| Abstrakt: |
AbstractSpatial variation in stream fish populations and habitats can be partitioned into many hierarchical levels (e.g., among drainage basins, streams within basins, and sites within streams). Studies of site level habitat relationships in more than one stream are common in fisheries research. Analyses of such data typically involve multiple regression to relate site level habitat features and fish population characteristics (e.g., biomass). Because sites within streams may not be independent, multiple-regression models should also include qualitative stream effects. As we show here with hypothetical and real examples, ignoring stream effects can lead to erroneous conclusions about the significance of site level habitat variables. Site and stream level effects may function independently or interactively in relation to fish populations. Alternatively, site and stream level effects may be confounded. An example with data on trout populations revealed that highly significant site level effects were only marginally significant after adding qualitative stream effects to the regression model. Further examination of the data revealed that consideration of variation among streams added much insight and complexity to understanding how site level effects may be related to trout populations. Inclusion of stream (or other large-scale) effects in regression models of site level habitat relationships may be a valuable method to more fully understand the spatial scale of habitat variability fish are responding to. |
Plný text