1Many ecological studies and applications involve measuring the height of grassland swards. An evaluation was made of the practicality, accuracy and comparability of using the sward stick, drop disc and direct methods of measurement. Each method proved to be appropriate for measuring swards that contain a wide range of heights, each was quick to perform, and recorder effects were negligible. Yet each had strengths and weaknesses.
2The sward stick gave the most variable results: it was considered the best method for recording the architecture of the sward surface, and hence invertebrate niches, but was poor for measuring short turf. The drop disc was the worst method for recording microheterogeneity in sward architecture and was completely unsuitable for measuring variation in short turf. But in medium-tall swards, it was considered to be the best method for measuring productivity, vertebrate herbivory and for large-scale monitoring of land managed for conservation and under agri-environment schemes. The direct method gave the most consistent and accurate results compared with an independent parameter, soil temperature. It was the only method suitable for measuring variation in short turf.
3A serious problem exists when research results, recommendations and assays involve measurements made by more than one method. The sward stick consistently gave higher absolute values than either of the other methods and, apart from in short turf (for which it is unsuitable), the drop disc gave values that were 73% and up to 40% lower, respectively, than those obtained using the sward stick and direct methods. This can lead to major misapplications of ecological results and recommendations in conservation and agri-environmental projects.
Several techniques have been devised to measure sward height and structure (Frame 1993). Most are variants of three common approaches. (i) The ‘direct measurement method’ involves placing a card or hand lightly on the vegetation at the level below which about 80% of the vegetation is estimated by eye to be growing (thus ignoring occasional tall stalks), then reading this height on a ruler (Hodgson, Taylor & Lonsdale 1971; ’tMannetje 1978). (ii) The ‘drop disc method’ (Holmes 1974; Anonymous 1986) involves letting a disc with a central slot drop down a vertically held ruler, and measuring the height above ground where it comes to rest; standard discs are 30 cm in diameter, weigh 200 g, and are dropped from a 1-m height. (iii) The ‘sward stick method’ (Barthram 1986) involves a 45-cm metal rule with 0·5-cm graduations, with a sleeve supporting a projecting 2 × 1-cm piece of clear perspex; the rule is held vertically, the sleeve lowered until the perspex touches the first piece of green non-flowering vegetation, and the height read on the rule.
All the above methods are quick to use and have been employed selectively on individual plant species or to measure general sward structure, often at multiple points within quadrats. Alternatively, the structure of the sward's surface is sometimes measured using the ‘point quadrat method’, in which a quadrat with downward-pointing pins is lowered onto the sward, and the heights at which up to 100 pins first contact the vegetation are recorded (Grant 1993). This time-consuming device was not evaluated separately from the quick methods discussed here, but is comparable to making 100 sward stick measurements in every 1-m2 sampled.
In a recent editorial, Ormerod & Watkinson (2000) suggest that ‘successful ecological management will become one of the most pressing necessities of our time’. Certainly, the importance of managing grassland swards to prescribed heights and structures has long been recognized in conservation and agriculture (Smith 1980; Anonymous 1986; Davies et al. 1993; WallisDeVries, Bakker & Van Wieren 1998; Hopkins 2000; Jefferson & Robertson 2000; Rook & Penning 2000), as has the need for a simple, quick and objective method to assess whether prescribed objectives are being achieved (Diack, Burke & Peel 2000). In Britain, this led the former Nature Conservancy Council (NCC) and English Nature (EN) to adopt the drop disc method as the standard for the numerous management prescriptions and scientific studies under their influence, whereas DEFRA (Department of Environment, Food and Rural Affairs) adopted the sward stick method for all agri-environment schemes (Peel & Jefferson 2000). Neither organization recommended the direct method for projects involving many or untrained recorders (e.g. farmers), due to its perceived lack of objectivity.
This raises the question of how comparable are the absolute measurements of sward height obtained by each method. Are land managers unwittingly being misdirected when advised to use one method when this differs from that used in making the original research on which recommendations were based?
Here, we describe how well absolute measurements of turf height made at the same points in semi-natural grassland were correlated when measured by each of the three methods. We also compare the cost and speed of making measurements using each method, and whether recorder effects bias or obscure the results. Finally, we compare how closely the measurements of turf made by each method are correlated with an independent parameter, ground temperature, which is known to be strongly influenced by sward height (Thomas 1993).
The sward stick, drop disc and direct methods were tested on three calcareous and neutral grassland swards in Dorset, UK, in May and July 1999. All were Sites of Special Scientific Interest being managed under conservation management agreements, with heterogeneous swards ranging from < 1 cm to about 30 cm tall, containing a diversity of plants ranging from prostrate species such as Thymus polytrichus A. Kern ex Borbás to tall clumps of coarse grasses, mainly Brachypodium pinnatum L.
Random points were selected for measurement, stratified to sample the full range of sward heights available, and a measurement of sward height made using each method at precisely the same spot on each occasion. The drop disc method was always used last because sparse or fine turf can be left slightly compressed after making this measurement; the other two methods were alternated. To test recorder effect, 50 independent measurements using each method were made of the same points by four different recorders, after initial ‘training’ in a group for 5 min. On 50 other occasions, during a 3-h period of constant (±1 °C) air temperature, the temperature of the top 1 cm of soil beneath the turf was recorded using a thermistor probe immediately before each triplet of sward height measurements was made. A square root transformation of sward height measurement was used for correlations with soil temperatures. All analyses and curve fitting were made using minitab 13.
Recorder effects, speed of measurement and cost
No significant bias was detected for any individual recorder using any method. Although, intuitively, the direct method seems subjective, very close agreement was obtained when four independent surveyors made blind measurements of the same spot using this method (Table 1; mean r2direct = 97%; mean maximum value by a recorder/mean minimum value by a recorder = 1·10). Even less variation was obtained when recorders used a drop disc (mean r2dropdisc = 98%; mean maximum recorder/mean minimum recorder = 1·02) but slightly greater variation was found using the sward stick (mean r2swardstick = 89%; mean maximum recorder/mean minimum recorder = 1·12).
Table 1. Results of blind tests when four recorders made independent measurements of sward height at the same sampling points using the same method. Figures give r2 values for each pair of recorders; n = 50 per recorder for each method. D-disc = drop disc; Sward = sward stick; Direct = direct measurement
All methods took negligible time compared with other aspects of the sampling, although the direct method, with 11·8 ± 1·3 SD recordings per minute, was 40% quicker than either the drop disc or sward stick (all dictated results). Sward sticks are more expensive, currently listed at > £100 each in the UK. The direct method involves a ruler of < £1 and the home-made dropped disc a standard 1-m rule costing about £5.
Comparison of results using the three methods
The results of measuring exactly the same points within swards by each of the three methods were closely correlated (r2 = 77–83%) over a sward range of about 0–30 cm (Fig. 1), with closer correlations being obtained when results were compared with those measured by the direct method. The sward stick consistently resulted in significantly higher absolute values of sward height and the drop disc in significantly lower values than the other two methods, with sward stick measurements being 73% higher, on average, than those measured using drop discs (Fig. 1c).
These relationships were more complex in short turf, where the results using the drop disc method were not correlated at all with those using either of the other methods (Fig. 1a,c). This occurred because, in short turf, the drop disc tended to record variation in microtopography rather than in vegetation height, on sites (such as those studied) where the terrain was not absolutely even. This was because the 707-cm2 disc came to rest on the highest points beneath it, and the random points where the vertical rule was placed were often a few centimetres lower than the highest point within the disc's 30-cm diameter. In practice, we obtained no sward measurement of < 2 cm tall using the drop disc method, compared with 23% using the direct method. Only 1·4% of measurements using the sward stick were also < 2 cm, but this method consistently overestimated turf height compared with the other two (Fig. 1b,c). In other words, compared with the direct method, the dropped disc overestimated turf height in swards < 4 cm tall but underestimated turf height in all swards of > 10 cm tall (Fig. 1a). In contrast, a weak correlation (P < 0·05) was obtained when turf < 5 cm tall was measured using the sward stick and direct methods (Fig. 1b).
Comparison of the three methods with an independent variable, soil surface temperature
The temperature at the soil surface varied by about 30 °C beneath the heterogeneous swards during the 3-h testing period in July (when air temperature varied by < 2 °C). As expected (Thomas 1983, 1993), the results of all three measurements of the square root of sward height were closely (and negatively) correlated with soil temperature over the full range of points sampled (Fig. 2). The closest correlation (r2 = 86%, n = 50, P = 0·0001) was obtained using the direct method. We next considered the correlation between soil temperature and tall and short turf separately, using the coolest and the warmest 15 °C temperature spans as separate variables. All three methods provided results that were closely correlated with soil temperature in the range 25 °C–40 °C (Fig. 2), corresponding to swards > 3 cm tall (drop disc) and > 5 cm tall (sward stick). However, over the warmer half of the spectrum (40–55 °C), only the results obtained using the direct method were significantly correlated with soil temperatures (r2 = 63%). Indeed, over the hottest 10 °C range of soil temperatures measured (45–55 °C), sward height as measured by the drop disc showed a weak positive correlation (r2 = 7%, n = 8, P < 0·05) with soil temperature. This was because an uneven microterrain also results in warmer soil surfaces (Thomas 1983) and, for reasons given above, the drop disc was more apt to overestimate short sward heights in rugged terrain.
These results indicate that the three methods that are most commonly used to measure turf heights in applied and pure ecology are all appropriate when the swards encompass a wide range of heights. We found negligible recorder effects in any method, and none was sufficiently time consuming to influence its choice over another. However, the values obtained using the different methods were not the same, and each method had some distinct advantages and drawbacks.
The sward stick samples maximum sward height over the smallest area (2 cm2) of the three methods, and thus gives the most variable results (Figs 1, 2 and Table 1). The benefit is that quite subtle variation in the architecture of the sward surface is measured. Although not tested independently, we conclude that this is the most appropriate quick method for recording structural heterogeneity within swards, especially if several measurements are made at fixed points within a 1-m2 quadrat. For example, this type of method is ideal for measuring the diversity of niches available to web-spinning spiders (Gibson, Hambler & Brown 1992) and other small animals. On the other hand, we found the sward stick to be unsuitable for recording variation within short vegetation (Fig. 2; see the 40–55 °C range of soil temperatures). In separate tests using herbage dry weight as an independent parameter, Dowdeswell et al. (2000) also found that the correlation with turf height was ‘poor’ (r2 = 39–60%) using this method, in this case especially ‘at heights > 15 cm [when] the relationship was largely lost’. Although the sward stick is clearly adequate for the general monitoring of grazing levels in improved swards for which it was designed (Bircham 1981), the greater variance in its results means that a larger number of samples is required to obtain the same reliability as either of the other methods. It is also rather more expensive for projects involving many recorders, and communal training in the field is required to achieve consistency between recorders.
The direct method samples average sward height over an area of about 10 cm2. It gave the most consistent results for measuring tall and all turf heights compared with an independent variable (soil temperature), and was the only method suitable for measuring variation in short (< 3–5 cm) turf. Despite the perceived subjectivity of this method, only 3% of the variation between the four sets of 50 blind measurements made by the four recorders was attributable to ‘recorder effects’, although communal training in the field is necessary if several recorders are involved. We therefore recommend the continued use of this method in ecological research, and consider it to be the only suitable method where sward height is measured as a surrogate for microclimate.
The drop disc measures average sward height over the largest sample area (707 cm2). It is therefore the least suitable method for measuring microheterogeneity in sward architecture, and it proved completely unsuitable for measuring variation in short turf; indeed, it can generate misleading results (Fig. 2). However, it scored highly in all other situations. It is objective, consistent, cheap and simple to use, and does not require group training beforehand. Moreover, in other independent tests Castle (1976) and O’Riordan (2000) generally obtained much closer correlations (r2 = 80–90% and 70–80%, respectively) between herbage dry weight and sward height measurements made using the drop disc than obtained (on other sites) by Dowdeswell et al. (2000) using the sward stick, although, like us, Castle (1976) found the drop disc to be less suitable with well-grazed vegetation. Overall, we conclude that the drop disc is the most suitable method for measuring productivity and vertebrate herbivory, and for large-scale monitoring of sward height on nature reserves and on land managed under agri-environment schemes.
A serious problem exists when research results, recommendations or assays involve measurements made by more than one method. The sward stick consistently gives higher absolute values than either other method and, apart from in short turf (for which it is unsuitable), the drop disc gives the lowest values (Fig. 1). Diack, Burke & Peel (2000) recorded rather greater discrepancies between sward stick and drop disc measurements in swards of fine grasses. This has important implications in applied ecology and conservation, and must be taken into account when turf heights based on other people's research are prescribed or assessed. For example, the widely cited turf height recommendations for the optimum niches of various UK butterfly species, made by J.A. Thomas in BUTT (Butterflies Under Threat Team; Anonymous 1986), were based on measurements made using the direct method, yet the same NCC publication advocates the use of only the drop disc to monitor whether desired sward heights have been achieved. Our new results suggest that this discrepancy will have resulted in site managers grazing too lightly for endangered ‘short turf species’ (e.g. Lysandra bellargus and Hesperia comma) and too heavily for species inhabiting the tallest niches (e.g. Thymelicus acteon) (Bourn et al. 2000). The opposite problem exists with the DEFRA-recommended sward stick, where prescribed sward heights for butterflies (again derived from BUTT; Anonymous 1986) are 3 cm too low at all turf heights < 10 cm tall and about 25% too low for swards > 10 cm tall. In this case, use of the sward stick would result in a grazing regime that could be up to 40% too heavy to achieve the stated conservation aims of agri-environment schemes. This would have a catastrophic effect on sensitive short-turf invertebrates (Thomas 1983; Thomas et al. 1986, 1998), many of which are of high conservation value across northern Europe (Thomas & Morris 1994), whereas most ‘tall sward species’ are more robust. An even greater discrepancy occurs between measurements made using sward stick (advocated by DEFRA) and drop disc (advocated by EN).
In conclusion, ecologists, conservationists and land managers measure sward height and structure for different reasons, and these differences are reflected in the development of three different methods. Our study suggests that great care should be taken over selecting an appropriate method for each purpose, and that even greater care should be taken when comparing data obtained using the different methods. It is important that future publications state which method was used to measure sward height. At present, more than half the papers published do not. In some cases, we suspect the omission occurred because authors feared criticism for having used the apparently subjective direct method. We hope that future practitioners will be reassured by our results.
We thank R. Hobson for help with field tests, R. T. Clarke for statistical advice, and M. S. Warren, J. M. Bullock and R. F. Pywell for their valuable comments. K. Stewart was funded by a BBSRC research grant, and other aspects of the work were part funded by English Nature, Countryside Council for Wales and Scottish Natural Heritage.