Reasons for performing study: Sandy beaches are often considered good training surfaces for trotter horses. However, their biomechanical effects on locomotion are insufficiently documented. Events at hoof impact have mostly been studied under laboratory conditions with accelerometers, but there is lack of data (acceleration, force, movement) on events occurring under every day practical conditions in the field.
Objectives: To investigate hoof landing and stride parameters on different tracks (from wet to dry) of a sand beach and on an asphalt road.
Methods: The right front hoof of 4 trotter horses was equipped with a triaxial accelerometer and a dynamometric horseshoe. Acceleration and force recordings (10 kHz) were synchronised with a high speed movie (600 Hz). Horses were driven on a sand beach where 3 tracks of decreasing water content had been delimited (from the sea to the shore): firm wet sand (FWS), deep wet sand (DWS) and deep dry sand (DDS). Firm wet sand and DWS were compared at 25 km/h and DDS compared to an asphalt road at 15 km/h. Recordings (10 strides) were randomly repeated 3 times. Statistical differences were tested using a GLM procedure (P<0.05).
Results: Main significant results were 1) a decrease in the amplitude of the vertical deceleration (and force) of the hoof during impact on a softer surface (about 59% between DWS and FWS and 95% between DDS and asphalt), 2) a decrease in the longitudinal braking deceleration (and force) on softer grounds (50% for DWS vs. FWS and 55% for DDS vs. asphalt), 3) a decrease in the stride length and an increase in the stride frequency on a softer surface.
Conclusions and clinical relevance: Drier sand surfaces reduce shock and impact forces during landing. For daily training, it should, however, be realised that improved damping characteristics are associated with a shorter stride length and a higher stride frequency.
The distal limb is subjected to repetitive shocks and vibrations, which have been suggested to be important sources of mechanical stress. Several studies (Radin et al. 1973; Serink et al. 1977) have demonstrated that large deceleration peaks in combination with high loading rates may contribute to subchondral bone damage, followed by degenerative changes in the joint. The amplitude of this deceleration peak depends on the nature of the 2 solid materials that collide. Among the external factors that may affect the amplitude of this shock, the type of ground surface is of great interest, because adaptation and selection of appropriate training surfaces can potentially be used to reduce those stresses. Trainers have empirically selected a wide variety of training surfaces but these choices are lacking scientific evidence. In France (considering the geographical proximity between a number of training centres and the sea), sand beaches are considered as very good training surfaces, particularly for trotter horses. However, their actual effects on locomotion are insufficiently documented.
Accelerometer studies of hoof landing have already been performed in several conditions. Barrey et al. (1991) have studied hoof impact at slow trot (4 m/s) on a number of different surfaces. They found that impact intensity was related to density and composition of the track. More recently, Gustas et al. (2001) have shown, using triaxial accelerometers mounted on the third metacarpal bone and the hoof, that the time lapse of the horizontal retardation of the hoof is also an important factor in the attenuation of the impact. This suggests that parameters that affect horizontal braking are important to the orthopaedic health of the horse. At different speeds, Gustas et al. (2006b) have documented that the absolute length of the hoof-braking period is between 30 and 50 ms and is independent of speed. The total period of concussion is approximately the period of muscle latency, which made the authors conclude that the events at concussion cannot be modulated by muscle contraction. At slow trot, Gustas et al. (2006a) have shown that the initial vertical deceleration of the hoof was higher on a sandpaper surface compared to a 1 cm layer of sand. These experiments need to be repeated at high speed, under common training conditions and on surfaces that are often used in practice for their potential benefit to the locomotor apparatus. For this purpose, measurements need to be done ‘in the field’, with instruments and protocols adapted for the measurement of biomechanical parameters outside the laboratory.
The aim of the present study was to investigate hoof landing and stride parameters on different tracks of a sand beach. For this purpose, these parameters were measured in horses trotting on 3 delimited lanes with different water content (from the sea to the shore) and compared to the results obtained on asphalt.
Materials and methods
Horses: Four French Trotter horses (2 geldings and 2 females, body mass: 550 ± 22 kg, age: 10 ± 7-years-old) were used.
Accelerometer: The right front hooves were equipped with one triaxial piezo-electric accelerometer (model 356B20, PCB1). This device (volume: 10.2 mm3, weight: 4 g, sensitivity: 0.1 mV/ms2, frequency range: 2–10,000 Hz) was rigidly fixed to the dorsal hoof wall by use of a metal hull screwed in the horn. The z axis of the accelerometer was aligned with the dorsal line of the hoof. Measurement of the hoof angle was performed to mathematically realign the z axis with the vertical (perpendicular to the sole) axis of the hoof and the x axis with the longitudinal (palmaro-dorsal) axis. The transversal (medio-lateral) Y axis was unchanged (Fig 1). Accelerations directed forward, lateral and downward were denoted positive. A bell-boot was attached to the other forelimb to compensate for weight asymmetry.
Dynamometric horseshoe: The right front hoof of each horse was also equipped with an instrumented shoe composed of 4 triaxial piezoelectric force sensors (Model 260A11, PCB1) sandwiched between 2 aluminium plates (Robin et al. 2009; Chateau et al. 2009b). The assembly between the instrumented shoe and hoof was achieved with a third thin (4 mm) aluminium shoe classically nailed into the hoof wall. Total weight of the dynamometric horseshoe (DHS) with sensors was 490 g and total height 22 mm. A noninstrumented horseshoe with matching height and weight was attached to the left front hoof. Measured forces were expressed in the same reference frame as the accelerometer data (positive × in the direction of the movement (palmaro-dorsal), positive y in the medio-lateral direction and positive z perpendicular to the solar plane of the shoe, pointing downwards). The coordinates of the centre of pressure (CoP) were computed as the point at which the moment of the vertical forces applied on each transducer equalled zero.
Speed measurement: The speed of the horse was recorded by use of a third, smaller, wheel equipped with a hub dynamo (Nabendynamo, Schmidt2) fixed behind the sulky. A digital speedometer (BC 506, Sigma3) was also used for real time control of the speed by the driver during the experiment.
High-frequency movie: A high-frequency camera (Phantom v5.1, Vision Research4) was fixed to a car following the right side of the horse during the experiment. The high-frequency movie (600 Hz) was synchronised with the acceleration and force measurements using a light-emitting diode (LED) signal that was synchronously recorded with the accelerometer and DHS signals.
Data acquisition: The wire was secured to the limb and connected to an analogue-to-digital converter (NI-USB 6218, NI5) plugged in a hardened computer (Toughbook CF-30, Panasonic6). The data acquisition devices were fitted in a box on the sulky (Fig 1). Data acquisition was performed at 10 kHz.
The instrumented horses were driven by the same experienced driver in a straight line on a sand beach (Varaville, Normandie, France) and on an asphalt road (Asph). Three tracks (from the sea to the shore) of the sand beach were selected and delimited with parallel and tightened cables (2.9 m apart). At the end of each experiment (performed on different days for each horse), 3 samples of sand were taken (to a depth of 15 cm) at the entrance, in the middle and at the exit of each of the 3 tracks. The water content of the 3 samples was then measured and averaged. According to these measurements, the 3 tracks were classified as: Firm Wet Sand (FWS) with a water content of 19.0 ± 0.8%, Deep Wet Sand (DWS) with a water content of 13.5 ± 3.7% and Deep Dry Sand (DDS) with a water content of 3.0 ± 0.6%. Recordings on each track were repeated 3 times and randomly alternated. Depending on the surface, 2 target speeds were selected. The 2 wet portions (FWS and DWS) were compared with a target speed set at 25 km/h. For the DDS and the asphalt road, the target speed was set at 15 km/h. These speeds were chosen for being comfortable for the horses and easily reproducible between the 2 compared surfaces.
For each trial, 10 consecutive strides were analysed. Hoof landing was investigated by evaluating events during the hoof braking phase which was considered as the period between the first contact of the hoof with the ground and its complete stabilisation. This hoof braking phase begun first with an impact spike and was then followed by a longitudinal sliding of the hoof. The acceleration and vertical force (vs. time) curves were used to determine: hoof first contact, peak of deceleration and force at impact spike, end of the hoof sliding phase (return to zero acceleration in the X direction). AccZimp and FZimp were, respectively, the peaks of vertical deceleration and force at impact spike (beginning of the braking phase). AccXbrake and FXbrake were, respectively, the peaks of longitudinal deceleration and force at the end of the sliding period (just before the immobilisation of the hoof). In this study we considered only the x (longitudinal) and z (vertical) components of the ground reaction force (GRF) and acceleration of the hoof.
Estimation of hoof velocity just before the first contact of the hoof and longitudinal sliding of the hoof during the braking phase was calculated by integration of the accelerometer signal. The constant of integration was obtained from the assumption that hoof velocity was zero at the end of the braking phase of the hoof (Burn 2006). Power spectral density of the accelerometer signal (during the phase of hoof braking) was calculated for each stride using a Fast Fourier Transform (Matlab, The MathWorks)7 (Gustas et al. 2006b). Stride length was computed multiplying stride duration by velocity. Data were calculated by averaging 120 strides on each surface (10 strides per trial and 3 trials per horse).
Data were analysed using the General Linear Model procedure in SAS (SAS Institute)8. The model included horse as a repeated effect and speed as a covariate to account for the effect of variations in speed on dependent variables. Least square mean differences were used for pairwise comparisons between surfaces (FWS vs. DWS at 25 km/h and DDS vs. asphalt at 15 km/h). Significance was set at P<0.05.
Stride and speed parameters
The speed of runs aimed for was 25 km/h (6.94 m/s) on FWS and DWS and 15 km/h (4.16 m/s) on DDS and asphalt. The mean of the measured speed for the 4 horses was, however, slightly more than the target speed on FWS (7.20 ± 0.53 m/s) and slightly less on DWS (6.65 ± 0.72 m/s). On DDS, speed was 4.64 ± 0.47 m/s and on asphalt 4.23 ± 0.30 m/s. Despite a slightly decreased speed on asphalt, the length of the stride was significantly increased on asphalt vs. DDS with about 5 cm (Table 1). In contrast, an increase in stride frequency was observed on sand (DDS) compared to asphalt. This phenomenon was also observed when comparing 2 firmer sand tracks with different water content. On FWS, despite a slight increase of speed, the stride frequency was decreased compared to DWS (Table 1). Stride length was on the other hand increased on FWS compared to DWS (about 37 cm).
Table 1. Mean (s.d.) (n = 120 strides) acceleration, velocity, displacement of the hoof, ground reaction force at impact and stride parameters for 4 horses trotting on 3 different tracks of a sand beach with decreasing water content: firm wet sand (FWS), deep wet sand (DWS), deep dry sand (DDS) and on an asphalt road (Asph)
Water content (%)
Vx Hoof (m/s)
Vz Hoof (m/s)
Braking duration (ms)
Hoof sliding (x) (cm)
Stance duration (ms)
Stride length (m)
Stride freq (Hz)
Results are significantly different between FSW and DWS (P<0.05).
# Results are significantly different between DDS and Asph (P<0.05). AccZimp and FZimp are, respectively, the first peaks of vertical deceleration and force at the beginning of the braking phase. Vx and Vz Hoof are, respectively, the longitudinal and vertical velocities of the hoof just before the collision with the surface. AccXbrake and FXbrake are, respectively, the peaks of longitudinal deceleration and force at the end of the sliding period (just before the immobilisation of the hoof).
During landing, the trajectory of the CoP started at the lateral side of the hoof before moving dorsally and medially from the lateral heel to the centre of the hoof (Fig 2). The peak of vertical deceleration of the hoof (impact spike) concurred with this first dorsal displacement of the CoP.
Just before hitting the ground, the hoof was moving forward and downward with a velocity ranging from 1.09 ± 0.87 m/s to 2.30 ± 0.61 m/s longitudinally and 1.15 ± 0.46 m/s to 2.62 ± 0.59 m/s vertically. Interestingly, at approximately the same speed (15 km/h), the vertical velocity of the hoof just before impact was greater on DDS compared to asphalt whereas longitudinal velocity was greater on asphalt compared to DDS (Table 1).
The deceleration peak of the hoof during impact was mainly vertical (Fig 3). Large differences were observed between the different tracks of the sand beach (Table 1). At approximately 25 km/h, the decrease (about 5.5%) in water content between the 2 sand tracks (DWS vs. FWS) resulted in a decrease (about 59%) in the vertical deceleration peak at impact. At approximately 15 km/h, this difference was increased (about 95%) when comparing the vertical deceleration measured on DDS with the one measured on an asphalt road (Table 1).
The deceleration peak occurred synchronously with a peak in vertical force. The amplitude of this peak was 3082 ± 820 N on FWS, reduced to 1442 ± 565 N on DWS and 2971 ± 870 N on asphalt, reduced to 1040 ± 385 N on DDS. The time at which this phenomenon occurred was quite variable between strides (standard deviation of 2.9% for an event occurring at 4.5% of the stance phase on asphalt). Considering the high frequency of this spike and the variability of its time of occurrence, the between-strides averaging process tended to level the peak, thus minimising the representation of this event on the mean plots (Fig 3).
Longitudinal braking of the hoof and sliding
Just after impact, the hoof continued to move forward before complete stabilisation. This period can be considered as a longitudinal braking phase during which the hoof is in full contact with the ground, slides forward and progressively decelerates. Whereas vertical deceleration of the hoof was rather sudden, longitudinal deceleration was a more gradual process with a peak of maximal (negative) longitudinal deceleration occurring between 15–20% of the stance phase (Fig 4). After this event, the hoof longitudinal acceleration returned to the baseline. The duration of this period of hoof braking (between impact and zero acceleration in the X direction) ranged from 32.6 ± 5.3 ms on FWS to 44.2 ± 12.8 ms on DWS. At approximately the same speed, longitudinal braking duration was shorter on FWS compared to DWS and shorter on Asphalt compared to DDS (Table 1).
The peak of maximal longitudinal deceleration (just before hoof immobilisation) was reduced by about 50.3% on DWS compared to FWS and by about 54.6% on DDS compared to Asphalt (Table 1). This peak of maximal longitudinal deceleration (just before hoof immobilisation) occurred simultaneously with a peak of longitudinal braking force (Fig 4). The amplitude of this maximal (longitudinal) braking force was reduced by about 57.7% on DWS compared to FWS and by about 58.9% on DDS compared to Asphalt (Table 1).
This period of progressive horizontal hoof deceleration from impact to zero-acceleration was characterised by a forward translation (sliding) of the hoof over the surface. Double integration of the longitudinal component of the accelerometer signal enabled estimation of the amplitude of this movement. Forward sliding of the hoof was not significantly different between FWS and DWS (respectively 4.6 ± 1.7 cm vs. 4.1 ± 2.0 cm). At about 15 km/h, forward sliding of the hoof was, however, significantly decreased on DDS (2.1 ± 1.4 cm) compared to Asphalt road (5.2 ± 3.2 cm).
During the period of hoof braking and longitudinal sliding, oscillations were observed in the accelerometer data. Those were most easily visible on the mean curve of acceleration measured on the Asphalt road (Fig 3). Analysis of the spectral density of this signal showed an increased magnitude of the power spectral density at higher frequencies (Fig 5) on Asphalt compared to DDS. Slight differences (increased magnitude and frequency) were also seen on FWS compared to DWS (Fig 5).
Combined analysis of acceleration and force exerted on the foot during the beginning of the stance phase is a useful tool for a better understanding of the mechanisms of landing. Because previous studies (Radin et al. 1973; Serink et al. 1977) have demonstrated the link between shock and vibration and subchondral bone damage, extrinsic factors (such as type of surface or shoeing) that can be influenced are of interest as potential means to reduce these phenomena. The analysis must be performed in the field, on everyday used surfaces. This precludes the use of a force-plate. Dynamometric horseshoes form an interesting alternative technique to solve this problem. Few studies (Ratzlaff et al. 1997; Robin et al. 2009; Setterbo et al. 2009) have, however, thus far used dynamometric horseshoes in the field to test the effect of various ground surfaces. The dynamometric horseshoe used in the present study is a custom-built device, validated earlier (Chateau et al. 2009b). Although it is lighter than a previous 3D dynamometric horseshoe described in the literature (Roland et al. 2005), it still has the drawback of being heavier and thicker than standard shoes used for trotter horses and this characteristic must be taken into account when interpreting the results. With this device, the 3D components of the GRF are expressed in a reference frame linked to the hoof itself, which has the advantage of keeping an anatomical meaning of the measurements, even if slight changes in hoof orientation occur on soft ground.
In the present study, accelerometer data were used in combination with GRF measurement. The relationship between force and acceleration (sum of external forces equals mass times acceleration) gives an indication of the involved mass during the first impact of the hoof, which was less than 5 kg. As already observed by Thomason and Peterson (2008), primary impact is characterised by a large acceleration and low forces. During this event, the hoof is acting as a passive mass that collides with the ground.
A previous study on the repeatability of measurements in a similar set-up (Chateau et al. 2009a) revealed wide variation in acceleration peaks between successive strides within the same trial, which have been also observed in other studies (Barrey et al. 1991; Gustas et al. 2004, 2006a,b, Ratzlaff et al. 2005). For horses of similar body mass, large variations between horses in the amplitude of the shock were also observed, suggesting that the individual locomotion pattern of each horse heavily influences the dynamics of the foot during impact. This is in line with the suggestion by Burn and Usmar (2005) that, although the initial shock might be principally determined by the direct effect of the track surface material decelerating the foot, it is also possible that the dynamics of locomotion are altered too, thus affecting the landing velocity of the foot. They demonstrated that hoof landing velocity is affected by changing the properties of the surface such that the more deformable surfaces elicited a higher hoof landing velocity. The results of the present study show that the relative contribution of the vertical and horizontal components of the hoof velocity just before impact is also influenced by the characteristics of the surface. On asphalt, the horizontal component of the hoof velocity predominates whereas vertical velocity is reduced. The opposite is seen on deformable dry sand (predominant vertical velocity). Concurrently, longitudinal sliding of the hoof was increased on asphalt compared to DDS, which can be interpreted as a combined effect of an increase in hoof longitudinal velocity and the more slippery nature of asphalt vs. sand. A change in hoof velocity direction at impact also goes with alteration of the stride characteristic. On asphalt, stride frequency was decreased and stride length was increased compared to DDS, suggesting that the horse modifies its locomotor pattern to minimise vertical impact of the hoof and reduce the amplitude of vertical deceleration by compensation in the longitudinal direction. Gustas et al. (2006b) have pointed out that the absolute length of the hoof-braking period is between 30 and 50 ms, which corroborates our results. They concluded that the events at concussion can therefore not be modulated by muscle contraction. Nevertheless, it should be stressed that anticipation and changes in the locomotor pattern that alter the way the hoof is landing can modify hoof-ground interaction at the beginning of the stance phase. It can, therefore, be hypothesised that on irregular, inhomogeneous or unpredictable surfaces, the mechanism of anticipation might be affected, which could lead to high and injurious stress levels.
As expected, hardness of the track and percentage of water content in the sand tracks have a major effect on the magnitude of the vertical deceleration peak at impact. The decrease in the peak vertical deceleration (about 95%) and force on DDS compared to asphalt is very high. This decrease is still clearly significant (about 59%) when comparing 2 sand tracks (DWS vs. FWS) in which the difference in water content is only 5.5%.
Deceleration at impact spike is characterised by a high vertical peak, with considerably smaller horizontal decelerations. As already pointed out by Thomason and Peterson (2008), the horizontal motion of the hoof is not braked as instantly as the vertical motion, especially on hard ground surfaces. In our experimental setting (harness trotter), motion of the hoof ceased within 32.6–44.2 ms. During this period, the hoof is still moving forward. The amplitude of this longitudinal sliding of the hoof and the duration of the longitudinal braking phase were dependent on the type of surface. At the end of this period, longitudinal deceleration of the hoof is maximal, the hoof comes to a halt and the longitudinal braking force reaches a maximum that tends to brake the forward motion of the body while the hoof is not sliding anymore. On softer grounds (DDS compared to Asphalt or DWS compared to FWS), maximal longitudinal deceleration and braking force are decreased, suggesting a more gradual transition between forward sliding and hoof stop. As proposed by Johnston and Back (2006), allowing a natural time course for horizontal braking of the hoof seems to be important in preventing injury. We indeed observed increased longitudinal deceleration and braking force when the duration of the hoof braking period decreased. The same effects (higher longitudinal deceleration and braking force) have been measured on other kinds of surfaces, such as crushed sand tracks compared to softer all-weather waxed tracks in trotter horses (Robin et al. 2009; Chateau et al. 2009a) but this effect seems to be less obvious during cantering exercise (Setterbo et al. 2009).
High-frequency vibrations (up to about 1200 Hz on asphalt) during this deceleration phase may be evoked by the hoof grinding over the surface. As expected, those vibrations are drastically reduced on sand tracks.
The present results indicate that dryer sand surfaces reduce the amplitude of the shock and impact force in both vertical and horizontal direction during landing. These observations suggest that the distal limb is subjected to reduced mechanical stress during the initial part of the stance phase on drier areas of a sand beach compared to areas closer to the sea. For practical daily training of trotter horses, it should, however, be realised that drier and more penetrable surfaces induce, at the same speed, a decrease in stride length and an increase in stride frequency. In fact, improved damping characteristics are often associated with a loss of efficiency during the propulsion phase. This study focuses on landing but it is obvious that the analysis of the remainder of the stance phase (Crevier-Denoix et al. 2010) is needed for a comprehensive understanding of the effects of differences in water contents of sand tracks on the locomotor apparatus.
The authors thank the Conseil Régional de Basse-Normandie, the Haras Nationaux and the Fonds Unique Interministériel for their financial support to this project, and the Pôle de compétitivité Filière Equine for their technical assistance.
The authors also thank the farriers of the Haras du Pin and J. Jecker for their contribution, as well as the City Hall of Varaville.