The performance of fish as indicators of habitat conditionmaybe
improved by accounting for natural spatial and temporal variability
when designing and analyzing tests of anthropogenic impacts.
Our experimental protocol and sampling scheme were designed to
compensate for the anticipated large scale (km) spatial variability in
fish abundance by confining comparison of trimming to untrimmed
plots to “within proximal pairs”. The underlying assumption of this
design was that the abundance of fish would be more similar in
plots in close proximity to one another and, accordingly, the difference
in fish abundance within pairs would be similar for all
pairs prior to trimming. This assumption proved to be incorrect
as, like Laegdsgaard and Johnson (1995), we found that mangrove
fish community assemblages can vary considerably, even over short
distances. The difference in fish abundance between paired sites
each month varied greatly from pair to pair (mean monthly CV of
differences = 114.2%). Moreover, the differences were not consistent
in magnitude or direction from month to month. The detection
of impacts is particularly challenging when pre-disturbance spatial
and temporal variability interact in this way (Underwood and
Chapman, 2003).