A walkthrough based on...
Kopp D, Bouchon-Navaro Y, Louis M, Legendre P, Bouchon C (2012) Spatial and Temporal Variation in a Caribbean Herbivorous Fish Assemblage. J Coast Res. 28(1): 63–72.
While not directed towards microbial ecology, this walkthrough illustrates the use of principal coordinates of neighbour matrices (PCNM), which has yet to become familiar to microbial ecology.
The walkthrough below is a simplified illustration of the central procedures used by Kopp et al. (2012). These authors investigated the spatial and temporal variation in a herbivorous fish assemblage found on 10 reefs around the Caribbean island of Guadeloupe. Reefs were differentiated by varying depth and cover availability and exposure to anthropogenic influences. Contextual data describing environmental parameters as well as temporal and spatial distribution were gathered. Data on the fish assemblage was collected by visual inspection along band transects. The authors were able to link several environmental variables with both spatial and temporal patterns exhibited by the assemblage.
The authors conducted two surveys, one during the dry season and another during the rainy season. One reef, located within a marine protected area (MPA), was surveyed on a monthly basis to allow temporal pattern detection. Another reef site, located within an MPA and on an easily accessible reef slope, was surveyed along a depth gradient with 5 intervals to allow insight in assemblage variation with depth. A specific set of contextual parameters were measured in both the temporal and spatial studies
Data preparation
While most environmental parameters reflected the conditions at the time of sampling, rainfall values were the average for the month before a sample was taken. The authors note that lagged effects of rainfall may influence the fish assemblages under study. Species data were Hellinger transformed prior to analysis.
Hypothesis testing
Mann-Whitney U tests (non-parametric tests, similar to t-tests, which test whether two populations have comparable rank composition) were used to assess whether assemblages found on flat reefs were the same as those found on sloped reefs.
Direct gradient analysis
Redundancy analysis (RDA), a form of constrained analysis, was used to assess the influence of contextual explanatory variables on the census data. Results were visualised as correlation triplots. A stepwise forward selection procedure (see "Variable selection in MLR" under multiple linear regression) was used to select variables which "best" explained variation in the response data, with significance assessed by permutation. RDA with forward selection was also applied to data sets with synthetic descriptors (e.g. species richness) as response variables and environmental descriptors as explanatory variables.
Temporal analysis with PCNM
The authors applied principal coordinates of neighbour matrices (PCNM) to their fish data after separating data pertaining to juvenile and adult fishes. The PCNM variables that passed forward selection were used as explanatory variables in RDAs. PCNM variables were used either alone or as part of a submodel which included multiple PCNM variables.
Additionally, the association between environmental variables and temporal variation in the fish assemblage data was estimated by Kendall rank correlations.
The illustration below is a simplified representation of the procedures used by Kopp et al. Please consult the manuscript for a more complete explanation.