To develop a map that would predict where saltmarsh sparrows live and reproduce in Connecticut, I compared models to test a) whether field data or remote-sensing data most effectively characterized within-marsh conditions that relate to sparrow occurrence, and b) whether including landscape-level variables improved model fit. The best sparrow presence model used a variable derived from raw spectral reflectance values associated with plots where sparrows did not occur, while the best nest presence model used a combination of vegetation structure descriptions. A second nest model, built using high resolution remote sensing data that organized marsh characteristics into high and low marsh categories, had enough support for state-wide application.
When the models were tested using new data, model performance, assessed by determining the area under a receiver operating curve and the model deviance, was significantly better than expected by chance alone. A large proportion of the saltmarsh area in Connecticut was predicted to have a high probability of being occupied by sparrows, yet a much smaller proportion of marsh was predicted to have a high probability of having nests. While detailed delineation of plant communities in the marsh provided good predictions of sparrow nesting, they poorly predicted presence. On the other hand, because areas of nesting activity are not well-identified by species presence models, a distribution model that describes only species presence would provide misleading information about where the most important areas for reproduction lie.
","abstract_html":"<p>To develop a map that would predict where saltmarsh sparrows live and reproduce in Connecticut, I compared models to test a) whether field data or remote-sensing data most effectively characterized within-marsh conditions that relate to sparrow occurrence, and b) whether including landscape-level variables improved model fit. The best sparrow presence model used a variable derived from raw spectral reflectance values associated with plots where sparrows did not occur, while the best nest presence model used a combination of vegetation structure descriptions. A second nest model, built using high resolution remote sensing data that organized marsh characteristics into high and low marsh categories, had enough support for state-wide application.</p> <p>When the models were tested using new data, model performance, assessed by determining the area under a receiver operating curve and the model deviance, was significantly better than expected by chance alone. A large proportion of the saltmarsh area in Connecticut was predicted to have a high probability of being occupied by sparrows, yet a much smaller proportion of marsh was predicted to have a high probability of having nests. While detailed delineation of plant communities in the marsh provided good predictions of sparrow nesting, they poorly predicted presence. On the other hand, because areas of nesting activity are not well-identified by species presence models, a distribution model that describes only species presence would provide misleading information about where the most important areas for reproduction lie. </p>","abstract_has_math":false,"creators":["Meiman, Susan T."],"institution":null,"degree_name":"Master of Science","degree_level":null,"degree_discipline":"Ecology and Evolutionary Biology","degree_department":null,"school":null,"contributors":["Daniel L. Civco; Kent E. Holsinger","Chris S. Elphick"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2011,"date_issued":"2011-04-29T07:00:00Z","date_published":"2011-04-29T07:00:00Z","updated_at":"2026-07-24T06:31:35Z","subjects":["Ammodramus caudacutus","Bayesian hierarchical models","habitat use","occupancy modeling","remote sensing","salt marsh","saltmarsh sparrow"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.lib.uconn.edu/gs_theses/72","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Daniel L. Civco; Kent E. Holsinger","Chris S. Elphick"]},{"key":"dc:creator","label":"Author","values":["Meiman, Susan T."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2011-04-29T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Ecology and Evolutionary Biology"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Ammodramus caudacutus","Bayesian hierarchical models","habitat use","occupancy modeling","remote sensing","salt marsh","saltmarsh sparrow"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.lib.uconn.edu/gs_theses/72"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["To develop a map that would predict where saltmarsh sparrows live and reproduce in Connecticut, I compared models to test a) whether field data or remote-sensing data most effectively characterized within-marsh conditions that relate to sparrow occurrence, and b) whether including landscape-level variables improved model fit. The best sparrow presence model used a variable derived from raw spectral reflectance values associated with plots where sparrows did not occur, while the best nest presence model used a combination of vegetation structure descriptions. A second nest model, built using high resolution remote sensing data that organized marsh characteristics into high and low marsh categories, had enough support for state-wide application.
When the models were tested using new data, model performance, assessed by determining the area under a receiver operating curve and the model deviance, was significantly better than expected by chance alone. A large proportion of the saltmarsh area in Connecticut was predicted to have a high probability of being occupied by sparrows, yet a much smaller proportion of marsh was predicted to have a high probability of having nests. While detailed delineation of plant communities in the marsh provided good predictions of sparrow nesting, they poorly predicted presence. On the other hand, because areas of nesting activity are not well-identified by species presence models, a distribution model that describes only species presence would provide misleading information about where the most important areas for reproduction lie.
"]},{"key":"dc:title","label":"Title","values":["Modeling Saltmarsh Sparrow Distribution in Connecticut"]}]}],"canonical_facts":{"dc:contributor":["Daniel L. Civco; Kent E. Holsinger","Chris S. Elphick"],"dc:creator":["Meiman, Susan T."],"dc:date.available":["2011-04-29T07:00:00Z"],"dc:description.abstract":["To develop a map that would predict where saltmarsh sparrows live and reproduce in Connecticut, I compared models to test a) whether field data or remote-sensing data most effectively characterized within-marsh conditions that relate to sparrow occurrence, and b) whether including landscape-level variables improved model fit. The best sparrow presence model used a variable derived from raw spectral reflectance values associated with plots where sparrows did not occur, while the best nest presence model used a combination of vegetation structure descriptions. A second nest model, built using high resolution remote sensing data that organized marsh characteristics into high and low marsh categories, had enough support for state-wide application.
When the models were tested using new data, model performance, assessed by determining the area under a receiver operating curve and the model deviance, was significantly better than expected by chance alone. A large proportion of the saltmarsh area in Connecticut was predicted to have a high probability of being occupied by sparrows, yet a much smaller proportion of marsh was predicted to have a high probability of having nests. While detailed delineation of plant communities in the marsh provided good predictions of sparrow nesting, they poorly predicted presence. On the other hand, because areas of nesting activity are not well-identified by species presence models, a distribution model that describes only species presence would provide misleading information about where the most important areas for reproduction lie.
"],"dc:identifier":["https://digitalcommons.lib.uconn.edu/gs_theses/72"],"dc:subject":["Ammodramus caudacutus","Bayesian hierarchical models","habitat use","occupancy modeling","remote sensing","salt marsh","saltmarsh sparrow"],"dc:title":["Modeling Saltmarsh Sparrow Distribution in Connecticut"],"thesis:degree_discipline":["Ecology and Evolutionary Biology"],"thesis:degree_name":["Master of Science"]},"updated_at":"2026-07-24T06:31:35Z"}