In our
study, we used replicate paired control and impact plots, with pairing
based on proximity (45–100m separation) and similarity in
mangrove forestry measures (Ellis and Bell, 2004b). This design
was intended to minimize pre-disturbance differences in fish abundance
and species composition. Replicate pairs of plots were used
to account for spatial variability in the effect of the impact event. In
this way, we could be more certain that any post impact changes
should correctly be attributed to the impact event rather than a
random change in either a single control or impact site.