Spatial and temporal variation in littoral-zone fish communities: A new statistical approach

Through the use of principal coordinates analysis, a technique is presented that allows the quantitative partitioning of the total variation within a data set into its spatial and temporal components. The graphical nature of this technique, in combination with the quantitative results, aids in the examination of specific hypotheses regarding the structure of the variation. These hypotheses are tested against null or model matrices of the structure of the variation with statistical significance assessed using Mantel's test. This approach is applied to a data set of species' abundances from the littoral-zone fish communities of three Ontario lakes sampled three separate times. The variation was partitioned and the influence of different similarity measures was explored. Among-lake (spatial) variation accounted for 67-81% of the total variation. The time of sampling represented 1-3% of the total. There was significant among-lake variation for all three sampling periods whereas only one lake showed a significant temporal component. Together, both lake and time of sampling accounted for 74-86% of the variation, the remainder being within-lake variation (i.e., among-site or site-by-time interactions). The structure of the variation was sensitive to the similarity measure used as a result of differing emphasis on particular attributes (e.g., relative versus absolute abundance).