Page 34 - Lloreta_alii_2001

from Cadix (N = 6), Dar-El-Beida (N = 6), Cagliari (N = 6), and Palermo (N = 6); inset

gives a zoom of the p-values between 0 and 0.01. Median, first and third quartile, and

1.5 by inter-quartile range are plotted. The p-values of the regressions with NAO and

LOD appeared non-significant and widely distributed between 0 and 1, whereas those

with temperature were all close to a p-value = 0.

Figure 6. Box-plots of the p-values of the correlation analyses calculated between the

BFT and environmental time series on: (a) different time periods (i.e., same as Figure 5,

but for correlation) and (b) a common period (restricted to the years 1868-1960).

Table 1. Main characteristics of the trap catch time series: index, area (country or

region), name, period with continuous data ≥ 80 years, total number of observations of

the time series (N), median and coefficients of variation (C.V., as the standard

deviation*100 and scaled by the mean).

Table 2. Main characteristics of the environmental time series: type of variable, source

(instrumental or reconstructed), period and bibliographic reference.

Table 3. Results from the analyses between environmental and BFT time series:

number of observations used in the regressions and correlations (N), slope and p-value

of the linear regressions and generalised least squares models (GLS) and uncorrected

and Holm’s corrected probabilities of the Spearman correlation coefficient. Bold values

indicate significant probabilities at the 5% level. When the trap catches time series was

discontinuous, GLS model was fitted on the longest period available.

35