Kinematic characteristics of hoof landing in jumping horses at elite level
Reasons for performing study: Biomechanical events of the distal limb during the landing phase of a jump have been proposed to be risk factors for injury, indicating need for further characterisation of the landing and the hoof-surface interaction. This is essential also for valid testing of arena surfaces when simulating actual conditions.
Objectives: To investigate the hoof landing characteristics for different limbs of elite showjumping horses during the landing phase when jumping 1.30–1.50 m competition fences on 2 different arena surfaces.
Materials and methods: A single, fixed high speed camera (1000 frames/s) was aimed at landing spots after different fences during 2 competitions. A total of 64 hoof landings were recorded on one sand and one turf surface (using studs on the turf). Hoof movements were tracked from calibrated video sequences. Landing velocities, landing angles, maximal vertical and horizontal deceleration and timing of maximal deceleration peaks were calculated and compared between leading/trailing fore-/hindlimbs. All outcomes were analysed for limb, using generalised linear models and controlling for effects of surface and obstacle.
Results: Landing speed differed among limbs (P<0.02 for all speeds and models). The leading hoof approached the ground more acutely angled to the horizontal plane than the trailing comparing fore- (P<0.001) and hindlimbs (0.05≥P>0.01), respectively. Differences in landing and braking kinematics were also found between surfaces and obstacles; however, these effects were hard to separate because of the nonexperimental design.
Conclusions: The landing and braking characteristics of the hooves varied substantially between hind-, fore-, trailing and leading limbs. Developing mechanical testing devices for arena surfaces, this fairly wide range of biomechanical events must be considered, in order to simulate the horse-surface interaction.
Musculoskeletal injuries are common in horses (Egenvall et al. 2006), causing reduced welfare for the animals, societal costs and reduced performance in equestrian sports. This may be of particular concern in the jumping discipline, as the duration of competitive life in Warmblood horses has been shown to be shorter for jumping than for dressage (Ducro et al. 2009). Injuries to horses during international showjumping competitions in recent years have drawn attention to the need for competition arenas to meet demands not only concerning characteristics that enhance performance, but also characteristics that reduce the risk of injury. In showjumping, the horse's body is subjected to great physical challenges. A recent study shows that the conformation of the distal limb is important for the duration of the competitive life of the showjumper (Ducro et al. 2009). The hoof's interaction with the ground, and the forces affecting the tissues of the distal limbs as a result of it, is thought to be a crucial event in the occurrence of injuries. As suggested by Kane et al. (1996) the interaction between hoof and surface is the main factor relating track properties to limb pathology. Similar findings have been presented by Folman et al. (1986) concerning human subjects.
In many human sports, safety and performance of surfaces are tested by equipment that mimics player-surface interaction. In equine sports, however, it has proven difficult to construct a testing device that simulates true horse-surface interaction, partly due to scarcity of data. Peterson et al. (2008) constructed a track testing system that simulated the loads and loading rates in Thoroughbred racing. Hoof-surface interaction in the jumping horse needs to be further characterised in order to enable future surface testing for this discipline.
Kinematic studies of jumping horses have been made (Leach and Ormrod 1984; Thoulon 1991, Santamaría et al. 2002, 2004a,b, 2005; Cassiat et al. 2004; Catinaud 2005; Powers 2005), often with a limited number of animals. Hindlimb kinematics during push-off in the jump stride have been studied by van den Bogert et al. (1994), hoof placement and limb contact variables by Leach et al. (1984), Clayton and Barlow (1991), Deuel and Park (1991) and kinetics and joint moments have been studied by Schamhardt et al. (1993) and Meershoek et al. (2001). However, there is lack of detailed information on hoof landing velocities and deceleration during hoof braking on landing after a fence.
The purpose of this study was to investigate the hoof landing and hoof braking characteristics of the leading/trailing fore-/hindlimbs of elite showjumping horses that jump 1.30–1.50 m competition fences on 2 different surfaces using high-speed video. These characteristics are needed to perform mechanical simulations of hoof landings in order to assess quality of arena surfaces.
Materials and methods
High speed video recordings of the landing distal limbs of competing horses were made during one international (FEI CSI**) (Arena 1) and one national elite-level (Arena 2) showjumping meeting in Sweden during the outdoor season.
Approximately 200 horses were filmed. After selection based on the possibility for valid analysis, films of 39 Warmblood horses (18 mares, 17 geldings and 4 stallions) were used in the study. The mean ± s.d. age was 10.4 ± 2.3 years. From these 39 horses, 64 hoof landings were captured, from a total of 42 jumps. In 12 jumps 2 (n = 10) or 3 (n = 2) hooves were caught on film. Three horses appeared in 2 classes and contributed with 2 observations per class. This resulted in 2 within-limb repetitions. Each horse was ridden by only one rider.
Arena 1 was an outdoor sand arena originally constructed 30 years ago of 0–8 mm sand. It had been used extensively, mucked out irregularly and no recent analysis of composition performed. The base consisted of hardly packed fine-grained ground moraine. The arena was maintained by watering and dragging during the competition with a frequency of approximately every 50th competitor.
Arena 2 was an outdoor turf arena fundamentally rebuilt 10 years ago. The natural foundation was flattened during reconstruction and a layer of soil with a high sand content was put over it and broad-type grass strips rolled out on top. The arena was considered by the maintainer to have a quite shallow rooting system and therefore had not been sand dressed. At the time of the meeting it was quite wet after a few days of rain. To enable the horses to get a better grip they were all equipped with 2–4 large studs in each shoe. There was minimal manual maintenance of take-off and landing spots on the arena during the competition.
Fences were chosen based on accessibility since the camera needed to be placed outside the course but close enough to the landing spot to get a detailed picture of the distal limbs at landing. One upright 130 cm, one oxer 140 cm and 2 triple bars at 140 and 150 cm were included. See Table 1 for more detailed information about obstacles in the separate recording situations.
Table 1. An overview of the recording situations, number of observations and type of data gathered in each case
|Fence type and height (cm)||Oxer 140||Triple bar 150||Upright 130||Triple bar 140|
|Fence width (cm)||140||170||-||160|
|Fence number on course||10 A||9 A||9 C||6|
A high-speed digital camera (Fastec Imaging TroubleShooter 1000)1 with a recording rate of 1000 frames/s and a resolution of 640 × 480 pixels was aimed at the presumed landing spot after the chosen fence. The field of view (zoom) was adjusted to fit the landing area, including the main part of the metacarpus/metatarsus of the landing limbs. The height over ground of the camera aperture was 98 cm and horizontal distance to expected centre of landing area 320 cm. It was tilted approximately 17° downwards. Calibration was made by filming a folding ruler giving a vertical (y, later lateromedial position of horse) and horizontal (x, later lengthwise position of horse) measurement at 3 different distances from the camera in the region of interest. The compact flash card of the camera was used to export the data into a computer.
Data processing and analysis
Digital video files were imported to a software for markerless tracking (Qualisys Video Analysis)2. Four points on the hooves were marked and tracked by pattern recognition algorithms in an automatic or semiautomatic manner. The 4 points were placed on proximodorsal, distodorsal, proximopalmar/plantar, distopalmar/plantar positions on each hoof. Marker position output was on subpixel level. The routine of markerless tracking uses several larger areas to determine the point of interest, resulting in a resolution well below 0.5 pixels. The size of the field of view resulted in an x-scalefactor in the centre of the image of 2 mm/pixel giving a resolution of 1 mm. If the hooves during the video sequence dug deeply into the surface, the areas for pattern recognition of the distal markers were changed to more proximal parts on the hooves. However, the marker position remained over the original part, making the assumption of the hoof being a rigid structure. During inspection of the films, each hoof was registered as a leading forelimb, trailing forelimb, leading hindlimb or trailing hindlimb. The x/y position files obtained from the tracking software were imported into MatLab3. By using the ruler measurements at different distances from the camera a set of calibration factors were produced to accommodate for different landing positions (in depth, camera y-direction). The y-position of the hoof at impact determined the distance of the hoof from the camera and after appropriate calibration, output was produced in metric values. A Butterworth lowpass digital filter with cut-off frequency 200 Hz and of order 4 was then applied forward and backward. Above this frequency the power spectrum showed a constant low level.
The start of hoof impact was defined manually by scrutinising the video files and graphical plots for the first change in movement direction, rotation or speed of the hoof at first interaction with the ground. The best-fit rigid-body transformation for the hoof was calculated using the method of Söderkvist and Wedin (1993) (Matlab scripts based on scripts available at http://isbweb.org/software/movanal/kinemat/index.html) with x and y-values from above and the z-value fixed. Hoof position when coming to stand still during stance phase was used as reference for the rigid body movements. This resulted in 3 DOF (degrees of freedom) data, 2 translations and one rotation. Based on visual anatomical landmarks from the films the position for the centre of rotation of the hoof was determined. The offset between the measured centre of the hoof markers and the actual centre of rotation of the hoof were adjusted for, thereby giving true x, y-translations and rotation in the x-y plane for the hoof. Landing speeds (total, vertical and horizontal components) were calculated as an average from position data over 5 ms preimpact. The angle between the horizontal plane and the movement path of the landing hooves was determined by calculating a continuous slope (regression) over 6 consecutive data points, consecutively moving through each point of the data series. The average of 5 data points, −8 to −3 in relation to impact, was used to calculate the landing angle. From the moment of first contact with the ground to the end of hoof braking, the maximal value for vertical deceleration (Max Dec Vertical) and the maximal value of horizontal deceleration (Max Dec Horizontal) were calculated. The word deceleration is used synonymously for negative acceleration. The index values for peaks were recorded and the difference between them calculated, describing the temporal relation between horizontal and vertical deceleration (timing difference) where each index step is 1 ms. The kinematic variables used to describe hoof landing and braking were: total speed, vertical velocity, horizontal velocity, landing angle, maximal vertical deceleration and maximal horizontal deceleration and timing difference.
Generalised linear models were developed using the MIXED procedure (SAS4) using the 7 kinematic variables as outcome variables. The normal distributions of the data were determined with the Shapiro-Wilks test and if not normal the data were log-transformed (for the Timing difference first adding a constant to make the data have the same sign before logging and in case where the values were negative, the absolute values were used to log-transform the outcome [vertical speed]). If this did not achieve normal distribution the best fitting distribution within the ladder of powers (1/y2, 1/y, natural logarithm of y, the square root of y or y2) was sought and deemed subjectively by similar means and medians, a ‘small’ standard deviation and the best values of skewness and kurtosis. Fixed effect categorical variables were limb, arena and obstacle type. Obstacle type was simply the 4 combinations used (Table 1, ‘Fence type and height’). Fixed effect variables were kept when P<0.1 and considered significant when P<0.05 (in the modelling process keeping those with a univariable P value of 0.15 and further reducing when the multivariable P≥0.10). Obstacle/arena effects were modelled as nested and kept if P<0.10, if not then tried as separate effects. Two-way interactions between variables in the models were tested and kept if P<0.05. Horse was added as random effect. For transformed variables least square means and 95% confidence intervals were back transformed to the original scale. Pairwise comparisons of least square means were done as performed by SAS (option PDIFF).
For 2 of the outcome variables the data were normal; for 4 variables log transformation achieved normality according to Shapiro-Wilks test (for the variable timing difference after adding a constant (i.e. ‘adding’ 8 ms and then subtracting when back transforming) to make possible log transformation); and for 3 variables square root transformation was considered to yield the most normal result (Tables 2 and 3).
Table 2. Least square means and 95% confidence intervals (95% CI) (back transformed where appropriate) for hoof landing and hoof braking are developed from general linear models with horse as random effect, and limb, arena and obstacle as fixed effects, where the nested obstacle/arena effect was ≥0.10 (n = 64 hooves). Results are presented when group P values are <0.1 and pairwise comparisons only when group P values are <0.05
|Horizontal||Normal||Limb||LdF||3.3||(2.7–3.9)||<0.0001||–|| *** || *** || ** |
|velocity (m/s)|| || ||TrF||0.6||(-0.3–1.5)|| || *** ||–|| *** || *** |
| || || ||LdH||5.9||(5.0–6.8)|| || *** || *** ||–||NS|
| || || ||TrH||4.9||(4.2–5.6)|| || ** || *** ||NS||–|
|Landing||Exp. Log||Limb2||LdF||41.1||(36.7–46.0)||<0.0001||–|| *** || ** ||NS|
|angle (°)|| || ||TrF||74.9||(62.8–89.4)|| || *** ||–|| *** || *** |
| || || ||LdH||29.7||(25.0–35.4)|| || ** || *** ||–|| * |
| || || ||TrH||38.7||(33.9–44.1)|| ||NS|| *** || * ||–|
| || ||Arena||Sand arena||46.4||(42.2–51.0)||0.08|| || || || |
| || || ||Turf arena||40.6||(36.2–45.4)|| || || || || |
|Max Dec||Exp. Log||Obstacle||Oxer 140||790||(698–894)||0.09|| || || || |
|Vertical (m/s2)|| || ||Triple bar 150||844||(662–1077)|| || || || || |
| || || ||Triple bar 140||967||(786–1190)|| || || || || |
| || || ||Upright 130||663||(552–797)|| || || || || |
Table 3. Least square means and 95% confidence intervals (95% CI) (back transformed where appropriate) for hoof landing and hoof braking are developed from general linear models with horse as random effect, limb as fixed effect, and obstacle/arena as a fixed, nested effect (n = 64 hooves). Results are presented when group P values are <0.1 and pairwise comparisons only when group P values <0.05
|Exp. Log||Limb||LdF|| ||5.6||(5.2–6.0)||<0.0001||–|| ** || ** || *** || || || || |
| || || ||TrF|| ||4.4||(3.9–5.0)|| || ** ||–|| *** || *** || || || || |
| || || ||LdH|| ||6.8||(6.1–7.6)|| || ** || *** ||–||–|| || || || |
| || || ||TrH|| ||7.1||(6.5–7.8)|| || *** || *** ||–||–|| || || || |
| || ||Obstacle/arena|
|Oxer 140 (1)||Sand||6.2||(5.8–6.6)||0.009|| || || || ||–||–|| * || * |
| || || ||Triple bar 150 (2)||Sand||6.8||(5.9–7.7)|| || || || || ||–||–|| ** || * |
| || || ||Triple bar 140 (3)||Turf||5.3||(4.7–5.9)|| || || || || || * || ** ||–||–|
| || || ||Upright 130 (4)||Turf||5.4||(4.9–5.9)|| || || || || || * || * ||–||–|
|Limb||LdF|| ||4.3||(4.8–3.9)||0.016||–||–|| * ||–|| || || || |
| || || ||TrF|| ||4.2||(5.0–3.5)|| ||–||–||–|| *** || || || || |
| || || ||LdH|| ||3.4||(4.0–2.8)|| || * ||–||–||–|| || || || |
| || || ||TrH|| ||5.0||(5.6–4.4)|| ||–|| *** ||–||–|| || || || |
| || ||Obstacle/arena|
|Oxer 140 (1)||Sand||4.7||(5.1–4.3)||0.018|| || || || ||–||–|| * || * |
| || || ||Triple bar 150 (2)||Sand||4.8||(5.7–4.0)|| || || || || ||–||–|| ** || * |
| || || ||Triple bar 140 (3)||Turf||3.5||(4.2–2.9)|| || || || || || * || ** ||–||–|
| || || ||Upright 130 (4)||Turf||3.9||(4.5–3.4)|| || || || || || * || * ||–||–|
|Exp. Log||Limb||LdF|| ||460||(400–529)||<0.0001||–|| ** || ** || * || || || || |
| || || ||TrF|| ||292||(231–368)|| || ** ||–|| *** ||–|| || || || |
| || || ||LdH|| ||723||(586–894)|| || ** || *** ||–||–|| || || || |
| || || ||TrH|| ||585||(492–695)|| || * ||–||–||–|| || || || |
| || ||Obstacle/arena|
|Oxer 140 (1)||Sand||490||(431–556)||0.090|| || || || ||–||–||–||–|
| || || ||Triple bar 150 (2)||Sand||464||(360–597)|| || || || || ||–||–||–||–|
| || || ||Triple bar 140 (3)||Turf||608||(491–752)|| || || || || ||–||–||–||–|
| || || ||Upright 130 (4)||Turf||412||(343–496)|| || || || || ||–||–||–||–|
|Oxer 140 (1)||Sand||5.7||(3.7–7.8)||0.012|| || || || ||–|| * ||–||–|
| || || ||Triple bar 150 (2)||Sand||12.4||(7.7–17.7)|| || || || || || * ||–|| ** || ** |
| || || ||Triple bar 140 (3)||Turf||3.3||(0.3–6.7)|| || || || || ||–|| ** ||–||–|
| || || ||Upright 130 (4)||Turf||2.9||(0.3–5.8)|| || || || || ||–|| ** ||–||–|
Hoof landing characteristics
Mean total landing speed for the hooves were in the range of approximately 4–7 m/s, differing between limbs. The forelimbs approached the ground with lower total landing speeds than the hindlimbs (Table 3). Significant differences were seen in horizontal landing speed, being lowest in the trailing fore compared with the other limbs. Landing angles also differed significantly between the limbs. The leading limb approached the ground with a more acute angle (i.e. more shallow trajectory) towards the horizontal plane than the trailing limb (Table 2).
Hoof braking characteristics
No between-limb differences were found in maximal peaks for vertical deceleration. The values for maximal horizontal deceleration, on the other hand, showed that the hind hooves came to a more abrupt horizontal braking than the fore hooves (Table 3).
Arena and obstacle effects
The time difference between vertical and horizontal peak decelerations differed among arenas and obstacles, but the effects were not fully separable due to the study design. There were also differences in total landing speeds and vertical velocities between obstacle and surface combinations.
During the landing phase of a jump, the rotational movement around the horse's centre of gravity, produced by the hindlimb push at take-off, is reversed by the forelimb landing (Clayton and Barlow 1991). In this study, the different limbs showed individual characteristics in the landing. The total landing speed of the trailing forelimb differed from other limbs, being composed almost entirely of vertical movement. It had practically no (in some cases negative) horizontal velocity at landing (Table 2). This agrees with earlier results from Clayton and Barlow (1991) who showed that the angle of the metacarpus of the trailing fore was almost vertical in the landing. It also concurs with the studies from Schamhardt et al. (1993) and Meershoek et al. (2001) who showed that the horizontal retardatory ground reaction force in the trailing fore was very low. The trailing fore has been shown to carry the highest loads in the landing and have greater flexor joint moments of the coffin and the fetlock joints than the leading fore. It also has the highest flexor tendon load of all limbs (Meershoek et al. 2001).
The hindlimbs have an extended position in mid jump and then swing forward below the trunk of the horse as the horse approaches the ground in the landing. This might explain why the hind hooves were found to have high landing speed with high peaks of maximal horizontal breaking. The leading/trailing relationship was similar to the front hooves with the leading approaching the ground by a more acute angle and a higher horizontal velocity.
The main findings of this study concern limb differences in hoof landing and braking characteristics of jumping horses. Between-limb differences need to be taken into account when simulating the biomechanics of hoof surface interaction in tests of arena surface for this discipline.
Surface and obstacle effects
Due to the unbalanced, nonexperimental design of this study, the obstacle and surface findings are associated with uncertainty and should be regarded as indicative (e.g. only some interactions could be evaluated). These factors need to be further explored in the future.
It is, however, interesting to note that in the maximal horizontal deceleration only one almost significant effect of obstacle/arena is seen (admittedly borderline in the group P value and the confidence intervals just overlap): when an upright fence is compared with a triple on the same arena (turf in Table 3). It is logical to expect an obstacle effect when comparing the 2 most different (lengthwise) obstacles within one arena due to changes in jumping ballistics.
On the sand surface, the horizontal deceleration peak occurred later in relation to the vertical compared with timing of peak hoof braking on the turf surface. This is seen as a significant difference for Max Dec Horizontal between the combination of a triple bar 150 cm and sand in relation to the other combinations (Table 3). Our interpretation of this finding in relation to surface is that the sand surface was stiff but had a looser top layer and therefore allowed the hooves to slide easily at impact, but the surface only deflected slightly upon full loading while horizontal deceleration commenced earlier at the turf due to the studs. The shorter sliding of the hoof has earlier been shown to give less attenuation of peak decelerations and increase the rate of distal-proximal loading during hoof braking (Gustås et al. 2001).
The significant differences seen in the speed variables (Total and Vertical speed, Table 3) were all between arenas and not within arenas. Although not statistically possible to distinguish fully from obstacle effects it is still of interest to speculate on adaptation to different surfaces. The horse is suggested to adapt its movements somewhat in relation to the surface response. Burn and Usmar (2004) showed that alteration of track properties gave rise to a change in hoof landing velocities at the trot. The change was said to be due to a combination of the direct effect of the surface material and altered foot kinematics at impact.
There are several events in hoof landing and hoof braking characteristics that need further investigation. The surface deflection upon loading, especially penetration of the heels into the surface during loading of the limb, which in this study seemed more pronounced on the turf arena, warrants further investigation. This event is suggested by Crevier-Denoix et al. (2009) to be the main reason for the more progressive superficial digital flexor tendon loading and a relative delay in the events of the stance phase seen in both ground reaction force and pattern of tendon loading. A larger number of horses in different situations and improved technical conditions (better detail in footage and refined calibration techniques, possibly 3D data) is required to obtain more knowledge of the surface effect on the hoof landing in jumping horses. There are also other demanding events such as take-off and sharp turns that should be studied. The variation in relation to fence height, fence type, fence placement and arena characteristics also need further investigation.
Methodology and estimated errors
Quantification of hoof landing characteristics based on a single camera and markerless tracking were chosen for this study due to the challenges when gathering relevant data from competition situations. Recording events during competitions provides no means of interaction with the objects (i.e. no markers, no timing, no repetitions). Accordingly, a large number of horses were filmed and selection was strongly based on recordings that could be analysed. The criteria for selection were made on several factors such as lighting, obscuring and contrast. This resulted in over 400 files being reduced to about 70, which may have resulted in some selection bias, since horses landing far away from the presumed landing spot (in the periphery of the picture or out of focus) will have been excluded from the final analysis. However, we have no reason to believe that using the selected recordings would have overstated differences.
We excluded files where the hoof landed in the periphery of the field of view to minimise the effect of spherical aberration in the lens. The camera was tilted slightly towards the ground. This gave rise to a trigonometric error in the estimated height measurements. Since the measurements were concentrated to relative small movements of the hoof when in, or very close to in contact with the ground, these errors were considered negligible.
Corrections for out of plane movement were not made in this study. Plane of movement and viewing direction of camera were assumed to be perpendicular. In Situations 1 and 2 (Table 1) the riders had a long approach to the fence and the landing was between 2 obstacles at a straight line. In Situation 3 (Table 1) the obstacle was the last in a triple combination and the departure route towards the next obstacle a straight line. In the last recording situation (Table 1) the route to the fence was on a straight line and they turned around a fence about 15 m away from the landing with a straight line to the turning point. In all situations it was beneficial for the riders to approach and depart from the jump stride in a straight line. A 10° maximal deviation from the line of perpendicular movement to the camera view would produce an error of ≤1.5% (cos10°) in the horizontal coordinates.
The accuracy of reading the folding ruler (graded in mm) in the picture during calibration procedure was estimated to be ±5 mm, rendering an error in spatial terms of <2%. Summing up the errors, calculated and estimated, would in the worst case still be <5%, which would not affect the conclusions in this study.
Strengths and weaknesses of the study
The main weakness of this study is the unbalanced nonexperimental study design, which does not enable full separation of surface effects from obstacle effects. Further, the horses in this study are all considered to have an above average jumping technique and also experience from jumping training and competitions. Because earlier studies of Schamhardt et al. (1993) and Meershoek et al. (2001) suggest substantial differences in external loads in the landing between experienced and unexperienced horses, this should be considered when interpreting the results and making assumptions about other horse populations. Further, factors likely to contribute to kinematic and kinetic differences in the hoof surface interaction should be elucidated, i.e. the horse's technique, use of studs and characteristics of the surface.
The landing and braking characteristics of the hoofs varied substantially between hind-, fore-, trailing and leading limbs. This shows a complexity of the conditions with which the tissues of the limbs need to be/are able to cope. The different kinematic landing profile of each limb poses a challenge for the creation of standardised testing equipment that mimics true conditions of the hoof track interaction.
This work has been supported by grants from the Fédération Equestre Internationale (FEI), the World Horse Welfare and the Swedish foundation for Equine Research.
Conflicts of interest
The authors have declared no potential conflicts.
1 Fastec Imaging, San Diego, California, USA.
2 QUALISYS AB, Gothenburg, Sweden.
3 The Math Works Inc., Natick, Massachusetts, USA.
4 SAS Institute Inc., Cary, North Carolina, USA.