The design of our bottom sampling which we described in detail earlier was pretty simplistic. We used depth and sediment characteristics to divide up the New York and New Jersey seascapes into bottom habitat patch types. Using a patch model to classify habitat based on bottom features is usually flawed in the sea particularly in the temperate coastal ocean where dynamic water column features like temperature and currents affect everything from the physiology to the movements of cold blooded and often nearly neutrally buoyant animals. But you have to start sampling systematically based on the questions you want to ask and what little you know about the system in the beginning. The information we had to design our field study with was sonar measurements of depth and sediment type and an intensive study of oceanography (Chant et al.2008, Schofield et al., 2008, Moline et al.2008).
Our big questions in the ECOS research program are: What are the dominant physical and biological processes controlling the abundance and health of animals and their assembly into communities in the coastal ocean? What are the relative importances and scales of operation of those dominant processes? And finally, how can we use information about those processes to identify sweet spots in the ocean likely to sustain healthy marine communities which should therefor be conserved?
In our inshore surveys we sampled at two time scales (year and season) and three spatial scales (seascapes [10s kms] , depth strata within seascapes [3-5 kms], patches of sediment within depth strata within seascapes [100s of m]). We assume that scales of community variation match the scales of operation of the important physical and biological processes causing the variation 1. So we can use the nested survey design to identify scales of community variation falling within the limits of the study resolution and get clues about the operational scales some of the driving processes. (Our study resolution doesn't include variation on time scales of hours to weeks or decades, or over spatial scales more than a few10s of kilometers. We can identify meter to sub-meter scale spatial changes in the distributions and habitats of some animals visible in imagery we collected with the underwater video sled. There are other longer and larger scale surveys we can turn to to define the context for our study). Once we identify the dominant scales of community variation we can use our scale matching assumption and measurements of water column and bottom features made with satellites and radar by MARCOOS and with water quality sensors, acoustics, and underwater video by us to winnow down the likely candidate processes.
More specifically we can use our nested beam trawl survey to ask:
-Are dissimilarities in the fish and invertebrate community bigger in time or in space?
-Is the species turnover bigger between years or between seasons?
-How dissimilar are the communities in the two seascapes?
-Are differences in the communities in the two seascapes bigger than those in depth strata within the seascapes or sediment patches within depth strata within seascapes.
-Does the community vary simultaneously in space and time and at what scales do those spatial dynamics occur?
Answering the last question should give us clues about the nature of dynamic water column features strongly affecting the distributions and abundances of the animals.
To answer these questions I used the vegan library in R software to calculate ecological distances between samples and to visualize the relationships between community structure and the survey design factors (year, season, seascape, depth strata, patches of fine or medium sand) using non metric multidimensional scaling (metaMDS; e.g. Figs 3 & 4). I then used partial redundancy analysis (pRDA) to estimate proportions of the variation in the fish and invertebrate community to the nested factors we used to design the study (Fig. 2).
Before running these analyses I removed species occurring in less than 3 samples from the data and standardized abundances by dividing numbers of the remaining fish and invertebrates by the number of meters of bottom trawled to collect each sample. I then double square root transformed these standardized abundances to shrink the range and make the analyses sensitive to the rare as well as common species. Finally I used the bray-curtis index of similarity (%) in species composition as the index of ecological distance between the samples for the multidimensional scaling.
Before running these analyses I removed species occurring in less than 3 samples from the data and standardized abundances by dividing numbers of the remaining fish and invertebrates by the number of meters of bottom trawled to collect each sample. I then double square root transformed these standardized abundances to shrink the range and make the analyses sensitive to the rare as well as common species. Finally I used the bray-curtis index of similarity (%) in species composition as the index of ecological distance between the samples.
1 Is the scale matching assumption always valid? Just as the variability of the atmosphere is buffered as it it translated across the oceans surface, don't the organisms use their physiologies and behaviors to lower “pitch” of the ocean?