7 ToR d Developing EcoQO on changes in the proportion of large fish
7.3 Summary of OSPAR QSR2010 analysis of univariate community
7.3.2 Metric redundancy
Figure 7.3.1.3. Variation in the LFI calculated for 8 subareas within the North Sea for which Q1 IBTS data were available. See Figure for area key.
7.3.2 Metric redundancy
The need to select an indicator of state that responded primarily to fishing mainly influenced the selection of size structure as the basis of the fish community EcoQO (ICES, 2001). Prior to the 2002 Bergen North Sea Ministerial Conference, other aspects of composition, structure and functioning of communities had received most atten‐
tion in policy and scientific fora. Earlier work by WGECO concluded that a suite of
indices was necessary to capture adequately all changes in the composition, structure and functioning of fish communities (ICES, 1994; 1995; 1996). Consequently, many other metrics describing various aspects of the fish community composition and structure have also been applied to groundfish survey data (Greenstreet and Hall, 1996; Greenstreet et al., 1999; Jennings et al., 1999; Piet and Jennings, 2005) and related to different fishing activity scenarios (Greenstreet and Rogers, 2006). Many other po‐
litical commitments, such as OSPAR Annex V and the EC Marine Strategy Frame‐
work Directive, place great emphasis on the conservation and/or restoration of biodiversity. This begs the question, if the EcoQO is achieved, and the proportion by weight of demersal fish in the North Sea once again exceeds 30%, will other aspects of fish community composition, structure and function be equally safeguarded? This is essentially a question of metric redundancy; how many metrics are actually re‐
quired to capture adequately the types of change in the composition, structure and function of the North Sea demersal fish community that might be relevant to policy or management?
To address this issue, fifteen univariate community metrics (Table 7.3.2.1) were ap‐
plied to the Q1 IBTS data for the whole North Sea (Figure 7.3.2.1) and a principal components analysis (PCA) was carried out (Table 7.3.2.2). This analysis was then repeated, but for each of the eight subregions depicted in Figure 7.3.1.2. Figure 7.3.1.3 provides an example of variation in the proportion of large fish metric in each of the eight subregions, while Figure 7.3.2.2 provides an example of variation in each of the 15 univariate metrics in one subregion, the Kattegat and Skagerrak. Results of the PCA analysis carried out for each subregion are provided in Table 7.3.2.3.
At both the whole North Sea scale (Table 7.3.2.2), and at the subregional scale (Table 7.3.2.3), redundancy was clearly apparent among the fifteen metrics: the biomass and overall productivity metrics (often with the abundance metric) always associated with one factor; the two species richness metrics always associated with a second fac‐
tor; and the three species evenness metrics always associated with a third factor.
These three factors combined accounted for between 52% and 70% of the total vari‐
ance. At the whole North Sea scale the proportion of large fish indicator was associ‐
ated with the same factors as the abundance, biomass and overall productivity metrics (Table 7.3.2.2). However, at the subregional scale this same grouping of met‐
rics never occurred. Instead, the proportion of large fish indicator was twice the sole metric linked to a factor; twice it combined with mean weight; on three occasions it was linked to the same factor as the two species richness metrics; and once it com‐
bined with the age at maturity life‐history trait metric (Table 7.3.2.3). Associative be‐
haviour of the specific productivity, mean individual weight, and four life‐history trait metrics was quite variable, although L∞ and Lmat were always linked to the same factor, and only once linked to the same factor as the age at maturity metric. Regional variability of the association between indices may partially be as a result of smaller amount of data available at the subregional scale, compared with the North Sea scale.
This could enhance uncertainty in the empirical estimates of the covariance‐matrices underlying the PCA. To conclude, the proportion of large fish indicator on its own would be insufficient to monitor the health of the North Sea fish community, but it would perform a key role in any suite of “surveillance” metrics.
Table 7.3.2.1. Descriptions, abbreviations and derivations of the fifteen univariate community metrics applied to the groundfish survey data.
METRIC ABBREVIATION METRIC CALCULATION TERMINOLOGY
Biomass B
∑ ∑ ∑
Abundance N
∑ ∑
="
Where P is the total daily growth production and B total biomass of the fish community(see above).
Large fish indicator LFI
B
Mean weight of fish W
BN
W = Where B is the total biomass and N the total number of fish in the sample (see above).
Species count S S Where S is the count of the number of species in the sample.
Margalef’s species
= −
Where S is the total number of species and N the total number of individuals in the sample (see above)Pielou’s evenness J
LogS
log
Where Ns is the number of individuals belonging to species s, N is the total number of individuals of all species in the sample, and where S is the total number of species recorded in the sample (see above).Hill’s N1 diversity N1
Hill’s N2 dominance N2
∑
="
Where Ns is the number of individuals belonging to species s, N is the total number ofindividuals of all species in the sample, and where S is the total number of species
Mean age at maturity Amat
N
LFI (B>40cm/Btotal)
40
Hill's N1
2 3 4 5 6
Hill's N2
28
1980 1990 2000 2010 Year
1980 1990 2000 2010 Year
1980 1990 2000 2010 Year
Figure 7.3.2.1. Trends in fifteen indicator metrics applied to the IBTS Q1 groundfish survey data for the whole North Sea. See Table 7.3.2.1 for explanation of metrics (y axis labels).
Table 7.3.2.2. Summary of principal components analysis results for whole North Sea.
FACTORS TOTAL
VARIANCE EXPLAINED
1 2 3 4 5 6
Variance Explained 30.5% 19.2% 15.2% 13.1% 11.2% 9.0% 98.1%
Associated Metrics N1 N2
1
LFI (B>40cm/Btotal)
0
Hill's N1
1
Hill's N2
30
1980 1990 2000 2010 Year
1980 1990 2000 2010 Year
1980 1990 2000 2010 Year Figure 7.3.2.2. Trends in fifteen indicator metrics applied to the IBTS Q1 groundfish survey data
for the Kattegat and Skagerrak. See Table 7.3.2.1 for explanation of metrics (y axis labels).
Table 7.3.2.3. Summary of principal components analysis results for eight subregions of the North Sea.
SUB-REGION
FACTORS TOTAL
VARIANCE EXPLAINED
1 2 3 4 5 6
North‐
western shelf
Variance explained 20.2 24.6 16.9 7.3 9.5 19.3 97.9%
Metrics B
P
Variance explained 18.0 35.9 13.3 14.5 8.1 7.9 97.7%
Metrics B
P
Variance explained 17.9 36.7 15.0 7.4 13.1 6.8 96.8%
Metrics B
P
Amat Lmat L∞
LFI
Norwegian Deeps
Variance explained 13.3 36.2 15.5 15.2 16.0 96.2%
Metrics B
P
Variance explained 16.9 29.5 16.3 16.8 14.0 94.0%
Metrics B
P
Variance explained 20.7 19.4 18.5 17.7 8.6 12.3 97.0%
Metrics B
P
Eastern Variance explained 21.2 17.6 13.2 24.8 11.7 13.2 96.0%
SUB-REGION FACTORS TOTAL
central basin Metrics B P
Variance explained 20.2 24.6 16.9 19.3 7.3 9.5 97.9%
Metrics B
P
tively steady decline in the biomass of large fish was evident from 1983 through to 2001. But between 1983 and 1995, the biomass of small fish doubled, and it was this rapid expansion in the biomass of small fish, combined with the decline in large fish, that was responsible for the initial sharp drop in the proportion of large fish indicator at the start of the time‐series (see Figure 7.1). During the 1990s, much of the year‐to‐
year variation in the indicator was driven by variation in the biomass of small fish.
Marked peaks in small fish biomass were evident in 1993, 1996 and 2001, coinciding with especially low points in the indicator trend. From 2001, a decline in the biomass of small fish, combined with an increase in the biomass of large fish, particularly since 2006, has been responsible for the recovery in the proportion of large fish indi‐
cator.
One point of note is that variation in the biomass of small fish was heavily influenced by changes in the biomass of some species (e.g., whiting) that never (or rarely) grow to a length where they eventually influence variation in the biomass of large fish (Figure 7.4.1). Changes in the abundance of such species will always therefore repre‐
sent environmentally driven noise in proportion of large fish indicator.
A contoured surface demonstrating variation in the proportion of large fish indicator with varying combinations of biomass of fish both greater than 40cm in length and smaller or equal to 40cm in length was generated. The observed trajectory of the pro‐
portion of large fish indicator was then plotted on top of this, revealing three clear phases and allowing the process generating them to be identified (Figure 7.4.2). Dur‐
ing the “decline phase”, from 1983 to 1992, indicator values were driven down by both a decrease in the biomass of large fish and increasing small fish biomass. Be‐
tween 1992 and 2000, large fish biomass altered little, and variation in the indicator was influenced primarily by changes in the biomass of small fish. A further decline in the biomass of large fish combined with a substantial increase in small fish biomass drove the indicator down to its lowest point in 2001. The “recovery phase”, from 2001 to the current time, was characterized by declining small fish biomass combined, par‐
ticularly since 2006, with an increase in the abundance of large fish.
Given that the indicator is a ratio metric, it is clear that variation in the indicator value, as illustrated in the discussion above, is influenced by variation in the abun‐
dance of fish both larger and smaller than the 40 cm bound. Relating variation in the indicator value directly to variation in the biomass of fish both larger and smaller than 40 cm, however, confirmed its greater sensitivity to the former (Figure 7.4.3).
The relationship for the small fish component of the community was heavily reliant on the data collected in years when small fish biomass was high. Excluding the three