Local and Landscape Predictors of Fish-assemblage Characteristics in the Great Swamp, New York

We used local and landscape models to predict fish assemblages in the Great Swamp, NY, a region undergoing rapid development. Fish were surveyed across 17 sites. Fish-species richness, diversity, percent intolerant species, and IBI metrics for fish species richness, benthic insectivores, terete minnows, and dominant species were calculated. Local stream features were characterized and surrounding land cover/use was quantified at four different scales (reach, segment, network, and watershed). Regression analysis and multinomial cumulative logit models were used to predict how fish assemblages varied according to habitat characteristics. Within the local variables, pool variability predicted diversity and epifaunal substrate/available cover predicted percent intolerant species and IBI metrics for fsh species richness, benthic insectivores, and terete minnows. The scale of analysis influenced which landscape-level predictors measuring percent wetland, forest cover, or residential land use best explained diversity, percent intolerant species, and IBI metrics for benthic insectivores, terete minnows, and dominant species. Although no single model (local or landscape) best predicted assemblages, the land cover/use at the segment (100-m buffer for 1 km upstream) scale provides sufficient information about fish assemblages to support this scale as optimal for regional land-use planners. Our findings show that forest cover should be maintained to protect fish assemblages in the Great Swamp and development that alters stream-habitat heterogeneity should be curtailed.