Sampling Attributes of Puerto Rico Stream Fishes: Bias, Selectivity, and Environmental Influences

Show full item record

Title: Sampling Attributes of Puerto Rico Stream Fishes: Bias, Selectivity, and Environmental Influences
Author: Brown, Christin Hambrick
Advisors: Kenneth H. Pollock, Committee Member
Thomas J. Kwak, Committee Chair
James F. Gilliam, Committee Member
Abstract: Puerto Rico, an island in the Caribbean Sea, is known for its marine sport and commercial fisheries, but the freshwater habitats of the island also support a substantial number of fishes, which provide recreational and subsistence fishery values. There are about 80 fish species that inhabit Puerto Rico freshwaters. Of those, there are fewer than 10 native fish species that reside within the rivers, and they are of primary management concern. Management of these stream fish resources would be enhanced by an understanding of gear catchability, a standardized sampling method, and accurate population estimates. My primary objectives for this study were to (1) quantitatively describe gear efficiency and selectivity relationships to estimate stream fish populations in Puerto Rico; (2) evaluate population models among species using electrofishing catch results analyzed with mark-recapture and removal methods to identify the most suitable parameter-estimating model; (3) use these findings to develop a standardized stream fish sampling protocol to be applied island-wide; and (4) develop empirical, hierarchical models that describe relationships between fish catchability and instream habitat and water quality parameters for each native fish species. In my first research component, I compared two fish sampling gear types (electrofishing and seining) and four models for estimating fish population parameters (Petersen mark-recapture and removal estimators of 2–4 sampling passes) to provide the quantitative basis for development of a standardized sampling protocol for Puerto Rico stream fish. I found electrofishing more efficient and logistically feasible for collecting fish in these environments. I determined that three- and four-pass removal models were more accurate than the Petersen mark-recapture model or a two-pass removal model, and that accuracy was similar between three- and four-pass removal models. I investigated variations of models that account for assumption violations and found model Mb, that adjusts for fish behavioral effects, to provide the overall best and most parsimonious fit for estimating population parameters. Based on these findings, I propose a standard fish sampling protocol for Puerto Rico wadeable streams that includes sampling stream reaches from 100 m to 200 m long, using the appropriate electrofishing gear (backpack or barge electrofishers) and conducting three sampling passes of equal effort. A Zippin-type, maximum-likelihood estimator will then be used to calculate estimates of fish population densities. I sampled fish in 81 wadeable stream reaches island-wide, totaling 105 sampling occasions, using the standardized sampling protocol with backpack or barge electrofishers. I estimated fish catchability using the standard maximum-likelihood removal estimator. At each sampling location, I measured seven instream habitat and 13 water quality parameters. I employed a correlation matrix to reduce 20 environmental parameters to seven, then developed hierarchical regression models and used AIC model selection to quantify the most parsimonious relationships between catchability and environmental variables. Mean catchability among six fish species ranged from 0.30 to 0.55. I found no trend relating environmental parameters to variation in catchability among benthic and water-column species. The most influential environmental parameters on fish catchability were mean column velocity, mean stream width, and percent cover. Catchability was negatively correlated to mean column water velocity and mean stream width and positively to percent cover. Turbidity was not closely associated with electrofishing catchability within the range of my sampling. The regression models that I developed can be used to better understand environmental variables that influence electrofishing catchability and may be applied to more efficiently estimate fish populations. Because these models correct for bias associated with varying sampling conditions, they can be utilized with single-pass electrofishing data to estimate stream fish populations. These models will enable fisheries researchers and managers in Puerto Rico to conduct fish population estimates with a single field sample, saving time and expense, with minimal bias. More complete, quantitative estimates of the fish community may then form the basis for improved stream fish and ecosystem management.
Date: 2008-10-16
Degree: MS
Discipline: Fisheries and Wildlife Sciences

Files in this item

Files Size Format View
etd.pdf 799.2Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record