Monitoring programs for freshwater fish are commonly rely on
fixed-site sampling, wherein non-randomly selected permanent
stations within particular habitat types are sampled repeatedly
through time. The Long Term Resource Monitoring Program fish
sampling sites were selected based on subjective judgments that
particular sites were representative of their habitat type.
Unfortunately, fixed-site sampling cannot produce data that can be
used to verify whether specific sampling sites are truly
representative of habitats. Because fixed-site sampling designs
are not spatially randomized, they cannot support statistically
valid conclusions about patterns or trends in the larger study
area. Fixed-point sampling designs can only be used to make
inferences about the specific set of sampling sites. Fixed-site
sampling is often justified based on the supposition that, for
example, fish catches are much more variable among different site
locations than within any specific site. Although this
supposition is commonly accepted, it has not been critically
tested. To examine this supposition, I partitioned the 1990 LTRMP
fish data into unique combinations of species, study reach,
sampling gear, habitat class, and time period and estimated
within-site and among-site variance from each partition.
Among-site variation was actually less than within-site variation
in over 50% of the partitions. The hypothesis that among-site
variation was less than or equal to within-site variation could
not be rejected. The LTRMP therefore adopted a stratified random
fish sampling design in 1993.