Long—term studies provide useful insight into the functioning of populations and communities. Gaining a long—term perspective may be important in providing back—ground information upon which BACI (Before–After—Control–Impact)—type impact studies can be designed or in identifying many types of impact, particularly those of a large—scale and diffuse nature. Although marine soft—bottom communities commonly exhibit patchy distributions, in many long—term studies few replicates, from each site, are collected on each occasion. To illustrate potential problems encountered when spatial variation confounds temporal patterns, we use data collected from a monitoring program of the intertidal sandflat communities of Manukau Harbour, New Zealand. Randomization techniques were used to generate possible temporal sequences, based on a sample size similar to that used in many long—term studies. In the generated sequences the timing and size of annual density changes were often different from the observed sequence, and indicators of trends in density were often missed. In order to provide accurate descriptions of temporal variability in the soft—bottom communities of estuaries and coastal embayments, low bias and highly precise density estimates from each sampling occasion are required.