Abstract

Morphometric data from fish are typically generated using one of two methods: direct measurements made on a specimen or extraction of distances from a digital picture. We compared data on 12 morphometrics collected with these two methods on the same collection of Cisco Coregonus artedi from Lake Ontario, North America, to assess the degree of bias in measurements made directly on a specimen- vs. an image-based method. We also assessed the degree of reproducibility within the image-based method by evaluating the amount of variation between different analysts for each morphometric method. Our results indicate specific morphometrics may be more prone to bias across the two methods and between analysts. Four of 12 morphometrics evaluated showed significant deviation from a 1:1 relationship that would be expected if the imaged-based method produced accurate specimen-based measurements. Pelvic fin length and pelvic–anal fin distance had the highest between-analyst variation for image-based landmarks, indicating low reproducibility for these metrics, compared with pectoral fin or total length, which had lower between-analyst variation. Although some morphometric measurements can be accurately obtained with either method, and therefore potentially used interchangeably in studies on Cisco morphology, our findings highlight the importance of considering method bias in morphometric studies that use data collected by different methods.

Introduction

Morphometrics—the study of the relative size of anatomically identical features across multiple groups—as applied to fisheries research has been, and remains, the primary way of discriminating species and identifying morphotypes. Across freshwater lakes in the northern temperate zones of North America and Europe, morphometrics is widely used to describe diversity among and within fish species in the coregonine group (family Salmonidae, subfamily Coregoninae, genus Coregonus). In the Great Lakes of North America, where coregonines are important as a commercial fishery resource (Ebener et al. 2008; Stockwell et al. 2009) and for occupying unique trophic niches within lake food webs (Eshenroder and Burnham Curtis 1999; Sierszen et al. 2014; Blanke et al. 2018; Rosinski et al. 2020), there is growing interest among resource management agencies to conserve and rehabilitate remnant populations to provide sustainable fisheries and restore ecosystem function (Stockwell et al. 2009; Bronte et al. 2017; Schmitt et al. 2020). A key knowledge gap for developing best practices aimed at restoring the diversity of coregonines is an adequate understanding of their morphological diversity. Such an understanding requires accurate descriptions of forms through high-quality morphological datasets.

Discrimination of coregonines in the Great Lakes, for example, traditionally relied on morphological measurements as the basis for separating different forms (Koelz 1929; Woodger 1976; Eshenroder et al. 2016). Morphological measurements are time consuming and typically collected on specimens in the laboratory rather than in the field. Furthermore, personnel require specialized training to ensure accuracy, and specimens should be archived in cases where it may be necessary to revisit samples to collect additional data or assess reproducibility. However, archiving fish in preservatives can modify shape, making it impossible to obtain the same values for some measurements as were obtained before preservation (Fruciano et al. 2020). The advent of digital imagery and computer software allows for the extraction of morphometric data between landmarks that are synonymous with the start points and endpoints used with direct caliper measurements. Important advantages of this image-based approach compared with traditional measurement techniques are that specimens do not need to be in hand at the time of measurement, nor do they need to be archived if there is reason to revisit a sample, as long as a proper image is captured. However, using different methods of data collection in the same dataset could lead to misinterpretation of results if they do not generate identical data with low bias between methods. Thus, it is critical to understand whether specimen measurements differ from those obtained from using digital images, particularly as the use of image-based techniques becomes more common in morphological studies of coregonines.

Studies on Cisco Coregonus artedi morphology use measurements obtained by both techniques on the sample individual (e.g., Piette-Lauzière et al. 2019), but no validation exists for comparisons of morphometric features across methodologies. Although, in theory, identical linear measurements from digital images also should be obtainable from live measurements made on the specimen, image distortion effects can impact apparent size and shape of body features differently (Muir et al. 2012; Collins and Gazley 2017). Whether this is true for all standard morphometrics as applied to Cisco, or highly biased for some measurements, but not others, requires evaluation to ensure measurements are similar across methods before they can be used interchangeably.

Our aim was to explore between-method reliability in morphometric data collection of Cisco. Specifically, we compared morphometric data collected by direct caliper measurements performed on specimens and an image-based approach performed separately by two analysts. Our motivation for comparing these methods with Cisco was to determine whether they can be used interchangeably, thereby facilitating new and improved studies to understand coregonine diversity and help guide conservation efforts that require accurate morphological data.

Methods

Cisco (n = 99) were collected from Lake Ontario, North America. We carried out all sampling and handling of fish in accordance with guidelines for the care and use of fishes by the American Fisheries Society (Jenkins et al. 2014). Upon collection, each fish was individually bagged and frozen and later thawed in the laboratory for morphometric measurements and imaging. We made fish collections by using two gear types to capture a wide range of body sizes. We collected Cisco during the spawning period by using gill nets in Chaumont Bay, New York, during November–December 2018 and from midwater trawling (Holden et al. 2019) in the eastern basin of Lake Ontario during September 2018. Larger Cisco were more common in gill net collections, whereas smaller Cisco were obtained from midwater trawl samples.

We measured 12 morphometrics by using two different methods (Figure 1A). We selected 12 morphometrics based on those commonly used in previous studies on Cisco morphs (e.g., Muir et al. 2013; Yule et al. 2013; Eshenroder et al. 2016; Piette-Lauzière et al. 2019). After we thawed the fish in the laboratory, we pinned the caudal, dorsal, and ventral fins of each fish, whereas paired fins were laid along the body. We captured a photograph on the left side of each fish for digital extraction of morphometric measurements by using a single-lens reflex camera, following recommendations of Muir et al. (2012), to minimize effects of distortion. Before capturing the photograph, we placed foam pads under the anterior and posterior ends of the body to help force the lateral view into a single horizontal plane (Figure 1B). We included a ruler in each image for scale. Next, we made measurements directly on the specimens by using a digital caliper (accurate to ±0.01 mm) by a single measurer (B.P.O.), except for total length and standard length that we measured to the nearest millimeter by using a measuring board. We refer to these direct measurements as the “specimen-based” dataset.

Figure 1.

(A) Drawing of Cisco Coregonus artedi indicating morphometric measurements used to compare specimen-based vs. image-based measurements in this study. (B) Digital image of a Cisco collected from Lake Ontario in 2018 with landmarks used to extract image-based measurements (U.S. Geological Survey, Lake Ontario Biological Station specimen ID: 50-2018-876-2). Morphometric measurements are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

Figure 1.

(A) Drawing of Cisco Coregonus artedi indicating morphometric measurements used to compare specimen-based vs. image-based measurements in this study. (B) Digital image of a Cisco collected from Lake Ontario in 2018 with landmarks used to extract image-based measurements (U.S. Geological Survey, Lake Ontario Biological Station specimen ID: 50-2018-876-2). Morphometric measurements are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

For the “image-based” morphometric dataset, we placed landmarks on each fish image to calculate distances between landmarks, consistent with the 12 morphometrics in the specimen-based dataset. Our goal in applying these landmarks was to use the same criteria in placement of the start point and endpoints used in the specimen-based measurements. We placed some landmarks at homologous structures (e.g., insertion of a fin ray), whereas we placed other semilandmarks (i.e., those that do not represent a homologous point) by predefined criteria, sometimes dependent on the location of homologous landmarks (Tables S1 and S2, Supplemental Material). For example, pelvic–anal fin distance was measured along the lateral line in the specimen-based dataset. To repeat this measurement from an image, we placed semilandmarks along the lateral line that were vertically in-line with the y-coordinates of the pelvic and anal fin landmarks.

We used the ‘digitzeImages' function from the R package StereoMorph (Olsen and Westneat, 2015) to place digital landmarks. We used a known distance along a ruler placed in each image to convert landmark coordinate units from pixels to millimeters. We calculated distance measurements between digitized landmarks by using the following formula:
formula
where d is the linear distance calculated between the starting (x1, y1) and ending (x2, y2) landmark coordinates. We carried out all calculations in R v.3.5.3 (R Core Team 2019) on scaled landmark coordinates.

To assess whether measurements extracted from images were consistent with specimen-based measurements, we used linear regression to test whether the observed slopes for each morphometric feature were significantly different from 1. We constructed regression models with the assumption that specimen-based values were the independent variable used to predict image-based values; therefore, our interpretation of bias refers to image-based deviation from the specimen-based method. We compared observed slopes of each regression model with a value of 1 by using the ‘slope.test' function from the R package smatr for ordinary least squares regression (Warton et al. 2012). We applied a standard level of significance followed by a Bonferroni correction for multiple comparisons (α = 0.05/12) to evaluate significant deviations from the hypothesized 1:1 slope.

We evaluated the level of reproducibility of image-based morphometrics by comparing values generated by two analysts (B.P.O. and J.D.S.) for each morphometric feature extracted from digital images. We performed this comparison in a blind format, meaning that landmarked images were not shared among analysts during the process. Analysts shared only raw images, instructions of landmark criteria, and a reference image of where to place each landmark to ensure that the placement of landmarks by the second analyst was not influenced by the first analyst. We evaluated the level of reproducibility between the two independent analysts for each image-based morphometric feature by comparing absolute values of Cohen's effect size (h) calculated between the two measurers for each morphometric feature (Cohen 1992). We ranked each morphometric feature by its mean value of h and applied thresholds defined in Cohen (1992) to identify morphometrics with negligible (<0.2), small (0.2–0.5), medium (0.5–0.8), and large (>0.8) levels of difference between analysts. We considered morphometrics with negligible values to be highly reproducible and those with large values as less reproducible.

Results

Comparison of morphometric data collected by the same measurer with specimen-based vs. image-based measurements revealed four morphometrics with linear regression slopes that were significantly different from 1 (Figure 2; Table 1): pectoral fin length (PCL); body depth; standard length; and orbital length. The amount of variability explained by linear regression ranged from relatively low values for orbital length measurements (R2 = 0.67) to high values for body depth and standard length measurements (both R2 = 0.99). Orbital length was the only variable with significant regression slope of less than 1, indicating that image-based measurements underestimate the same measurement obtained on the specimen by using a caliper. A similar analysis that compared the same specimen-based measurements of the first analyst to image-based measurements of the second analyst revealed slopes there were not significantly different from a value of 1 for each of the 12 morphometrics (Table S3, Supplemental Material).

Figure 2.

Comparison of specimen-based vs. image-based measurements for each morphometric measurement collected on Cisco Coregonus artedi collected from Lake Ontario in 2018. Dashed gray line represents a hypothetical 1:1 relationship. Black line indicates linear regression for those morphometrics where the observed slope was significantly different from a value of 1 following Bonferroni correction for multiple comparisons (α = 0.05/12 = 0.0042). Morphometric measurements are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

Figure 2.

Comparison of specimen-based vs. image-based measurements for each morphometric measurement collected on Cisco Coregonus artedi collected from Lake Ontario in 2018. Dashed gray line represents a hypothetical 1:1 relationship. Black line indicates linear regression for those morphometrics where the observed slope was significantly different from a value of 1 following Bonferroni correction for multiple comparisons (α = 0.05/12 = 0.0042). Morphometric measurements are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

Table 1.

Summary of ‘slope.test' statistics comparing whether slope values of linear regressions differed from a test value of one for image-based observations as a function of specimen-based observations for Cisco Coregonus artedi collected from Lake Ontario in 2018. Bolded P values denote observations that were significant following Bonferroni correction for multiple comparisons (α = 0.05/12 = 0.0042).

Summary of ‘slope.test' statistics comparing whether slope values of linear regressions differed from a test value of one for image-based observations as a function of specimen-based observations for Cisco Coregonus artedi collected from Lake Ontario in 2018. Bolded P values denote observations that were significant following Bonferroni correction for multiple comparisons (α = 0.05/12 = 0.0042).
Summary of ‘slope.test' statistics comparing whether slope values of linear regressions differed from a test value of one for image-based observations as a function of specimen-based observations for Cisco Coregonus artedi collected from Lake Ontario in 2018. Bolded P values denote observations that were significant following Bonferroni correction for multiple comparisons (α = 0.05/12 = 0.0042).

Comparison of image-based measurements by using effect size estimates to assess reproducibility between two analysts showed a range from negligible to large effect sizes across morphometrics (Figure 3). Ranked by mean effect size, pectoral fin length (h = 0.05), total length (0.08), and head length (0.15) had negligible effect sizes and therefore had relatively high reproducibility between analysists. Pelvic–anal fin distance (1.73) and pelvic fin length (1.41) showed the highest effect sizes, indicating the largest degree of difference between measurers among the 12 morphometrics.

Figure 3.

Mean (±95% CI) Cohen's effect size (h) between two analysts' image-based morphometric measurements of Cisco Coregonus artedi collected from Lake Ontario in 2018. Cut-offs values were set a 0.2, 0.5, and 0.8 for qualitative thresholds to rank the degree of difference between analysts from negligible to large, for each morphometric. Morphometric abbreviations are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

Figure 3.

Mean (±95% CI) Cohen's effect size (h) between two analysts' image-based morphometric measurements of Cisco Coregonus artedi collected from Lake Ontario in 2018. Cut-offs values were set a 0.2, 0.5, and 0.8 for qualitative thresholds to rank the degree of difference between analysts from negligible to large, for each morphometric. Morphometric abbreviations are as follows: BDD = body depth; CPL = caudal peduncle length; DOH = dorsal fin height; HLL = head length; MXL = maxillary length; OOL = orbital length; PAD = pelvic-anal fin distance; PCL = pectoral fin length; POL = preorbital length; PVL = pelvic fin length; STL = standard length; TTL = total length.

Discussion

Our results suggest that specific morphometrics when applied to Cisco are prone to bias across two commonly used methods. Although morphometric measurements by caliper and digital images of fish are often used interchangeably in datasets in the literature, we highlight the importance of considering multiple methods to generate the same dataset across different groups of samples because of the apparent variability revealed by our analyses. Four of 12 morphometrics evaluated show significant deviation from a 1:1 relationship that would be expected if specimen-based measurements were accurately reproduced from the image-based method. Among the eight remaining morphometrics that did have consistent values between specimen- and image-based methods, the level of reproducibility by using an image-based method between two analysts in a blind evaluation ranged from negligible to large. Total length and head length were the most consistent measurements that also had negligible differences between image-based measurers. By contrast, pelvic–anal fin distance showed consistency between the two methods, but ranked as the metric with the highest between-analyst difference, indicating the measurement may be poorly reproduced by two digital analysts.

Surprisingly, paired fin measurements (PCL and pelvic fin length) varied despite these fins being features with a generally similar shape and size (Eshenroder et al. 2016). The PCL was the metric with the best reproducibility score between analysts, whereas pelvic fin length was the second least reproducible metric. Both fin measurements also showed different results between methods. The image-based measurements for PLC tended to be greater than those obtained by caliper, whereas pelvic fin length measurements were not significantly different between the methods. The observed variation might be affected by differences in fin placement against the body or level of pigment that likely affect the ability of a measurer to locate the distal ends of fins from an image. When a fin is laid against a light background, as shown for PCL in Figure 1, the distal end is easier to locate than in pelvic fin length, which has lighter pigment. In addition, its distal end is close to the ventral margin of the body, creating difficulty in distinguishing the end of the fin from the body. To better identify structures, researchers also could mark structures on the fish with dissecting pins or place small pieces of paper behind the ends of paired fins to increase contrast between the structure and its background.

Historical descriptions of Cisco are all via linear measurements made directly on specimens, well before advances in digital image technology made it possible to carry out image-based morphometrics. Representative specimens are deposited in museum collections and continue to be revisited in morphometric (Yule et al. 2013; Eshenroder et al. 2016) and ecological (Blanke et al. 2018; Schmidt et al. 2009, 2011) studies; however, the application of new methods to generate similar data with proper evaluation of comparability, and trade-offs with each, is typically overlooked in Cisco taxonomy studies. One advantage to a digital image approach when dealing with rare or conservation-focused species, where mortality is a concern, is to take a picture in the field and then live release the fish, rather than rely on direct caliper measurements that may not be possible to obtain. All fish examined in this study represented thawed specimens; however, other studies may present images of live-released fish (sensuMuir et al. 2012). Aside from measurement impracticality with a number of hand measurements on a live fish compared with a quick picture, there are also other trade-offs to consider. By using caliper measurements, returning a live fish also means the sample is no longer available should there be an error identified that requires redoing the original measurements. Alternatively, a picture can be revisited infinite times, is easily shareable, and facilitates assessment of measurement error when there are multiple measures, as we demonstrated in this study.

Adult-sized Cisco dominated our body-size range measurements among samples, which limited our ability to test for potential effects of body size on measurement variation. Adult size classes are typically used in Cisco morphometric studies, whereas juvenile Cisco morphometrics are far less explored in the literature. Future studies that incorporate a wider distribution of body sizes that encompass younger fish, and varying life stages, would be useful toward evaluating potential influences of size or life stage on measurement outcomes, going beyond our simple comparison that did not explicitly test for juvenile and adult differences.

We only measured a subset of 12 commonly used morphometrics for Cisco. Muir et al. (2013) included 24 linear measurements in a study on Cisco from Great Slave Lake (Northwest Territories, Canada), and Eshenroder et al. (2016) summarized 15 body features to describe different forms of ciscoes that included angular measurements and meristic counts. Although meristic counts on internal features such as gillrakers are not possible from a lateral-view fish image, an evaluation of additional metrics in lateral view beyond those considered here would be of interest, given that more than half of our morphometrics showed variation between measurement methods. Because many contemporary morphometric studies obtain direct caliper measurements and photographs of the specimen, researchers can relatively easily explore potential variation by adding a few comparison measurements. This approach may be an important step for quality assurance or outlier detection in early-stage analyses.

Our results identified morphometrics that can be reliably obtained from hand measurements and digital images, while also identifying those that we consider to be more suspect to bias. Findings from this study support an ongoing effort to refine techniques that will help resolve coregonine taxonomy from contemporary and historical collections (Svärdson 1949; Eshenroder et al. 2016, 2020; Ackiss et al. 2020). Our identification of reliable morphometrics that are consistently obtained with alternative methods should benefit situations that call for additional data collection to inform new analyses.

Supplemental Material

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. List of digital landmarks and descriptions of their placement along the body of Cisco Coregonus artedi collected from Lake Ontario in 2018. Landmarks were used to extract 12 image-based morphometrics. See Figure 1 for an image of a Cisco with digitally placed landmarks.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S1 (20 KB DOCX).

Table S2. List of definitions for morphometric abbreviations and start-end landmarks used to calculate distance between points for 12 image-based morphometrics from Cisco Coregonus artedi collected from Lake Ontario in 2018.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S2 (20 KB DOCX).

Table S3. Summary of one-sample test statistics comparing whether slope values differed from a test value of one for linear regressions of image-based morphometric measurements (performed by one analyst) as a function of specimen-based caliper measurements (performed by a different analyst) on Cisco Coregonus artedi collected from Lake Ontario in 2018. None of the P values were significant following Bonferroni correction (α = 0.05/12 = 0.0042).

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S3 (20 KB DOCX).

Reference S1. Bronte CR, Bunnell DB, David SR, Gordon R, Gorsky D, Millard MJ, Read J, Stein RA, Vaccaro L. 2017. Report from the Workshop on Coregonine Restoration Science. No. 2017-1081. U.S. Geological Survey.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S4 (903 KB PDF).

Reference S2. Eshenroder RL, Vecsei P, Mandrak NE, Yule DL, Gorman OT, Pratt TC, Bunnell DB, Muir AM. 2016. Ciscoes (Coregonus, subgenus Leucichthys) of the Laurentian Great Lakes and Lake Nipigon. Great Lakes Fishery Commission Miscellaneous Publication 2016-01.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S5 (32.53 MB PDF).

Reference S3. Holden JP, Connerton MJ, Weidel BC. 2019. Hydroacoustic and midwater trawl assessment of pelagic planktivores 2018 in New York State Department of Environmental Conservation, editor. Annual Report 2018, Bureau of Fisheries Lake Ontario Unit and St. Lawrence River Unit to the Great Lakes Fishery Commission's Lake Ontario Committee. Albany, New York: New York State Department of Environmental Conservation.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S6 (8.11 MB PDF); also available at https://www.dec.ny.gov/docs/fish_marine_pdf/lourpt18.pdf.

Reference S4. Koelz WN. 1929. Coregonid fishes of the Great Lakes. Bulletin of U.S. Bureau of Fisheries 43:297–643.

Found at DOI: https://doi.org/10.3996/JFWM-20-029.S7 (220.75 MB PDF).

Archived Material

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any archived material. Queries should be directed to the corresponding author for the article.

To cite this archived material, please cite both the journal article (formatting found in the Abstract section of this article) and the following recommended format for the archived material.

Smith CT, Reid SB, Godfrey L, Ardren WR. 2011. Data from: Gene flow among Modoc sucker and Sacramento sucker populations in the upper Pit River, Journal of Fish and Wildlife Management, 2(1):72–84. Archived in Dryad Digital Repository: https://doi.org/10.5061/dryad.8433.

Data A1. Data from this paper are publicly available at O'Malley BP, Schmitt JD, Holden JP, Weidel BC. 2020. Morphometric measurements of Cisco (Coregonus artedi) from Lake Ontario 2018: U.S. Geological Survey data release, https://doi.org/10.5066/P92B534W.

Acknowledgments

We thank Michael Connerton and Les Resseguie of the New York Department of Environmental Conservation; Taylor Brown of Cornell University, and several staff from the U.S. Geological Survey (USGS) and Ontario Ministry of Natural Resources and Forestry (OMNRF) that helped collect Cisco. We thank Chris Olds from the U.S. Fish and Wildlife Service and Andrew Muir and Randy Eshenroder from the Great Lakes Fishery Commission for providing morphometric training. We appreciate comments by Bryan Maitland, anonymous reviewers, and the Associate Editor that improved the quality of this manuscript. Funding support was provided by the Great Lakes Restoration Initiative Coordinated Science and Monitoring Initiative to USGS and provincial funding to implement OMNRF priorities under the Canada-Ontario Agreement on Great Lakes Water Quality and Ecosystem Health and Ontario's Great Lakes Strategy.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.

References

Ackiss
AS,
Larson
WA,
Stott
W.
2020
.
Genotyping-by-sequencing illuminates high levels of divergence among sympatric forms of coregonines in the Laurentian Great Lakes
.
Evolutionary Applications
13
:
1037
1054
.
Blanke
C,
Chikaraishi
Y,
Vander Zanden
MJ.
2018
.
Historical niche partitioning and long-term trophic shifts in Laurentian Great Lakes deepwater coregonines
.
Ecosphere
9
:
e02080
.
Bronte
CR,
Bunnell
DB,
David
SR,
Gordon
R,
Gorsky
D,
Millard
MJ,
Read
J,
Stein
RA,
Vaccaro
L.
2017
.
Report from the Workshop on Coregonine Restoration Science. No. 2017-1081. U.S. Geological Survey
(see Supplemental Material, Reference S1).
Cohen
J.
1992
.
A power primer
.
Psychological Bulletin
112
:
155
159
.
Collins
KS,
Gazley
MF.
2017
.
Does my posterior look big in this? The effect of photographic distortion on morphometric analyses
.
Paleobiology
43
:
508
520
.
Ebener
MP,
Kinnunen
RE,
Schneeberger
PJ,
Mohr
LC,
Hoyle
JA,
Peeters
P.
2008
.
Management of commercial fisheries for lake whitefish in the Laurentian Great Lakes of North America
.
Pages
99
143
in
Schechter
MG,
NJ,
Leonard
Taylor
WW,
editors.
International governance of fisheries ecosystems: learning from the past, finding solutions for the future
.
Bethesda, Maryland
:
American Fisheries Society
.
Eshenroder
RL,
Burnham Curtis
MK
.
1999
.
Species succession, and sustainability of the Great Lakes fishery
.
Pages
145
184
in
Taylor
W,
Ferreri
CP,
editors.
Great Lakes fishery policy, and management: a binational perspective
.
East Lansing, Michigan
:
Michigan State University Press
.
Eshenroder
RL,
Olds
CM,
Kao
YC,
Davis
CL,
Kinney
DN,
Muir
AM.
In press
.
Status of Cisco (Coregonus artedi) ecomorphs in Lake Huron, 1917–2016, with speculations about phenotypic plasticity in shorthead cisco
.
Advances in Limnology
.
Eshenroder
RL,
Vecsei
P,
Mandrak
NE,
Yule
DL,
Gorman
OT,
Pratt
TC,
Bunnell
DB,
Muir
AM.
2016
.
Ciscoes (Coregonus, subgenus Leucichthys) of the Laurentian Great Lakes and Lake Nipigon
.
Great Lakes Fishery Commission Miscellaneous Publication 2016-01
(see Supplemental Material, Reference S2).
Fruciano
C,
Schmidt
D,
Ramirez Sanchez
MM,
Morek
W,
Avila Valle
Z,
Talijancic
I,
Pecoraro
C,
Schermann Legionnet A
.
2020
.
Tissue preservation can affect geometric morphometric analyses: a case study using fish body shape
.
Zoological Journal of the Linnean Society
188
:
148
162
.
Holden
JP,
Connerton
MJ,
Weidel
BC.
2019
.
Hydroacoustic and midwater trawl assessment of pelagic planktivores 2018 in New York State Department of Environmental Conservation, editor. Annual Report 2018, Bureau of Fisheries Lake Ontario Unit and St. Lawrence River Unit to the Great Lakes Fishery Commission's Lake Ontario Committee
.
Albany, New York
:
New York State Department of Environmental Conservation (Supplemental Material, see Reference S3)
.
Jenkins
JA,
Bart
HL
Jr,
Bowker
JD,
Bowser
PR,
MacMillan
JR,
Nickum
JG,
Rose
JD,
Sorensen
PW,
Whitledge
GW,
Rachlin
J.
2014
.
Use of fishes in research committee (joint committee of the American fisheries society, the American institute of fishery research biologists, and the American society of Ichthyologists and Herpetologists). Guidelines for the use of fishes in research
.
Bethesda, Maryland
:
American Fisheries Society
.
Koelz
WN.
1929
.
Coregonid fishes of the Great Lakes
.
Bulletin of U.S. Bureau of Fisheries
43
:
297
643
(see Supplemental Material, Reference S4).
Muir
AM,
Vecsei
P,
Krueger
CC.
2012
.
A perspective on perspectives: methods to reduce variation in shape analysis of digital images
.
Transactions of the American Fisheries Society
141
:
1161
1170
.
Muir
AM,
Vecsei
P,
Pratt
TC,
Krueger
CC,
Power
M,
Reist
JD.
2013
.
Ontogenetic shifts in morphology and resource use of Cisco Coregonus artedi
.
Journal of Fish Biology
82
:
600
617
.
Olsen
AM,
Westneat
MW.
2015
.
StereoMorph: an R package for the collection of 3D landmarks and curves using a stereo camera set-up
.
Methods in Ecology and Evolution
6
:
351
356
.
Piette-Lauzière
G,
Bell
AH,
Ridgway
MS,
Turgeon
J.
2019
.
Evolution and diversity of two Cisco forms in an outlet of glacial Lake Algonquin
.
Ecology and Evolution
9
:
9654
9670
.
R Core Team.
2019
.
R: a language and environment for statistical computing
.
Vienna
:
R Foundation for Statistical Computing
.
Available: https://www.R-project.org/ (April 2021).
Rosinski
CL,
Vinson
MR,
Yule
DL.
2020
.
Niche partitioning among native ciscoes and nonnative rainbow smelt in Lake Superior
.
Transactions of the American Fisheries Society
149
:
184
203
.
Schmidt
SN,
Harvey
CJ,
Vander Zanden
MJ.
2011
.
Historical and contemporary trophic niche partitioning among Laurentian Great Lakes coregonines
.
Ecological Applications
21
:
888
896
.
Schmitt
JD,
Vandergoot
CS,
O'Malley
BP,
Kraus
RT.
2020
.
Does Lake Erie still have sufficient oxythermal habitat for Cisco Coregonus artedi?
Journal of Great Lakes Research
46
:
330
338
.
Schmidt
SN,
Vander Zanden
MJ,
Kitchell
JF.
2009
.
Long-term food web change in Lake Superior
.
Canadian Journal of Fisheries and Aquatic Sciences
66
:
2118
2129
.
Sierszen
ME,
Hrabik
TR,
Stockwell
JD,
Cotter
AM,
Hoffman
JC,
Yule
DL.
2014
.
Depth gradients in food-web processes linking habitats in large lakes: Lake Superior as an exemplar ecosystem
.
Freshwater Biology
59
:
2122
2136
.
Stockwell
JD,
Ebener
MP,
Black
JA,
Gorman
OT,
Hrabik
TR,
Kinnunen
RE,
Mattes
WP,
Oyadomari
JK,
Schram
ST,
Schreiner
DR,
Seider
MJ.
2009
.
A synthesis of Cisco recovery in Lake Superior: implications for native fish rehabilitation in the Laurentian Great Lakes
.
North American Journal of Fisheries Management
29
:
626
652
.
Svärdson,
G.
1949
.
The coregonid problem. I. Some general aspects of the problem
.
Report: Institute of Freshwater Research, Drottningholm
29
:
89
101
.
Warton
DI,
Duursma
RA,
Falster
DS,
Taskinen
S.
2012
.
Smatr 3–an R package for estimation and inference about allometric lines
.
Methods in Ecology and Evolution
3
:
257
259
.
Woodger
CD.
1976
.
Morphological variations as induced by environment in coregonids
.
Environmental Biology of Fishes
1
:
101
105
.
Yule
DL,
Moore
SA,
Ebener
MP,
Claramunt
RM,
Pratt
TC,
Salawater
LL,
Connerton
MJ.
2013
.
Morphometric variation among spawning cisco aggregations in the Laurentian Great Lakes: Are historic forms still present?
Advances in Limnology
64
:
119
132
.

The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.

Author notes

Citation: O'Malley BP, Schmitt JD, Holden JP, Weidel BC. 2021. Comparison of specimen- and image-based morphometrics for Cisco. Journal of Fish and Wildlife Management 12(1):208–215; e1944-687X. https://doi.org/10.3996/JFWM-20-029

Supplementary data