Proc. 56th Southern Pasture and Forage Crop Improvement Conference, Springdale, AR April 21-22, 2001

The Conduct of Grazing Trials:

Rationale, Treatment Selection, and Basic Measurements

L.E. Sollenberger les@gnv.ifas.ufl.edu and J.C. Burns

University of Florida, Gainesville, FL, and

USDA-ARS North Carolina State University, Raleigh, NC

INTRODUCTION

The great majority of ruminant livestock in the southern USA obtain a high proportion of their nutrients from grazed pasture. Therefore, research aimed at improving the productivity and profitability of these systems has focused on plant-animal interactions under grazing. Grazing experiments are required to define input-output relationships that cannot be quantified in laboratory, greenhouse, or clipping studies, but there are many challenges to proper conduct of grazing studies. These include 1) availability of sufficient and suitable livestock, land, and equipment, 2) difficulties associated with measurement of key variables including herbage mass, diet selection, intake, and changes in animal liveweight, and 3) the resultant limitations in power of statistical tests and ability to establish causal relationships. The objective of this paper is to establish a rationale for grazing trials, describe the types of trials that may be conducted, establish a rational process for treatment selection, and outline a minimum set of measurements that should be part of most grazing trials. This information is intended to provide a framework for a subsequent paper which considers the merits of more detailed measurements of pasture and animal responses in grazing trials (Burns and Sollenberger, 2001).

WHY CONDUCT GRAZING TRIALS?

Grazing experiments provide animal-based comparisons of potential new cultivars, supply guidelines for practical grazing management, and may provide basic information on the biology of grassland ecosystems. Plant-animal interactions on pasture are very complex, and grazing experiments can define input-output relationships that cannot be quantified satisfactorily in laboratory, greenhouse, or clipping studies.

Measurement of animal performance is often the key objective in grazing trials, but quantifying it alone is seldom sufficient to provide an understanding of the biological relationships underpinning the animal response. Often, a thorough characterization of pasture mass, botanical composition, and nutritive value are required to explain the responses observed. If results of a grazing trial are to be extrapolated beyond a particular experiment or environment, understanding the relationship of animal response to pasture characteristics is critical.

TYPES OF GRAZING TRIALS

Grazing trials may be categorized as demonstration or formal. After a brief discussion of demonstration trials, the focus of this paper will be on formal trials.

The objective of demonstrations is to illustrate animal response to several pasture treatments or grazing methods for a producer or extension audience, but explaining the biology of the response is less important. Demonstrations often have relatively large numbers of animals per pasture, and treatments (pastures) will generally not be replicated. Demonstrations often follow formal grazing trials as a means of extending the results to a broader audience or demonstrating a technology on a scale more nearly approaching that of a farm. For example, research conducted on replicated small pastures (e.g., 1 ha) has documented the potential of two novel pasture systems for livestock production and has explained that potential based on forage nutritive value, forage intake, and seasonal distribution of forage production. It may be appropriate at this juncture to scale up these pasture systems and a control treatment to 20 ha each, assign cattle, and quantify the animal response to the three treatments across a growing season or longer period of time. In some cases, demonstration trials result accidentally when a researcher fails to recognize that pasture is the experimental unit in a grazing study and does not replicate it. This issue will be discussed more fully later.

Formal grazing trials are typically replicated experiments that are designed to test hypotheses. Formal grazing trials may fit into categories of “what happened” trials and “why did it happen” trials. In the former case, the primary question to be answered is which treatment is best. The biological reason for the response is of less importance to the researcher. In the “why did it happen”case, understanding the biology of the response is an integral part of the trial, and relationships between animal production response and pasture characteristics, or animal production and more detailed animal measures are likely to be explored in depth.

In most formal grazing trials, the pasture is the experimental unit (Sollenberger and Cherney, 1995). An experimental unit is defined as the smallest unit of experimental material to which a treatment can be allocated independently of all other units. When a treatment is applied, for example a forage cultivar or a N fertilizer rate, it is applied to an entire pasture, thus, pasture is the smallest unit of experimental material. When weight gain or plasma urea nitrogen are to be measured, the individual animal is the sampling unit and differences between or among animals within a pasture can be tested, but despite the fact that data are obtained for each animal, animal is not the experimental unit. Likewise in feeding trials in pens, the experimental unit is the pen not the animal. In grazing trials replication generally occurs in both space (another land area) and time (another year), with replication in space considered true, or a valid replication of the experiment. Some researchers have conducted non-replicated (in space) trials, repeated them for several years, and then considered year to be a replicate in the analysis. If environment (rainfall, temperature) varied among years there may be a treatment by year (replicate) interaction. Because this interaction is the error term for testing treatment effects in these studies, a large mean square for this term in the model makes detecting treatment effects difficult.

Others have proposed non-replicated, multiple stocking rate or grazing level experiments as a means of reducing cost of grazing trials while describing the animal performance response to a range of pasture conditions (Bransby et al., 1988). This approach remains controversial and will be discussed in more detail later.

METHODOLOGY CHOICES IN FORMAL GRAZING TRIALS

There are a range of methodology options to be considered when designing grazing trials. Choices include use of fixed or variable stocking rate, use of single or multiple levels of the measure of grazing intensity (i.e., stocking rate, herbage mass, stubble height, or herbage allowance), and use of replicated versus non-replicated experiments.

Fixed or Variable Stocking Rate

 

The question of whether to use fixed or variable rates has generated heated and passionate discussion throughout the last four decades. Most USA researchers including Gerry Mott and Roy Blaser argued for the use of variable stocking rates, while those from Australia, including Raymond Jones, argued just as strongly for fixed stocking rate experiments. The paper of Wheeler et al. (1973) attempted to add perspective and balance to the debate. They pointed out that both fixed and variable stocking methods have a place and that the controversy over the validity of the two methods “appears to have stemmed from the view that in all applications one was valid and the other invalid”. So the logical question to ask is when should these approaches be used. In discussing this issue, we will describe the two approaches, indicate when they have traditionally been used, and outline their advantages and disadvantages.

General Description and Traditional Applications

Fixed stocking rate. Fixed stocking rate refers to the practice of assigning to pasture a constant number of animals per unit land area for an extended period of time, e.g., a grazing season, year, or during the course of an experiment. It does not imply that continuous stocking is the grazing method used. Indeed fixed stocking rates can be applied to rotationally grazed pastures. In reality stocking rate is rarely truly fixed, at least in terms of liveweight per unit land area, because of changes in animal weights during the course of the trial. When using a fixed stocking rate approach, it is recommended to include more than one rate because the optimum may vary from one year to the next due to differences in temperature or rainfall. Prior experience with the forage is desirable in fixed stocking rate experiments, otherwise the range of stocking rates chosen may not include the optimum. Fixed stocking rate studies have traditionally been used when 1) forage does not show large seasonal fluctuations in growth, 2) excess forage can be preserved on the stalk for later consumption, 3) economically viable animal systems do not require consistent high levels of animal performance, and 4) long-term or rangeland studies are conducted (Wheeler et al., 1973)

Variable stocking rate. In a variable stocking rate trial, stocking rate is adjusted to maintain an equilibrium state between herbage mass and the requirements of the animal during the course of a season, year, or experiment. Most often a sward canopy characteristic such as herbage mass, leaf herbage mass, or canopy height is used to determine if stocking rate should be adjusted. Tester animals remain on the pasture throughout the entire period of the experiment, and animal performance is measured using these animals. The testers are carefully selected to 1) represent the target population of animals about which information is needed, and 2) be as uniform as possible with respect to age, weight, breed, conformation, and previous treatment. In some cases, however, testers that do not represent the target population may be chosen because they provide a sensitive biological assay. For example, forages for use in cow-calf systems may be evaluated using weaned steers as testers because the steers’ liveweight response is sensitive to changes in forage nutritive value. Put-and-take animals are moved on and off pastures as needed to maintain the desired pasture attribute (grazing pressure, height, mass, etc.). They need not be as uniform as the testers nor necessarily like the testers because their primary use is to determine average stocking rate (others use the term carrying capacity). Variable stocking rate can also be imposed by actually or effectively adjusting the size of the pasture. The actual size can be adjusted by harvesting surplus forage from a given area of the pasture and reducing the land area available for grazing. Size can be adjusted effectively by bringing in forage harvested from another area of land. Variable stocking rates have their greatest application where 1) there are marked seasonal changes in forage growth, 2) excess forage must be harvested mechanically or used in the field by the grazing animal because it cannot be preserved in the field for later use, 3) animals consistently need to be maintained in an above-maintenance condition, 4) in planted pasture versus rangeland studies, 5) in short-term rather than year-round studies, and 6) in temperate-humid areas as opposed to more arid regions (Wheeler et al., 1973).

Mott (1960) indicates that stocking rate adjustments should occur frequently in variable stocking rate trials under intensive management in humid areas or under irrigation. We suggest that adjustments in animal number correspond with the growth curve of the particular pasture being tested. As an example, in Florida where there is typically a single peak in growth rate of warm-season forages, it is our goal to add a put and take animal only once and remove it only once, i.e., it is added as growth rates increase during spring/early summer and it is removed as growth slows in late summer/autumn.

Advantages and Disadvantages

Fixed stocking rate. Advantages of fixed stocking rates are 1) once stocking rates are determined, researcher subjectivity does not enter into stocking decisions, 2) management of experiments is simplified and may require less observation and fewer trained personnel, and 3) the results may be more applicable to existing production systems in many areas of the world. Disadvantages are that 1) stocking rates selected may not be near optimum for the forage being evaluated, 2) seasonal and annual changes in environmental conditions cause large fluctuations in herbage mass and allowance for a given treatment, 3) forage quality is often not measured because quantity may limit gain during parts of the trial, 4) extrapolation of results from these studies to other environments is difficult unless herbage mass and nutritive value are quantified and reported, and 5) typically, at least three stocking rates per forage entry should be evaluated in a fixed stocking rate experiment causing the number of experimental units to be high.

Variable stocking rate. Advantages include 1) the likelihood of a single pasture management treatment (if only one can be evaluated) being in the optimum range is increased relative to a fixed rate approach, 2) large fluctuations in forage mass due to year or seasonal differences can be accommodated thus allowing forage quality to be measured, and 3) this approach is very useful when area of pasture or resources for a given experiment are severely restricted and only one level of grazing management can be imposed on each forage. Disadvantages include 1) variable stocking may be highly subjective if no previous studies were done to delineate the optimal range of grazing management for the forages being studied or if optimal levels of grazing management are better understood for one forage being tested than for a second, 2) if only one management treatment is used, there is no opportunity to detect the existence of genotype by treatment interaction, 3) the approach may produce results that are less practical for producers who lack interest in managing pastures or who cannot vary stocking rate across the season, and 4) experiments require regular observation by trained personnel and periodic adjustment of animal numbers.

Single or Multiple Levels of the Pasture Management Treatment Factor

Whether we are using fixed or variable stocking rates, the power of an experiment is increased if multiple levels of the management factor are imposed. Unfortunately in many situations, the primary determinant of whether single or multiple levels of the management factor are used is not the scientist’s preference or what will maximize knowledge gained, but instead it is the resources (land, cattle, budget) available to conduct the experiment. The fact that many grazing trials contain only one level of management speaks more to the limitations of funding than to what is the best experimental practice.

When during an evaluation program should a multiple-level experiment be conducted? This approach is likely to be impractical early in evaluation of a particular forage or forages. Instead a small pasture, plant response to defoliation stress study should precede it. This study can evaluate a wide range of management treatments much more economically, assess the merits of the forage for further evaluation or cultivar release, and narrow the range of management treatments for which a measure of animal response is needed. If animal response data under grazing are needed early in an evaluation program, then a single-level, variable stocking rate trial seems like the most cost-effective alternative. Later, after it has been well established that a forage has a niche in the production system, there may well be justification for a multiple-level experiment using either variable or fixed stocking. If the choice is made to use multiple levels, how many are needed? To have some reasonable power of interpretation, it is best to have at least three or preferably four levels of the treatment factor (e.g., stocking rate, herbage mass, canopy height). This allows one to bound the range of reasonable levels of the treatment plus have some intermediate level or levels that may be closer to the biological optimum. Further, the resulting response curve is valuable for identifying the economic optimum. This type of experiment is very costly and should not be undertaken without strong justification.

Replicated or Non-Replicated Levels of the Pasture Management Treatment Factor

Land replication has two important roles in experiments. It allows the estimation of experimental error, and this estimate becomes the basic unit of measurement for determining if means are different. Secondly, increasing numbers of replicates increases the precision of estimates of treatment effects and the power of the test for differences. Thus, it is clearly desirable to replicate treatments. The high cost of each additional experimental unit in a grazing trial has led some to propose that given the choice between land replication and evaluating multiple levels of a grazing factor, it is preferable to choose multiple levels (Bransby et al., 1988).

Without a true estimate of experimental error, how can these data be analyzed? Regression analysis is used, and deviation from linear regression is considered to be the error term. This is essentially a covariance approach (termed “heterogeneity of slopes” in Freund and Littell, 1981). For example, in a non-replicated experiment two species are being tested at three or more stocking rates (or herbage masses) per species. Animal performance is measured, and regressed on stocking rate. Then the slopes of the regression lines for the two species can be compared. If the slopes are different, then there is a species by stocking rate interaction, and we can do no further testing. If the slopes are not different, then the intercepts can be compared to determine if one species is superior to the other over the entire range of stocking rates imposed. It should be noted here that in regression analyses the extreme points (highest and lowest stocking rate in this case) are strongly influential and can greatly alter the slope. Any perturbation, such as armyworm invasion or periodic flooding, on either extreme treatment would cause large changes in the resulting slope and could lead to wrong conclusions. In contrast, if treatments are replicated, analysis of variance is straight forward and routine. If multiple stocking rates are used and there is a treatment by species interaction, species effects within a level of stocking rate can be compared statistically. Further, replication of pastures guards against extraneous perturbation greatly altering the conclusions reached. It is important to keep in mind that the issue is not which is better, replicated or non-replicated. Replicated trials are clearly superior. The question is, given limited resources which is more important, replication or data from multiple levels of stocking rate or herbage mass.

Of major concern in non-replicated trials is pasture to pasture variation in production, botanical composition, or nutritive value. If this variation is great, it may mask the effect of stocking rate or herbage mass treatments and result in non-significant regressions and an inability to draw conclusions from the trial. The result is a demonstration that probably has little utility. This is most likely to occur in areas where there are large differences in soil characteristics among experimental units. Such differences often can be controlled by blocking in a replicated trial.

CHOICE OF THE MANAGEMENT FACTOR IN

VARIABLE STOCKING RATE EXPERIMENTS

Mott (1960) stated that “in a grazing trial it is necessary to adjust stocking rate to provide equal grazing pressures on all treatments and replications, since failure to do so may bias performance per animal and per acre”. Grazing pressure is defined as “the relationship between the number of animal units or forage intake units and the weight of forage dry matter per unit area at any one point in time” (The Forage and Grazing Terminology Committee, 1992). Thus, a measure of grazing pressure is kg of animal liveweight per kg of forage dry matter. Equal grazing pressure is somewhat difficult to achieve in grazing trials because it requires frequent measurement of forage mass. Also, because pastures and consequently animal numbers are relatively small, any change in animal number will have a rather large effect on grazing pressure. For these reasons, some scientists conducting variable stocking rate studies have chosen to use other management factors in comparisons of forage species. These include herbage mass, leaf herbage mass, and stubble height. In the following section, these factors along with grazing pressure and herbage allowance will be discussed.

Grazing Pressure or Herbage Allowance

One advantage of using grazing pressure or herbage allowance is that they incorporate aspects of both the plant and animal. Thus, they provide a more complete picture of the balance between herbage mass and stocking rate, and they can be used meaningfully in studies making comparisons across species. Grazing pressure is not synonymous with stubble height or herbage mass as is sometimes seen in the literature. Herbage allowance, the inverse of grazing pressure (kg of forage per kg of animal weight), is often more easily understood and likely should be used instead of grazing pressure.

Herbage Mass

Herbage mass or leaf herbage mass are very useful measures in grazing trials, but comparing forage species at the same herbage mass may or may not be rational. The more nearly alike are the growth habit and growth rates of the species being compared in a grazing study, the more likely that maintaining a constant herbage mass on all pastures will result in meaningful comparisons. If, however, the growth habits (e.g., Pensacola bahiagrass and Mott elephantgrass) or growth rates are very different, then comparisons at the same herbage mass may not provide useful data because they will likely favor one species over another or, at the very least, result in quite different herbage allowances. For example, work in Florida compared Florakirk and Tifton 85 bermudagrasses at the same herbage mass. Even though they are relatively similar in growth habit, Tifton 85 has a faster growth rate, was stocked at a higher rate, and as a result had lower herbage allowance when herbage mass was the same as for Florakirk (Pedreira et al., 1998). Thus, use of constant herbage mass for comparisons across species must be done with great caution.

 

Another issue relative to use of herbage mass is to what height should it be measured. It has been argued by many that clipping to soil level provides the most repeatable results. This may be true, but for some C4 grasses it may cause weakening of the stand or provide opportunity for weed invasion at sampling sites. Also, herbage in the bottom 5 to 10 cm may never be grazed in some warm-season grass pastures. We propose that pastures be sampled to the lowest depth at which grazing occurs on the pasture. So, if average pasture height under continuous stocking or average postgraze stubble height under rotational grazing is 20 cm, but some areas of the pasture are grazed to 5 cm, then determining herbage mass above a 5-cm height may be reasonable.

Canopy Height

Similar to the discussion for herbage mass, use of a fixed canopy height across species may bias results in favor of one species or another. This is most likely to occur when the growth habits of species being compared are very different. In general, canopy height as a treatment is best used in multiple levels within a cultivar to determine the optimum grazing management. Canopy height can be a useful pasture descriptor when different species are being compared, however, the heights chosen may need to vary among species, should be based on the results of previous research/experience, and should account for differences in growth habit, herbage mass, bulk density, and grazing tolerance among species.

A MINIMAL SET OF MEASUREMENTS IN GRAZING TRIALS

In their comprehensive review, Burns et al. (1989) suggest that herbage mass, green leaf mass, diet nutritive value, herbage density, and botanical composition be measured in all grazing trials. They provide other lists of responses that are suggested for “additional explanatory purposes” and “maximum explanatory purposes”. As they indicate, there is a large menu of plant and animal responses that could be quantified in a grazing trial. We suggest that the list chosen for a particular study be driven by the nature of the problem being studied and the researchers’ objectives. In the following paragraphs, we will briefly address the choice of responses to be measured as related to the objectives of the study.

“What Happened” Trials

In this case, the primary question is which treatment is best, i.e., which forage results in greatest animal performance. Implied in this objective is less concern with explaining why. Thus the list of responses measured will likely be less comprehensive. The danger to the researcher asking which is best is that grazing trials are notorious for their inability to answer that simple question. Why? Because there are few treatments, few replicates, and relatively large coefficients of variation associated with the response variables of interest. So, the researcher who is content to measure only average daily gain, average stocking rate, and gain per hectare in a liveweight gain study faces the very real prospect of being unable to distinguish among treatments as well as not having supporting data to explain the responses observed. What are the minimum responses for studies of this type? At least monthly measures of herbage mass (continuous stocking) or pregraze and postgraze herbage mass (rotational grazing) are needed as well as canopy height and an estimate of nutritive value of the diet consumed. The latter can be approximated using a hand plucking technique (Sollenberger and Cherney, 1995). If the sward consists of more than one species, then botanical composition should be measured at least seasonally, e.g., once in spring, once in summer, and once in fall. Although these pasture data may go beyond the original objective of telling us which treatment is best, they can be obtained relatively cheaply compared to the overall cost of the experiment and will allow a much more meaningful synthesis of the results. For example, with these data we can plot the relationship between herbage mass and gain, herbage bulk density and gain, herbage allowance and gain, and herbage nutritive value and gain. All of these additional response variables have explanatory power that may enable us to understand the animal response.

“Why Did it Happen” Trials

Some grazing trials are undertaken with the intent of gaining a deeper understanding of the mechanisms at play in forage-livestock systems. Knowing what happened is only a starting point in these trials. In addition to the responses already mentioned, detailed characterization of sward canopy structure and chemical composition, animal grazing behavior, and intake are often a part of this type of trial. Burns and Sollenberger (2001, this volume) develop the rationale and describe the processes involved in gathering these more detailed response data.

SUMMARY AND CONCLUSIONS

Grazing experiments provide animal-based comparisons of potential new cultivars, supply guidelines for practical grazing management, and may provide basic information on the biology of grassland ecosystems. Measurement of animal performance is often the key objective in grazing trials, but quantifying it alone is seldom sufficient to provide an understanding of the biological relationships underpinning the animal response. There is a large menu of plant and animal responses that can be quantified in grazing trials, but we suggest that the list chosen for a particular trial be driven by the nature of the problem being studied and the researchers’ objectives. When the primary question is which treatment is best, i.e., which forage results in greatest animal performance, the list of responses measured will likely be less comprehensive than when the primary question is what are the mechanisms driving animal response. When the focus is on which treatment is best, measurements should include at least monthly measures of herbage mass, canopy height, and diet nutritive value. If the sward consists of more than one species, then botanical composition should be measured at least seasonally. Although these pasture data may go beyond the original objective of telling us which treatment is best, they allow meaningful synthesis of the results. In a companion paper, Burns and Sollenberger (2001) discuss responses to be measured in more mechanistic studies.

 

REFERENCES

Bransby, D. I., B. E. Conrad, H. M. Dicks, and J. W. Drane. 1988. Justification for grazing intensity experiments: Analysing and interpreting grazing data. J. Range Management 41:274-279.

Burns, J.C., H. Lippke, and D.S. Fisher. 1989. The relationship of herbage mass and characteristics to animal responses in grazing experiments. p. 7-19. In G.C. Marten (ed.) Grazing research: Design, methodology, and analysis. CSSA Spec. Publ. No. 16, Madison, WI.

Burns, J.C., and L.E. Sollenberger. 2001. The conduct of grazing trials: Measurements explaining why animal response differences occur. In D. Lang (ed.) Proc. Southern Pasture Forage Crop Improvement Conf.

Freund, R. J., and R. C. Littell. 1981. SAS for linear models. SAS Institute, Inc., Cary, NC.

Mott, G. O. 1960. Grazing pressure and the measurement of pasture production. p. 606-611. In Proc. 8th Int. Grassl. Cong., Surfers Paradise, Australia.

Pedreira, C.G.S., L.E. Sollenberger, and P. Mislevy. 1998. Performance of yearling cattle grazing ‘Tifton 85′ and ‘Florakirk’ bermudagrasses. Agron. Abst. p. 150.

Sollenberger, L.E., and D.J.R. Cherney. 1995. Evaluating forage production and quality. p. 97-110. In R.F Barnes, D.A. Miller, and C.J. Nelson (eds.) Forages: The science of grassland agriculture (vol. 2), Iowa State Univ. Press, Ames, IA.

The Forage and Grazing Terminology Committee. 1992. Terminology for grazing lands and grazing animals. J. Prod. Agric. 5:191-201.

Wheeler, J.L., J.C. Burns, R.D. Mochrie, and H.D. Gross. 1973. The choice of fixed or variable stocking rates in grazing experiments. Exp. Agric. 9:289-302.