Digital analysis of the longest dinosaur trackway (CA6) from the Lower Jurassic (Hettangian–Sinemurian) Rotzo Formation, Coste dell’Anglone, Trentino-Alto Adige, Northern Italy

Digital tracksite analysis

The Coste dell’Anglone (CA) tracksite preserves at least 20 dinosaurian trackways on a single track-bearing horizon. A rock fall divides this surface into two sections: A1 being the larger section containing 13 trackways (CA1–13); and A2 being the smaller section with six trackways (CA14–20). The original mapping method was constructed using AutoCAD software after compiling documenting and drawing all tracks on acetate overlays, measuring them, and georeferencing them using GPS systems (Leica® 500 and Leica® 1200) and a Total Station (Leica® TC2002) (see Petti et al., 2011 for further details). The current study used the Coste dell’Anglone tracksite map was sourced from the published literature (Petti et al., 2011: Plate 1) and imported into the open-source software Blender (version 4.1). The image was scale-corrected using the map’s 20 m scalebar and oriented with north aligned to Blender’s +Y axis. The perimeter of the A1 and A2 track surface sections were traced to create a multi-vertex polyline, and their surface areas and perimeter lengths calculated using a custom-built Python script (available as a Blender add-ons from the SuperHive Market, https://superhivemarket.com/creators/paleotrack-solutions). The scaled and oriented map was then used to determine trackway parameters and trackmaker biometrics (see below).

Digital trackway analysis

A multi-vertex polyline was created in Blender to mark the track positions within the CA6 trackway. Vertices were placed sequentially along the trackway at the proximal-most portion of each track and was selected for vertex placement because it was the most consistently preserved feature of the footprints across the CA6 trackway (and thus a reliable reference point for trackway parameter measurements). These vertex positions then served as the location points of foot-to-ground contact for the animated virtual dinosaur-leg armature (see below). In those instances of missing tracks (i.e., tracks 30, 62, 116, and 117), their position was estimated as being equidistant between the adjacent footprints with a lateral offset of 0.10 m from the trackway midline (chosen as a value between the mean and median trackway width value as shown in Table 1). This offset was implemented in a custom Python script within Blender. This approach ensured continuity in the trackway and allowed for consistent analysis of the reconstructed sequence.

A custom Python script was developed to measure interfootprint distances and perform calculations required to derive trackway parameters and trackmaker biometric data, following methods previously described (Murphey et al., 2024; Romilio 2024; Romilio & Shao 2024). Length measurements measured the distance between the point at the start (x₁, y₁) and the end (x₂, y₂) of the respective trackway parameter using the Equation (1).

Pace angulation and step angle (the latter corresponding to the ‘angle of divergence’ sensu Thulborn, 1990) were determined based on the methods outlined by Thulborn (1990: Fig. 4.11). When a moving trackmaker reduces the width between its legs, pace angulation increases to 180º, while step angle decreases towards 0º. Pace angulation is defined as the angle between two successive pace lengths (Equation 2), whereas step angle is the angle formed between the pace and the stride (Equation 3), as illustrated by Thulborn (1990; Fig. 4.11)

Trackway width was considered the distance between footprints measured perpendicular to the stride orientation, using Equation (4).

For both pace angulation and trackway width, footprint symmetry was explicitly considered. Although left footprints are normally expected to lie to the left of the trackway midline (and right footprints to the right), this was not strictly assumed. Instances of crossover—where a left footprint occurs on the right side of the trackway midline, or vice versa—were identified as pace angulations greater than 180° and as negative trackway widths. Such crossovers occurrences were recorded and visualised as stride length versus trackway width.

Step length is defined as the distance between successive footprints, measured parallel to the stride, using Equation (5).

The trackway length is defined here as the total distance covered by the trackmaker along the trackway trajectory and was calculated by summing the step lengths.

To evaluate potential asymmetry in limb function for the CA6 trackmaker alternating (i.e., right versus left) pace, step, and stride lengths were each compared using a two-sample t-test. The average and standard deviation of each group of lengths were calculated, and a t-statistic and corresponding p-value were derived. A p-value of less than 0.05 was considered statistically significant, indicating asymmetry movement.

Digital Trackmaker Analysis

In ichnology, the standard approach in assessing trackmaker biometrics includes estimated hip height based on extrapolations from footprint length. To represent this, a line composed of two vertices was created in Blender, corresponding to a footprint length of 0.2133 m, based on the published average footprint length within the CA6 trackway (Petti et al., 2011: Table 1). Using a footprint multiplier of 4.5 (Thulborn, 1990), the trackmaker’s estimated hip height was calculated to be 0.96 m.

The velocity of the bipedal trackmaker was then estimated using a formula adapted by Ruiz & Torices (2013) from Alexander’s (1976) original equation (Equation 6):

When calculating step-velocity, the stride value in the velocity equation was replaced with twice the step length.

To estimate the duration the trackmaker spent on-site, the total trackway length was divided by the calculated average speed. Cadence (steps per unit time) was determined by dividing the speed by the step length, and the step duration (time per step) was obtained as the reciprocal of cadence.

Relative stride, used to determine the trackmaker’s gait, was calculated by dividing the stride length by hip height, following the method of Alexander (1976). A relative stride length less than 2 indicates a walking gait, values greater than 2.9 suggest running, and intermediate values represent trotting (i.e., jogging in bipeds). Trackmaker speeds at these gait transitions were calculated using the velocity equation, substituting values of 2 and 2.9 for stride length to model walk-to-trot and trot-to-run transitions, respectively.

Relative step, calculated as the step length divided by hip height, provided additional insight into gait. Values below 1 indicated walking, values above 1.45 indicated running, and intermediate values corresponded to jogging (equivalent to quadrupedal trotting).

To estimate the body mass of the theropodan trackmaker, the Equation (7) from Weems (2006) is used:

The body mass estimates were used to calculate the trackmaker’s terrestrial theoretical maximum speed by using Equation (8) from Hirt et al., (2017):

where M represents the body mass (kg), e is the base of the natural logarithm, and the constants 25.5 and 22 are derived from empirical scaling relationships.

Trackmaker biometrics

Trackmaker biometrics encompass a range of calculations derived from an inferred hip height of an unobserved trackmaker. These included relative stride, relative step, speed, cadence (step frequency), step duration, time on tracksite, and speed estimates.

The hip height of the CA6 trackmaker was estimated by multiplying the average footprint length of 0.2133 m (Petti et al., 2011) by a factor of 4.5 (Thulborn, 1990). The custom-built Python script developed to calculate the trackway parameters (see above) also concurrently estimated the trackmaker biometrics.

Tracksite parameters

Using the digital approach, the section A1 is estimated to be approximately 2,212 m² with a perimeter of approximately 263 m, and section A2 to be approximately 770 m² with a perimeter of approximately 125 m.

Trackway parameters

The digital analysis of the CA6 trackway (Fig. 1-4; Table 1) revealed a total trackway length of 78 meters, composed of 123 footprints, of which four are missing (i.e., ~97% complete), 122 paces, 122 steps, and 121 strides. The trackway’s orientation was predominantly south-southeast, averaging 164º (medium 163º, SD 10º) with values ranging from 133º to 188º relative to north (Fig. 1). Pace lengths varied from 0.519 m to 0.816 m, with both the average and median at 0.654 m and 0.651 (SD 0.053 m). Step lengths were of a similar range, ranging from 0.516 m to 0.793 m, with an average and median of 0.639 m and 0.637 (SD 0.052 m), respectively (Fig. 2A). Strides measured between 1.103 m and 1.500 m, with an average of 1.279 m and median 1.280 m (SD 0.064 m) (Fig. 2). Twelve crossover events (left footprint appears on the right side of the midline, and vice versa) were recorded within the CA6 trackway (at track positions 2, 7, 8, 14, 43, 60, 80, 81, 94, 104, 121, and 122) accounting for 9.9% of the total number of prints within the trackway. Trackway width (Fig. 3) fluctuated between -0.095 m to 0.404 m, with an average of 0.115 m (median 0.108 m, SD 0.079 m) and pace angulation (Fig. 4) ranged from 116º to 197º, with an average of 161º and median of 162º (SD approx. 13º).

Fig. 1 A line graph plots west to east versus north to south, and a polar plot with scatter dots in south east region.

Fig. 2 A scatter plot showing pace, step, and stride lengths along a trackway, and a Violin plot shows scattered dots in a region to show stride density.

Fig. 3 Three graphs: A) A scatter dot graph plots footprint sequence versus width. B) A violin plot, and C) A scatter dot graph plot the stride versus width.

Fig. 4 Two Graphs: A) A scatter dot graph plots footprint sequence versus degrees, and B) A violin plot with scatter dots shows angulation density.

Comparative analyses of pace and step lengths from the right and left sides of the trackway (Fig. 3) revealed minor, yet consistent, differences of a few centimetres that were statistically significant across the entire trackway. The average pace length from right-to-left was 0.643 m (SD = 0.055 m) was statistically different, to the left-to-right pace length of 0.665 m (SD = 0.049 m; t = -2.401, p = 0.018). In terms of step lengths, the right-to-left measurements averaged 0.627 m (SD = 0.052 m), while the left-to-right measurements averaged 0.651 m (SD = 0.049 m), with this difference also proving statistically significant (t = -2.513, p = 0.013). Conversely, the stride lengths between the right and left sides were not significantly different. The average stride lengths were 1.279 m (SD = 0.061 m) for right-to-right and 1.278 m (SD = 0.068 m) for left-to-left, with a negligible difference of one millimetre (t = 0.026, p = 0.979).

Trackmaker biometrics

The estimated biometrics for the CA6 trackmaker (Fig. 5; Table 2) were derived from an average footprint length of 0.2133 m (Petti et al., 2011: Table 1). Using a footprint multiplier of 4.5, the trackmaker’s estimated hip height was calculated to be 0.96 m.

Fig. 5 Two scatter dot plots show footprint sequence versus relative unit and footprint sequence versus velocity in kilometres per hour.

The relative stride averaged 1.33 (median 1.33, SD 0.07) which ranged from 1.15 to 1.56 across the trackway, consistent with a walking gait (values <2). Relative step lengths also indicate a walking trackmaker (values <1), averaging 0.67 (median 0.66, SD 0.05), and varying between 0.54 and 0.83. Utilising Alexander’s speed equation (Alexander, 1976), modified with a 0.226 multiplier (following Ruiz & Torices, 2013), the velocity based on stride ranged from 3.1 km/hr to 5.3 km/hr, averaging 4.0 km/hr (median 4.0 km/hr, SD 0.3 km/hr), while step-based velocity showed the same average at 4.0 km/hr (median 4.0 km/hr, SD 0.6 km/hr) but exhibited a wider range from 2.8 to 5.8 km/hr. Cadence varied from approximately 91 to 121 steps/min, with an average of 105 steps/min (median 105 steps/min, SD 6 steps/min). Step duration ranged from 0.50 s to 0.66 s, with an average of 0.57 s (median 0.57 s, SD 0.03 s).

Utilising Weems’ (2006) method to calculate body mass and the Hirt et al. (2017) method to estimate maximum speed, the CA6 trackmaker was estimated at approximately 93.6 kg, with a theoretical maximum speed of 63.4 km/hr (17.9 m/s). The calculated average velocity of 4.0 km/hr suggests that the CA6 trackmaker was moving, on average, at about 6.2% of its maximum speed. With a step-based speed range of 2.8 to 5.8 km/hr, the trackmaker may have moved between 4.4% and 9.1% of its theoretical maximum speed at the time of CA6 trackway registration.

Tracksite surface areas

A previous study by Petti et al. (2011) estimated the Coste dell’Anglone tracksite surface areas of sections A1 and A2 as approximately 2,600 m² and 700 m² (respectively). By applying this digital approach, these estimates can be refined, calculating the surface area of section A1 to be approximately 2,212 m² with a perimeter of 263 m, and section A2 to be 770 m² with a perimeter of 125 m. These refined values illustrate the capability of novel digital approaches to enhance the precision of surface area calculations and introduce perimeter measurements for dinosaur tracksites, particularly useful for those with irregular geometries.

Traditional methods for measuring tracksite surface areas typically rely on simplified length and width estimates, which can misrepresent the true dimensions of complex surfaces. Such methods, while simple and rapid to apply, lack the precision needed to fully capture the intricate geometries of track-bearing surfaces. In contrast, digital approaches enable users to trace the exact tracksite outline, allowing for more accurate and efficient calculations. This improved precision to be important for several reasons.

Firstly, accurate surface area measurements are essential for understanding the spatial extent of fossil tracksites, particularly when comparing tracksite characteristics across regions or stratigraphic intervals. Precise area measurements allow palaeontologists to assess the density of tracks more reliably (Murphey et al., 2024; Romilio, 2025).

Secondly, exact surface area data is crucial for monitoring the preservation and degradation of tracksites over time. The CA6 trackway comprises 123 footprints, of which four are missing (approximately 97% trackway completeness), so current inference about footprints within the trackway acknowledges both preserved and absent prints. Although this study is constrained to published mapping and includes no new in situ measurements, researchers with site access could repeatedly capture footprint outlines and calculate their surface areas using the method in Romilio. Those data would establish a baseline preserved area—either for the same footprint through time or as the summed area of all existing impressions—and any subsequent degradation would be detected as a decrease in measured preserved area.

Digital analysis of trackway parameters

Petti et al. (2011; Table 1) provided foundational data for all trackways at the Coste dell’Anglone tracksite. The current study provides additional datasets via re-analysing a single trackway from this site, albeit the longest. This approach not only validates the original findings but also highlights the potential of digital methods to uncover nuanced details and enhance the reproducibility of scientific analyses.

Petti et al. (2011) reported the CA6 trackway parameters as an orientation to the southeast at 164º, mean pace length of 0.7083 m, mean stride length of 1.3838 m, and pace angulation ranging from 122º to 180º. The results of the current study (Table 2) largely corroborate and expand on these findings, with an average orientation to the south-southeast of 164º, with a range between 133º and 188º. This study found the average pace length to be slightly smaller at 0.65 m, ranging from 0.53 m to 0.79 m, as with the stride length, with an average of 1.28 m with a range of 1.10 m to 1.50 m. Some differences from the current study may in part be attributed to the scaling of the original map data, underscoring the importance of digit reassessment of published maps that have included scalebars that are as large as possible, particularly for larger track-bearing surfaces with long trackways. Another notable difference is that the current study resolves a broader pace angulation range (116°–197°), including values exceeding the 180° upper bound reported by Petti et al. (2011: Table 1). This arises because crossover events are explicitly account for—instances where a footprint occurs on the other side of the trackway midline—which can produce pace angulations above 180°, whereas the original study apparently did not consider such occurrences. Interestingly, trackway crossovers are infrequently reported in the ichnological literature (Romilio & Salisbury 2014; Lallensack et al., 2016), but it is possible that this apparent scarcity may result from methodological choices in data processing rather than reflecting a genuinely rare trackmaker behaviour (pers. obs.).

The CA6 trackway, with 123 footprints, is approximately 97% complete, with only four prints missing (i.e., tracks 30, 62, 116, and 117). While trackway completeness has not been routinely quantified in ichnology, this straightforward metric provides a clear standard for assessing the reliability of trackway datasets. Higher completeness indicates greater reliability, making highly complete trackways particularly valuable for studying locomotor patterns. In contrast, trackways with significant gaps may require more cautious interpretations. Additionally, this metric could serve as a practical tool for monitoring site degradation and aiding conservation efforts by tracking changes in completeness over time.

Our findings also highlight the use of digital tools in identifying subtle but statistically significant locomotor asymmetries. By building on the foundation of high trackway completeness, enhanced digital analysis capabilities facilitate a detailed examination of sequential pace (and step) lengths. This analysis revealed a subtle, yet statistically significant, asymmetry between the right (mean = 0.643 m, SD = 0.055) and left (mean = 0.665 m, SD = 0.049) pace lengths (see Results for this consistent trend for step lengths). The computed t-statistic of -2.401 and the corresponding p-value of 0.018 verify the hypothesis that the observed difference, albeit minor, is statistically unlikely to have occurred by chance.

Despite visible overlaps in pace lengths, where individual right paces occasionally exceed those on the left, the overall statistical evidence supports a consistent difference, reflecting a possible intrinsic physiological condition of the trackmaker. For instance, the asymmetry could denote habitual behaviours, such as favouring one side over another (McCrea et al., 2015), although it may indicate compromised functionality in one limb, potentially suggestive of a limping condition due to injury (Lockley et al., 1994). Substrate factors are also considered: the palaeo-surface at the time of track registration was a marginal marine tidal flat with only a gentle westward inclination (Petti et al., 2011), so large-scale slope is unlikely to be the primary driver of the persistent left/right pace asymmetry. Nonetheless, small-scale heterogeneities in sediment consistency, water content, or microbial-mat stiffness could differentially affect foot penetration and withdrawal, introducing minor deviations in pace length. Such substrate-driven variability would more commonly produce scatter rather than the systematic left/right bias observed here, so while it may contribute marginally, the asymmetry is best interpreted as potentially multifactorial with physiological or behavioural causes being the primary cause.

An important aspect of this understanding is afforded by the novel digital methods that traditional methods might overlook; for example, they allowed detection of crossover in 9.9% of paces. Additionally, integrating statistical analyses may offer justification for evaluating insights into the biology of extinct trackmakers.

Digital Analysis of Trackmaker Biometrics

The large datasets provided in this study include traditional stride-based analyses such as relative stride (to determine gait) and speed measurements, as well as non-traditional step-based metrics such as relative step, step-based velocity, cadence, and step duration (Table 2) (Romilio & Shao 2024) that are considered here to provide a more nuanced understanding of trackmaker movements at each stage along the trackway.

By combining traditional and non-traditional approaches, a greater resolution in interpreting movement patterns and behavioural dynamics is achieved. While the CA6 trackmaker has been calculated to have maintained a consistent walking gait along the trackway, the study reveals variations within these parameters. Specifically, subtle changes in speed (associated with pace, step, and stride), straddle (trackway width, pace angulation, and step angle), and orientation over the length of the trackway can be identified.

Although average parameter values are often reported and remain important in trackway analyses, interpreting them without the context of variability can misleadingly suggest uniform behaviour by the trackmaker at the time of track registration (Romilio & Shao 2024). This issue becomes even more significant with longer trackways, as they provide more extensive datasets that reveal subtle variations and offer greater opportunities to insights into the dynamic and variable nature of the trackmaker’s locomotor behaviour. The findings here clearly demonstrate that such variability is present in the CA6 trackway.

Our digital approach also integrates estimations of body mass (Weems, 2006) in combination with theoretical maximum speed calculations (Hirt et al., 2017), where the CA6 trackmaker is predicted to have weighed 93.6 kg and having a top speed of 63.4 km/hr. While inherent uncertainties exist in applying such models to an unobservable extinct taxon, an aim is to provide a broader framework for understanding its potential locomotor capabilities.

Notably, the calculated average velocity of 4.0 km/hr represents approximately 6.2% of the trackmaker’s maximum theoretical speed, with speed variations ranging from 2.8 to 5.8 km/hr that equate to a range of between 4.4% and 9.1% of its theoretical maximum. These relatively low percentages of maximum speed are consistent with a steady walking gait, as opposed to more dynamic behaviours such as trotting or running or sprinting. This reinforces the interpretation of the CA6 trackway as a record of routine locomotion rather than high-performance activity as noted elsewhere (Weems 2006).