H. Franklin - Georges Bank Historical Data

Click here for an Adobe PDF file version of this poster

Retrospective Analysis of Variability in Zooplankton Composition on Georges Bank and the Northwest Atlantic

Heidi M. Franklin¹, Stephen M. Bollens¹,and Andrew R. Solow²
¹San Francisco State University, 1600 Holloway Ave., San Francisco, CA 94132; ²WoodsHole Oceanographic Institute, Woods Hole, MA 02543

ABSTRACT - INTRODUCTION - DATA - METHODOLOGY - RESULTS - SUMMARY - ACKNOWLEGEMENTS

ABSTRACT

The goal of the project described in this presentation is to identify trends in interannual to interdecadal variations in the relative abundance or composition of zooplankton communities on and around Georges Bank. We have applied newly developed statistical techniques to MARMAP zooplankton data covering the period 1977-1987 to identify trends in zooplankton composition on Georges Bank and in three adjacent regions- the Gulf of Maine, Southern New England, and the Mid-Atlantic Bight. We grouped the MARMAP zooplankton data into 6 groups (5 dominant taxa by sub-region and 'other'), binned the data into annual means, and performed compositional trend analyses on the data. Preliminary results from Gulf of Maine (GOM) show a significant compositional shift over the 11 year period. The Georges Bank trend exhibits a similar pattern to GOM but the trend was not significant. We present these results in relation to historical data on predator abundance (Atlantic herring and mackerel). We found a high negative correlation of predators vs. M. lucens for GOM which could be a function of size-selective predation. Future work will focus on examining atmospheric and physical oceanographic data which relate to factors previously hypothesized to control zooplankton community structure; and interdecadal analyses.

INTRODUCTION

The goal of the Georges Bank-Northwest Atlantic GLOBEC program is to develop an understanding of the physical and biological processes controlling the abundance of marine animals in space and time. This project is a retrospective analysis of historical data focusing on interannual to interdecadal variations in the relative abundance or composition of zooplankton communities on and around Georges Bank.

We use relative abundance (compositional) data to identify trends. Some of the advantages with compositional data analysis are:

While total abundance may stay constant, species composition may change dramatically;

Changes in composition can provide additional information about causes of biological changes; and
Composition may be spatially less variable (less patchy)than abundance (e.g. Mackas, 1984) and therefore, more accurately estimated from relatively sparse spatial sampling.

The goals of this project are:

Document and chart interannual trends in composition of zooplankton communities on a regional basis within the Northwest Atlantic;
Determine whether these trends are consistent with hypothesized biological and physical causes (trends in predation, mesoscale advective events, and local stratification and wind-mixing); and
Extend the analysis as feasible to the interdecadal time scale.

Figure 1. (a) Locations of MARMAP sampling stations (b) Schematic surface layer circulation on and around Georges Bank (includes the four sub-regions).

DATA

The zooplankton data in this presentation were collected as part of the MARMAP program. The data were kindly provided to us by K.Sherman and J. Goulet of NMFS. There are 11 years of data from 1977 to1987. Table 1 shows the 4 sub-regions and the 5 dominant taxa for each region plus an 'other' group. The 4 geographic sub-regions and their surface currents are shown in Figure 1.

The predator data consist of NMFS/NEFSC stock size estimations of mackerel, Scomber scombrus, and herring, Clupea harengus, two known predators of zooplankters generally, and larger copepods in particular. These data were kindly provided by W. Overholtz of NEFSC/NMFS.

METHODOLOGY

This presentation focuses on the interannual trends of zooplankton variability and comparisons with predator abundances over an11-year period for 4 sub-regions. Analyses of temporal trends in the composition of zooplankton communities on and around Georges Bank are based on methods in Solow (1994) and are as follows. Suppose that a community contains p groups and let Z_i(t) be the relative abundance of group i in period t. The composition is defined by Z(t) = (Z₁(t) Z₂(t)... Z_p(t)). The relative abundances were transformed into X(t)= (X₁(t) X₂(t) ... X_p(t)) where (1):

We want to construct a scalar index series of the form (2).

that best captures any common trend in the individual components. Principal component analysis (PCA) uses "weights",w1, w2, ..., wp, to maximize the variance of the index series Y(t) subject to the constraint that the weights have unit sum of squares. Here, instead of maximizing the variance, we use weights to maximize the smoothness of Y(t) as measured through lag-one autocorrelation.

The weights that maximize the lag-one autocorrelation of Y(t) are given by the elements of the eigenvector of the matrix C^-1Vcorresponding to the smallest non-zero eigenvalue. Here, C is the p-by-p sample covariance matrix of X(t) and V is the p-by-p sample covariance matrix of the first differences of the elements of X(t). As in PCA, this method can be used to construct a set of p-1 orthogonal index series with lag-one autocorrelation decreasing from the maximum possible. The weights in these index series are given by the elements of the non-zero eigenvectors of C^-1V. As in PCA, the orthogonality constraint can give rise to behavior in the higher order index series that should not be given physical or biological interpretation. A smoothed version of the components of X(t)can be constructed by projection on to one or more of these index series and these can be transformed back to the compositional scale with the inverse transformation of (1).

It is necessary to determine how many index series to retain for analysis. Solow (1994) outlined a sequential randomization procedure for assessing the significance of the index series. For example, to assess the significance of the first index series, its lag-one autocorrelation is compared to the distribution of lag-one autocorrelations generated by randomly permuting the time-ordering of the compositional time series. If the first index series is retained, then the significance of the second index series can be assessed.

GEORGES BANK	GULF OF MAINE	SOUTHERN NEW ENGLAND	MID ATLANTIC BIGHT
Calanus finmarchicus	Calanus finmarchicus	Centropages typicus	Centropages typicus
Pseudocalanus spp.	Pseudocalanus spp.	Calanus finmarchicus	Pseudocalanus spp.
Centropages typicus	Metridia lucens	Pseudocalanus spp.	Temora longicornis
Paracalanus parvus	Centropages typicus	Sagitta elegans	Calanus finmarchicus
Paracalanus parvus	Paracalanus parvus	Metridia lucens	Sagitta elegans
other	other	other	other

Table 1. MARMAP zooplankton data, 1977-1987 :5 most abundant taxa.

RESULTS

The compositional (relative abundances) data are shown in Figure 2. The first index series derived from the X(t) transformed relative abundances for the sub-regions are shown in Figure 3. The lag-one autocorrelations and the significance levels are presented in Table 2. The estimated trends for each taxon for the 4 sub-regions are shown in Figure 4. These are the results of projecting the components of X(t), the zooplankton taxa, onto the first index series and then transforming back to the compositional scale using the inverse of equation (1).

Figure 2. Five dominant taxa compositional abundances for each sub-region. Note that the taxa are not the same in the four regions.

Figure 3. Index series for the four sub-regions from 1977 - 1987 (SNE, 1977 - 1986).

SUB-REGION	LAG-ONE AUTOCORRELATIONS	SIGNIFICANCE LEVEL
GEORGES BANK	0.566	0.324
GULF OF MAINE	0.738	0.031
SOUTH NEW ENGLAND	0.604	0.323
MIDATLANTIC BIGHT	0.500	0.512

Table 2. Lag-one autocorrelations and estimated significance levels of the index series shown in Figure 3.

Figure 4. Estimated compositional trends for the five dominant taxa by sub-region.

Figure 5. Herring and mackerel stock sizes for the 1977 - 1987 period from VPA (NEFSC,NMFS).

Figure 6. Correlation coefficients of predators and X(t) transformed zooplankton groupings. Predators versus first series index correlations represented by 'MAF1' on x-axis.

The Gulf of Maine (GOM) sub-region was the one region with a significant first index series. Since the annual mean compositional trend analysis results for Southern New England (SNE) and Mid Atlantic Bight (MAB) were not significant, we will not focus the Results and Summary on these two sub-regions. Although the Georges Bank(GB) lag-one autocorrelation (0.57)was not significant (0.32), we will include GB in our discussion for the following reasons: 1) GB is the focus region

of the GLOBEC program; 2) GB is adjacent to the GOM; and3) the shapes for GOM and GB are quite similar - relatively flat from 1977to 1982, then a decrease until 1985 for GOM and 1987 for GB. It is important to note that the index series are sign-indeterminate and can show that a compositional shift is occurring but the time series do not show which zooplankton groups are shifting or in which direction. In order to determine which groups are shifting, correlations and projections are used.

By correlating the zooplankton taxa with the index series for GOM and GB, for GOM there is a high correlation with Metridia lucens(0.97), and moderate correlations with Paracalanus parvus (0.58).For GB, there is a high correlation with Centropages hamatus (0.90)and Pseudocalanus (0.73). From the graphical representation of the projected taxa (Figure 4), GOM shows a compositional shift away from M. lucens starting in 1982 and towards the 'Other' group. For GB, the projected groups show a compositional shift away from Pseudocalanus and towards C. hamatus over the 11 year time series.

The next step of our analyses is to look at the relationships between predators and zooplankton groups. The planktivorous fish stock sizes are shown in Figure 5. The predator-zooplankton correlations consist of comparing each separate zooplankton group abundances and the total zooplankton abundances versus the mackerel stock sizes, herring stock sizes, and the sum of the mackerel and herring. We also correlated the predator groupings with the first index series (MAF1). The correlation coefficients for these comparisons are shown in Figure 6. For GOM, M. lucens is highly negatively correlated (~-0.8)with mackerel, herring, and their sum. For GB, C. hamatus shows a moderate correlation (~0.6) with all three predator groupings; and, a moderate negative correlation (~-0.6) for Pseudocalanus with the mackerel and sum total groupings.

SUMMARY

A significant compositional shift occurs in the GOM region. This shift appears similar to that of the GB region although the GB trend was not significant. It should be emphasized that the dominant taxa used are specific to each subregion and thus were not the same for all regions. The changes in GOM could be related to the increase in the stock sizes of the planktivorous fish. M. lucens, a relatively large-bodied copepod, is negatively correlated with the stock sizes of fishes, possibly as a result of increased predation.

The next steps for this study are to analyze the compositional shifts for 5 or more taxa common to all subregions; and examine zooplankton data with finer temporal resolution (e.g. monthly and seasonal). Also, we expect to expand the analyses to interdecadal time scales as well (Atlantis,1939-41; GLOBEC, 1994-1997). Lastly, we are going to examine atmospheric and physical oceanographic data (e.g. North Atlantic Oscillation index, winds, and stratification) to investigate factors driving compositional shifts.

ACKNOWLEDGEMENTS

We would like to thank NSF/NOAA under the Georges Bank/NW Atlantic GLOBEC program for funding for this project.

Page maintained by Sean Avent. 6/29/99.