Seasonal and Geographical Differences in Total and Pathogenic Vibrio parahaemolyticus and Vibrio vulnificus Levels in Seawater and Oysters from the Delaware and Chesapeake Bays Determined Using Several Methods

The first step in our analysis was to examine simple correlation between the methods. The COPP assay for TV correlated moderately well with total V. vulnificus and V. parahaemolyticus as determined by both direct plating (DP) and MPN-PCR in water (r values = 0.50 to 0.69) (Table 1). In oysters, similar relationships were observed for total V. vulnificus (r = 0.56 to 0.62), but poor correlation was noted for total V. parahaemolyticus (r = 0.23 to 0.34). Correlations with virulence markers tdh, trh, and vcgC were all <0.45. Notably, the relationships between MPN-PCR and the COPP assay for tdh and trh in water were 0.32 and 0.38, respectively. Overall, there is a general trend of slightly stronger correlation of the COPP assay with direct plating than with MPN-PCR (Table 1).

DP and MPN-PCR showed variable relationships depending on the Vibrio species and the gene marker evaluated. Correlation was stronger for total V. parahaemolyticus (r = 0.66 and 0.67 in water and oysters, respectively) than for total V. vulnificus (r = 0.42 and 0.59, respectively) (Table 1). For the virulence markers, correlation was similar among all gene targets and in both matrices (oysters and seawater), ranging from 0.31 to 0.54.

The second step in our evaluation of the two methods was to further examine the performance of DP versus MPN-PCR to offer an explanation of the moderate to poor correlations observed. Sensitivity, or the measure of the proportion of instances when both methods provided positive results, was highest in oysters for total (tlh⁺) V. parahaemolyticus (87%) and total (vvhA⁺) V. vulnificus (85%) and was consistently higher in oysters (58%) than in water (37%) across all gene targets (Table 2). Generally poor results were obtained for all virulence markers in both matrices (19 to 51%). Again, with the exception of total V. parahaemolyticus and total V. vulnificus in oysters (13 and 15%, respectively), DP failed to detect all other Vibrio targets 42 to 81% of the time. Thus, for the remainder of this study, only MPN-PCR results were used for COPP assay model development, Vibrio species and strain enumeration, and environmental comparisons.

The final step for method development was the examination of the potential utility of the COPP assay to serve as an “indicator” of elevated abundances of the two Vibrio species and virulence markers that were evaluated. For the purpose of this initial evaluation, “elevated” was defined as the top 25%, or quantile, of the data for each species/marker, and COPP was examined alone and in combination with measurements of temperature and salinity at the depth of sample collection (bottom for oyster and surface for water). The results are presented in Table 3 and are ordered by the Akaike information criterion corrected for small sample size (AICc) (21), as described in Materials and Methods, with AICc values shown from closest to the “truth” for this candidate set of predictors to furthest from the truth. AICc weights, as shown in Table 3, provide the likelihood that a given model is the best, and concordance represents the classification error or, in this case, the proportion of times the model correctly classifies the dependent variable as elevated.

For oysters, the COPP assay alone never provided the best model (concordance = 56.9 to 81.5%), and with the noted exception of V. vulnificus (81.5% concordance), it generally provided moderate to low predictive power on its own (Table 3). However, in combination with bottom water temperature and salinity, all final models exceeded 73% concordance and were generally more than twice as likely to be the best model as competing models. The noted exceptions are models for tdh and trh, for which the addition of the COPP assay afforded little improvement over temperature and salinity alone (<1% concordance, ΔAICc < 2). However, there is little downside to incorporating COPP assay results, as these models demonstrated high concordances of 85.7 and 90.9%, respectively, for oysters.

Water models generally showed higher concordance than oysters, with all over 81% (Table 3). COPP alone was never the best-performing model, but for V. parahaemolyticus (tlh, tdh, and trh) it demonstrated potentially useful concordances of 91.3, 81.6, and 88.0%, respectively. In all cases, the top-performing model was at least four times more likely to be the best model over the other candidates. In two instances, individual models failed the Hosmer-Lemeshow (HL) goodness-of-fit test (Table 3), where HL (P) was ≤0.05; however, neither model was selected as the top performer, and these results are included only for reference.

The second goal of this study was to compare Vibrio dynamics between the estuaries in both water and oysters. To begin, we first characterized the physicochemical environment to determine the major gradients that exist between the two systems. Overall, 64.4% of the environmental variation in surface water characteristics could be explained by the first two principal components (PCs) (Fig. 1A), with PC1 (38.9%) representing a salinity/turbidity gradient and PC2 (25.5%) a temperature/oxygen/pH gradient. PC1 clearly separates the states, with Delaware sites being more turbid and having a greater salt content than those sampled in Maryland. Nearly identical results were obtained for bottom waters, with PC1 (salinity/turbidity) explaining 44.1% of the overall variation and PC2 (temperature/oxygen/pH) contributing 27.6% (71.7% overall) (Fig. 1B).

Surface and bottom water temperature and pH did not differ significantly between the states; however, all other water quality parameters measured did (P < 0.05) (Fig. 2). Dissolved oxygen was low during the summer and relatively high in the winter, with values ranging from 2.8 mg/liter (July) to 10.1 mg/liter (November). The annual mean dissolved oxygen values were 7.23 and 5.02 mg/liter, respectively, in Maryland and Delaware surface waters and 7.20 and 4.90 mg/liter, respectively, in bottom waters. The chlorophyll a values for water at both Maryland and Delaware sites ranged from 0.1 to 41.1 μg/liter; they were significantly higher in Maryland as noted from the principal-component analysis (PCA). The most striking differences separating the bays are in salinity and turbidity. Delaware sites were over 10 ppt saltier than Maryland sites on average and over four times more turbid.

In our initial analysis of environmental conditions in the two bays, a clear separation of Maryland and Delaware was noted along a salinity/turbidity gradient (PC1) (Fig. 1A and B). This difference in environmental conditions is confounded with the designation of “state” as a predictor variable, precluding determination of the variability in abundance associated with both. Indeed, evaluation of models including both PC1 and state as independent variables for multicollinearity demonstrated variance inflation in exceedance of an acceptable range. Therefore, the analysis presented below uses only PC2 as a covariate, and differences highlighted between the states may be due to either differences in turbidity/salinity or other factors unique to the water bodies—factors that were not measured.

All oyster samples were positive for total Vibrionaceae (TV) in both states, with concentrations typically ranging from 2.78 to 5.18 log CFU/g (Table 4; Fig. 3). State, site, and PC2 were significant (P < 0.05), with the effect size of PC2 (ω² = −0.14) and site (ω² = −0.13) greater than that of state (ω² = −0.02). Overall, Maryland oysters were marginally higher in TV than Delaware oysters (4.33 CFU/g versus 3.95 CFU/g, respectively; P = 0.04) (Fig. 3C), with the Slaughter Beach and Lewes sites generally having lower concentrations of TV in oysters than the other sites (Fig. 3). In comparison to physicochemical variables, TV in oysters correlated somewhat with temperature (r = 0.49) but not with any other factors (Table 5).

The vast majority of water samples were also positive for TV, although generally at concentrations more than a log lower than found in oysters (Table 4). The largest effect size was attributed to state (ω² = 0.56), with PC2 (ω² = 0.17) and site (ω² = 0.04) also contributing significantly (P < 0.05). In contrast to the case for oysters, TV in Delaware waters averaged over a log greater than in Maryland (2.90 and 1.71 log CFU/ml, respectively) (Fig. 3D). Correlation of TV with PC1 was very strong (r = 0.72, P < 0.0001), suggesting that state level effects may largely be associated with higher salinity and turbidity in Delaware. Further, TV correlated to an extent with salinity of the seawater (r = 0.46) and moderately with turbidity (r = 0.63), while a strong negative correlation with dissolved oxygen was observed (r = −0.70) (Table 5).

Frequencies of detection of V. parahaemolyticus in oysters as determined by tlh-positive samples were similar between the states, occurring in all Delaware samples and 92.4% of Maryland oysters (Table 4). Vibrio parahaemolyticus concentrations in oysters were marginally higher in Delaware than in Maryland (2.59 and 1.97 log MPN/g, respectively) (Fig. 3). The largest effect size was attributed to site (ω² = 0.20) and year (ω² = 0.09), with state (ω² = 0.05) and PC2 (ω²=0.03) also contributing to a lesser extent (P < 0.05). Site differences were largely attributable to consistently low levels in oysters from the Chester River, MD, compared to the other sites (Fig. 4). None of the water’s physicochemical factors correlated particularly well with V. parahaemolyticus in oysters (Table 5).

In water, V. parahaemolyticus was detected 100% of the time in Delaware and slightly less frequently (88.6%) in Maryland (P = 0.11, Fisher’s exact test). Difference among the states (1.86 versus 0.29 log MPN/g in Delaware and Maryland, respectively) was the only component of those examined that explained a significant proportion of variance (ω² = 0.63) (Fig. 4). Similar to the case for TV in water, correlation with PC1 was very strong (r = 0.75, P < 0.0001), suggesting that state differences may, in fact, be due to the salinity/turbidity gradient. Vibrio parahaemolyticus in the water correlated moderately with salinity and turbidity (r = 0.68 and 0.69, respectively) and correlated negatively with dissolved oxygen (r = −0.59) (Table 5).

A high detection frequency was also observed for total V. vulnificus in oysters (vvhA⁺), with 100% occurrence in both states (Table 4). Year, state, and PC2 were all significant (P < 0.05), with the largest effect size attributed to year (ω² = 0.16) and PC2 (ω² = 0.16). The Vibrio vulnificus concentration was higher in 2017 (3.39 log MPN/g) than in 2016 (2.49 log MPN/g) in both states. The average V. vulnificus concentration was nearly 1 log higher in Maryland samples than in Delaware samples (3.19 versus 2.31 log MPN/g, respectively) (Fig. 5).

Vibrio vulnificus in water responded similarly in terms of annual differences (2017 > 2016, P < 0.05) but was marginally higher in Delaware than Maryland (P < 0.05) (Fig. 5D). Overall, mean concentrations were 1 to 2 logs lower than those observed in oysters (1.57 and 1.14 log MPN/g in Delaware and Maryland, respectively). PC2 (ω² = 0.18) demonstrated the largest effect size, reflecting increasing abundance with temperature. The frequency of detection was 100% in Delaware and 98.7% in Maryland (Table 4). Vibrio vulnificus counts in seawater and oysters correlated somewhat with water temperature (r = 0.41 and 0.48, respectively) but poorly with all other variables (Table 5).

While overall concentrations of V. vulnificus were higher in Maryland oysters than in Delaware oysters (3.19 versus 2.31 log MPN/g), the prevalence (92.4 versus 93.5%) and concentration (1.06 versus 1.07 log MPN/g) of vcgC⁺ strains were nearly identical (Table 4). Of the candidate predictors, PC2 and site explained most of the variance, but the overall model was not significant (P > 0.05).

Potentially pathogenic (tdh⁺) V. parahaemolyticus strains were found in 96.8% of Delaware oyster samples, compared to 48.1% of Maryland oysters (P < 0.0001, Fisher’s exact test) (Table 4). State demonstrated the largest effect size (ω² = 0.39), with site also contributing (ω² = 0.06) (P < 0.05). Concentrations were low in oysters from both states, averaging 0.94 and 0.15 log MPN/g in Delaware and Maryland, respectively (Table 4). High correlation was noted for PC1 (r = 0.62, P < 0.0001) and specifically salinity (r = 0.69, P < 0.0001), suggesting that state differences may be attributable to environmental differences.

Potentially pathogenic V. parahaemolyticus (trh⁺ strains) was detected in 93.5% of Delaware oyster samples, compared to only 22.8% of Maryland oysters (P < 0.0001, Fisher’s exact test) (Table 4). In Delaware oysters, 93.5% had both pathogenicity markers (tdh and trh), compared to just 19.2% in Maryland (P < 0.0001, Fisher’s exact test). The same trend as tdh is apparent for trh in oysters, with low concentrations of the strains overall but significantly higher in Delaware than in Maryland (0.95 ± 0.07 and 0.07 ± 0.04 log MPN/g, respectively; P < 0.0001, ω² = 0.49) (Table 4). No other factor explained a significant portion of variance. As with tdh, trh correlated well with PC1 (r = 0.63, P < 0.0001) and salinity (r = 0.69, P < 0.0001) (Table 6), suggesting that higher salinity in Delaware may account for the observed differences.

Frequencies of detection of virulence genes in water samples followed a trend similar to that in oysters, with 90.3 and 79.7% vcgC positive in Delaware and Maryland, respectively (P < 0 0.05, Fisher’s exact test) (Table 4). State (ω² = 0.25), year (ω² = 0.04), and PC2 (ω² = 0.06) were all significant (P < 0.05), with Delaware having higher average concentrations than Maryland (0.39 versus 0.09 log MPN/g) (Table 4) and overall concentrations being slightly higher in 2017. A moderate correlation with turbidity was observed (r = 0.45, P < 0.05) (Table 6).

For V. parahaemolyticus in water, tdh⁺ and trh⁺ strains were present in a higher proportion of samples in Delaware (87.1% tdh⁺, 83.9% trh⁺) than in Maryland (13.9% tdh⁺, 3.8% trh⁺) (P < 0.0001, Fisher’s exact test) (Table 4). Both tdh⁺ and trh⁺ strains were more prevalent in 2017 than in 2016 (P < 0.05, Fisher’s exact test). As with oysters, state carried the largest effect size for both markers (ω = 0.20 and 0.31 for tdh and trh, respectively), with moderate correlations with PC1 and salinity (r = 0.40 to 0.51, P < 0.05). A weak year effect was also notable with tdh (ω² = 0.03), being greater in 2017 than in 2016 (P = 0.04).

Study sites and locations were selected to represent the range of salinities where Vibrio species may be present. The sites in Delaware represent the higher salinity range (generally 21 to 30 ppt), and the sites in Maryland represent the lower salinity range (10 to 18 ppt). For this project, three Delaware Bay study sites (Lewes, Bowers, and Slaughter Beach) in Delaware and five Chesapeake Bay sites (Broad Creek, Chester River, Manokin River, Oxford, and Tangier Sound) in Maryland were selected (Fig. 6) based on salinities, accessibility, and availability of oysters. Salinities at selected sites were representative of oyster-growing areas nationally.

The Delaware Bay is located on the east coast of the United States and is bordered by the states of Delaware and New Jersey. It is 84 km (52 miles) long and covers an area of approximately 2,025 km² (782 miles²). The shoreline is composed mainly of salt marsh lowlands. Salinity is highest at the mouth of the bay where it meets the Atlantic Ocean. The middle of the bay is brackish, and the northern part of the bay is freshwater with salinity of less than 1 ppt (35).

The Chesapeake Bay is located on the east coast of the United States within the states of Maryland and Virginia. The Chesapeake Bay is the largest estuary in North America. It is 314 km (195 miles) long and between 6 and 48 km (4 to 30 miles) wide, covering an area of 8,384 km² (3,237 miles²) with 4,470 km² (1,726 miles²) in Maryland. Salinities increase from north to south as one moves downstream from Maryland’s portion of the bay into Virginia, where the Chesapeake Bay meets the Atlantic Ocean.

Oyster and water samples were collected from three study sites in Delaware (monthly) and five sites in Maryland (every 3 weeks) from May through October for 2 years (2016 and 2017). The five sampling sites in Maryland are classified as approved by the National Shellfish Sanitation Program (NSSP) for commercial harvesting of oysters. Twelve market-size oysters and one container (1 liter) of seawater were collected from each study site, yielding a total of 110 oyster and 110 water samples. These samples were then divided into three subsamples (a total of 330 oysters and 330 water samples) for analysis of TV by the COPP assay and of TV and pathogenic Vibrio species by microbiological and molecular methods. During the collection of samples, temperature, salinity, turbidity, dissolved oxygen, chlorophyll a, and pH in the surface and bottom water were measured with a YSI 6600 multiparameter meter (Yellow Springs Instrument Co., Yellow Springs, OH). The depth of the oyster-harvesting areas in Maryland and Delaware ranged from 2.4 to 5.5 m and 0.25 to 1.65 m, respectively. Water samples were collected following the method of the American Public Health Association (36) in sterile 1-liter wide-mouth containers (Thermo Fisher Scientific, Waltham, MA). Immediately after harvest, oysters were bagged and chilled in an insulated chest with ice, and a sheet of bubble wrap was used to prevent direct contact of samples with the ice. The shipping temperature was monitored using a data logger (Dickson, Addison, IL) to verify that the temperature was less than 10°C (minimum temperature for Vibrio growth) during transport. The oyster and water samples were transported to the laboratories at the University of Maryland Eastern Shore and Delaware State University and analyzed within 24 h of collection (4, 37).

Total Vibrionaceae counts were determined according to the procedures described by Richards et al. (15). The COPP assay detects a lysyl-aminopeptidase produced by Vibrionaceae family members. In brief, the 12 oysters from each sample were subdivided into 3 sets of 4 oysters. Each set was blended, and 25 g was removed and homogenized with 225 ml of 0.1% peptone buffer (Thermo Fisher Scientific, Waltham, MA). The remaining homogenate was used for the MPN. Serial dilutions of each homogenate (to a final dilution of 10⁻⁶ to 10⁻⁹ depending upon the expected Vibrio counts at the time of sampling) and water sample (to a final dilution of 10⁻⁴) were prepared in phosphate-buffered saline (PBS) or 0.1% peptone buffer. One hundred microliters of each dilution, as well as 100 μl of undiluted water samples, was spread plated in duplicate onto tryptic soy agar (Becton, Dickinson, Sparks, MD) containing 1% NaCl (TSA-N) and incubated overnight at 37°C. After incubation, one countable plate (preferably with 20 to 200 colonies) from each subsample was subjected to the COPP assay by overlaying the colonies with a cellulose acetate membrane (Sartorius, Edgewood, NY) containing the fluorogenic substrate l-lysyl-7-amino-4-trifluoromethylcoumarin (MP Biomedicals, Solon, OH) (15). After precisely 10 min, the membrane was carefully removed from the plate and observed under long-wave UV light for fluorescence. The number of fluorescent foci, representing colonies of Vibrionaceae, was enumerated. Total Vibrionaceae organisms per gram of oyster and per milliliter of seawater were calculated.

For direct plating (DP), the same serial dilutions that were prepared for the enumeration of total Vibrionaceae (up to 10⁻⁶ to 10⁻⁹), as described in the previous section, were plated onto CHROMagar Vibrio (CHROMagar, Paris, France). One hundred microliters of each dilution was spread plated, and the plates were incubated at 37°C for 24 h. After incubation, presumptive colonies of V. parahaemolyticus (mauve) and V. vulnificus (green) were counted, and all or 10% of colonies (depending on the number of colonies on each plate) were confirmed using real-time PCR (38, 39). In brief, isolates were grown overnight in tryptic soy broth (Thermo Fisher Scientific, Waltham, MA) with 1% NaCl and then were boiled for 10 min. The aqueous DNA from cellular materials was separated by centrifugation at 10,000 × g for 2 min. Supernatants were used as PCR templates and were stored at –80°C until they were tested. All presumptive V. parahaemolyticus isolates were tested for tlh (total) and tdh and trh (pathogenic) hemolysin genes, while all presumptive V. vulnificus isolates were tested for vvhA (total) and vcgC (pathogenic) genes. Probes, primers, and cycling conditions for all genes are described in the section below. The presumptive colonies were then multiplied by the percentage of isolates that were confirmed by real-time PCR to be either V. parahaemolyticus or V. vulnificus, and the data were converted to CFU/g or CFU/ml of sample.

Three-tube most-probable-number (MPN) procedures were performed in alkaline peptone water (APW), and multiplex real-time PCR assays were used for enumeration of total (tlh) and pathogenic (tdh and trh) V. parahaemolyticus and total (vvhA) and pathogenic (vcgC) V. vulnificus in oyster and water samples (38, 39). In brief, oyster homogenates from the same sample, consisting of three subsets of four oysters each, prepared for the COPP assay and direct plating as mentioned in the previous sections were combined and reblended for 30 s. This homogenate (12 oysters) was serially diluted 10-fold to 10⁻⁶ to 10⁻⁹ (depending upon the expected Vibrio counts at the time of sampling) in 0.1% peptone. One gram of homogenized tissue and 1 ml of each dilution were inoculated into triplicate MPN tubes containing sterile APW consisting of 1% peptone and 1% NaCl, pH 8.5. Incubation was at 37°C for 18 to 24 h. PCR was performed on MPN tubes showing signs of microbial growth (increased turbidity) for tlh⁺, tdh⁺, and trh⁺ V. parahaemolyticus using PCR primers and probe sequences and PCR amplification conditions described in a previous study (39). An internal amplification control (IAC) was used to ensure PCR integrity to detect any false-negative samples (39). The same MPN tubes were screened for the vvhA and vcgC genes of V. vulnificus by PCR. Probes, primers, and cycling conditions for these genes were described by Baker-Austin et al. (38), except that the initial denaturation time was changed from 10 to 3 min. All primers and probes were manufactured by Thermo Fisher Scientific, except the IAC probe, which was synthesized by Integrated DNA Technologies (Skokie, IL). PCR was performed using iTaq Universal Probes Supermix (Bio-Rad Laboratories, Hercules, CA) containing iTaq DNA polymerase, deoxynucleoside triphosphates (dNTPs), and MgCl₂ according to the manufacturer’s instructions in 25-μl reaction volumes. All PCR was performed on an Applied Biosystems (ABI) cycler (7500 RT PCR system; Thermo Fisher Scientific). Positive and negative controls were used for each PCR cycle. After completion of the MPN–real-time PCR, the number of MPN-positive tubes that showed amplification of virulence genes was recorded, and an MPN index was generated using standard U.S. Food and Drug Administration (FDA) procedures and tables (https://www.fda.gov/food/foodscienceresearch/laboratorymethods/ucm109656.htm).

Pearson's correlation (Proc Corr, SAS v 9.4; SAS Institute, Cary, NC) was employed to examine the potential association of the COPP total Vibrio (TV) assay with the more specific methods for V. parahaemolyticus and V. vulnificus. Correlations were considered weak when the r values were between 0.40 and 0.49, moderate when r values were between 0.50 and 0.69, and high when r values were ≥0.70. All data were transformed using log(CFU + 1) for statistical analysis, and an alpha level of 0.05 was considered the minimum level for statistical significance.

To compare DP on CHROMagar Vibrio with MPN-PCR, the sensitivity (true positive rate), specificity (true negative rate), and accuracy (method agreement) of DP were calculated with standard formulae, and MPN-PCR was used as the standard method (24, 40). DP-negative and MPN-PCR-positive results were interpreted as DP false negative, whereas MPN-PCR-negative and DP-positive results were considered DP false positive. Samples that were positive and negative for total and pathogenic V. parahaemolyticus and V. vulnificus by both methods were considered true positive and true negative, respectively (24).

For the statistical methods component of this project, we used an information-theoretic approach (21) for model selection (i.e., Akaike information criterion corrected for small sample size [AICc]) to determine the potential for using the COPP TV assay and known environmental determinants of Vibrio growth to predict elevated occurrence of the species and genetic markers evaluated in this study. Models were constructed for 10 outcome variables (dependent variables): total V. parahaemolyticus (tlh), total V. vulnificus (vvhA), and pathogenicity markers vcgC, trh, and tdh in both water and oyster tissue as determined by the MPN-PCR method. The predictor variables (independent variables) for the candidate models were selected a priori and are combinations of the COPP assay and surface (water) and bottom (oyster) water temperature and salinity. Two categories were created to represent the top quantile of the data distribution (>75th quantile) and the remainder (<75th quantile). Logistic regression was used in all cases to model the categorical data using both Proc Logistic and Proc Genmod (logit link) to obtain the range of diagnostic statistics reported (SAS Institute, Cary, NC).

Under the information-theoretic approach, a candidate set of models is fit and then compared using a model selection criterion. The best model is determined by examining their relative distance to the “truth.” We used AICc weight as a measure of relative plausibility of the models within the candidate set. We interpret the AICc weight (w_i) of evidence that model i is the best approximating model, given the data and set of candidate models (21). While model selection was done based on AICc weight, tests of concordance (how often the model correctly categorized the dependent variable) and the Hosmer-Lemeshow goodness-of-fit test were also evaluated (α > 0.05 was considered an adequate fit).

For state comparisons of Vibrio abundance, principal-component Analysis (PCA) was first employed to characterize the physicochemical gradients that exist within and between Delaware and Maryland. This was done to illuminate major differences in water quality parameters that could explain variability in Vibrio levels between the states and to consolidate the number of independent variables. Principal component (PC1) and PC2 were evaluated in subsequent multivariate analysis of variance (ANOVA) (Proc GLM; SAS Institute, Cary, NC). Models were developed for each species and gene target with the explanatory variables state, year, site nested within state, and PC1 and PC2 as covariates, for the COPP TV assay and MPN-PCR method only. All values were log transformed prior to analysis, with model residuals examined to ensure no violations of normality and homogeneity of variance assumptions. Variance inflation was also examined for each model to assess multicollinearity. Tukey’s test was used for within-class comparisons in all cases. While P values are provided, we focused primarily on effect size of significant components as a means of understanding the relative contribution of each explanatory variable and in recognition of their increased use in other fields of science (41). Partial omega squared (ω²) was employed in all cases, following the criteria of Field (42), with 0.01 considered a small effect, 0.06 a medium effect, and ≥0.14 a large effect. Partial omega squared is preferred for multifactorial analyses due to the calculation of effect size relative to error variance rather than total variance. The latter diminishes the effect size of all factors as each new factor is entered (43). Percent data explored for comparison of detection were examined using frequency tables and Fisher’s exact test to account for small sample size (Proc Freq; SAS Institute, Cary, NC).