Michigan Stream Fish: Distribution Models, Future Predictions, and UrbanImpacts.

[摘要] To identify abiotic requirements necessary to maintain growth, survival, and reproduction of species, researchers often use models to search for patterns between species occurrences and environmental characteristics of sampled locations.Models are used to test hypotheses about processes that are important to organisms, and used to make species distributions and abundance predictions for management application.Several analytical methodologies were assessed for modeling associations between fish species and habitat characteristics.Brook trout presence/absence models were created using different techniques: multiple regression, logistic regression, neural networks, and classification trees.Results showed that all methods could be successful provided underlying assumptions were met.This analysis indicated classification trees were a technique uniquely suited to the creation of large numbers of interpretable models.Classification tree methods and landscape-scale habitat variables were used to create and validate presence/absence models and relative abundance models for Michigan stream fish.Ninety-three presence/absence models were on average 72% correct and 46 relative abundance models were on average 76% correct when tested against independent data.Water temperature and catchment area were determined to be dominant constraints on fish distributions.Classification trees were applied to land-use alteration and climate change scenarios to understand how fish communities of the Muskegon River system (Michigan, USA) would be structured through the year 2100.Models predicted cold-water species reduction due to water temperature warming, and walleye loss due to increased urban development.Warm-water species were expected to have substantial range increases. Classification trees were used to explore how managers might manipulate predictor variables to maximize probability of species presence. Urban development has been shown to have strong negative impacts on fish community quality.However, effort is needed into understanding why streams with similar urban levels have fish communities of significantly different quality.Univariate tests and covariance structure analysis were used to investigate how natural and anthropogenic features are related to variance of fish biotic integrity in urban streams.Urbanized streams with more natural land-cover, more point source discharges, better water quality, and that are adjacent to non-urbanized streams supported higher quality fish communities.

[发布机构] University of Michigan

[效力级别] Distribution Models

[关键词] Michigan Stream Fish;Distribution Models;Climate Change;Urban Impacts on Fish;Muskegon River;Landscape-scale Analysis;Natural Resources and Environment;Science;Natural Resources and Environment