Premium
Kernel Density Estimates of Linear Home Ranges for Stream Fishes: Advantages and Data Requirements
Author(s) -
Vokoun Jason C.
Publication year - 2003
Publication title -
north american journal of fisheries management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 72
eISSN - 1548-8675
pISSN - 0275-5947
DOI - 10.1577/m02-141
Subject(s) - statistics , kernel density estimation , range (aeronautics) , mathematics , sample size determination , kernel (algebra) , home range , smoothing , estimator , computer science , econometrics , ecology , combinatorics , biology , materials science , habitat , composite material
Kernel density estimates of linear home range can increase the information content of the home range estimate. Particularly in lotic systems, univariate kernel density estimates have desirable properties relative to simple reporting of linear ranges. Kernel density estimates are calculated from a set of relocation points (i.e., radiotelemetry contacts) that can be interpreted as a utilization distribution (UD). The UD estimates the amount of time spent at a given point within the home range. The amount of time that elapses between relocations (sampling interval), total number of relocations (sample size), and bandwidth (smoothing applied to the estimate) are important considerations when using kernel density estimates. Monte Carlo simulations using unimodal and bimodal distributions sampled randomly at sample sizes of 10‐100 and three automated bandwidth selection procedures (simple normal reference, Silverman's rule of thumb, and Sheather‐Jones plug‐in) suggested that at sample sizes of 30 or more kernel density estimates began to level off and the largest gains in accuracy were evident for all three bandwidth selection procedures. I present example kernel density estimates for flathead catfish Pylodictis olivaris and use them to illustrate methods of reporting data. A literature review found relatively few examples of home range delineation for stream fishes but confirmed that the most common practice was to report the distance between the most upstream and downstream relocations. Sample size information was often inadequately reported in the reviewed literature.