Reductionism and synthesis in the grazing sciences : models and experiments

Abstract

Proc. Aust. Soc. Anim. Prod. 1994 Vol. 20 INVITED REVIEW REDUCTIONISM AND SYNTHESIS IN THE GRAZING SCIENCES: MODELS AND EXPERIMENTS M.W. DEMMENTA and E.A. LACA B *Dept of Agronomy and Range Science, University of California, Davis CA , U.S.A. 95616 BDcpt of Range and Wildlife Management, Texas Tech University, Lubbock, TX, U.S.A. 79409-2125 SUMMARY The advancement of the grazing sciences can be accelerated by a balance of reductionist studies of mechanisms and the integration of these results to develop predictive models at hierarchical levels that are relevant for management. Foraging theory, including optimality approaches and new approaches that integrate nutrient components and spatial heterogeneity of plants, offers an integrating framework that will efficiently organize such research. We demonstrate that herbage mass availability has many dimensions that may show different relationships across or even within vegetation types. The shape and parameters of the relationship between intake rate and herbage availability depends on how these dimensions covary as herbage mass changes. These patterns of covariance are typically different depending on whether herbage mass increases by growth early in the season, growth late in the season (differentiation), defoliation of a vegetative sward, or defoliation of a reproductive or differentiated sward. A series of experiments quantify the relationship between sward structure and bite characteristics, determine bite dimensions as patches of vegetation are consumed, and integrate these mechanisms by an optimal foraging model to successfully predict behaviour in heterogeneous environments. Keywords: grazing, intake, optimal foraging, ingestive behaviour, heterogeneity. INTRODUCTION 0 The nature of research on grazing systems is changing from a focus at higher to lower levels of organization. Last year at the International Grasslands Congress this shift in emphasis was a prominent point of discussion in the grazing session at Palmerston North. There was some clear discomfort with the large number of papers focused on lower level phenomena centred on individual animals, bites and tillers and fewer at the whole pasture and herd level. Some participants questioned the role of the microfocus of the research in addressing management problems occurring at much higher levels in the hierarchy of organization within these systems. We address this question and try to give some examples from our work of how micro- and macro-level can be connected to advance understanding of grazing systems. A parallel debate about the appropriate level of focus is occurring in the plant sciences related to agriculture (Demment and Laca 1993). Over the last 30 years, in response to national and global requirements for food, research in the plant sciences in the United States has been concentrated on the processes controlling production. The constraints on production at the systems level have remained relatively consistent over this period and this consistency has allowed researchers to focus on processes at progressively lower levels within production systems. This research environment has encouraged 0 reductionist science. Because the behaviour of cropping systems beyond their production was not a critical issue, there was little urgency to integrate this rich literature at the process level to understand systems' behaviour. The recent concerns about the environmental impact and long-term performance of cropping systems 0 pose new questions and impose new constraints operating on croppingsy stems. These questions refocus attention to the systems level and test the capabilities of reductionist science to integrate basic science to solve complex problems in cropping systems. The ability of a discipline to respond at this level reflects the balance between reductionism and synthesis. Advances in grazing science have also suffered from an antithetical imbalance in research direction. In this science the emphasis has been on the phenomenological not the mechanistic, the higher not the lower level processes. Research has often measured response to treatments without a conceptual framework of processes that link the two. For example, stocking rates and animal production have been investigated thoroughly but without an understanding or even a conceptual model of how the animal interacts with the vegetation when it feeds, how this interaction affects the dynamics of the pasture, and . how animal density affects individual animal behaviour. Such studies are performed in a wide array of 6 Proc. Aust. Sot. Anim. Prod. 1994 Vol. 20 conditions and we attempt to extract fundamental principles and generalities from these collective data with little understanding of the key processes that link our treatments to the animals' response. Thus much of the inquiry at higher levels of organization is not synthesis, since it has not integrated lower level processes, but merely phenomenological work done on high level processes. The identification and quantification of fundamental mechanisms is critical to our understanding of phenomenological relationships and the initial step to understandin, the behaviour of complex systems. Without such 0 understanding or models, how do we choose the appropriate variables or contexts for comparison? We argue that the advancement of animal production on grasslands requires a balanced research focus that incorporates a major effort to understand the processes that control the plant-animal interface. This effort must identify and quantify processes and also integrate those processes across the hierarchies of organizations within these systems. The challenge is not just to determine how bite size is determined, both by the animal and the plant structure, but what determines how those bites are placed in space and time. The identification of the mechanics of grazing falls comfortably in the established methods and techniques, while the process of integration requires new paradigms and strong interdisciplinary interactions that require individuals to adjust and reshape their programs. In this paper we present examples of a model and experiments that demonstrate how we developed a conceptual framework for our work, identified fundamental mechanisms, and integrated our results at the bite level to explain and predict animal behaviour within different experimental environments. FUNCTIONAL RESPONSE 0 In order to develop a comprehensive understandin, of intake we must choose a suitable model that organizes our understanding of the relationships between processes and that provides a framework for their integration to intake. The functional response model is an example of a phenomenological representation of the relationship between intake and forage abundance. We develop a mechanistic model that represent the underlying processes that generate this phenomenon. This approach allows the identification of key processes and variables affectin, intake for experimental investigation. With a 0 conceptual basis for the integration of mechanisms to explain high level phenomena, this approach holds greater potential for generalized predictive capability than a solely empirical approach. Indices of animal performance and daily intake have been related to measurements of the forage resource such as total herbage mass/unit area (Poppi et al. 1987), mass of selected species or plant parts/unit area (Chacon and Stobbs 1976; Wickstrom et al. 1984; Forbes and Coleman 1993) and pasture height (Trudell and White 1981; Barthram and Grant 1984; Arnold 1987). In areas where management is intensive and animals graze in small pastures for short periods, animal performance has been related to measures of forage available/animal.unit time (herbage allowance, Gibb and Treacher 1976, 1978; Marsh 1977; Jamieson and Hodgson 1979). Typically, researchers have attempted to distil the complexities of sward structure to a single me'asure based on the simple but limited concept that more forage equals more food. In ecology, the relationship between food supply and intake rate has been termed functional response 0 (FR), and it was initially developed for insectivorous predators (Hollin, 1959). The functional response of herbivores has been used phenomenologically as a component of systems models to explain the behaviour of grazing systems (Noy-Meir 1978). Empirical derivations of the relationship have been useful qualitative tools for range and wildlife managers. In spite of the widespread use of the traditional FR model in agriculture and ecology, it has failed to provide a valuable management tool to make quantitative predictions for purposes of managing animal production and animal impacts on the vegetation. Often, these models are fitted statistically without much consideration of the underlying mechanisms, and the curves derived are quite site specific. For example, although we know that in strip-grazing systems one must use herbage allowance, and in setstocking situations one must relate intake to herbage mass/unit area, we lack a quantitative understanding of why this is the case. We do not know at what level of animal density one should switch from allowance to availability. Although widely used, the empirically developed FR has failed to generate concepts with generality and predictive power. Recent advances in the techniques used to study these plant-animal interactions (Black and Kenney 1984; Spalinger et al. 1988; Astrom et a2. 1990; Illius and Gordon 1990; Laca et al. 1992b) have emphasized the need for new mechanistic models to aid in designing experiments, deciding which variables to measure, and to identify mechanisms for correct interpretation of experimental results. We suggest that advancement in grazing research can be achieved by reductionist studies if these studies are part of a plan that includes formal integration and synthesis of the results at scales of space and time that are relevant for community processes (Demment and Laca 1993). While functional response is a mechanism that explains higher order phenomena, it is itself a Proc. Aust. Sot. Anim. Prod. 1994 Vol. 20 phenomenon to be explained. Models of grazing behaviour (Illius and Gordon 1987; Ungar and NoyMeir 1988) derived the functional response on the basis of lower-order variables such as bite weight and biting rate, which in turn can be analyzed into bite depth, bite area, bulk density, search time and handling time. These models are more complex and detailed than the traditional model, but they have the potential to explain a wider range of situations because they underscore the multidimensional nature of food availability. Additionally, these models explicitly take into account the major differences between the food-gathering process of large herbivores and the predators for which the original model was developed. Most large herbivores face a food environment with relatively high density of relatively low quality food, and they must spend a great portion of their time gathering and chewing fibrous materials from which to extract nutrients. The trade-off between rate of intake and passage or digestion, and diet quality is central in the foraging process (Demment and Greenwood 1988; Verlinden and Wiley 1989). Experimental results and mechanistic models of grazing behaviour have indicated the importance of spatial structure and heterogeneity of the vegetation in determining intake rate and diet selection (Black and Kenney 1984; Clark and Harris 1985; Demment et al. 1987; Arnold 1987; Demment and Greenwood 1988; Ungar and Noy-Meir 1988; Laca and Demment 1991). Natural grassland communities and seeded pastures exhibit spatial variability in both quantity and quality (Shiyomi et al. 1983, 1984). Spatial heterogeneity affects growth, utilization and intake (Penning et al. 1991a,b; Forbes and Coleman 1993) but this characteristic is not captured by measures of total or average herbage availability. Realistic models and experiments that explicitly take into account the spatial variability of vegetation quantity and quality are needed to explore the role of heterogeneity on intake and animal performance (Laca and Demment 1991; Laca et al. 1993). Table 1. Functional response model. Equations, variables and parameters 8 Proc. Aust. Sot. Anim. Prod. 1994 Vol. 20 We present a simple grazing model that relates intake rate while grazing to sward height, bulk density of the sward, percentage cover, and selectivity. The main goal is to show that the relationship between herbage mass/unit area and intake rate is strictly dependent on the spatial arrangement of the forage in the vertical and horizontal dimensions Intake rate can vary dramatically, even if average measures of forage availability remain constant, as a result of changes in other variables. Functional response model Equation 1 (Table 1) indicates that instantaneous intake rate of herbage DM (IIR) is a function of bite weight, time cost involved in searching for and locating each acceptable bite, time for bite formation and prehension, and chewing time before the bite is swallowed. We assumed that searching and chewing can overlap. Thus, only the maximum of these 2 time costs is relevant. This is contrary to the standard assumption of foraging models by which searchin g and handling are mutually exclusive behaviours (Stephens and Krebs 1986). However, observation of grazing ungulates clearly shows that these animals usually chew a mouthful while they move from one feeding station to the next. Bite formation time is a constant/bite, whereas chewin, time is a constant/g of DM ingested. Bite weight was determined 0 experimentally with large steers (750 kg liveweight) as a function of density and height of the vegetation at the site of the bite (Laca et al. 1992a). The functional response model involves no optimization. Selectivity (proportion of total available biomass that is acceptable) is treated as an independent variable whose effects are explored. The real value that selectivity takes in a given situation likely depends on the internal state of the animal (eg. satiation, physiological state) and on the quality and degree of nutritional differentiation of the pasture. The model pasture is a mosaic of vegetated areas and bare ground. Within vegetated areas the vegetation is homogeneous in quality, height and density, and all vegetated areas have the same height and density. Selectivity represents the proportion of vegetated area that is acceptable. The relationship of searching time and bite weight to herbage availability is critical to the model' s performance. Searching time and bite weight are functions of lower level variables that also are determinants of total herbage mass/area. Thus, bite weight and searching time are parametrically related to herbage mass (Table 1). Two contrasting search efficiencies were posed as extremes of a continuum to examine the relationship between searchin g time/bite and forage availability. Animals with a low efficiency of search would discriminate bites only when they are in very close range by using touch, taste, smell and or sight of features that require proximity. These animals would discriminate bare ground from vegetated areas at a rate of 20 mZ/minute,and they would discriminate acceptable from non-preferred vegetation at a rate of 3 m2/minute. Animals with high efficiency of search would locate acceptable vegetation at any distance by sight, and perform a 'linear' search moving from one acceptable bite to the next at a rate of 50 m/minute. Figure 1 shows the effects of bite weight and search time/bite on IIR for realistic ranges of values. At very low values of search time the response to bite weight is similar to the traditional functional response. Figure 1. Relationship between instantaneous intake rate of herbage dry mass as function of travel time/bite and bite weight. This surface is calculated from equation 1 from Table 1. 9 Proc. Aust. Sot. Anim. Prod. 1994 Vol. 20 It showsan increasing phase with slope inversely related to the sum of the time for bite formation plus search time, and an asymptotic phase determined by the chewing constraint. As search time increases IIR decreases and the asymptotic phase is limited to the range of very large bite weights. The effects of varying herbage mass on IIR were tested by increasing height or bulk density of model swards. Herbage mass was increased by varying height of a sward with a constant bulk density of 1000 0 b DM/m3 (Figure 2, solid line) or by increasing bulk density of a pasture with a constant height equal to 0.15 m (Figure 2, dashed line). Both swards have an equal ground cover of 0.90 and consist of totally acceptable forage. For this comparison the animal was assumed to have a low efficiency of search; a high efficiency of search would have produced similar results. The functional response obtained by varying height is different from that obtained by changing bulk density (Figure 2). Below 135 g/m' the solid line represents swards that are denser and shorter than those represented by the dashed line, while the converse is true for herbage availabilities above 135 g/m? At equal herbage mass/area, tall and sparse swards yield higher IIR than short and dense ones. The importance of this effect is greatest in the range of low availability, where IIR obtained in a tall sward would be considerably greater than in dense swards. The critical mechanism responsible for this difference is the relationship between bite dimensions and sward characteristics, which determines bite weight. Increments in height would result in larger bite weight because both bite area and depth increase, particularly when bulk density is low. However, increments in bulk density would tend to reduce bite volume, which counteracts the positive effects of density on bite weight (Figure 2, thin lines). These results indicate that herbage mass by itself is not a sufficiently good predictor of intake to warrant its use ooests that cattle should prefer tall swards to in heterogeneous environments. Additionally, the model su,, short ones, even if herbage availability is the same or somewhat lower in the tall swards. The effects of herbage availability were also studied by varying the percentage of ground covered by patches of vegetation of constant characteristics. The functional response was generated for animals with low and high efficiency of search, and for 2 degrees of selectivity (all herbage acceptable or only 30% of the available biomass selected (Figure 3)). At all levels of biomass, vegetated areas had a height of 0.20 m and a bulk density equal to 1000 g/m 3. Therefore, bite weight was constant, while search time/bite increased as area covered bv selected biomass decreased. Figure 2. Effects of changes in sward height and bulk density on predicted instantaneous intake rate. Dashed line: sward of constant height = 0.15 m in which herbage availability increases by increasing bulk density. Solid line: sward of constant bulk density = 1000 g/m3 in which herbage availability increases by increasing height. In both cases cover equals 90%, and the animal has a low efficiency of search. There was a remarkable lack of response of predicted IIR over a wide range of herbage mass, particularly when all herbage would have been acceptable. This lack of effects of herbage availability on IIR was demonstrated experimentally by Spalinger et al. (1988) with deer, and by Gross et al. (1993) with lemmings. When bite weight is constant, density of bites does not affect IIR unless it reaches extremely low values. The mechanism behind this response is that changes in bite density have minor effects on search time, particularly when bite weight is large. Animals compensate for the lower bite density by chewing while walking and by ignoring bare ground between vegetated areas. 10 Proc. Amt. Sot. Anim. Prod. 1994 Vol. 20 A more typical functional response is observed only when the animal has a low efficiency of search and a small fraction of the total forage available is selected. This underscores the importance of these 2 factors in determining the shape of the functional response usually observed in pastures and grasslands. Figure 3. Effects of changes in ground cover and proportion of selected biomass on predicted instantaneous intake rate and rate of movement. A = 1.00 and A= 0.30 indicate that 100% and 30% of the vegetated area is covered by acceptable forage, respectively. Dashed line: low efficiency of search. Solid line: high efficiency of search. For explanation of each graph ((a) to (d)) see text. The rate of movement of the animal relative to herbage availability exhibits 2 phases (Figure 3c and 3d). At high levels of herbage mass (ground cover) movement would be limited by the time it takes the animals to bite and chew the acceptable biomass encountered. Since the animal with high efficiency of search would encounter more acceptable herbage/unit distance, it would move at a slower rate than the animal with low efficiency of search. In this range there would be perfect compensation between herbage availability and movement rate. At low levels of herbage available movement would be limited by the search rate of the animal. Thus, when all herbage is acceptable, both types of search would result in similar steppin, rate (Figure 3c), but when there is a large proportion of non-preferred herbage, the 0 animal with low efficiency of search would move more slowly because it has to sort through the unacceptable herbage that is ignored by the animal with high efficiency of search (Figure 3d). Movement rate responds to herbage mass over a wider range of values than IIR. Thus, stepping rate is an easily observed behaviour that enables detection of reductions in the cover by acceptable herbage before they have an effect on IIR. The steep reduction of movement rate as herbage mass increases from low values was observed by Spalinger et al. (1988). They interpreted it to mean that animals were compensating by walking faster between bites or reducing the search path width.. Our model indicates an alternative and perhaps more parsimonious explanation. As herbage availability decreases to very low values, animals exhibit a greater overall movement rate because they spend a greater proportion of the total foraging time walking between bites, whereas rate of movement between bites while searching and search path width remain constant. The response of predicted IIR to increasing ground cover by non-preferred herbage was tested while keeping availability of selected herbage constant at 1.0 g/m? Both selected and rejected patches had a height of 0.20 m and a bulk density of 1000 g/m3 (Figure 4). A high efficiency of search would result in much greater IIR than a low efficiency of search, and IIR would not respond to the addition of non-preferred herbage. This result suggests that a change in searching mode can be an important mechanism to compensate for declining herbage availability. Searching paths and modes should be studied directly by manipulating the spatial arrangement of acceptable and non-preferred patches at various levels of total ground cover, and performing detailed measurements of movement patterns (Gillingham and Bunnell 1989). 11 Pi-oc. Amt. Sot. Arlirn. Prod. 1994 Vol. 20 Figure 4. Effects of increasing herbage mass on predicted instantaneous intake rate by addition of non-preferred biomass while availability of selected biomass is maintained constant at 1.0 g/m2. All patches have equal height (0.20 m) and bulk density (1000 g/m3). Dashed line: low efficiency of search. Solid line: high efficiency of search. When efficiency of search is low, the model predicts a decline in IIR with increasing herbage mass, which is opposite to the traditional functional response. This phenomenon, by which non-preferred herbage hinders the search for acceptable bites, has been cited in the literature to explain why deer foraging in herbaceous meadows attain a lower IIR than when browsing on bites of comparable weight in clear-cut areas (Collins and Urness 1983). The same mechanism is responsible for the grazing facilitation that takes place when large grazers or bulk feeders remove low quality herbage mass before the grassland is grazed by more selective ungulates (Bell 1970; McNaughton 1976; Gordon 1988). The tests performed with the model su,, o*est the following conclusions. First, the response of IIR to herbage availability depends strongly on which components cause the changes in herbage mass. Intake rate was affected very little by factors such as ground cover, which determine density of bites/unit area, but it was very sensitive to factors that affect bite weight, such as sward height and bulk density. Moreover, when the discrimination of bites involves a time cost, IIR can actually decrease with increasing herbage mass/area if ground cover of non-preferred vegetation increases sharply. Second, the increase in movement rate as bite density becomes very low does not necessarily involve an active increment of the search speed of the animal; it is a passive result of a greater proportion of the grazing time being allocated to searching. This is an example of how a model aids in the interpretation of quantitative results. MECHANISMS AND SYNTHESIS: FROM BITES TO INTAKE While the interaction between an herbivore and its food supply is a primary process in the function of Orasslands, the dynamics of the plant-animal interface remain poorly understood relative to its 3 importance. Two points argue for a reductionist approach. First, the lack of comprehensive, general 0 models of grazing systems resultin, from pasture level experiments indicates the inadequacy of present models and phenomenological approaches to characterize these systems. Perhaps we need to know much more about lower level processes and how they integrate to larger scales to identify the basis for generality. Second, system level models and experiments indicate that lower level processes at very small scales have strong effects on the dynamics of grasslands (J.A. Newman pers. comm.; Bergelson et ooests that the study of small scale processes may in itself be critical to understand al. 1993). This su,, systems' behaviour, and that pasture level studies are too general to elucidate key interactions. The aim of our research was to quantify the mechanisms that determine intake rate by large grazers. Intake rate was partitioned into its components (Hodgson 1985) and the relationships between local sward characteristics and intake components were determined under controlled conditions (Ungar et al. 1991; Laca et al. 1992a,b). Then, the information at the level of bite dimensions and bite weight was integrated over increasing spatial and temporal scales by using models and performing new experiments at the appropriate level (Demment and Laca 1993). 12 Proc. Aust. Sot. Anim. Prod. 1994 Vol. 20 Figure 5. Hierarchical levels, processes and scales of time and space covered in the research. Sward characteristics at the site of the bite determine bite weight. Bite weight determines the rate of forage removal from the feeding station, and forage removal modifies sward characteristics at the bite site. As a consequence, intake rate declines with increasing time at the feeding station. This pattern of decline in intake rate, together with larger scale characteristics of the feeding habitat such as a bundance of patches of different types, determine the pattern of selection and utilization at the habitat scale. We studied the grazing process at 3 scales defined by the interaction of grazing behaviour and forage distribution in space: bite selection and formation, feedin, station utilization, and habitat grazing pattern 0 (Figure 5). Experiments and models at each scale allowed us to quantify and determine the importance of heterogeneity of the forage at all 3 levels (Gri,, et al. 1991; Distel elt al. 1991; Laca and Demment 1991). 00s In the present article we focus on the value of our approach to interpret and predict grazing behaviour; for a more detailed treatment of the importance of sward heterogeneity see Demment and Laca (1993). Bite scale Bite weight is the unit of intake for grazers and is the most important determinant of intake rate (Stobbs 1973; Hodgson 1985; Spalinger et al. 1988; Spalinger and Hobbs 1992). Chacon et al. (1978) found that bite weight was well correlated with weight gain of steers. Our understanding of what determines bite weight has been obstructed by 2 problems. First, it has been difficult to determine the independent effects of sward characteristics such as height, density and quality, because these variables are strongly correlated in natural swards. Second, establishing the relation between bite weight and sward characteristics has been complex, because the characteristics of the sward in the areas actually orazed typically differ from the average characteristics usually measured (Penning et al. 1991a,b; Forbes &d Coleman 1993). These problems were overcome by further developing Black and Kenney' (1984) s methodology. Swards were constructed by hand, and variables such as height, density, and vertical and horizontal heterogeneity in plant parts were independently controlled (Ungar et al. 1991; Griggs et al. 1991; Laca et al. 1992a; Flores et al. 1993). This technique was combined with an integrated methodology to accurately describe and measure grazing behaviour by cattle (Laca et al. 1992b). Experiments at the bite scale using this new methodology revealed the mechanisms of bite formation and allowed quantification of the relationship between sward characteristics and bite weight. In homogeneous swards, where vertical and horizontal selection was not a factor, bite dimensions resulted from the interaction of sward height, stiffness of plant units, and harvesting behaviours of the animal. Cattle appeared to respond to sward density by adjusting the number and amplitude of sweeping tongue movements to gather, as effectively as possible, the largest feasible bite area. Amplitude and density of tongue movements declined with increasing sward density, as did bite area. The effectiveness of the 'attempted' bite area was determined by sward height. Tillers whose distance from the rooting point to the incisor edge was greater than tiller length escaped the bite area. On tall swards of low density animals attempted to obtain a bite area as large as possible, and bite area was limited by mouth size and maximum tongue extension. When there were no restrictions on bite depth, animals seemed to bite to approximately 13 Proc. Amt. Sot. Anim. Prod. I994 Vol. 20 one half of sward height for a range of heights between 4 and 30 cm (Ungar et al. 1991; Laca et aZ. 1992a; Flores et al. 1993). The causes of this behaviour are not completely understood, but it seems to be common in cattle and sheep (Milne et al. 1982; Barthram and Grant 1984; Wade et al. 1989; Burlison et al. 1991). In agreement with results by Burlison et al. (1991) sward height and bulk density had independent effects on bite weight. Therefore, bite weight cannot be predicted only on the basis of herbage mass; both sward height and density must be taken into account. Bite weight was very well predicted on the basis of sward height, bulk density, and number of chews/bite (R' = 0.72). Hand-constructed swards (HCS) presented a unique opportunity to study the ability of cattle to select different plant parts, vertically and horizontally, at the scale of 1 to few bites. The response to vertical heterogeneit