Evaluation of some recently developed techniques for estimating diet composition in browsing herbivores

D.G. Smith* - Research Fellow, Institute of Cell, Animal and Population Biology, University of Edinburgh, Ashworth Laboratories, The King's Buildings, Edinburgh, EH9 3JT, Scotland

I. J. Gordon - Programme Manager, Animal Ecology in Grazed Ecosystems, Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen, AB15 8QH Scotland

B. F. Mpendu - Graduate Student, University of Fort Hare, Private Bag X1314, Alice, South Africa, 5700

M.B. Lamani - Graduate Student, University of Fort Hare, Private Bag X1314, Alice, South Africa, 5700

L. Dziba - Graduate Student, University of Fort Hare, Private Bag X1314, Alice, South Africa, 5700

E. Kelly - Research Technician, University of Fort Hare, Private Bag X1314, Alice, South Africa, 5700

* - corresponding author.

Accuracy of estimates of botanical composition of herbivore diet may be improved through the use of recently developed techniques (alkane analysis, remote controlled oesophageal fistula valve). Three Boer goats allowed free choice of five forage species were used to compare seven methods of estimating diet composition with observed diet composition. The evaluated techniques either monitored intake [bite-count (IT1), bite mass corrected bite-count (IT2), relative bite mass corrected bite-count (IT3), micro-histological analysis of oesophageal extrusa collected at hourly intervals (IT4)] or faeces [alkane analysis (FT1), micro-histological faecal analysis (FT2), and in vitro corrected, micro-histological faecal analysis (FT3)]. Intake techniques had higher Kulczynski's similarity coefficients with measured values (70.7, 70.2, 65.5, and 78.6 respectively for IT1, IT2, IT3 and IT4) than did faecal techniques (58.9, 58.0 and 51.2 respectively for FT1, FT2 and FT3). Compared to IT2 and IT3, IT4 provided estimates of diet composition that was more similar to measured diet composition in terms of Kulczynski's similarity coefficients. However, estimates of diet composition were significantly different from measured values in 60% of cases for IT4 but only 40% of cases for IT2 and IT3. It is concluded that bite-count techniques (IT1 - IT3) can give as accurate estimate of diet composition as the more costly and technically demanding micro-histological analysis of oesophageal extrusa (IT4).

TABLE OF CONTENTS

Introduction

Materials and methods

Results

Discussion

Acknowledgements

References

INTRODUCTION

The botanical composition of diets consumed by free-ranging herbivores is commonly determined using micro-histological analysis of oesophageal extrusa, micro-histological analysis of faeces or by counting bites by direct observation (Gordon, 1995). Comparison between these techniques have shown poor correlations of diet composition estimates (Vavra et al. 1978, Sanders, et al. 1980, McInnis et al. 1983, Henley et al. 2001). Recently, two techniques have been developed that may allow more reliable estimates of the composition of diets consumed by grazing animals. The first technique estimates diet composition by comparing the n-alkane concentrations of consumed herbage with that in the faeces using a least square optimisation procedure (Dove and Moore 1995). The second technique is based on the remote control of an oesophageal valve (Raats 1993), that enables more frequent sampling of oesophageal extrusa from free-ranging animals.

The use of faecal samples enables estimation of the composition of the diet consumed by free-ranging animals, without any direct interference (Putman 1984). Existing faecal techniques rely on micro-histological analysis of plant epidermal fragments. These methods are laborious and are prone to error because of the differences in the digestibility of different component species and of their plant parts. Furthermore, the unequivocal identification of some species (especially dicots) is difficult because of the absence of a unique combination of epidermal features that enable identification (Henley et al. 2001). The alkane technique has been successfully used to estimate diet composition of animals which fed pasture plants (Mayes and Dove 2000), however, it has not been evaluated in either animals fed principally on browse species or, in animals whose diets which contain more than four components (Leury et al. 1999).

Comparison between estimates of grass/browse ratios determined by counting bites and either hourly sampling or daily sampling of oesophageal extrusa showed that increasing sampling frequency of oesophageal extrusa improved reliability (Raats et al. 2001). This still requires validation against values of diet composition measured in animals fed known diets.

Direct comparison of micro-histological analysis of oesophageal extrusa, micro-histological faecal analysis, bite count and alkane techniques is difficult because they differ, both in the methods used to identify the forage species consumed and in the sampling period over which the estimation of diet composition is conducted. Discrepancies between results obtained using the various techniques are thought to arise from differences in digesta passage rate between species and differences in epidermal survival following digestion of forage species (Holechek et al. 1982). Comparison between faecal techniques and intake techniques is made more difficult because the interval between forage consumption and the appearance of material in the faeces is usually unknown.

The objective of the experiment reported here was to compare the botanical composition of goat diets estimated by conventional bite count techniques and micro-histological faecal analysis with that estimated by increased sampling frequency of oesophageal extrusa and alkanes techniques in order to determine the most reliable method for use in free-range animals.

MATERIALS AND METHODS

The experiment took place at the University of Fort Hare, Honeydale Research Farm, Eastern Cape, Republic of South Africa, during March 1999. Three, individually penned Boer goats (Capra hircus) (mean live weight 49.3 kg s.e. _ 1.01), one of which was fitted with an oesophageal fistula, were given free access to a choice of five fodder species for 6 hours per day (0900 - 1500 hours), the usual grazing time allowed in semi-arid Africa. Animals were given a 10 day adaptation period to adjust to the diet and the pen environment. The fistulated goat was surgically prepared with a large oesophageal fistula (1050 mm² aperture size - suitable for use with a remote control oesophageal fistula valve) according to the technique described by Raats (1993).

Forages offered

The five forages (four browse and one grass hay) were presented to the animals using a cafeteria design (Qinisa and Boomker 1998), so that the intake of each individual species could be measured. The browse species offered were Acacia karroo Hayne (AK), Grewia occidentalis L. (GO), Rhus longispina (RL) and Scutia myrtina (SM) (Burm. F.), the grass hay (GH) offered comprised mainly Themeda triandra Forssk. All browse species were collected daily from an area of False-Thorn bush-veldt (Acocks 1975) in close proximity to the research farm. In order to avoid any differential preference effects arising through local environmental/soil conditions, collections always took place in the same area. Each forage was offered at approximately 35% of the estimated total daily dry matter intake (D. Cuddeford, pers. comm.), to avoid excessive consumption of one species. Browse species were presented to the animals using a metal frame placed at the front of each pen, to which the branches of browse were fixed at their base in a vertical position using wire in order to mimic the habit of the plants. Grass hay was offered in a trough situated next to the wire frame. Hay and fresh browse were placed in front of the animals initially at 0900 hours, further browse was attached to the wire frame or hay placed in the trough at 1200 hours. All forage was removed from the pen at 1500 hours, and not replaced until 0900 hours the following day.

Compared Techniques

A total of seven indirect techniques for estimating diet composition were evaluated against direct measurements of diet composition. Four intake-based techniques (IT1 - IT4) and three faecal-based (FT1 - FT3) techniques were tested.

IT1. (Bite count)

Direct observation of the goats was carried out in all three animals over a 5-day period (15 replicates) using the bite count method described by Risenhoover (1989). Observers were trained in the bite count method for a period of 7 days, during the adaptation phase of the experiment.

Observation sessions began at 0900 hours (when food was given to the goats) and finished at 1500 hours (when food was withdrawn). Starting on the hour, each goat was observed in a randomly determined sequence for two 5-minute periods per hour. Observers recorded animal identity, time of observation, plant species and number of bites of each forage species during the 5-minute period. Observers distinguished bites from exploratory mouthing by the occurrence of both distinct head movement and the sound of a bite; the use of auditory cues to identify prehension has been shown to give the best indicator of a bite (Ungar 1996). In the case of hay, bites were readily identified from the goat's tugging head movements as the forage was removed from the rack. A 5-minute transition period between observations allowed time for the observer to move to the next goat giving the animal time to settle before recording started.

IT2. (Bite mass corrected bite count)

Estimates of diet composition using the bite mass corrected bite count method were carried out for the fistulated goat over a 5-day period (5 replicates). Bite mass for each forage species was determined daily on a dry weight basis by counting bites during oesophageal extrusa collection whilst the fistulated goat was offered only a single forage species. The oesophageal extrusa was then dried until constant weight in an oven at 60_C, and then weighed. Dry bite weight was determined by calculating the dry weight of forage consumed and dividing it by the number of bites.

Estimation of diet composition using the bite mass corrected bite count method was calculated using the following equations:

(1 - adapted from Hodgson 1982)

(2)

Where RI_s is the relative daily intake of a forage species (g DM), OB_s is the number of observed bites of a forage species, BS_s is mean bite mass (g DM) of a forage species, P_s is the daily proportion of a forage species in the diet and RI is the daily relative intake of each forage species measured in the fistulated goat (g DM).

IT3. (Relative bite mass corrected bite count)

In practise only a limited number of animals involved in a study are fistulated and relative bite mass measurements of each forage species measured in fistulated animal are applied to other non-fistulated animals. In order to test the effectiveness of this method estimates of diet composition using the relative bite mass corrected bite count method were made for the two un-fistulated goats over a 5-day period (10 replicates). Relative bite mass was calculated by expressing the bite mass (measured in the fistulated goat) of each forage species as a proportion of the forage species with the largest measured bite mass. The same procedure used for calculating IT2 was then followed substituting bite mass BSs, with relative bite mass (RBS_s).

IT4 (Micro-histological analysis of frequently collected of oesophageal extrusa)

Daily diet composition were estimated using micro-histological analysis of oesophageal extrusa corrected hourly from 0900 - 1500 hours for the fistulated goat over a 5 day period (5 replicates); oesophageal extrusa is more usually collected once daily (Raats, 1996). Oesophageal extrusa was collected in a bag placed around the goat's neck after the plug and spatula that sealed the fistula had been removed, simulating the action of a remote controlled oesophageal valve. After a period of 10 minutes the bag containing the extrusa sample was removed and the plug and spatula replaced. The collection procedure was carried out at 0900, 1000, 1100 1200, 1300 and 1400 hours. Oesophageal extrusa was frequently not obtained during the 1200,1300 and 1400 hours collections because the goat often did not consume forage at these times. After collection, the extrusa was placed between two filter papers then dried overnight in a force draught oven at 60ºC. Once dry, all the extrusa from a single sampling day were pooled and milled through a 1mm screen and a 2 g sub sample retained for micro-histological analysis.

Two microscope slides of each oesophageal extrusa sample were prepared using a method adapted from Storr (1961) (details described below). From each study day 120 oesophageal extrusa epidermal fragments were identified (20 views per pair of slides from each of the 6 daily extrusa samples collected from each animal), allowing forage species that comprised more than 5% of the diet to be determined (Stewart 1967). The total number of oesophageal extrusa fragments recorded for the animal over the 7 day collection period was 840.

FT1. (Faecal odd-chain alkane concentrations)

Diet composition was estimated using faecal odd-chain alkane concentrations in all three goats over a 5 day period (15 replicates). Faecal odd-chain alkanes (C₂₇ - C₃₅) were used to estimate the botanical composition of the diet. Each daily faecal grab-sample and each of the forage species samples were analysed for their alkane content using the extraction and analytical method described by Duncan et al. (1999). Peak areas obtained from gas chromatography were used to calculate the concentration of odd-chain alkanes present in both forage and faecal samples. The alkane concentrations were then used to estimate diet composition using the `Eat-What' Diet Selection Calculator Version 1.2 which automates the least square optimisation procedures described by Dove and Moore (1995).

FT2. (Micro-histological faecal analysis)

Diet composition estimation using the micro-histological faecal analysis technique was carried out for all three goats over a 5-day period (15 replicates). Faecal pellets were collected rectally from each goat at 0900 hours and 1500 hours on each day of the 7-day faecal collection, in order that intake and faecal samples could be synchronised. These faecal pellets from the morning and afternoon collections were pooled to provide one faecal sample from each animal per study day; giving a total of seven samples per animal. The faecal samples were dried at 60°C and then a 10 g of each sample was milled through a 1mm screen and a 2 g sub-sample retained for micro-histological analysis (Alipayo et al. 1992, Holechek 1982).

Five microscope slides from each faecal sample were prepared by the method described below. From each of the seven study days 100 faecal epidermal fragments were identified, allowing forage species that comprised more than 5% of the diet to be determined (Stewart 1967); a total of 700 microscopic observations were recorded per animal.

FT3. (In vitro corrected faecal analysis)

Diet composition was estimated using in vitro corrected faecal analysis in all three goats over a 5 day period (15 replicates). In order to estimate diet composition using the in vitro corrected faecal analysis the relative survival of the epidermis of the forage species samples in the digestive tract had to be determined.

A single calibration mixture comprising of equal proportions of AK, GH, GO, RL and SM was prepared by mechanically mixing 5g samples of each ground (1mm screen) forage species together. Once thoroughly mixed, twenty 0.5g samples of the calibration mixture and a 0.5g reference sample of each forage species were digested in vitro according to the method described by Tilley and Terry (1963). Residue from the reference samples of each species were used to prepare reference slides of digested epidermal fragments of each species from which photomicrographs and annoted drawings were produced. The residue from each in vitro digested calibration sample was used to prepare one microscope slide (20 slides in total) for micro-histological analysis using a method adapted from Storr (1961) (see below). In addition twenty slides using undigested calibration mixture were prepared along with a reference slide for each undigested forage species. Using a binocular microscope at 200x magnification, epidermal fragments from twenty separate fields-of-view per slide were identified (using photomicrographs and drawings prepared from the reference samples in addition to published dichotomous keys) and recorded (twenty fields-of-view per slide x twenty slides = 400 observations in total for both undigested and digested samples).

The following equation was used to calculate an in vitro correction factor for each species:

(3)

Where IVCF_s is the in vitro correction factor for a species, DF_s is the number of observed fragments of a digested species per slide and UF_s is the number of expected fragments of a species per slide determined from the undigested samples. The mean frequency of positively identified epidermal fragments in faecal samples (FT2) were then divided by the appropriate species in vitro correction factor to provide in vitro corrected faecal analysis data (FT3).

Measurements

Intake of forage

Branches of browse and the hay were weighed before being offered to the animals. Refused material from each forage species were re-weighed (green material was stripped from the woody branches and the two portions weighed separately) when the wire frames and troughs were removed from the pens at 1500 hours. Evaporation losses of woody material were determined by weighing a leaf-stripped branch of each browse species at 0900 hours when the forages were offered. The leaf-stripped branches were then reweighed at 1500 hours; dry matter losses per gram of refused woody material were calculated for each browse species and the leaf-stripped branch refusal weights corrected for the evaporation loss. In order to provide the leaf-stripped braches with the same environment as the branches that were offered to the animals, the leaf-stripped branches were attached to a wire frame placed as close as possible to the animal pens, but beyond the reach of the goats. Dry matter of refused green-leaf was determined separately. In the case of hay, refusal dry matter was determined from material remaining in the trough at 15:00 when the forages were removed. Dry matter intake of each forage species was calculated from the amount of fresh food consumed and the dry matter content of the green material. The dried samples of leaf collected on a daily basis from the four browse species, grass hay, and refused food from all five forage species were pooled on a weekly basis.

Table 1

Comparison of Acacia karroo (AK), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM and grass hay (GH),) in terms of dry matter (DM) content (_ s.e.), organic matter (OM) content, crude protein (CP), neutral detergent fibre (NDF), acid detergent fibre (ADF), in vitro dry matter digestibility (IV-DMD) and alkane concentrations (C₂₅ - C₃₆) and bite mass (_ s.e.). Units are in g/kg DM unless stated.

¹ Measured in one fistulated goat; ² C₃₄ used as an internal standard, no estimation possible

The dry pooled samples were then ground through a mill fitted with a 1 mm screen and retained for analysis of n-alkane composition, organic matter (OM), neutral detergent fibre (NDF), acid detergent fibre (ADF), crude protein (CP) and in vitro dry matter digestibility (IV-DMD) of each sample from each species (Table 1). Proximate analysis (OM, NDF, ADF and CP) were carried out according to the methods described by the Association Of Official Analytical Chemists (1990), alkane analysis was carried out according to the methods described by Salt et al. 1994, IV-DMD was determined using the method described by Tilley and Terry (1963).

Rate of passage

In order to measure the rate of passage of each forage species to determine the lag time for correlating diet consumed with faecal estimates of diet composition, the goats were given small quantities of each fresh forage species labelled with a different even-chain alkane; GH was labelled with C₂₆ alkane, RL with C₂₈ alkane, GO with C₃₀ alkane, SM with C₃₂ alkane, and AK with C₃₆ alkane. A 6 - 10 g sample of fresh leaf from each forage species was labelled with approximately 600 mg of the appropriate even-chain alkane by dissolving the alkane in warm (70_C) heptane and then spraying the resulting solution onto the fresh leaf material; the alkane was then fixed to the forage by drying in a stream of cool air (20_C) for 10 minutes. Each marked forage species was then roughly chopped into 1 - 2 cm lengths using a food processor and a 1 g sub-sample taken, dried and retained for alkane analysis. The remaining marked forage was mixed and compressed into a crude pellet then stored until dosing in refrigerated, airtight containers, in order to minimise wilting. After the 10 day adaptation period, one crude pellet, composed of five labelled forage species was fed to each goat at 00 hours on day one of the total faeces collection (see below).

Faecal samples were gathered over a 7 day period at 0, 6, 9, 12, 16, 20, 24, 28, 32, 38, 44, 53, 65, 77, 89, 111, 137 and 161 hours after dosing the labelled forages (Mayes et al. 1997). Faecal sampling was facilitated by placing the animals on a slatted wooden floor, under which plastic sheets were arranged to funnel all urine and faecal pellets to a collection sieve that separated faeces from urine. At each collection time, all fresh faeces voided since the previous collection were weighed, and a ~ 100 g sub-sample removed and dried at 60_C until constant weight. The dried sample was then re-weighed and the dry matter content of the faecal sample determined. Each dry faecal sample was ground through a mill fitted with a 1mm screen and retained for alkane analysis according to the method of Duncan et al. (1999). In order to compare the gut-retention time of each forage species, the time taken to recover 50% of the total output of each dosed, even-chain alkane over the 7-day collection period was calculated by plotting the accumulated output against time after dosing. This procedure was used in preference to the curve- fitting modelling techniques of Grovum and Williams (1973) and Dhanoa et al. (1985) because each forage contained small amounts of the same even-chain alkanes as those dosed and there was considerable variation in the amount of each forage consumed, it was not possible to accurately estimate a baseline level of natural even-chain alkane concentration in the faeces.

Micro-histological analysis

Slides of oesophageal extrusa samples (IT4), faecal analysis samples (FT2), in vitro digested calibration mixture (FT3)were prepared using a method adapted from Storr (1961). Reference slides of the five forage species and in vitro digested samples of the five forage species were prepared by the same method (Storr 1961) using ground samples (1 mm screen) of each species.

Samples (~0.5 g) were treated with 69% nitric acid at 70ºC for 2 minutes then rinsed several times with hot water (80ºC). After draining the samples were dehydrated by progressively passing them through an alcohol series (30%, 50%, 70% and 100% ethanol). Samples were then stained by placing in a 1% solution Safranin stain in ethanol for 10 minutes. A small quantity of each sample was then placed on a microscope slide. The sample was then de-waxed with xylene and mounted under a cover slip using DSX (Merek Ltd). The number of slides prepared from each sample depending on the type of sample (5 slides were prepared from each faecal sample, two from each oesophageal extrusa sample and one for each calibration and reference sample. In order to assist in the identification of fragments samples from each of the five forage species (both digested and undigested) were used as reference material in order to construct dichotomous keys, photo- and electron micrographs.

The slides prepared from faeces and oesophageal extrusa were examined following systematic transects using a binocular compound microscope at between 40x and 200x magnification. Transects were at least one field-of-view apart to avoid counting the same plant fragment twice. From each slide the first twenty fragments of sufficient size that were encountered were identified using the key, reference slides and species description (Green 1987). To reduce bias toward small readily identifiable fragments, only particles larger than half the field-of-view area at 200x magnification were identified (Stewart 1967).

Data analysis

Statistical analysis of measured daily intake of each forage species goats was carried out using one-way analysis of variance (Minitab for Windows - version 12). Rate of passage data were statistically analysed using two way analysis of variance with forage species as treatment factor and each goat as the block factor (Minitab for Windows - version 12).

Daily measured proportions of each forage species were compared with proportions estimated using techniques IT1, FT1, FT2 and FT3 using the Mann-Whitney two-sample test (Minitab for Windows - version 12). The same analysis was carried out for techniques IT2 and IT4 using data from the fistulated goat only; data were compared with daily measured proportions of each species in the fistulated goat. Species proportions calculated by IT3 from the two un-fistulated goats were compared with measured species portions for these goats using the Mann - Whitney test.

Kulczynski's Similarity Coefficient (Oosting 1956) comparing measured diet composition with those estimated for each technique was calculated on a daily basis for each goat.

RESULTS

In all three goats there was little daily variation in the quantity of each species consumed (Table 2). There was no significant difference between goats in the proportion of each forage species consumed.

Table 2

Dry matter intake (g DM / day) and proportion in diet of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) in three Boer goats (s.e. is shown in parenthesis).

There were no significant differences between forage species in time taken to recover 50% of the dosed even-chain alkanes, although there were highly significant (P< 0.001) differences between animals (Table 3). Because time taken to recover 50% of the dosed markers was between 43 - 51 h after dosing, daily faecal measurements were correlated with daily intake measurements that had been recorded 2 days (48 hours) previously; for example day one feed intake measurements were equated with day three faecal measurements.

Table 3

Time taken (with s.e.) to recover 50% of total even-chain alkane faecal output over a 7-day collection period in Boer goats dosed once with C₃₆ labelled Acacia karroo (AK), C₂₆ labelled grass hay (GH), C₃₀ labelled Grewia occidentalis (GO), C₂₈ labelled Rhus longispina (RL) and C₃₂ labelled Scutia myrtina (SM).

The alkane concentration differed considerably between forage species, RL had the highest total (C₂₅ - C₃₆) alkane concentration (3035 mg/kg DM), whilst AK had the lowest total (C₂₅ - C₃₆) alkane concentration (158 mg/kg DM) (Table 1). These large differences in alkane concentrations between species affected the least square optimisation procedure used to determine diet composition from the alkane concentration of the faeces (Mayes and Dove 2000), because the contribution of alkanes derived from AK in the faeces were diluted by those species with higher alkane concentrations. Correction-factors to adjust faecal samples for differences in epidermal survival between species were calculated from the in vitro calibration mixture (Table 4).

Table 4

Actual frequency per slide (_ s.e.), predicted frequency per slide and correction factors of epidermal fragments of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) recorded from twenty microscope slides prepared from an in vitro digested calibration mixture containing equal proportions of each species (number of views per slide = 20).

The four techniques used to estimate diet composition that could be applied to all three goats (IT1, FT1, FT2 and FT3) were compared to measured diet composition (Figure 1). In all four techniques, variation within estimates of diet composition was greater than for measured values (Figure 1). With the exception of RL, FT1 showed the greatest amount of daily variation in the estimated proportion of forage species (AK was never detected using FT1) (Figure 1).

The IT2 and IT4 techniques could only be applied directly to the animal fitted with the oesophageal fistula. Comparison between the proportion of each forage species in the diet as estimated with IT2 or IT4 and that measured directly are shown in Figure 2. Estimates of diet composition using the IT4 were significantly different from measured values in three out of five cases (Table 5), whilst estimates of diet composition using the IT2 were significantly different in only two cases (Table 5).

Use of relative bite mass to correct bite count (IT3) estimates of diet composition was applied in the case of the two un-fistulated goats (Figure 3). Estimates of diet composition using IT3 were significantly different from measured values in two out of 5 cases (Table 5).

Figure 1

The median proportion of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) in the diets of three Boer goats over a 5 day period (n= 15) either measured directly or estimated by bite count (IT1), faecal alkane concentration (FT1), faecal analysis (FT2) or in vitro corrected faecal analysis (FT3). Inter-quartile ranges are indicated by error bars.

Figure 2

The median proportion of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) in the diet of a oesophageal fistulated Boer goat over a 5 day period (n= 5) either measured directly or estimated by bite mass corrected direct observation (IT2) or frequently sampled oesophageal extrusa (IT4). Inter-quartile ranges are indicated by error bars.

Table 5

Kulczynski's Similarity Coefficients (with s.e.) and statistical comparison (Mann - Whitney) of measured proportions with estimated proportions of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) in the diet of three Boer goats using seven estimation techniques; bite count (IT1), bite mass corrected bite count (IT2), relative bite mass corrected bite count (IT3), oesophageal extrusa micro-histological analysis (IT4), faecal alkane concentration (FT1), faecal micro-histological analysis (FT2) and in vitro corrected faecal micro-histological analysis (FT3).

¹ Estimated in two un-fistulated goats; ² Estimated in one fistulated goat

Figure 3

The median proportion of Acacia karroo (AK), grass hay (GH), Grewia occidentalis (GO), Rhus longispina (RL), and Scutia myrtina (SM) in the diet of two un-fistulated Boer goat over a 5 day period (n= 10) either measured directly or estimated by relative bite mass corrected direct observation (IT3). Inter-quartile ranges are indicated by error bars.

DISCUSSION

Measured versus estimates

In the present study all estimation techniques showed considerably more daily variation in the proportion of each species consumed than did the measured values (Figures 1, 2 and 3). The discrepancy between measured and estimated values probably arose from sampling error, and illustrates the importance of taking sufficient samples or observations in the field when estimating diet composition (Holechek et al. 1982).

Faecal methods versus intake methods

Typically, non-sacrificial estimates of free-ranging herbivore diet composition are made either by monitoring at the point of intake (IT1 - IT4) or the point of exit from the faeces (FT1 - FT3). A fundamental difference in the estimates of botanical composition provided by intake and faecal techniques is that faecal methods give estimates of mean diet composition over a period of days whilst intake-monitoring methods give an instantaneous estimate of diet composition at one point in time (Sanders et al. 1980). Although many studies have compared these two approaches only a few studies have compared diet composition measured in penned animals with estimates of diet composition using both intake and faecal techniques (e.g. McInnis et al. 1983). Most comparative studies do not have an independent means of measuring actual diet composition (e.g. Henley et al. 2001, Mofareh et al. 1997, Johnson and Pearson 1981).

Faecal methods

Micro-histological analysis of faecal samples (FT2 and FT3) is known to have several methodological shortcomings, for example forage species passed in the faeces are often not proportional to those consumed (Holechek et al. 1982). This may result from destruction of some plant species during preparation (Vavra and Holechek 1980) or from species becoming unidentifiable in the faeces because of the processes of mastication and digestion (Slater and Jones 1971). In the present study these shortcomings may have been compounded by the use of concentrated nitric acid (Storr 1961) to prepare the samples for microscopic analysis. The use of nitric acid was selected for use in this study because Holechek (1982) reported that the grasses tended to be over estimated by a factor of 2.2 to 2.6 in diets comprising mainly of shrubs when less corrosive reagents (such as sodium hydroxide or bleach) were used. Storr (1961) reported a near 1:1 relationship between the proportion of browse and forb species measured in the diet of quokka (Setonix brachyurus) compared to that estimated in the faeces using a technique which has been widely used Africa to study the diet composition of free-ranging herbivores (e.g. Stewart 1967, Planton 1987, Schuette et al. 1998).

The use of in vitro digestion to provide correction factors for micro-histological faecal analysis (FT3) resulted in the poorest estimate of diet composition amongst those tested with all estimates of forage species been significantly different from measured values and having the lowest Kulczynski's Similarity Coefficient (Table 5).

The shortcomings of micro-histological faecal analysis (FT2 and FT3) could be overcome to a large degree by the alkane technique (FT1), because alkanes are digested only to a small and predictable extent (Mayes and Dove 2000), thus eliminating any bias towards a particular forage species. However, during this study, the low concentration of alkanes (C₂₅ - C₃₆) in AK, resulted in its estimated proportion in the diet being close to nil by FT2 (Table 5). The effect of this low-alkane species on the estimated proportion of the other species consumed, was difficult to predict because the presence of species with both high and low alkane concentration within the same faecal sample is likely to give unstable EATWHAT results (R. W. Mayes, pers. comm.). The dissimilarity between the FT1-estimated proportions and the measured proportions (Kulczynski's Similarity Coefficient = 58.9) reflects the overall reduction in the accuracy of the technique when there is a large between species variation in alkane concentration.

Henley et al. (2001) stated that a further difficulty encountered with faecal methods is that, because of differential passage rates, it is not possible to assign the ingestion of food items to a specific point in time. This study has shown that there was little difference in passage rate between species (Table 3) with a range of dry matter digestibility between 0.55 - 0.63 (Table 1).

Intake methods

The bite count technique (IT1) is an easily applied technique that requires little equipment, although operator training in plant identification is required. IT1 is more readily applied to browsing herbivores than to grazers because the distant identification of browse species is easier than for grasses (Henley et al. 2001). In the present study, although IT1, FT1 and FT2 had a similarly number (3 out of 5) of estimates of forage species that were statistically different from measured values, the estimates of diet composition using IT1 were more similar to the actual values (Table 5).

When bite count was corrected for bite mass (IT2) the agreement between estimated and measured species composition increased to three out of 5 species and the level of statistical significance for the remaining two species decreased (Table 5). Application of relative bite mass measured in the oesophageal fistulated animal to the other animals (IT3) also provide more reliable estimates of diet composition than bite count alone (IT1). Free et al. (1971) used a similar bite count correction technique and found no significant difference between bite count and micro-histological analysis of oesophageal extrusa techniques. Accurate measurement of bite mass requires the collection of oesophageal extrusa which is greatly facilitated under free-range conditions by using remote controlled oesophageal fistula valves.

In the present study estimates of diet composition made by micro-histological analysis of oesophageal extrusa technique (IT4) were more similar to measured values than any other of the estimation techniques tested. Generally the micro-histological analysis of oesophageal extrusa is considered a more accurate estimate of diet composition than the bite count technique, particularly when some forage species represent only a small portion of the diet (below 2%) (Ortega et al. 1995). However, studies using micro-histological analysis of oesophageal extrusa without the aid of remote controlled oesophageal fistula valves techniques are limited to sampling for short periods (typically 30 minutes per day), usually in the morning after a period of fast (e.g. Ortega et al. 1995, Mofareh et al. 1997).

Micro-histological analysis

Micro-histological analysis was used with techniques IT4, FT2 and FT3. When used to analyse faecal samples (FT2 and FT3), agreement with measured diet composition was poor, being statistically indistinguishable in only two out of 5 species and having, in the case of FT3, a Kulczynski's Similarity Coefficient of only 51% (Table 5). IT4 provided estimates of diet composition that were not statistically different from measured values in two out of 5 cases but had the highest Kulczynski's Similarity Coefficient (79%) of any of the techniques tested.

In this study Kulczynski's Similarity Coefficient between the IT4 and FT3 techniques was 45%; other studies have reported greater similarity between these two techniques. Johnson and Pearson (1981) report 90% similarity between diet composition determined by micro-histological analysis of either oesophageal extrusa or faeces, whilst similarity between the two methods was reported to be 77% by Mofareh, Beck and Schneberger (1997). These differences are probably due to the high proportion of browse species in the diets fed during the present study.

Most micro-histological evaluation studies have been carried out in conditions where grass and forbs are the major components of the diet (Sanders et al. 1980, Vavra et al. 1978, McInnis et al. 1983). Vavra and Holechek (1980) report that similarity between measured composition and that estimated by micro-histological analysis was higher for grass species than for browse or forb species. The value of micro-histological analysis in diets with a high proportion of woody shrubs was also questioned by Holechek and Valdez (1985), although Alipayo et al. (1992) reported more similarity between micro-histological analysis of oesophageal extrusa and faeces (in diets consisting mainly of browse species) when there was a high proportion of current growth in the browse material. Henley (1993), in a study carried out in the same area as the present study, reported a greater abundance of diagnostic features in grass species than in dicotyledonous species (both browse and forbs). The present study has also added to doubt about the reliability of micro-histological in herbivore species that are principally browser, such as goats.

In the present study identification of dicotyledonous (dicots.) species was impaired by both the scarcity of diagnostic features on the epidermis of the dicots. and, in the case of the faecal techniques, the low relative survival rate of epidermal fragments of forage species such as AK following digestion (Table 4).

Micro-histological analysis requires a substantial amount of operator training (Holechek and Gross 1982) and the construction of a micro-histological reference collection, therefore, the IT4, FT2 and FT3 are more financially and technically demanding than the IT1 and FT1. Furthermore, the IT4 technique requires surgical preparation of animals with oesophageal fistula, restricting the number of animals that can be studied and hence limiting sample sizes.

Technique of choice

Micro-histological analysis techniques (IT4, FT2 and FT3) cannot be recommended as appropriate methods for the study of goat feeding behaviour in the False Thorn Veld of the Eastern Cape because of doubts about the reliability of analysis in diets that consist mainly of browse species, the relative expense of this identification method and the high demand for operator training.

Without further refinement the alkane technique (FT1), such as the simultaneous use of other marker compounds (Mayes and Dove 2000), cannot be reliably used to estimate diet composition were some of the forage species consumed have very low alkane concentrations.

In this study the bite count (IT1) technique provided estimates of diet composition that were as good as those provided by micro-histological analysis of oesophageal extrusa (IT4). Where bite count was corrected for bite mass (IT2 and IT3), estimates of diet composition were better than that obtained using micro-histological analysis of oesophageal extrusa (IT4). Applying relative bite mass correction factors (IT3) derived from fistulated animals to non-fistulated goats of similar size provided good estimates of diet composition. IT3 has several advantages over the other techniques tested in the present study.

1. Measurements can be replicated in many animals, using bite mass data collected in only a few oesophageal fistulated animals.

2. Intensive observations can be carried throughout the day (but not during the night) allowing more accurate inferences drawn about the diet as a whole, and not just providing instantaneous estimates of consumed forage.

3. With the aid of remote controlled fistula valves bite mass measurements can be readily obtained and replicated.

4. Operator training for bite count observations is minimal and requires little specialist expertise.

5. The costs associated with the establishment and maintenance of fistulated animals is optimised because data can be applied to other animals of the same species and similar size.

A major disadvantage of all techniques using bite count is that they require a lot of labour to make visual observations in the field and measurements cannot be repeated in order to check for errors. The precision of the technique is also limited to identifying forage species that make up more than 2% of the diet (Ortega et al. 1995), although this is probably adequate for most rangeland utilisation studies.

In this study, the reliability (65 - 70%) of bite count techniques for estimating diet composition has been established. Frequently, in the continent of Africa bite count techniques are the only ones available to researchers, hopefully this study will improve confidence in bite count methods and encourage the study of herbivores in some of the world's most vulnerable rangeland habitats.

ACKNOWLEDGEMENTS

The authors would like to thank the UK Royal Society under the RS-NRF Set Joint Programme for their financial assistance with the preparation of this paper. The comments of Professor Andrew Illius and Dr. Derek Cuddeford during the preparation of the manuscript were greatly appreciated, together with the advice and assistance of Dr. R. W. Mayes and Mr. C. S. Lamb of MLURI.

REFERENCES

Acocks, J.P.H. 1975. Botanical Survey of South Africa, No. 40: Veld types of Southern Africa. Pretoria: Dept. Agriculture. pp 300 -301.

Alipayo, D., Valdez, R. , Holechek, J.L. and Cardenas, M. 1992. Evaluation of micro-histological analysis for determining ruminant diet composition. Journal of Range Management, 45, 148-152.

Association Of Official Analytical Chemists 1990. Official Methods of Analysis of the Association of Analytical Chemists (15th ed.). (pp 10 -55) Association of Analytical Chemists, Virginia, USA.

Dhanoa, M.S., Siddons, R.C., France, J. and Gale, D.L. 1985. A multi-compartment model to describe marker excretion patterns in ruminant faeces. British Journal of Nutrition, 53, 663-671.

Dove, H. and Moore, A.D. 1995. Using a least-squares optimisation procedure to estimate diet composition based on the alkanes of plant cuticular wax. Australian Journal of Agricultural Research, 46, 1535-1544.

Duncan, A.J., Mayes, R.W., Lamb, C.S., Young, S.A. and Castillo, I. 1999. The use of naturally occurring and artificially applied n-alkanes for estimating short-term diet composition and intake in sheep. Journal of Agricultural Science (Cambridge), 132, 233-246.

Free, J.C., Sims, P.L. and Hansen, R.M. 1971. Methods of estimating dry-weight composition in diets of steer. Journal of Animal Science, 32, 1003-1007.

Gordon, I.J. 1995. Animal-based techniques in grazing ecology research. Small Ruminant Research, 16, 203-214.

Green, M.J. 1987. Diet composition and quality in Himalayan musk deer based on faecal analysis. Journal of Wildlife Management, 51, 880-892.

Grovum, W.L. and Williams, V.J. 1973. Rate of passage of digesta in sheep 4. Passage of marker through the alimentary tract and the biological relevance of rate constants derived from the changes in concentration of marker in faeces. British Journal of Nutrition. 30, 313-329.

Henley, S.R. 1993. An evaluation of three techniques applied to the study of the diet selection in free-ranging large herbivores. B.Sc. Thesis, University of Natal, Pietermaritzburg, Republic of South Africa.

Henley, S.R., Smith, D.G. and Raats, J.G. 2001. Evaluation of three techniques for determining diet composition. Journal of Range Management, 54, 582-588.

Hodgson, J. 1982. Ingestive behaviour. In: Leaver J.D. (ed) Herbage intake handbook. Hurley: British Grassland Society. pp. 113-136.

Holechek, J.L. 1982. Sample preparation techniques for micro-histological analysis. Journal of Range Management, 35, 267-269.

Holechek, J.L. and Gross, B.D. 1982. Training needs for quantifying simulated diets from fragmented range plants. Journal of Range Management, 35, 644-647.

Holechek, J.L. and Valdez, R. 1985. Magnification and shrub stemmy material influence on fecal analysis accuracy. Journal of Range Management, 38, 350-352.

Holechek, J.L., Vavra M. and Pieper, R.D. 1982. Botanical composition determination of range diets: a review. Journal of Range Management, 35, 309-315.

Johnson, M.K. and Pearson, H.A. 1981. Esophageal, fecal and exclosure estimates of cattle diets on a Logleaf Pine-Bluestem range. Journal of Range Management, 34, 232-234.

Leury, B.J., Siever-Kelly, C., Gatford, K.L., Simpson, R.J. and Dove, H. 1999. Spray-topping annual grass pasture with glyphosate to delay loss of feeding value during summer. IV. Diet composition, herbage intake and performance in grazing sheep. Australian Journal of Agricultural Research, 50, 487-495.

Mayes, R.W. and Dove, H. 2000. Measurement of dietary nutrient intake in free-ranging mammalian herbivores. Nutrition Research and Reviews, 13, 107-138.

Mayes, R.W., Giraldez, J. and Lamb, C.S. 1997. Estimation of Gastrointestinal passage rates of different plant components in ruminants using iostopically-labelled plant wax hydrocarbons or sprayed even-chain n-alkanes. Proceedings of the Nutrition. Society, 56, 187a.

Mcinnis, M.L., Vavra, M. and Krueger, W.C. 1983. A comparison of four methods used to determine the diet of large herbivores. Journal of Range Management, 36, 302-307.

Mofareh, M.M., Beck, R.F. and Schneberger, A.G. 1997. Comparing techniques for determining steer diets in northern Chihuahuan Desert. Journal of Range Management, 50, 27-32.

Oosting, H.J. 1956. The study of plant communities. San Francisco: Freeman. pp. 110-112

Ortega, I.M., Bryant, F.C. and Drawe, D.L. 1995. Contrast of esophageal-fistula versus bite-count techniques to determine cattle diets. Journal of Range Management, 48, 498-502.

Planton, H. 1987. The dietary preferences of domestic ruminants (cattle, sheep, goats) on sahelian and sudano-sahelian rangelands. IV. Diet description by micro-histological analysis of plant samples collected by shepherds, or in oesophageal boli and faeces of cattle and sheep as a means of characterizing the diet. Revue d'Elevage et de Medecine Veteinaire des Pays Tropicaux, 42, 245-252.

Putman, R.J. 1984. Facts from Faeces. Mammal Review, 14, 79-97.

Qinisa, M.M. and Boomker, E.A. 1998. Feed selection and water intake of indigenous wethers under stall-feeding conditions. South African Journal of Animal Science, 28, 173-178.

Raats, J.G. 1993. Development and Testing of a Remote Controlled Oesophageal Fistula Valve for Goats. Ph.D. Dissertation, University of Natal, Faculty of Agriculture.

Raats, J.G., Smith, D.G. and Mogorosi, C.K. 2001. An evaluation of a remote controlled oesophageal fistula valve for sampling forage ingested by free-ranging Boer goats. Small Ruminant Research (submitted).

Risenhoover, K.L. 1989. Composition and quality of moose winter diets in interior Alaska. Journal of Wildlife Management, 53, 568-577.

Salt, C.A., Mayes, R.W., Colgrove, P.M., and Lamb, C.S. 1994. The effects of season and diet composition on the radiocaesium intake by sheep grazing heather moorland. Journal of Applied Ecology, 31, 125-136.

Sanders, K.D., Dahl, B.E. and Scoot, G. 1980. Bite-count vs. fecal analysis for range animal diet. Journal of Range Management, 32, 146-149.

Schuette, J.R., Leslie, D. M., Lochmiller, R. L. and Jenks, J. A. 1998. Diets of hartebeest and roan antelope in Burkino-Faso: Support of the long-face hypothesis. Journal of Mammalogy 79, 426-436.

Slater, J. and Jones, R.J. 1971. Estimation of the diets selected by grazing animals from microscopic analysis of faeces. Journal of the Australian Institute of Agricultural Science, 37, 238-239.

Stewart, D.R.M. 1967. Analysis of plant epidermis in faeces, a technique for studying the plant preferences of grazing herbivores. Journal of Applied Ecology, 4, 83-111.

Storr, G.M. 1961. Microscopic analysis of faeces, a technique for ascertaining the diet of herbivores mammals. Australian Journal Biological Science, 14, 157-164.

Tilley, J.M.A. and Terry, R.H. 1963. A two stage technique for the in vitro digestion of forage crops. Journal of the British Grassland Society, 18, 104.

Ungar, E.D. 1996. Ingestive behaviour. In: Hodgson J., Illius A.W. (eds) The ecology and management of grazing systems. Wallingford, UK: CAB International. pp. 185-210

Vavra, M. and Holechek, J.L. 1980. Factors influencing micro-histological analysis of herbivore diets. Journal of Range Management, 33, 371-374.

Vavra, M., Rice, R.W. and Hansen, R.M. 1978. A comparison of esophageal fistula and fecal material to determine steer diets. Journal of Range Management, 31, 11-13.