefore would result in distinct method-oriented stressor-response relationships. Patterns described by linear models focused on the changes of fish species richness with the use of the selected predictors; while patterns described by nonlinear models tended to systematically link multiple variables without the identification of major predictors. Based on the results of our analysis, we recommend that a more effective watershed management strategy should consider landscape as well as riverine habitats as a whole and maintain long-term monitoring programs as a key element to river conservation.