Morphometric analysis for nonlethal sex determination in brook trout: a new tool for research and management. A Final Independent Project Report

Morphometric analysis for nonlethal sex determination in brook trout: a new tool for research and management A Final Independent Project Report by Amanda E. Holloway In partial fulfillment of the requirements for the degree of Master of Science in Environmental Science and Policy (Concentration in Ecological Management) The Johns Hopkins University Spring 2012 Project Mentor Robert Hilderbrand Associate Professor, University of Maryland Center for Environmental Science Appalachian Lab, Frostburg, MD Introduction Brook trout (Salvelinus fontinalis) are an iconic species that has been considered recreationally and aesthetically important throughout its native range for centuries. Brightly-colored and charismatic, brook trout are also ecologically important - they are recognized as a bioindicators for water temperature and quality in the southern regions of their historical range (Waco and Taylor 2009). Brook trout require water that is

relatively cold (seldom exceeding 25 o C) and well-oxygenated. As such this species typically inhabits streams that are surrounded by forests. Brook trout face many threats and have already been extirpated from many of their native streams in Maryland (Stranko et al. 2008). The only native trout species found in Maryland, they have decreased from a population of millions to a few hundred thousand (MD DNR 2005). Habitat threats include water temperature increases due to climate change, land use changes, run-off (urban, agricultural and mining) and habitat fragmentation (Heft 2006, Letcher et al 2007). Brook trout also face competition and predation by stocking of non-native brown trout (Salmo trutta) and rainbow trout (Oncorhynchus mykiss). The western Appalachian region, with its cool streams in less disturbed, mostly forested watersheds, contains most of the state s extant brook trout. Brook trout can be found in other streams throughout Maryland, but those populations are considered greatly reduced (Hudy et al. 2005). However, even strong populations are in danger of decline. As a species of "Greatest Conservation Need' in Maryland (MD DNR 2005), conservation efforts and management plans for the species have been instituted, including fishing restrictions, conservation assessments, and life history research. Sexspecific life history differences are not known, but are potentially important to effective management. However, no reliable non-lethal methods for sexing brook trout have been established outside of the spawning season when adults are full of eggs or milt. A nonlethal approach to distinguishing male and female brook trout is essential to reduce stress on the threatened native populations of brook trout while obtaining important life

history information. Anecdotally, there are distinctive physical differences between male and female brook trout within populations, though this has not been quantified to-date. Sexual dimorphism has been studied extensively in fish and, typically, becomes more pronounced with size in sexually mature fishes (Beacham and Murray 1986). In species such as threespine sticklebacks (Gasterosteus aculeatus) and Mediterranean blennies (Blenniidae sp.), body size has shown significant relationship to dimorphism (Cooper et al 2011, Lengkeek et al. 2008). The relationship between head size and sex was examined in threespine sticklebacks (Gasterosteus aculeatus), using a geometric morphometric approach based on anatomical landmarks, most of which were located in the head region of the fish. Through photo-analysis, males could be distinguished from females through larger head, eye and mouth size (Cooper et al 2011). For some species, such as the California Sheephead (Semicossyphus pulcher), whose life history involves transition from female to male, morphometrics do not provide adequate basis for sex determination (Loke-Smith et al. 2010). Although sexual dimorphism has not been rigorously examined in brook trout, other salmonids have been studied for morphological differences. Jahunen et al. (2009) conducted photo-analysis of Arctic charr based on a truss network of twenty-eight measurements and found that mature males have more robust (greater length and depth) bodies and heads than females and juveniles. Pacific salmon species were shown to exhibit differences in kype presence and adipose fin size compared between sexes (Beacham and Murray 1986); length and height of adipose fin was on average 1%-48% (variable between species) larger in males than females and suggesting that

adipose fin size could be used to externally determine sex in Pacific salmon species. Further work with Oncorhynchus species has found that 87-97% (variable between species) of individuals could be correctly sexed using adipose fin size and/or jaw length (Beacham and Murray 1986). Similarly, Merz and Merz (2004) found that ratios of snout length to fork length in chinook salmon (Oncorhynchus tshawytscha) led to 96% accuracy in determining sex in handled fish, and ratios of adipose fin length to fork length led to 86% accuracy for fish measured from video images at a fish passage facility. When combining these two ratios with head length measurements, accuracy increased to 92% in determining sex of fish measured through video images. While nonlethal approaches to sex determination via morphology have been successfully developed for other salmonids (Beacham and Murray 1986, Merz and Merz 2004), no such techniques exist for brook trout. Thus, biologists lack effective tools to determine sex for brook trout. To address this concern, I measured a suite of morphological characters in hopes of identifying distinct and quantifiable differences between male and female brook trout that could allow for non-lethal determination of sex. Here I present those metrics most effective at determining sex and which of those metrics may be applied to rapidly sex brook trout in the field. Although brook trout are exhibiting widespread declines and are the focus of many conservation efforts, we still lack basic management tools. Developing a nonlethal approach to determine sex via morphology will open new doors to brook trout research and has the potential to improve the biological information on which management decisions are based. Metrics for determining sexual dimorphism could be used to

examine sex-specific vital rates, including growth, survival, and movement. Sex ratios may also be evaluated using this approach. This information may have considerable implications for population dynamics, but is currently overlooked by managers because it is too difficult to obtain. Although this tool was developed based on individuals collected in a single watershed, two distinct geographic morphotypes were determined, and the results show that this tool should be applicable across a much broader geographic area. The development of this non-lethal tool for sex identification in brook trout will further basic and applied science and, consequently, aid in the conservation of the species. Methods Data Collection In October 2011, backpack electrofishing was used to collect brook trout in tributaries within the Savage River watershed in western Maryland (Figure 1). This work was conducted in close collaboration with MD DNR and the UMCES Appalachian Laboratory, in support of an ongoing collaborative research project. The collection took place during spawning season when sex of brook trout is most evident and can be accurately determined. The fish were lightly anesthetized with tricaine methanesulfonate (ms-222) buffered with 0.2 mm NaHCO, ph = 7. The total length (mm) of the fish was measured, and the width of head (mm) was measured with calipers. The fish were assessed through manual gamete expression to accurately determine sex. They were then placed on a light-colored background and photographed on their left side with a digital camera attached to a tripod at a specified height. The target sample size of 25 fish of each sex was calculated based on a two-tailed powered analysis for a moderate

effect size. We collected individuals - 6 adult males, adult females, 46 adults of unknown sex (15 only of which full measurements were taken). Fish were collected from the upper reaches of the Big Run system and from the lower portion of Big Run (Figure 1). Individuals <100 mm total length were not used for analysis, as these fish were assumed to be sexually immature and we could not determine their sex. Figure 1. Sample sites in Savage River watershed. The squares represent presence of Morphotype 1 and the circle represents presence of Morphotype 2. Image Analysis I analyzed brook trout images of known sex using ImageJ, an open-source imaging software. I measured twenty-seven distances between anatomical features in form of a truss network (Figure 2a, b). This method has been successfully used to examine

morphologic variation in other salmonids (Janhunen 2009). A suite of metrics were derived from the truss network (Table 1). Figure 2a. Truss network of morphometrics applied to Arctic charr (Salvelinus alpinus) (Janhunen et al. 2009) Figure 2b. Truss network of morphometrics applied to brook trout. Table 1. Measurement descriptions modeled after truss network measurements used by Jahunen et al. Metric Description M1 crown to snout tip M2 snout tip to anterior pectoral fin M snout tip to gill juncture M4 gill juncture to crown M5 gill juncture to anterior pectoral fin

M6 M7 M8 M9 M10 M11 M12 M1 M14 M15 M16 M17 M M19 M20 M21 M22 M2 M M25 M26 M27 crown to anterior pectoral fin crown to anterior pelvic fin crown to anterior dorsal fin anterior pectoral fin to anterior dorsal fin total pectoral width anterior pectoral fin to anterior pelvic fin anterior dorsal fin to anterior pelvic fin anterior pelvic fin to posterior dorsal fin anterior pelvic fin to anterior anal fin anterior dorsal fin to posterior dorsal fin anterior dorsal fin to anterior anal fin posterior dorsal fin to anterior anal fin posterior dorsal fin to posterior anal fin posterior dorsal fin to anterior adipose fin anterior anal fin to anterior adipose fin anterior anal fin to posterior anal fin posterior anal fin to anterior adipose fin posterior anal fin to dorsal insertion of caudal fin posterior anal fin to ventral insertion of caudal fin anterior adipose fin to ventral insertion of caudal fin anterior adipose fin to dorsal insertion of caudal fin dorsal insertion of caudal fin to ventral insertion of caudal fin Statistical Analysis After processing the images, each measurement was regressed against fish length in order to determine if the raw data or log10 transformed data were most appropriate (Janhunen et al. 2009) for analysis. The standardized residuals were calculated and used in Analysis of Variance (ANOVA) to test for differences between sexes for each measurement. All variables used in the analyses met the homogeneity of variances assumptions of regression and ANOVA. Results were considered statistically significant at P<0.05. Finally, I used discriminant functions analysis to predict sex from the morphological measurements and used jackknifing to produce a cross-classified error rate on the predictions. As utilizing the full suite of 27 measurements is not practical for field application, I re-ran the discriminant functions analysis on the seven

measurements with the highest separations to determine if fewer variables could be used to accurately assess sex. All analyses were conducted using the software program R (a free software environment for statistical computing and graphics). Results All morphometric measurements were significantly related to length, regardless of whether the raw or log10 transformed data were used (Table 2). Except for measurements M and M5, two distinct morphotypes were present with each forming a distinct group as demonstrated in Figure. Morphotype 1 consisted of fish captured from the upper portions of the Big Run system, while morphotype 2 consisted of those captured in the lower portion of Big Run. Fish from the two morphotypes were separated by approximately 4 km of stream (Figure 1). Truss network measures that showed strong separation of the morphotypes were measurements M1,M 6, M7, M8, M9, M16, M17, M, and M2 (Appendix A). As there were two distinct groups that emerged, I ran regressions for each morphological measurement on each morphotype where it was obvious from regression plots that the two morphotypes were present. This was necessary because the standardized residuals used in the ANOVA analysis would be biased if they were calculated on both morphotypes combined. Table 2. Regression table of results for log10 Transformed Measurements metric slope intercept df t-value p-value M1 0.0000 1.996221 90 17.4 <2e-16 M2 0.00577 2.081042 90 21.1 <2e-16 M 0.00266 2.028842 90 15.77 <2e-16 M4 0.006 1.962 90 19.5 <2e-16 M5 0.0050 1.4097 90 10.62 <2e-16 M6 0.0012 1.914 90.48 <2e-16

M7 0.0009 2.764 90 17.52 <2e-16 M8 0.000887 2.199012 90 17.01 <2e-16 M9 0.0014 2.191284 90 17.16 <2e-16 M10 0.00005 1.94958 90 12.95 <2e-16 M11 0.00966 2.217496 90 15.16 <2e-16 M12 0.00287 2.079204 90 17.67 <2e-16 M1 0.00205 2.0415 90 16.79 <2e-16 M14 0.002 2.058806 90 1.06 <2e-16 M15 0.0027448 1.9 90 1.05 <2e-16 M16 0.00087 2.26107 90 17.54 <2e-16 M17 0.0059 2.0544 90 17.61 <2e-16 M 0.0041 2.174958 90 15.98 <2e-16 M19 0.00095 2.04128 90 12.59 <2e-16 M20 0.0056 2.011721 90 17.27 <2e-16 M21 0.002612 1.855225 90 11.9 <2e-16 M22 0.0006 1.869091 90 17.0 <2e-16 M2 0.000088 2.0779 90 17.27 <2e-16 M 0.00804 1.9956 90 17.16 <2e-16 M25 0.0022 2.19559 90 14.98 <2e-16 M26 0.0021 2.1222 90 1 <2e-16 M27 0.00142 1.99445 90.4 <2e-16 Significant differences in the standardized residuals existed between sexes for the measures M2, M, M4 and M6 (Table ). All four measurements (M2, M, M4, and M6) represent the geometric relationship of anatomical landmarks in the head of the fish (Table 1, Figure 2). All four measures tend to be longer for males, at a given length, than females, suggesting that males have longer snouts and deeper, more robust heads than females. Specifically, the significance of M suggests that the distance from snout along the top jaw line could provide accurate determination of sex that is conserved across morphotypes.

LT8 2.5 2.6 2.7 2.8 2.9.0 Standardized Residuals -2-1 0 1 2 4 LT 2. 2.4 2.5 2.6 2.7 2.8 Standardized Residuals -2-1 0 1 2 Regression for Transformed M Std Residuals for Transformed M 72 59 74 55 90 105 4 50 8 89 41 45 22 16 92 88 49 10 6 1 100 104 998 26 21 11 17 56 81 52 10 67 74 72 77 61 66 59 65 1 62 64 7 289 14 85106 57 25 79 76 94 91 6 97 2 107 8 7 8 2 84 86 9 109 87 47 40 20 96 6 4 65 62 7957 64 76 55 6 90 7 61 8 105 91 77 89 66 100 8 6 14 45 104 107 4 85 16 99 50 98 84 97 25 106 92 1 22 41 94 1 7 28 88 17 21 26 109 2 9 56 49 86 15 10 54 87 0 40 8 2 11 20 81 101 96 52 9 47 54 0 15 101 7 19 7 4 19 6 5 5 120 140 1 0 200 220 0 120 140 1 0 200 220 0 Regression for Transformed M8 Std Residuals for Transformed M8 (Morphotype 1) 8 67 6166 77 7 59 64 74 62 6572 57 8 7 79 76 1 40 55 50 10 90 22 88 5 4 92 19 49 41 105 45 54 89 8 15 1016 0 6 1 7 81 11 26 21 104 56 99 17 98 100 52 28 6 106 9 25 94 9 2 45 86 14 972 91 6 40 7 10 96 107 8 4 2084 87 109 10 1 91 90 55 6 104 81 96 25 106 86 92 50 8 1 88 9956 107 22 6 17 20 4 14 21 972 47 45 98 2611 28 15 49 84 7 85 94 9 101 19 105 4 100 54 2 89 8 16 5 9 41 52 109 87 0 10 120 140 1 0 200 220 0 120 140 1 0 200 220 0 Figure. a) Linear regression for the log-transformed values of M shows no clear distinction between individuals caught in the lower portion of Big Run and individuals caught in upper reaches of Big Run. b) Standard residuals for the log-transformed values of M7 shows no clear distinction between individuals caught in the lower portion of Big Run and individuals caught in upper reaches of Big Run. c) Linear regression for the log-transformed values of M8. Points 57-80 represent the group of individuals caught in the lower portion of Big Run while the rest represent the individuals caught in upper reaches of Big Run. d) Standard residuals for the log-transformed values of M8 of individuals that fall into morphotype 1.

Table. ANOVA table of results comparing the standardized residuals of the log10 transformed measurements in brook trout of known sex. metric morphotype df f-value p-value metric morphotype df f-value p-value M1 1 1, 90.08 0.08 M15 1 1, 90 2.67 0.1064 M1 2 1, 90 4.8126 0.04089 M15 2 1, 90 2.1045 0.162 M2 1 1, 90 8.54 0.0046 M16 1 1, 90 1.55 0.2211 M2 2 1, 90 11.46 0.00227 M16 2 1, 90.20 0.08925 M combined 1, 90 5.58 0.02514 M17 1 1, 90 0.81 0.52 M4 1 1, 90 5.6696 0.0200 M17 2 1, 90.15 0.0845 M4 2 1, 90 5.6696 0.0200 M 1 1, 90 0.1 0.2 M5 combined 1, 90 0. 0.67 M 2 1, 90 2.9727 0.1009 M6 1 1, 90 5.972 0.01741 M19 1 1, 90 0.281 0.59 M6 2 1, 90 7.2669 0.0142 M19 2 1, 90 1.8065 0.1948 M7 1 1, 90 1.07 0.1902 M20 1 1, 90 2.4865 0.1194 M7 2 1, 90 4.465 0.04805 M20 2 1, 90 5.1412 0.0522 M8 1 1, 90 1.4692 0.26 M21 1 1, 90 0.1156 0.749 M8 2 1, 90 5.2526 0.051 M21 2 1, 90 1.5857 0.222 M9 1 1, 90 2.101 0.1517 M22 1 1, 90.944 0.051 M9 2 1, 90.09 0.05 M22 2 1, 90 5.8058 0.02628 M10 1 1, 90.5417 0.06406 M2 1 1, 90.122 0.08165 M10 2 1, 90. 0.08742 M2 2 1, 90 5.112 0.0264 M11 1 1, 90 0.1 0.722 M 1 1, 90 2.9997 0.087 M11 2 1, 90 1.7 0.27 M 2 1, 90 5.5614 0.02 M12 1 1, 90 7.1919 0.009155 M25 1 1, 90 2.21 0.141 M12 2 1, 90 5.52 0.02651 M25 2 1, 90 4.9012 0.0927 M1 1 1, 90 5.197 0.02652 M26 1 1, 90 2.950 0.905 M1 2 1, 90 4.252 0.0514 M26 2 1, 90 4.22 0.054 M14 1 1, 90 0.0015 0.9695 M27 1 1, 90 0.6202 0.47 M14 2 1, 90 0.1966 0.6625 M27 2 1, 90 2. 0.11 The discriminant functions analysis provided a reasonable method to predict sex of individuals from the morphological measurements. Good separation existed between sexes (Figure 4), with relatively low misclassification error rates (Table 4); classification accuracy for the full dataset was 85% (Table 4). Accuracy decreased only slightly with jackknifing. The discriminant functions analysis ran on the jackknifed data had an accuracy rate of 85% with 5 individuals randomly removed and % with 10 individuals removed from the full data set (Table 4). Simplifying the model from the full

0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 27 variables to the best seven variables (Table 5) reduced classification accuracy from 85% to 84% (Table 4). Classification accuracy remained high on the reduced variable dataset after performing cross-classification error analysis with jackknifing based on leaving out 5 individuals (% accuracy) or 10 individuals (84% accuracy). Variables with the highest separation between sexes were in decreasing order: M, M2, M19, M14, M12, M21, and M6). Variables were related to the head(m2, M, M6) and body depth(m12), which are typically longer in males, and posterior length(m14, M19, and M21), which are typically longer in females. Regardless of the specific model implementation, all errors were females that were misclassified as males (Table 4). -4-2 0 2 4 group F -4-2 0 2 4 group M Figure 4. Histogram depicting separation between females and males across all fish collected in the Big Run system.

Table 4. Cross classified error assessment of the full model containing 27 measurement variables and a reduced model containing the seven most discriminating measurement variables. Accuracy is shown for the full dataset as well as jackknifing where 5 individuals were withheld or with 10 individuals withheld. Predicted female Predicted male Full dataset, full model Female 14 Male 0 54 Jackknifed = 5, full model Female 14 Male 0 54 Jackknifed = 10, full model Female 21 17 Male 0 54 % accuracy 85 85 Allocated to female Allocated to males Full dataset, reduced model Female 2 15 Male 0 54 Jackknifed = 5, reduced model Female 21 17 Male 0 54 Jackknifed = 10, reduced model Female 2 15 Male 0 54 % accuracy 84 84 Table 5. Discriminant Function Analysis loadings for full model of 27 measurement variables and a reduced model of the seven most discriminating measurement variables. metric LD1 loadings group means (female) group means (male) Full model M1-22.88661 2.4909 2.502699 M2 51.6518 2.6121 2.676 M 15.4027415 2.572 2.598625 M4-4.095005 2.4411 2.505 M5-0.0811412 1.985272 1.999642 M6-27.8649984 2.449941 2.44 M7.484481 2.85197 2.8554 M8-2.9954 2.41 2. M9 55.586658 2.9158 2.9942

M10-0.0022717 0.8025 6.86 M11-4.621401 2.7 2.78 M12 65.212122 2.91 2.6916 M1-8.74489797 2.51 2.58977 M14-5.0152571 2.5679 2.56 M15 6.167404176 2.4 2.445 M16 0.001454255 62.6509 616.6677 M17-0.0025495 404.4089 98.088 M -27.612287 2.657 2.658099 M19.7666909 2.548077 2.51121 M20-15.52877 2.501257 2.5006 M21-0.0192892 209.817 196.86 M22 10.1087989 2.48964 2.6961 M2 5.52527 2.55 2.5867 M 11.877987 2.481 2.498854 M25-0.009112 457.6997 459.1995 M26 0.0950.8922.9906 M27-0.024126 276.892 27.5961 Reduced model M 9.06212215 2.572 2.598625 M2 27.107 2.6121 2.676 M19-8.77614 2.548077 2.51121 M14-12.410922 2.5679 2.56 M12 11.15868 2.91 2.6916 M21-0.02014005 209.811 196.86 M6 -.22662649 2.449941 2.44 Discussion Distinct morphological differences existed between male and female brook trout. Of the 27 truss network measurements used, retaining only 7 that focused on the head, depth of body and the posterior length produced a model with good separation between sexes and minimal loss of accuracy (84% accurate). Males had longer snouts, more robust heads and deeper bodies, while females had longer post-dorsal lengths. Sexual dimorphism appears common in salmonids. Mature male Artic charr have more robust (greater length and depth) bodies and heads than females (Jahunen et al. 2009). Similarly, Pacific salmon were correctly sexed with 87-97% accuracy using jaw length

(Beacham and Murray 1986), whereas sex of chinook salmon was predicted with 92% accuracy by combining head length measurements with snout and adipose fin length to fork length ratios (Merz and Merz 2004). Although sex determination classification accuracy for Big Run brook trout was lower than the previously mentioned studies, it remains high enough to be a useful tool. Sexual dimorphism in salmonids typically increases with an individual s age or size where males tend to develop more divergent head shape (Beacham and Murray 1986, Janhunen et al. 2009). Thus, most classification errors should comprise small males classified as females because they have not yet achieved secondary sexual characters. However, all misclassified fish in Big Run were females classified as males. Size did not appear to affect results as the lengths of misclassified females spanned almost the entire range from -206 mm and with weights ranging from 1-97 grams. Female salmonids can have greater variability in morphological measurements and ratios (Merz and Merz 2004), which could cause overlap with more male-like features for some female brook trout. Classification errors were distributed across both morphotypes, suggesting that the morphological variation in females can be widespread and is not restricted to a single subpopulation. Morphometric measurements on misclassified fish were checked and found to be error free. Two distinct morphotypes were present in the Big Run system. One morphotype was present in the upper reaches of Big Run and the second was found exclusively in the lower portion of the system. The presence of two morphotypes allowed for better determination of measurements most strongly associated with sex and least related to

population variation. Sex-specific morphological differences are conserved in salmonids exhibiting morphological variation related to population divergence (Janhunen et al. 2009). Distinct separation between the sexes was apparent and sexual dimorphism can be determined empirically, regardless of morphotype. Brook trout, and salmonids in general, exhibit great phenotypic plasticity and morphological variation. Multiple morphotypes may occur in the same system due to genetic or behavioral separation caused by physical or other barriers. Populations of brook trout separated between two lakes exhibited morphological variation both in groups that were genetically distinct and those that did not show significant genetic variation (Dynes et al. 1999). Similarly, populations of rainbow trout (Oncorhynchus mykiss) isolated by waterfalls have been found to be genetically and morphologically divergent (Currens et al. 1990). Genetic divergence is not a likely cause for the morphological variation in the Big Run population, as there are no physical barriers, and gene flow is probable. However, behavioral segregation could possibly occur if habitat or life history strategy influences body shape. Life history strategies are frequently associated with phenotypic expression. Migratory and resident brook trout were correctly classified with 87% accuracy based on discriminant functions analysis of five traits, including measurements of body depth and pectoral and pelvic fin lengths (Morinville and Rasmussen 2008). Migratory brook trout exhibit more streamlined bodies and shorter fins than resident brook trout (Morinville and Rasmussen 2008). Similarly, body depth decreased in coho salmon (Oncorhynchus

kisutch) and sockeye salmon (Oncorhynchus nerka) with increased migration distances (Fleming and Gross 1989, Hendry and Quinn 1997). Environmental variation, such as water depth or resource availability could explain morphological divergence in the Big Run system. One morphotype was found in the headwaters where there are multiple, small tributaries and shallower water, whereas the second was found in the lower portion of Big Run where the water is deeper and discharge greater. Salmonids, including coaster brook trout and sockeye salmon, that breed in deeper waters have deeper bodies than those who breed in shallower stream environments (Huckins et al. 2011, Hendry and Quinn 1997, Blair et al. 199). Similarly, brook trout were assigned to geographic groups with 69%-% accuracy based on fin shape, posterior body length and caudal peduncle height, with differences consistent to the respective foraging requirements of benthic and pelagic environments (Dynes et al. 1999). Morphotypic variation based on benthic and pelagic foraging types has also been found in Artic charr (Salvelinus alpinus) (Arbour et al. 2006, Snorrason et al. 1994). Collection of morphological data required only a few additional seconds to fish processing. The tools used are basic making this an accessible method. Post processing with ImageJ was similarly straightforward, but does allow for human error. However, repeating measurements on the same fish resulted in very little inconsistency. The high classification accuracy rates indicate discriminant functions analysis is a reliable and repeatable method for determining sex. The truss network accounts for all

of the major anatomical landmarks on a fish and provides a thorough method to determine morphological variability. Brook trout are an ecologically important, declining, and well studied species but significant gaps remain in our understanding of their life history. The morphological differences that I found between the sexes allow a quick, nonlethal way of determining the sex for each brook trout captured. The method is easily implemented with minimal additional handling or equipment, but is over 80% accurate. Although more validation is required for small, immature fish, the tool should be useful for researchers and managers alike. Determining sex-specific life history attributes such as growth rates, survival rates, and movement as well as sex ratios may provide important insights in more effective management of brook trout and the watersheds they inhabit.

Literature Cited Arbour, J. H., Hardie, D. C., Hutchings, J. A. (2011). Morphometric and genetic analyses of two sympatric morphs of Arctic char (Salvelinus alpinus) in the Canadian High Arctic. Canadian Journal of Zoology, 89(1), 19-0. Beacham, T., Murray, C. (198). Sexual dimorphism in the adipose fin of pacific salmon (Oncorhynchus ). Canadian Journal of Fisheries and Aquatic Sciences, 40(11), 2019-20. Beacham, T., Murray, C. (1986). Sexual dimorphism in length of upper jaw and adipose fin of immature and maturing pacific salmon (Oncorhynchus). Aquaculture, 58(-4), 269-276. Blair, G. R., Rogers, D.E., Quinn, T.P. (199). Morphology of Sockeye Salmon in the Kvichak River System, Bristol Bay, Alaska. Transactions of the American Fisheries Society, 122 (4). 550-559 Cooper, I. A., Gilman, R. T. and Boughman, J. W. (2011). Sexual Dimorphism and Speciation on Two Ecological Coins: Patterns from Nature and Theoretical Predictions. Evolution, 65: 255 25. Currens, K. P., Schreck, C. B., Li, H.W. (1990). Allozyme and Morphological Divergence of Rainbow Trout (Oncorhynchus mykiss) above and below Waterfalls in the Deschutes River, Oregon. American Society of Ichthyologists and Herpetologists, 1990(). - 746. Dynes J., Magnan P., Bernatchez L., Rodriguez M.A. (1999). Genetic and morphological variation between two forms of lacustrine brook charr. Journal of Fish Biology 54: 955 972. Fleming I.A., Gross M.R. (1989). Evolution of adult female life history and morphology in a Pacific salmon (coho, Oncorhynchus kisutch). Evolution 4: 141 157. Heft, A.A., editor. (2006). Maryland brook trout fisheries management plan. Maryland Department of Natural Resources, Annapolis. Available: www.dnr.state.md.us/fisheries. Hendry, A.P., Quinn, T.P. (1997). Variation in adult life history and morphology among Lake Washington sockeye salmon (Oncorhynchus nerka) populations in relation to habitat features and ancestral affinities. Canada Journal of Fishery and Aquatic Science. 54: -84.

Huckins, C., Baker, E., Fausch, K., Leonard, J. (2008). Ecology and life history of coaster brook trout and potential bottlenecks in their rehabilitation. North American Journal of Fisheries Management, 28(4), 121-142. Hudy, M., Theiling, T., Gillespie, N., Smith, W. (2005). Distribution, status, and perturbation to brook trout within the eastern United States. Technical report to the Eastern Brook Trout Joint Venture. Janhunen, M., Peuhkuri, N., Piironen, J. (2009). Morphological variability among three geographically distinct arctic charr (Salvelinus alpinus l.) populations reared in a common hatchery environment. Ecology of Freshwater Fish, (1), 106-116. Lengkeek, W., Didderen, K., Côté, I. M., van der Zee, E. M., Snoek, R. C., Reynolds, J. D. (2008). Plasticity in sexual size dimorphism and Rensch s rule in Mediterranean blennies (Blenniidae). Canadian Journal of Zoology, 2008, 86:117-11, 10.119/Z08-10. Letcher B.H., Nislow K.H., Coombs J.A., O'Donnell M.J., Dubreuil T.L. (2007). Population Response to Habitat Fragmentation in a Stream-Dwelling Brook Trout Population. PLoS ONE 2(11): e119. doi:10.1/journal.pone.000119 Loke-Smith, K.A, Sundberg, M.A., Young, K.A., Lowe, C.G. (2010). Use of Morphology and Endocrinology to Predict Sex in California Sheephead: Evidence of Altered Timing of Sex Change at Santa Catalina Island, California. Transactions of the American Fisheries Society, Vol. 19, Iss. 6. Maryland Department of Natural Resources (2005). Maryland Wildlife Diversity Conservation Plan. Maryland Department of Natural Resources, Annapolis. Available: www.dnr.state.md.us Merz, J., Merz, W. (2004). Morphological features used to identify chinook salmon sex during fish passage. Southwestern Naturalist, 49(2), 197-202. Morinville G.R.,Rasmussen J.B. (2008) Distinguishing between juvenile anadromous and resident brook trout (Salvelinus fontinalis) using morphology. Environmental Biology of Fishes 81: 1 4. Snorrason, S.S., Sku lason, S., Jonsson, B., Malmquist, H. J., Jo nasson, P. M., Sandlund, O. T., Lindem, T. (1994). Trophic specialization in Arctic charr (Salvelinus Alpinus) (Pisces; Salmonidae): morphological divergence and ontogenetic niche shifts. Biological Journal of the Linnean Society 52(1), 1.

Stranko, S.A., Hilderbrand, R.H., Morgan, R.P., Staley, M.W., Becker, A.J., Roseberry- Lincoln, A., Perry, E.S., Jacobson, P.T. (2008). Brook trout declines with land cover and temperature changes in Maryland. North American Journal of Fisheries Management. 28:122-122. Waco, K., Taylor, W. (2010). The influence of groundwater withdrawal and land use changes on brook charr (Salvelinus fontinalis) thermal habitat in two coldwater tributaries in Michigan, U.S.A. Hydrobiologia, 650(1), 101-116.

LT1 2. 2.4 2.5 2.6 2.7 2.8 Appendix A. Linear regression plots for log10 transformed values of 27 measurement variables. Regression for Transformed M1 61 66 64 67 6 81 17 56 26 11 79 107 6 797 47 84 14 91 8 86 96 20 4 87 109 40 57 76 62 5972 7 65 74 106 2 10 94 289 25 2 85 6 9 77 1 4 5 50 45 8 22 41 89 90 0 105 49 10 55 92 15 88 54 1016 19 52 99 104 98 100 7 21 1 8 120 140 1 0 200 220 0

LT2 2.4 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M2 67 6 11 1 79 5776 59 72 62 65 64 7 6 8 107 797 91 84 20 86 109 87 40 4 74 85 106 14 94 9 2 25 47 2 61 10 66 77 28 1 9 22 4 8 90 41 55 50 5 105 10 192 45 100 99 17 104 98 56 7 52 21 26 81 96 15 101 54 0 19 49 88 120 140 1 0 200 220 0

LT 2. 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M 10 67 6 74 72 59 65 62 64 7 14 85106 57 25 79 76 94 91 6 97 107 2 8 7 2 8 84 86 9 109 87 47 77 61 66 1 289 55 90 105 4 50 8 89 41 45 22 16 92 88 49 10 100 104 998 17 56 52 1 26 21 11 81 40 20 96 4 6 54 0 15 101 19 7 5 120 140 1 0 200 220 0

LT4 2. 2.4 2.5 2.6 2.7 Regression for Transformed M4 90 55 6 1 56 100 98 26 99 104 81 21 11 17 61 67 77 72 7 64 74 1066 59 65 79 62 5776 85 281 9 2 25 94 6 97 106 91 14 2 9 107 47 8 6 7 8 86 84 96 109 4 20 40 8 105 4 50 89 22 41 5 92 49 88 16 45 10 19 15 0 54 52 7 101 120 140 1 0 200 220 0

LT5 1.8 2.0 2.2 Regression for Transformed M5 6 2 9 5 2 61 67 4 7 66 15 16 92 101 45 8 54 19 22 10 41 49 4 90 88 11 7 17 6 81 56 1 100 21 98 26 99 52 104 20 79 40 57 76 8 84 96 91 109 107 62 59 65 72 7 87 6 86 97 47 85 64 74 14 25 106 894 28 10 77 9 1 105 89 0 55 50 120 140 1 0 200 220 0

LT6 2.2 2. 2.4 2.5 2.6 2.7 Regression for Transformed M6 6 1 11 100 21 26 56 81 55 99 104 4 98 22 17 8 7 90 16 550 52 89 41 15 105 92 49 88 45 19 101 54 0 67 61 77 59 74 72 64 66 65 7 62 10 76 79 28 2 9 85 1 57 14 94 2 106 8 6 25 97 9 107 47 6 91 7 8 86 84 87 20 4096 109 4 120 140 1 0 200 220 0

LT7 2.7 2.8 2.9.0.1 Regression for Transformed M7 77 61 74 67 66 50 5 55 4 22 10 49 41 90 88 8 105 19 45 54 892 0 1016 15 6 1 81 2611 21 100 7 998 104 56 17 52 59 62 72 7 64 65 57 76 79 28 1 9 894 106 47 25 856 86 9 6 97 91 7 14 2 96 107 87 84 4 40 20 8 109 2 10 120 140 1 0 200 220 0

LT8 2.5 2.6 2.7 2.8 2.9.0 Regression for Transformed M8 67 6166 77 57 7 59 64 62 6572 74 8 79 76 1 55 50 10 90 22 88 5 4 92 19 49 41 105 45 54 89 8 15 1016 0 6 1 7 81 11 26 21 104 56 99 17 98 100 52 28 6 106 9 25 94 9 2 45 86 14 972 91 6 40 7 10 96 107 8 4 2084 87 109 120 140 1 0 200 220 0

LT9 2.5 2.6 2.7 2.8 2.9.0 Regression for Transformed M9 64 7 72 59 62 65 74 66 61 77 67 7957 76 1 28 9 6 25 6 14 106 94 2 9 972 47 96 97 858 6 867 10 4 40 109 20 87 8 84 4 50 5 10 88 55 49 90 22 0 89 8 19 41 45 92 105 54 15 101 16 56 04 98 99 52 100 17 81 21 26 11 1 120 140 1 0 200 220 0

LT10 2.2 2. 2.4 2.5 2.6 2.7 Regression for Transformed M10 77 92 41 55 8 105 4 45 89 90 49 88 10 50 15 22 1011654 0 19 6 98 99 56 100 04 52 17 21 1 81 2611 57 79 8 64 2585 94 2 106 97 107 14 47 9 867 9109 96 20 8 4 84 40 76 62 59 65 72 6 7 74 6 67 2 6166 10 28 1 9 5 120 140 1 0 200 220 0

LT11 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M11 77 61 66 74 57 76 64 72 7 62 59 65 67 7 6 81 26 1 21 11 79 1 94 289 86 106 47 6 8 96 87 25 797 2 4 107 9 9109 6 14 20 2 40 84 85 10 8 5 50 49 88 19 4 0 54 41 10 22 45 90 101 55 8 105 89 92 15 16 100 104 98 52 9956 17 120 140 1 0 200 220 0

LT12 2.4 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M12 61 7957 74 5972 7 64 65 62 67 77 66 28 1 4 55 22 41 50 89 8 90 5 0 88 45 105 49 16 92 19 6 98 56 21 81 2611 1 99 100 104 2 17 96 40 76 2 91 85106 10 94 2 25 9 97 6 14 6 107 47 208 109 86 84 4 87 7 8 9 15 54 101 120 140 1 0 200 220 0

LT1 2.4 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M1 6 57 79 76 91 7 65 72 64 59 62 74 25 61 67 77 1 66 28 9 9 2 10 6 97 2 14 85694 106 45 16 92 15 54 4 55 8 10 49 90 88 50 89 0 41 10522 5 19 56 98 52 104 17 99 7 100 1 81 26 21 11 96 40 20 8 107 109 4 84 87 86 7 47 8 101 120 140 1 0 200 220 0

LT14 2. 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M14 61 6 1 66 57 67 77 79 65 7 64 76 72 59 74 62 1 9 8 28 47 10 94 106 94 96 87 25 9 6 86 7 2 4 6 2 40 85 20 107 2 104 56 17 2611 81 21 91 84 8 109 49 50 90 10588 41 4 10 55 5 92 22 54 19 101 89 0 45 8 100 99 98 15 16 120 140 1 0 200 220 0

LT15 2.1 2.2 2. 2.4 2.5 2.6 Regression for Transformed M15 61 76 64 65 7 74 62 72 67 77 1 16 10 8 49 50 54 19 22 88 55 89 90 0 41 5 45 4 15 92 101 105 6 26 1 52 10056 21 98 104 99 11 17 81 7 57 59 79 20 96 8 84 40 4 87 86 7 2 25 9 10 97 85 109 6 6 14 47 94 91 107 8 106 2 28 66 9 120 140 1 0 200 220 0

LT16 2.6 2.7 2.8 2.9.0 Regression for Transformed M16 61 79 5776 64 59 72 7 62 65 74 67 77 66 28 1 9 6 1 47 2 10 85106 94 25 972 14 6 6 8 9 96 91 107 86 7 4 87 40 84 109 20 8 4 550 22 55 19 49 88 41 90 89 105 0 10 54 92 45 8 1016 10056 7 52 98 99 104 17 21 26 11 81 15 120 140 1 0 200 220 0

LT17 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M17 61 4 5 55 22 50 19 49 88 41 90 89 0 105 92 54 10 101 45 8 6 7 56 21 100 1 52 99 104 98 81 11 17 26 67 66 77 74 59 79 72 64 7 5776 62 65 28 2 1 9 47 106 10 2 85 96 97 4 6 14 25 86 7 97 94 87 9 40 84 8 20 109 15 16 120 140 1 0 200 220 0

LT 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M 61 79 74 67 66 77 19 22 5 4 88 55 50 41 0 10 49 90 92 89 105 101 54 6 21 100 7 17 11 52 104 56 1 81 26 99 98 57 76 59 72 64 7 62 65 47 10 2 28 9 2 85106 96 87 1 4 6 7 94 86 97 84 14 9109 107 25 9 8 40 20 16 45 8 15 120 140 1 0 200 220 0

LT19 2. 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M19 79 59 64 74 61 66 77 67 101 5 7 6 21 100 104 9956 11 1 52 17 26 98 7 5776 62 47 72 10 65 87 2 6 8 2 85106 96 9 6 86 84 4 7 94 109 8 97 107 14 91 28 40 9 25 1 20 105 19 55 50 49 88 4 54 92 90 41 89 0 22 10 45 81 15 16 8 120 140 1 0 200 220 0

LT20 2. 2.4 2.5 2.6 2.7 Regression for Transformed M20 57 79 76 72 7 64 59 65 62 74 67 77 61 66 28 1 6 91 107 6 97 4 87 96 867 2084 40 109 8 9 25 6 2 47 14 85 106 894 2 10 9 4 5 0 50 55 89 49 88 8 10 19 90 41 22 1654 105 45 92 56 98 7 52 104 17 99 100 21 26 1 81 11 15 101 120 140 1 0 200 220 0

LT21 2.0 2.1 2.2 2. 2.4 2.5 Regression for Transformed M21 77 57 65 64 72 7 74 62 67 61 66 0 10 19 54 22 16 8 88 41 50 892 49 4 90 5 101 105 55 45 6 1 81 21 26 100 52 98 17 11 56 104 99 7 79 76 59 107 87 281 9 47 9 109 2 10 2 94 96 84 85 86 14 25 6 6 20 91 97 106 4 7 8 8 40 15 120 140 1 0 200 220 0

LT22 2.1 2.2 2. 2.4 2.5 2.6 Regression for Transformed M22 79 5776 72 59 65 7 62 64 74 61 67 77 66 28 1 6 25 106 2 6 9 2 91 10 85 8 94 4107 97 6 14 47 40 20 96 86 7 9 109 84 22 4 7 56 98 100 52 104 17 81 21 2611 1 8 87 15 5 16 19 49 90 88 50 89 0 10 41 55 8 105 45 54 92 99 101 120 140 1 0 200 220 0

LT2 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M2 66 67 61 77 6 99 26 21 11 1 72 7 74 59 65 64 62 76 57 79 6 1 97 9 10 9 2 14 28 94 47 106 8 91 25 85 86 6 20 109 7 2 96 107 8 4 40 84 5 52 104 17 4 98 50 55 7 56 100 90 22 41 92 88 45 89 0 105 49 15 10 19 54 101 16 8 81 120 140 1 0 200 220 0

LT 2. 2.4 2.5 2.6 2.7 Regression for Transformed M 67 66 77 5776 72 65 64 7 62 59 97 74 6 61 22 55 4 50 5 54 19 88 41 45 92 90 101 105 0 15 8 49 21 26 99 11 1 6 52 104 17 100 98 81 7 56 10 79 9 1 14 9 87 106 94 28 47 2 20 9109 96 86 8 8 6 7 2 85 25 107 40 4 84 89 16 120 140 1 0 200 220 0

LT25 2.5 2.6 2.7 2.8 2.9 Regression for Transformed M25 72 67 6166 77 6 1 79 76 57 65 62 59 7 74 64 972 96 86 14 109 91 20 4 87 107 7 40 8 84 6 85 47 25 8 9 106 6 2 94 10 1 289 22 4 50 88 19 55 10 92 105 90 19 8 49 41 5 45 0 54 17 100 99 52 98 7 104 56 21 11 26 81 101 15 120 140 1 0 200 220 0

LT26 2.4 2.5 2.6 2.7 2.8 Regression for Transformed M26 76 79 72 65 7 62 59 64 74 61 77 66 67 10 9 22 50 892 4 105 55 19 88 16 10 5 90 41 6 100 52 17 98 99 104 7 56 81 21 1 11 26 57 972 94 106 96 109 86 4 6 2 97 14 85 8 20 87 40 6 7 25 8 47 84 28 9 1 45 0 49 15 54 8 101 120 140 1 0 200 220 0

LT27 2.2 2. 2.4 2.5 2.6 Regression for Transformed M27 6 79 20 57 76 96 84 4 8 40 91 72 59 65 7 64 62 6 86 2 94 47 85 8 87 7 6 74 25 9 106 94 67 61 66 2 10 77 1 28 9 7 17 26 1 109 107 88 4 19 10 49 90 105 55 41 45 8 16 54 22 0 5 50 98 56 52 99 100 104 21 81 11 892 101 15 120 140 1 0 200 220 0