Size metrics. Choice of size metrics was based on a study in 78 German lakes (Xxxxxxx et al., 2011, see part 1 of this report). For all European lakes included, we calculated geometric mean length (cm), variance of the length data, skewness and kurtosis of the length-frequency distribution, number of size classes, maximum length (cm), interquartile range (cm), length at the 95th percentile (cm) and size diversity. Size diversity (Xxxxxxxx et al., 2008) combines several aspects of other size metrics into a single comparable value and was tested to be sensitive to a broad range of lake descriptors at the regional scale (Xxxxxxx et al., 2011). For fish- length data, size diversity is high when the catch consists of many different size classes (single size classes can still dominate the catch) or if the abundances between the size classes are relatively equal (Xxxxxxx et al., 2011). For details of all other size metrics see Xxxxxxx et al. (2011). Size metrics related to fish weight could not be calculated for the European dataset, because of missing FW data in 61% of the lakes. Additionally, we did not include normalised length spectra, as for almost one third of the lakes the R2 of the linear regression models was < 0.5.
Size metrics. Many size metrics were strongly correlated and highly variable (Table 7) and BRT analyses indicated high sensitivity of skewness, kurtosis, number of size classes and interquartile range towards sampling effort (relative influence of number of nets > 10%; Table 8; Fig.13). Consequently, we selected only three weakly correlated (Xxxxxxxx’x r <0.34) metrics for further analyses that were obviously less sensitive to sampling effort, namely geometric mean length (GM), maximum length (Lmax) and size diversity (SD). Variability in these three metrics was substantial between the 728 lakes. Average geometric mean length of all fish assemblages was 14.1 cm (±2.6; 9.6-32.0 cm) (standard deviation; range). Maximum length averaged 54.4 cm (±15.9; 19-120 cm) and average size diversity was 1.8 (± 0.3; 0.4-2.7) (Table 7). Table 7: Correlation matrix of the nine size metrics. The white matrix contains Xxxxxxxx’x r values; the grey matrix contains the corresponding P-values. Furthermore minimum, mean and maximum values of the size metrics are given. GM VAR SKEW XXXX NSC IQR Lmax L95 SD GM --- 0.308 -0.392 -0.451 0.024 0.465 -0.043 0.566 0.169 VAR <0.001 --- -0.475 -0.311 0.537 0.576 0.525 0.786 0.785 SKEW <0.001 <0.001 --- 0.936 -0.704 -0.678 -0.335 -0.590 -0.664 XXXX <0.001 <0.001 <0.001 --- -0.508 -0.673 -0.252 -0.434 -0.558 NSC 0.514 <0.001 <0.001 <0.001 --- 0.324 0.549 0.473 0.534 IQR <0.001 <0.001 <0.001 <0.001 <0.001 --- 0.165 0.606 0.778 Lmax 0.249 <0.001 <0.001 <0.001 <0.001 <0.001 --- 0.193 0.341 L95 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 --- 0.577 SD <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 --- Minimum 9.6 3.7 -0.3 -1.8 8.0 1.0 19.0 12.0 0.4 Mean 13.6 30.9 1.0 -0.3 27.5 6.0 53.0 23.0 1.8 Maximum 32.0 198.0 3.8 15.1 54.0 22.0 120.0 50.0 2.7 GM = geometric mean length; VAR = variance; SKEW = skewness of the length-frequency distribution (LFD), XXXX = kurtosis of the LFD; NSC = number of size classes; IQR = interquartile range; Lmax = maximum length; L95 = length at the 95th percentile of the LFD; SD = size diversity. Table 8: Contribution of predictors (relative influence in %) of the boosted regression tree analyses to the nine size metrics ranked in order of decreasing median contribution. For full names of the size metrics see Table 7. Tmax = maximum temperature; Tamp = amplitude temperature; Xxxx = average temperature; CLC = Xxxxxx Land Cover. Predictor GM VAR SKEW XXXX NSC IQR Lmax L95 SD Median Latitude 15.51 13.74 7.86 5.65 12.98 7.2...
