Sanchez-Rubio and Perry: Meteorological and hydrological regimes and their influence on recruitment of Brevoortta patronus 
397 
line between Mississippi and Alabama to as far west 
as northcentral Texas. Anderson (2007) used mitochon- 
drial DNA to denote a single population in the west- 
ern GOM and more recent data indicate a single unit 
stock throughout the northern GOM (Vanderkooy and 
Smith 4 ). 
Fishery-dependent data from landings of the reduc- 
tion purse-seine fishery for Gulf menhaden by age dur- 
ing the period 1964-2010 were obtained from Smith 
and Vaughan. 12 Descriptions of the fishery-dependent 
data and the method used to generate that time series 
are provided in Vaughan et al. (2007). Total landings of 
Gulf menhaden, numbers and proportions of fish by age 
per vessel-ton-week (vtw, defined as the net tonnage of 
a vessel multiplied by the number of weeks that vessel 
unloaded fish at least one day (Smith, 1991), were used 
in this study. 
Analyses 
Annual meteorological data (air temperature and 
north-south and east-west wind stress), hydrological 
data (precipitation, PDSI, river flow, sea level, SST, 
and Mississippi River N:P ratio) and biological data 
(fishery-independent age-0 abundance of Gulf menha- 
den in the central region and of menhaden species in 
western region; fishery-dependent numbers and propor- 
tions of Gulf menhaden in commercial landings) were 
compared among decadal regimes associated with the 
couplings of AMO cold and NAO negative (average re- 
gime: 1964-1970), AMO cold and NAO positive (wet 
regime: 1971-1994), and AMO warm and NAO nega- 
tive (dry regime: 1995-2010) phases and among inter- 
annual regimes associated with ENSO warm events 
(wet years: 1963, 1965, 1968-1969, 1972, 1976-1977, 
1982, 1986-1987, 1991-1994, 1997, 2002, 2004, 2006, 
2009) , neutral events (average years: 1962, 1966-1967, 
1971, 1978-1981, 1985, 1989-1990, 1995-1996, 2001, 
2003), and cold events (dry years: 1964, 1970, 1973- 
1975, 1983-1984, 1988, 1998-2000, 2005, 2007-2008, 
2010) . Nonparametric rank-sum tests (Kruskal-Wallis 
H- test, x 2 statistic; Mann- Whitney U-test, Z statistic) 
were performed with SPSS Statistics, vers. 20.0, on the 
meteorological, hydrological, and biological responses 
imposed by the coupling of AMO and NAO phases and 
by ENSO events. To adjust the P-values for multiple 
comparisons, the alpha level of each individual test 
was adjusted downward by using the Bonferroni cor- 
rection method. Linear associations among abiotic vari- 
ables were checked with the variance inflation factor 
to quantify the severity of collinearity. This test was 
performed with linear regression analysis in SPSS, 
vers. 20.0. To linearize relationships and approximate 
normality, annual river flows from the western region 
were logarithmically transformed and annual north- 
12 Smith, J. W., and D. S. Vaughan. 2011. Harvest, effort, 
and catch-at-age for Gulf menhaden. Southeast Data, As- 
sessment, and Review SEDAR 27-DW05, 28 p. [Available 
at website.] 
south wind values were cube-root-transformed. All 
abiotic variables were standardized to a mean of 0 and 
variance of 1 by subtracting the mean and dividing by 
the standard deviation. 
Two multivariate techniques were used to determine 
the relationships between abundance of menhaden and 
climate-related meteorological and hydrological vari- 
ables. Principal component analysis (PCA) was used 
on the correlation matrix of abiotic variables to reduce 
data. The PCA transformed the original set of variables 
into a smaller set of orthogonal linear combinations of 
abiotic variables that account for a major portion of 
the variance in the original set (Chatfield and Collins, 
1980; Dillon and Goldstein, 1984). A high number of 
abiotic variables were reduced to few principal compo- 
nents by using the factor procedure in SPSS Statistics, 
vers. 20.0. The scree plot was examined to determine 
where the curve started to flatten between principal 
components, and only the components with eigenval- 
ues higher than 1 were retained for interpretation. The 
component scores that were retained were correlated 
against annual abundances of age-0 Gulf menhaden 
in BPL and seine hauls in the central region, annual 
abundances of age-0 menhaden species in seine hauls 
in the western region, and annual landings (numbers 
and proportions) of Gulf menhaden by age from off- 
shore waters of the area studied in the northern GOM. 
Because abundance determined with BPL surveys 
in the central region was not linear, those annual val- 
ues were cube-root-transformed. Covariability between 
principal components and biotic variables was exam- 
ined by performing a Pearson correlation analysis in 
SPSS Statistics, vers. 20.0. To adjust the P-values for 
multiple correlation analyses, the alpha level of each 
individual test was adjusted downward with the Bon- 
ferroni correction method. In the case of significant 
correlations, it was assumed that the variables with 
larger eigenvectors (coefficients of structure or correla- 
tions) for the axis were the ones that most influenced 
the abundance of Gulf menhaden. Further covariabil- 
ity analysis was performed between the actual climate- 
related meteorological and hydrological variables and 
data on fishery-independent catches and fishery-depen- 
dent landings of menhaden species. The alpha level of 
each individual test was also adjusted downward with 
the Bonferroni correction method. 
To identify models of climate-related meteorologi- 
cal and hydrological parameters that contributed to 
the variability in recruitment of menhaden from Ala- 
bama through Texas, the retained component scores 
were used as predictors of fishery-independent catches 
(abundances of age-0 Gulf menhaden per BPL or seine 
haul in the central region and abundances of age-0 
menhaden species per seine haul in the western re- 
gion) and fishery-dependent landings (numbers and 
proportions of age-1 Gulf menhaden per vtw) in an 
automatic linear modeling procedure in SPSS Statis- 
tics, vers. 20.0. A predictive model was developed by 
regressing fishery-independent recruits to age 0 (Po,t) 
and fishery-dependent recruits to age 1 (Pi ,t+l) on 
