Reasons for performing study: Although beach training is commonly used in horses, limb loading on beach sand has never been investigated. A dynamometric horseshoe (DHS) is well adapted for this purpose.
Objectives: To compare ground reaction force (GRF) and fetlock kinematics measured in harness trotters on 2 tracks of beach sand with different water content.
Methods: Two linear sand tracks were compared: firm wet sand (FWS, 19% moisture) vs. deep wet sand (DWS, 13.5% moisture). Four French trotters (550 ± 22 kg) were used. Their right forelimb was equipped with a DHS and skin markers. Each track was tested 3 times at 7 m/s. Each trial was filmed by a high-speed camera (600 Hz); DHS and speed data acquisition was performed at 10 kHz on 10 consecutive strides. All recordings were synchronised. The components Fx (parallel to the hoof solar surface) and Fz (perpendicular) of the GRF were considered. For 3 horses the fetlock angle and forelimb axis-track angle at landing were measured. Statistical differences were tested using the GLM procedure (SAS; P<0.05).
Results: Stance duration was increased on DWS compared to FWS. Fzmax and Fxmax (oriented, respectively, downwards and forwards relatively to the solar surface) and the corresponding loading rates, were decreased on DWS and these force peaks occurred later. Fxmin (backwards) was not significantly different between both surfaces; the propulsive phase (Fx negative) was longer and the corresponding impulse higher, on DWS compared to FWS. The forelimb was more oblique to the track at landing and maximal fetlock extension was less and delayed on DWS.
Conclusions: This study confirms that trotting on deep sand overall reduces maximal GRF and induces a more progressive limb loading. However, it increases the propulsive effort and likely superficial digital flexor tendon tension at the end of stance, which should be taken into account in beach training.
Beach training is often used in race and sport horses, either for regular training or rehabilitation. Although horses are most commonly exercised along the water edge, beaches offer surfaces with a variety of dynamic responses according to their water content. It is well known that ground surface material can modify the load transfer from the ground to the hoof and therefore the propagation of forces to limb bones, tendons and ligaments (Cheney et al. 1973; Thomason and Peterson 2008; Crevier-Denoix et al. 2009). Soft sand running has been recommended as a useful rehabilitative exercise in man since impact forces are reduced in compliant (soft) surfaces and muscle activation strategies to provide stability are emphasised (McMahon and Greene 1979). However, compliant surfaces have also been shown to result in increased eccentric muscle activity in the propulsive muscles, generating shin pain (Richie et al. 1993).
In spite of their potential interest for training or rehabilitation, the dynamic effects of beach sand surfaces on the equine locomotor system, such as the forces and loading rates applied to the limbs, have not been studied to date. Indeed, measuring the ground reaction force (GRF) in a horse moving at training speed under outdoors conditions is a technological challenge. Force platforms, that have to be sealed in the ground in a rigid frame, are not adapted to these measurement conditions, especially when several surfaces are to be compared. A dynamometric device mounted to the horse is a preferable alternative. Several models of dynamometric horseshoe (DHS) have been described in the literature most of them using uniaxial force sensors (Frederick and Henderson 1970; Barrey 1990; Ratzlaff et al. 1990; Roepstorff and Drevemo 1993; Kai et al. 2000). To date, to the authors' knowledge only two 3-dimensional (3D) DHS models have been developed and applied in equine biomechanical studies: based on strain gauge transducers (Roland et al. 2005) and piezoelectric sensors (Chateau et al. 2009). After experimental validation, both have been used to compare the effects of track surfaces on the 3D GRF, either in Thoroughbred racehorses (Setterbo et al. 2009) or in French trotters (Robin et al. 2009) under outside training conditions. One of the main advantages of the piezoelectric technology based DHS is its weight (490 vs. 860 g for the strain gauges device).
In man, it is well known that walking or running on sand requires a greater effort than on firm ground. Zamparo et al. (1992) suggested that the increase in energy cost when running on soft sand (compared with a firm terrain) was mainly attributable to a reduced capacity of the runner to store and recover elastic energy. Lejeune et al. (1998) demonstrated that the increase in energy cost during walking and running on sand was primarily due to an increase in the mechanical work (force × displacement) done on the sand, especially during the propulsion phase, and to a decrease in the efficiency of positive work done by the muscles and tendons. Pinnington et al. (2005) demonstrated that running on sand resulted in a greater stance to stride duration, a shorter stride length and a greater stride frequency compared with the firm surface values.
Given this scientific background, the objective of the present study was to compare, through GRF analysis and forelimb kinematics, the loading of the forelimb on 2 tracks of beach sand with different water content in harness trotters, under training conditions. It was hypothesised that GRF would decrease as the compliance of the surface increases and that deep sand would require a larger propulsion effort compared to a firmer surface.
Materials and methods
Two tracks of sand 100 m long were delimited on the Varaville beach (Normandy, France), using for each 2 parallel and tightened cables (2.9 m apart), equipped in their middle part with pairs of markers placed every meter. A firm wet sand (FWS) track was defined as parallel and as close as possible to the sea; a deep wet sand (DWS) track was located parallel to and at a distance of about 20 m from the FWS. At the end of the experiment, 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 2 tracks and hermetically stored. In the laboratory, the initial wet weight of each sample was measured (Talent TE210, Sartorius)1 before oven-drying it for 48 h at 105°C; then, the corresponding dry weight was measured. Both values were then used in the following equation: . The water contents of the 3 samples were averaged.
Two females and 2 geldings French trotters (mean ± s.d. body weight: 550 ± 22 kg; age 10 ± 7 years) were used. All horses were clinically sound, with no subjectively observed gait abnormalities.
After trimming of both front hooves by an experienced farrier, the right front hoof of each horse was equipped with an instrumented shoe composed of 4 triaxial piezoelectric force sensors (Model 260A11, PCB)2 sandwiched between 2 aluminium plates (Chateau et al. 2009). The assembly between the instrumented shoe and hoof was achieved with a third thin (4 mm) aluminium shoe classically nailed to the hoof wall. This support shoe and the instrumented system were fixed together by means of 3 bolts. Total weight of the device with sensors was 490 g and total height 22 mm. A noninstrumented horseshoe with equal height and weight was fixed to the left front hoof. The transducers were laid out along the horseshoe profile, symmetrically oriented relatively to the sagittal plane of the hoof, 2 towards the toe and the other 2 towards the heels. The origin of the coordinate system was at the centre of gravity of the 4 transducers in the sagittal plane; the associated reference frame was defined as follows: positive in the palmaro-dorsal direction, positive y in the medio-lateral direction towards the outside of the hoof and positive z was perpendicular to the shoe plane, directed downwards. The GRF was calculated as the sum of forces applied on each sensor. In this study we considered only the x and z components of the GRF, designated Fx and Fz, respectively, parallel and perpendicular to the solar surface of the hoof in the sagittal plane (Fig 1a).
The DHS was connected via wires, secured to the limb, to an analogue-to-digital converter (NI-USB 6218)3 plugged into a rugged computer (Toughbook CF-30, Panasonic)4. The data acquisition devices were placed in a box fixed on the sulky (Fig 1b). For each trial, data acquisition was performed at 10 kHz.
The speed of the horse was measured and recorded by a third, smaller, wheel equipped with a hub dynamo (Nabendynamo, Schmidt)5 fixed behind the sulky (Fig 1b). The speed was also controlled in real time by the driver by means of a digital speedometer (BC 506, Sigma)6.
During the tests the 4 horses were filmed with a high-speed camera (600 Hz, Phantom v5.1, Vision Research)7 mounted to a vehicle following the right side of the horse at a distance of about 7 m. The resolution of the camera was 1024 × 640 pixels and the field of view was approximately 4.5 m long by 2.8 m high. The high frequency movies were synchronised with the DHS and speed data using the lighting of a LED (Light-Emitting Diod) placed on the sulky right branch, the signal of which was retrieved on the same acquisition card as the other devices. Seven reflective markers were placed on the right forelimb, indicating the main limb joints and the hoof (Fig 1a).
The 4 horses were led by the same experienced driver. Recordings on each surface (tested on a straight line) were randomly repeated 3 times. The speed aimed at was 25 km/h (about 7 m/s), i.e. a speed that could be theoretically reached easily by all horses even on the deep sand (DWS).
Ten consecutive strides were analysed for each trial. From the DHS recordings, the 3D GRF was calculated. Temporal parameters of the stride were calculated using the vertical GRF data. To delimit the stance phase, the threshold was set to 50 N and the frames designed as first contact and toe-off, were, respectively, the frames just before and after this value. The stride length was calculated as stride duration multiplied by the horse's speed. Custom software8 was used to calculate peak forces, impulses (integral of force over time), stride frequency and temporal parameters of each force peak (beside the early impact peaks, there were 1 peak per stance in Fz: Fz max, and 2 in Fx: Fx max and Fx min; see Fig 2). The overall loading rate was calculated for Fz max and Fx max by dividing peak force (in Newton) by peak force time (in seconds). DHS and speed data were calculated by averaging 120 strides on each surface (10 strides per trial, 3 trials per horse).
In addition, for 3 horses and for at least 2 trials for each surface, the fetlock dorsal angle was calculated from the 2D coordinates of the markers placed on the carpus, fetlock and hoof. The orientation of the elbow-fetlock axis to the track was also determined at landing, using the markers placed every meter along the 2 tightened cables delimiting each track. To eliminate parallax errors, this angle was calculated using the 2D DLT (Direct Linear Transformation) technique (Abdel-Aziz and Karara 1971). Kinematic data were calculated by averaging 70 strides on each surface (10 strides/trial, 2–3 trials/horse).
Data were analysed by the General Linear Model procedure in SAS (SAS v.9.2)9. The models included horse as a repeated effect and speed as a covariate. Least square means differences were used for pair wise comparisons between track surfaces. These differences took into account potential effects of speed and/or horse on each dependent variable since speed and horse variables were included in the model. Significance was set at P<0.05.
Water content was 19% for the FWS vs. 13.5% for the DWS (Table 1). For Horse No. 4, DWS water content was higher than for the other horses; therefore the water content difference between FWS and DWS was less for this horse than for the other 3 (2% vs. about 7%).
Table 1. Least squares (LS) mean and standard error (s.e.) of stride parameters, ground reaction force and kinematic variables (adjusted for speed) for 4 horses (3 for the kinematic variables) trotting on 2 beach sand tracks with different water content: firm wet sand (FWS) and deep wet sand (DWS). Statistical comparison between tracks is based on LS means; the observed means (and standard deviations, s.d.) are only mentioned here for a descriptive purpose
Water content (%)
Observed mean (s.d.)
Observed mean (s.d.)
Except , all differences were significantly different between FWS and DWS (P<0.05). % sta.d.: % of stance duration.
In spite of the driver's attempt to reach exactly the same speed on the 2 surfaces, the 4 horses were in average slightly slower on DWS (mean ± s.d.: 6.65 ± 0.72 m/s) compared to FWS (7.21 ± 0.53 m/s). However, the statistical comparison between tracks (Table 1 and description below) is based on the least squares means that take in account these speed variations.
Stride frequency was increased on DWS compared to FWS (+2%, P<0.0001), whereas stride length was reduced (−2%, P<0.0001). Stance duration was increased on DWS (+4%, P<0.0001); the increase was even larger when considering the relative stance duration (+6%, P<0.0001).
Ground reaction force
The Fx and Fz vs. time (normalised to stance duration) graphs are presented in Figure 2. They both appeared different on the 2 surfaces, as illustrated also in Figure 4 (with Horse No. 2 as an example).
After an early impact peak (the analysis of which is dealt with in Chateau et al. 2010), the Fz (‘vertical’ force) rises progressively to a maximum, then decreases to zero at lift-off. Figure 2 and Table 1 show that the Fz max was lower (−13%, P<0.0001) and that it occurred later (+15%, P<0.0001) on DWS than on FWS; as a consequence, the Fz max loading rate was less on DWS (−26%, P<0.0001). The Fz impulse was also lower on DWS (−12%, P<0.0001).
During the stance phase, Fx initially has a positive phase, followed by a negative phase (positive/negative signs depend on the convention chosen for the DHS reference frame); these phases are generally designated ‘braking’ and ‘propulsive’ phases, respectively. Figure 2 and Table 1 show that the Fx0 time, separating the positive and negative phases, was less on DWS (−16%, P<0.0001), indicative of a longer ‘propulsive’ phase. Fx max was almost 2-fold lower on DWS compared to FWS (−45%, P<0.0001) and it occurred later (+39%, P<0.0001). There was inter-individual variability in Fx max time on DWS, but in all horses the Fx max loading rate was drastically lower compared to FWS (−53%, P<0.0001). Contrary to Fx max, Fx min was not significantly different between the 2 sand tracks (P = 0.155); Fx min time was nevertheless significantly delayed, although only slightly, on DWS (+1%, P = 0.014).
The total Fx impulse (absolute values) was significantly lower on DWS (−26%, P<0.0001). Besides, whereas Fx positive impulse was much lower (−52%, P<0.0001), Fx negative impulse was drastically higher (+59%, P<0.0001), on DWS compared to FWS. As a consequence, the net Fx impulse was positive on the FWS, meaning an overall net ‘braking’ effect, whereas it was about nil on the DWS.
Observation of both Fz and Fx plots for the 4 horses (Fig 2) reveals that at the end of the stance phase, especially at about 90% of stance duration, both GRF components were higher on DWS (maximal Fz difference of 300–950 N depending on the horse).
Figure 3 illustrates the angle-time graphs of the fetlock joint (dorsal angle) for the 3 horses on which this angle could be measured. The minimal dorsal angle was significantly higher (+4%; P<0.0001), i.e. the joint was less extended and this minimum occurred later (+9%, P<0.0001), on DWS compared to FWS (see also Fig 4b). However, as for the GRF components, between about 80 and 95% of the stance phase, the order reverses and the fetlock dorsal angle is smaller, i.e. the joint is more extended, on DWS than on FWS (Fig 3).
At landing, the horse's forelimb was significantly more oblique to the ground (less vertical) on DWS than on FWS (forelimb axis-track angle: −7%, P<0.0001), as shown in Figure 4a.
On beaches, horses are generally trained at the water edge. For practical and technical reasons, our FWS condition had to be slightly further from the sea (10–20 m in average) and therefore possibly had slightly lower water content than the surfaces used under real training conditions.
The dynamometric horseshoe affects the dynamics of the horse because its mass and thickness are greater than a regular training shoe. However, since the same device was used to test both sand tracks, it should not affect surface comparison.
Since the experimental horses were harness trotters, the sulky interacted with the ground surface, as discussed below. The differences observed here between both surfaces may therefore be more evident compared to what would have been observed in ridden horses.
Significant differences between the 2 beach sand surfaces tested here in 4 harness trotters were found for stride parameters, GRF and kinematic variables reflecting the forelimb loading during the stance phase. An average 30% decrease (13.5 vs. 19%) of the sand water content induced a slight increase in stride frequency and a decrease in stride length, indicative of some discomfort of the horses, combined with a poorer performance on DWS (Chateau et al. 2010). The higher duration of stance observed on the deep sand in the present study is in accordance with previous observations, both in man and the horse (Burn and Usmar 2005; Pinnington et al. 2005); it is explained by the yielding nature of the surface, offering poor support and reaction. The higher stance to stride duration is correlated with a relatively shorter swing phase on DWS, which can be attributed to a less effective propulsion (Lejeune et al. 1998). The shorter swing phase on the deep surface is likely responsible, at least partly, for the increased obliquity of the forelimb at impact on this surface (the horse has less time to retract its forelimb before landing). In turn, the increased obliquity of the limb, associated with an increased penetration of the hoof in DWS (Fig 4a), induce a more gradual rise of the GRF and a more progressive decrease in the fetlock dorsal angle, than on FWS (Figs 2, 3).
Generally speaking, a lower water content induces a significant decrease of the force peaks Fz max and Fx max, and of the corresponding loading rates applied to the forelimb. Although stance duration was greater, Fz and positive and total Fx impulses were lower on deep sand. These results demonstrate the damping effect, in all directions, of DWS compared with FWS. Fetlock maximal extension was also significantly decreased on DWS. Interestingly, in Horse No. 4 for which the difference in water content between the 2 tracks was the least, the differences in the GRF (both Fz and Fx) and fetlock angle plots on DWS vs. FWS were also the least.
Fz max peak was significantly delayed on DWS, which can be attributed to at least 2 factors. As discussed above, a kinematic delay exists from the beginning of the stance on DWS since the initial position of the forelimb at landing is more oblique to the track (by almost 5°); this delay persists throughout the stance. Besides, energy is lost as the distal limbs penetrate the deep sand, especially the hindlimbs as they start to support the body. Therefore, it takes more time and effort to get the centre of mass of the horse in the same position relative to the forelimb hoof at Fz max time, which is also correlated with the more backward orientation of the forelimb, on DWS compared with FWS (Fig 4b). This is likely aggravated by the presence of the sulky (and driver) that also interacts with the ground.
At the beginning of the stance phase, since more load is applied on the caudal part of the hoof, the heels tend to penetrate in the ground, all the most when the surface is more compliant (Fig 4a); heel penetration is therefore increased on DWS. As a consequence, since the reference frame of the dynamometric horseshoe is related to the hoof, Fx is slightly less (and Fz slightly greater) thus becoming negative earlier than if - for the same GRF - the hoof solar surface had been more horizontal. This is more notable the more the position of the hoof at landing is heel first, as illustrated by the differences between Horses Nos 1 and 2, the hoof surface of Horse No. 1 becoming horizontal earlier during the stance phase. These observations stress the advantage of using a dynamometric horseshoe (vs. a force platform) for comparison of ground surface effects: it better represents the forces effectively applied on the hoof and information relative to the hoof orientation can be deduced from Fx and Fz plots. However, high-speed films and kinematic data, as included in the present protocol, are required for confirmation and thorough analysis. On the basis of previous works (Denoix 1994; Crevier-Denoix et al. 2001), the lower heel position of the hoof can contribute to explain the larger fetlock dorsal angle on DWS compared with FWS in the first part of stance.
As the resistance the DWS offers is low, the horse receives less propulsion from the same amount of force generation than on FWS (Clayton 2004). Therefore, if the force generation had been the same, the absolute value of Fx (including Fx min) and Fz would have been lower on DWS. It is not the case since Fx min values are not significantly different, and at about 90% of the stance, Fz and Fx (in absolute values), are even larger on DWS compared with FWS. Furthermore, because of hoof rotation towards the toe during propulsion, likely larger on the more compliant DWS, the absolute value of Fx in the reference frame of the hoof is less compared to what would have been observed for the same GRF with a more horizontal position of the hoof and a fortiori in the reference frame of the track. This confirms that the propulsion force developed by the horse's forelimb is increased on DWS compared with FWS. This result is also correlated with the strong increase in the Fx negative impulse on DWS. As a consequence, whereas the net Fx impulse is positive on FWS (‘braking impulse’), as it is traditionally observed on the forelimbs (Clayton 2004), this impulse is close to 0 (even slightly negative according to the observed means, Table 1) on DWS. This clearly demonstrates the increased contribution of the forelimbs to propulsion on deep sand.
The efficiency of the propulsive effort is nevertheless obviously lower on DWS compared to FWS, as indicated by the dorsal fetlock angle that is less on DWS at 90% of the stance phase (Fig 3). This may be explained, at least partly, by the difference in fetlock maximal extension on the 2 surfaces. Indeed during the support phase, the equine palmar tendons, especially the superficial digital flexor tendon (SDFT) and suspensory ligament (SL), are highly strained and more when the fetlock is more extended. These tendons thus store elastic energy that is passively released during propulsion. On DWS, maximal fetlock extension is reduced compared with FWS; therefore the SDFT and SL are less strained and their passive contribution to propulsion is less. Consequently, active contribution of the digital flexor muscles to propulsion is more required on DWS. Incidentally, an increased muscular contraction from the flexor tendons could also contribute to explain the lesser fetlock maximal extension observed on DWS (at about 60% of the stance). However, measurement of tendon force is required to confirm, or not, this hypothesis.
It is well known that hard surfaces are a predisposing factor for SDFT and SL injuries, especially at high speeds (Williams et al. 2001). However it is also admitted that deep surfaces can induce tendon lesions as well. It appears from this study that when a horse runs on a deep surface, propulsion is a critical phase (especially at about 80% of stance): limb loading forces are still high (since Fz max is delayed compared to a firmer surface), the fetlock joint is still submaximally extended (and more extended than on a firmer surface at the same speed), and digital flexor muscles have to develop more force to push the limb off. In this situation, tendons, especially the SDFT, are very likely more stressed and therefore more prone to injury than on a firmer surface. Furthermore, because the forward rotation of the hoof during propulsion, likely increased on deep sand, induces a correlative decrease of the distal interphalangeal palmar angle (Chateau et al. 2006), the deep digital flexor tendon is relatively released, which puts even more stress on the SDFT to support the fetlock joint.
Through this study, clear kinetic and kinematic differences have been established in horses trotting on deep vs. firm sand beach at a moderate training speed: the loading of the limb is overall reduced and more progressive; however the propulsive muscular effort is increased, which may be accompanied by an increased tension on the SDFT at the end of stance. These data should contribute to improve safety and efficiency of equine training on beaches.
The authors thank the Conseil Régional de Basse-Normandie, the Fonds unique interministériel, The French Ministry of Agriculture, the Haras Nationaux and the FEDER 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 the Garde Républicaine, and J. Jecker, for their contribution, and the City Hall of Varaville.