Does watershed land use influence freshwater fish assemblages in San Francisco Bay Area streams?

Background/Question/Methods

There are approximately 70 small streams that flow directly into the San Francisco Bay (Bay), that vary in their surrounding landscape. Bay watersheds include high-density urban areas to rural agricultural landscapes. Broad scale spatial analyses utilizing GIS watershed tools can be useful in elucidating correlations between land use and stream fish communities. Using a large dataset of stream fishes collected at 275 sites throughout the Bay region from 1993-1999, we examined the relationship between large-scale land use and local (reach) scale variables with fish communities in Bay streams. At each sample site, all captured fishes were identified to species and enumerated. Additionally, several reach scale variables were collected, including canopy cover, width, sediment type, and water quality. Here we expanded on these data to delineate individual watershed basins and characterize the composition of land cover (e.g., % impervious surface) and the presence of hydrologic structures (e.g., canals, reservoirs) for individual watersheds, including at multiple scales (e.g., riparian zone to the watershed scale). We then used a combination of non-metric multidimensional scaling (NMDS), BIOENV, and multiple regression analyses to illuminate the relationships between large-scale land use and local factors in explaining fish assemblages in the greater Bay area.

Results/Conclusions

The Bay region encompasses a broad diversity of land use types, ranging from ranchland to densely urbanized cities. This region also contains a diverse fish fauna; 39 species were collected at varying abundances and distributions.  Preliminary results show that NMDS ordination was able to explain most of the variation in the fish community structure (R²=0.97) and that an acceptable stress was achieved (0.17). The best model explaining fish community structure included both landscape (e.g., elevation, canal density) and local (e.g., conductivity, canopy cover, width) scale variables (R²=0.35, P<0.0001, BIOENV analysis), while watershed land cover variables were generally less important. The importance of landscape variables varied across scales, notably with localized urbanization (1km radius of sample sites) explaining a larger proportion of the variation in fish community structure than variables calculated at the total watershed-scale. With only two exceptions, nonnative fishes were clearly separated from native fishes in NMDS space and were associated with high percentages of altered landscapes. Our results suggest that fish assemblages may be structured by multi-scale processes and that these should be considered for native fish conservation in mixed-urbanized regions. Future research includes community shifts across decadal changes in land use within the region.