Measurement error in geometric morphometrics

Low resolution scans can provide a sufficiently accurate, cost- and time-effective alternative to high resolution scans for 3D shape analyses

Background

Advances in 3D shape capture technology have made powerful shape analyses, such as geometric morphometrics, more feasible. While the highly accurate micro-computed tomography (µCT) scanners have been the "gold standard," recent improvements in 3D surface scanners may make this technology a faster, portable, and cost-effective alternative. Several studies have already compared the two devices but all use relatively large specimens such as human crania. Here we perform shape analyses on Australia's smallest rodent to test whether a 3D scanner produces similar results to a µCT scanner.

Methods

We captured 19 delicate mouse (Pseudomys delicatulus) crania with a µCT scanner and a 3D scanner for geometric morphometrics. We ran multiple Procrustes ANOVAs to test how variation due to scan device compared to other sources such as biologically relevant variation and operator error. We quantified operator error as levels of variation and repeatability. Further, we tested if the two devices performed differently at classifying individuals based on sexual dimorphism. Finally, we inspected scatterplots of principal component analysis (PCA) scores for non-random patterns.

Results

In all Procrustes ANOVAs, regardless of factors included, differences between individuals contributed the most to total variation. The PCA plots reflect this in how the individuals are dispersed. Including only the symmetric component of shape increased the biological signal relative to variation due to device and due to error. 3D scans showed a higher level of operator error as evidenced by a greater spread of their replicates on the PCA, a higher level of multivariate variation, and a lower repeatability score. However, the 3D scan and µCT scan datasets performed identically in classifying individuals based on intra-specific patterns of sexual dimorphism.

Discussion

Compared to µCT scans, we find that even low resolution 3D scans of very small specimens are sufficiently accurate to classify intra-specific differences. We also make three recommendations for best use of low resolution data. First, we recommend that extreme caution should be taken when analyzing the asymmetric component of shape variation. Second, using 3D scans generates more random error due to increased landmarking difficulty, therefore users should be conservative in landmark choice and avoid multiple operators. Third, using 3D scans introduces a source of systematic error relative to µCT scans, therefore we recommend not combining them when possible, especially in studies expecting little biological variation. Our findings support increased use of low resolution 3D scans for most morphological studies; they are likely also applicable to low resolution scans of large specimens made in a medical CT scanner. As most vertebrates are relatively small, we anticipate our results will bolster more researchers in designing affordable large scale studies on small specimens with 3D surface scanners.

Procrustes-based geometric morphometrics on MRI images: An example of inter-operator bias in 3D landmarks and its impact on big datasets

Using 3D anatomical landmarks from adult human head MRIs, we assessed the magnitude of inter-operator differences in Procrustes-based geometric morphometric analyses. An in depth analysis of both absolute and relative error was performed in a subsample of individuals with replicated digitization by three different operators. The effect of inter-operator differences was also explored in a large sample of more than 900 individuals. Although absolute error was not unusual for MRI measurements, including bone landmarks, shape was particularly affected by differences among operators, with up to more than 30% of sample variation accounted for by this type of error. The magnitude of the bias was such that it dominated the main pattern of bone and total (all landmarks included) shape variation, largely surpassing the effect of sex differences between hundreds of men and women. In contrast, however, we found higher reproducibility in soft-tissue nasal landmarks, despite relatively larger errors in estimates of nasal size. Our study exemplifies the assessment of measurement error using geometric morphometrics on landmarks from MRIs and stresses the importance of relating it to total sample variance within the specific methodological framework being used. In summary, precise landmarks may not necessarily imply negligible errors, especially in shape data; indeed, size and shape may be differentially impacted by measurement error and different types of landmarks may have relatively larger or smaller errors. Importantly, and consistently with other recent studies using geometric morphometrics on digital images (which, however, were not specific to MRI data), this study showed that inter-operator biases can be a major source of error in the analysis of large samples, as those that are becoming increasingly common in the 'era of big data'.

Delineating modern variation from extinct morphology in the fossil record using shells of the Eastern Box Turtle (Terrapene carolina)

Characterization of morphological variation in the shells of extant Eastern Box Turtles, Terrapene carolina, provides a baseline for comparison to fossil populations. It also provides an example of the difficulties inherent to recognizing intraspecific diversity in the fossil record. The degree to which variation in fossils of T. carolina can be accommodated by extant variation in the species has been disagreed upon for over eighty years. Using morphometric analyses of the carapace, I address the relationship between modern and fossil T. carolina in terms of sexual dimorphism, geographic and subspecific variation, and allometric variation. Modern T. carolina display weak male-biased sexual size dimorphism. Sexual shape dimorphism cannot be reliably detected in the fossil record. Rather than a four-part subspecific division, patterns of geographic variation are more consistent with clinal variation between various regions in the species distribution. Allometric patterns are qualitatively similar to those documented in other emydid turtles and explain a significant amount of shape variation. When allometric patterns are accounted for, Holocene specimens are not significantly different from modern specimens. In contrast, several geologically older specimens have significantly different carapace shape with no modern analogue. Those large, fossilized specimens represent extinct variation occupying novel portions of morphospace. This study highlights the need for additional documentation of modern osteological variation that can be used to test hypotheses of intraspecific evolution in the fossil record.

2D or Not 2D? Testing the Utility of 2D Vs. 3D Landmark Data in Geometric Morphometrics of the Sculpin Subfamily Oligocottinae (Pisces; Cottoidea)

We contrast 2D vs. 3D landmark-based geometric morphometrics in the fish subfamily Oligocottinae by using 3D landmarks from CT-generated models and comparing the morphospace of the 3D landmarks to one based on 2D landmarks from images. The 2D and 3D shape variables capture common patterns across taxa, such that the pairwise Procrustes distances among taxa correspond and the trends captured by principal component analysis are similar in the xy plane. We use the two sets of landmarks to test several ecomorphological hypotheses from the literature. Both 2D and 3D data reject the hypothesis that head shape correlates significantly with the depth at which a species is commonly found. However, in taxa where shape variation in the z-axis is high, the 2D shape variables show sufficiently strong distortion to influence the outcome of the hypothesis tests regarding the relationship between mouth size and feeding ecology. Only the 3D data support previous studies which showed that large mouth sizes correlate positively with high percentages of elusive prey in the diet. When used to test for morphological divergence, 3D data show no evidence of divergence, while 2D data show that one clade of oligocottines has diverged from all others. This clade shows the greatest degree of z-axis body depth within Oligocottinae, and we conclude that the inability of the 2D approach to capture this lateral body depth causes the incongruence between 2D and 3D analyses. Anat Rec, 301:806-818, 2018. © 2017 Wiley Periodicals, Inc.

Hind wing variation in Leptura annularis complex among European and Asiatic populations (Coleoptera, Cerambycidae)

The ability to quantify morphological variation is essential for understanding the processes of species diversification. The geometric morphometrics approach allows reliable description of variation in animals, including insects. Here, this method was used to quantify the morphological variation among European and Asiatic populations of Lepturaannularis Fabricius, 1801 and its closely related species L.mimica Bates, 1884, endemic for Japan and Sakhalin islands. Since the taxonomic status of these two taxa is differently interpreted by taxonomists, they are collectively called “Lepturaannularis complex” in this paper. The analysis was based on the measurements of hind wings of 269 specimens from six populations from Europe and Asia. The level of morphological divergence between most of continental European and Asiatic populations was relatively small and proportional to the geographic distance between them. However, distinct morphotype was detected in Sakhalin Is. and Japan. These data confirm the morphological divergence of the endemic L.mimica species. Obtained results highlight the potential of the geometric morphometric method in studying morphological variation in beetles.

Morphological convergence in 'river dolphin' skulls

Convergent evolution can provide insights into the predictability of, and constraints on, the evolution of biodiversity. One striking example of convergence is seen in the ‘river dolphins’. The four dolphin genera that make up the ‘river dolphins’ (Inia geoffrensis, Pontoporia blainvillei, Platanista gangetica and Lipotes vexillifer) do not represent a single monophyletic group, despite being very similar in morphology. This has led many to using the ‘river dolphins’ as an example of convergent evolution. We investigate whether the skulls of the four ‘river dolphin’ genera are convergent when compared to other toothed dolphin taxa in addition to identifying convergent cranial and mandibular features. We use geometric morphometrics to uncover shape variation in the skulls of the ‘river dolphins’ and then apply a number of phylogenetic techniques to test for convergence. We find significant convergence in the skull morphology of the ‘river dolphins’. The four genera seem to have evolved similar skull shapes, leading to a convergent morphotype characterised by elongation of skull features. The cause of this morphological convergence remains unclear. However, the features we uncover as convergent, in particular elongation of the rostrum, support hypotheses of shared feeding mode or diet and thus provide the foundation for future work into convergence within the Odontoceti.

Can skull form predict the shape of the temporomandibular joint? A study using geometric morphometrics on the skulls of wolves and domestic dogs

The temporomandibular joint (TMJ) conducts and restrains masticatory movements between the mammalian cranium and the mandible. Through this functional integration, TMJ morphology in wild mammals is strongly correlated with diet, resulting in a wide range of TMJ variations. However, in artificially selected and closely related domestic dogs, dietary specialisations between breeds can be ruled out as a diversifying factor although they display an enormous variation in TMJ morphology. This raises the question of the origin of this variation. Here we hypothesise that, even in the face of reduced functional demands, TMJ shape in dogs can be predicted by skull form; i.e. that the TMJ is still highly integrated in the dog skull. If true, TMJ variation in the dog would be a plain by-product of the enormous cranial variation in dogs and its genetic causes. We addressed this hypothesis using geometric morphometry on a data set of 214 dog and 60 wolf skulls. We digitized 53 three-dimensional landmarks of the skull and the TMJ on CT-based segmentations and compared (1) the variation between domestic dog and wolf TMJs (via principal component analysis) and (2) the pattern of covariation of skull size, flexion and rostrum length with TMJ shape (via regression of centroid size on shape and partial least squares analyses). We show that the TMJ in domestic dogs is significantly more diverse than in wolves: its shape covaries significantly with skull size, flexion and rostrum proportions in patterns which resemble those observed in primates. Similar patterns in canids, which are carnivorous, and primates, which are mostly frugivorous imply the existence of basic TMJ integration patterns which are independent of dietary adaptations. However, only limited amounts of TMJ variation in dogs can be explained by simple covariation with overall skull geometry. This implies that the final TMJ shape is gained partially independently of the rest of the skull.

Sharing is caring? Measurement error and the issues arising from combining 3D morphometric datasets

Geometric morphometrics is routinely used in ecology and evolution and morphometric datasets are increasingly shared among researchers, allowing for more comprehensive studies and higher statistical power (as a consequence of increased sample size). However, sharing of morphometric data opens up the question of how much nonbiologically relevant variation (i.e., measurement error) is introduced in the resulting datasets and how this variation affects analyses. We perform a set of analyses based on an empirical 3D geometric morphometric dataset. In particular, we quantify the amount of error associated with combining data from multiple devices and digitized by multiple operators and test for the presence of bias. We also extend these analyses to a dataset obtained with a recently developed automated method, which does not require human-digitized landmarks. Further, we analyze how measurement error affects estimates of phylogenetic signal and how its effect compares with the effect of phylogenetic uncertainty. We show that measurement error can be substantial when combining surface models produced by different devices and even more among landmarks digitized by different operators. We also document the presence of small, but significant, amounts of nonrandom error (i.e., bias). Measurement error is heavily reduced by excluding landmarks that are difficult to digitize. The automated method we tested had low levels of error, if used in combination with a procedure for dimensionality reduction. Estimates of phylogenetic signal can be more affected by measurement error than by phylogenetic uncertainty. Our results generally highlight the importance of landmark choice and the usefulness of estimating measurement error. Further, measurement error may limit comparisons of estimates of phylogenetic signal across studies if these have been performed using different devices or by different operators. Finally, we also show how widely held assumptions do not always hold true, particularly that measurement error affects inference more at a shallower phylogenetic scale and that automated methods perform worse than human digitization.

Estimating the magnitude of morphoscapes: how to measure the morphological component of biodiversity in relation to habitats using geometric morphometrics

Ecological indicators are currently developed to account for the different facets of loss of biological diversity due to direct or indirect effects of human activities. Most ecological indicators include species richness as a metric. Others, such as functional traits and phylogenetic diversity, account for differences in species, even when species richness is the same. Here, we describe and apply a different indicator, called morphoscape dimension, accounting for morphological variability across habitats in a geographical region. We use the case of ground beetles (Coleoptera: Carabidae) in four different habitats in the Po Plain in Northern Italy to exemplify how to quantify the magnitude of the morphological space (i.e. the dimension of the morphoscape) occupied by the species in each habitat using geometric morphometrics. To this aim, we employed a variety of metrics of morphological disparity related to univariate size, and more complex multivariate shape and form. Our 'proof of concept' suggests that metrics assessing size and form might largely tend to simply mirror the information provided by species richness, whereas shape morphoscape disparity may be able to account for non-trivial differences in species traits amongst habitats. This is indicated by the woodland morphoscape being on average bigger than that of crops, the most species-rich habitat, despite having almost 20% less species. We conclude suggesting that the analysis of morphoscape dimension has the potential to become a new additional and complimentary tool in the hands of conservation biologists and ecologists to explore and quantify habitat complexity and inform decisions on management and conservation based on a wide set of ecological indicators.

Semi-supervised determination of pseudocryptic morphotypes using observer-free characterizations of anatomical alignment and shape

Accurate, quantitative characterization of complex shapes is recognized as a key methodological challenge in biology. Recent development of automated three‐dimensional geometric morphometric protocols (auto3dgm) provides a promising set of tools to help address this challenge. While auto3dgm has been shown to be useful in characterizing variation across clades of morphologically very distinct mammals, it has not been adequately tested in more problematic cases where pseudolandmark placement error potentially confounds interpretation of true shape variation. Here, we tested the sensitivity of auto3dgm to the degree of variation and various parameterization settings using a simulation and three microCT datasets that characterize mammal tooth crown morphology as biological examples. The microCT datasets vary in degree of apparent morphological differentiation, with two that include grossly similar morphospecies and one that includes two laboratory strains of a single species. Resulting alignments are highly sensitive to the number of pseudolandmarks used to quantify shapes. The degree to which the surfaces were downsampled and the apparent degree of morphological differentiation across the dataset also influenced alignment repeatability. We show that previous critiques of auto3dgm were based on poorly parameterized alignments and suggest that sample‐specific sensitivity analyses should be added to any research protocol including auto3dgm. Auto3dgm is a useful tool for studying samples when pseudolandmark placement error is small relative to the true differences between specimens. This method therefore represents a promising avenue forward in morphometric studies at a wide range of scales, from samples that differ by a single genetic locus to samples that represent multiple phylogenetically diverse clades.

Mega-evolutionary dynamics of the adaptive radiation of birds

The origin and expansion of biological diversity is regulated by both developmental trajectories and limits on available ecological niches. As lineages diversify, an early and often rapid phase of species and trait proliferation gives way to evolutionary slow-downs as new species pack into ever more densely occupied regions of ecological niche space. Small clades such as Darwin's finches demonstrate that natural selection is the driving force of adaptive radiations, but how microevolutionary processes scale up to shape the expansion of phenotypic diversity over much longer evolutionary timescales is unclear. Here we address this problem on a global scale by analysing a crowdsourced dataset of three-dimensional scanned bill morphology from more than 2,000 species. We find that bill diversity expanded early in extant avian evolutionary history, before transitioning to a phase dominated by packing of morphological space. However, this early phenotypic diversification is decoupled from temporal variation in evolutionary rate: rates of bill evolution vary among lineages but are comparatively stable through time. We find that rare, but major, discontinuities in phenotype emerge from rapid increases in rate along single branches, sometimes leading to depauperate clades with unusual bill morphologies. Despite these jumps between groups, the major axes of within-group bill-shape evolution are remarkably consistent across birds. We reveal that macroevolutionary processes underlying global-scale adaptive radiations support Darwinian and Simpsonian ideas of microevolution within adaptive zones and accelerated evolution between distinct adaptive peaks.

Genetic linkage of distinct adaptive traits in sympatrically speciating crater lake cichlid fish

Our understanding of how biological diversity arises is limited, especially in the case of speciation in the face of gene flow. Here we investigate the genomic basis of adaptive traits, focusing on a sympatrically diverging species pair of crater lake cichlid fishes. We identify the main quantitative trait loci (QTL) for two eco-morphological traits: body shape and pharyngeal jaw morphology. These traits diverge in parallel between benthic and limnetic species in the repeated adaptive radiations of this and other fish lineages. Remarkably, a single chromosomal region contains the highest effect size QTL for both traits. Transcriptomic data show that the QTL regions contain genes putatively under selection. Independent population genomic data corroborate QTL regions as areas of high differentiation between the sympatric sister species. Our results provide empirical support for current theoretical models that emphasize the importance of genetic linkage and pleiotropy in facilitating rapid divergence in sympatry.

Morphometric Wing Characters as a Tool for Mosquito Identification

Mosquitoes are responsible for the transmission of important infectious diseases, causing millions of deaths every year and endangering approximately 3 billion people around the world. As such, precise identification of mosquito species is crucial for an understanding of epidemiological patterns of disease transmission. Currently, the most common method of mosquito identification relies on morphological taxonomic keys, which do not always distinguish cryptic species. However, wing geometric morphometrics is a promising tool for the identification of vector mosquitoes, sibling and cryptic species included. This study therefore sought to accurately identify mosquito species from the three most epidemiologically important mosquito genera using wing morphometrics. Twelve mosquito species from three epidemiologically important genera (Aedes, Anopheles and Culex) were collected and identified by taxonomic keys. Next, the right wing of each adult female mosquito was removed and photographed, and the coordinates of eighteen digitized landmarks at the intersections of wing veins were collected. The allometric influence was assessed, and canonical variate analysis and thin-plate splines were used for species identification. Cross-validated reclassification tests were performed for each individual, and a Neighbor Joining tree was constructed to illustrate species segregation patterns. The analyses were carried out and the graphs plotted with TpsUtil 1.29, TpsRelw 1.39, MorphoJ 1.02 and Past 2.17c. Canonical variate analysis for Aedes, Anopheles and Culex genera showed three clear clusters in morphospace, correctly distinguishing the three mosquito genera, and pairwise cross-validated reclassification resulted in at least 99% accuracy; subgenera were also identified correctly with a mean accuracy of 96%, and in 88 of the 132 possible comparisons, species were identified with 100% accuracy after the data was subjected to reclassification. Our results showed that Aedes, Culex and Anopheles were correctly distinguished by wing shape. For the lower hierarchical levels (subgenera and species), wing geometric morphometrics was also efficient, resulting in high reclassification scores.