S. D. Fuhlendorf, Rangeland Ecology and Management, Department of Plant and Soil Sciences, Oklahoma State University, 368 AGH, Stillwater, OK 74078–6028, USA (fax +405 744 5269; e-mail email@example.com).
1Management of rangelands has long operated under the paradigm of minimizing spatially discrete disturbances, often under the objective of reducing inherent heterogeneity within managed ecosystems. Management of grazing animals has focused on uniform distribution of disturbance, so that no areas are heavily disturbed or undisturbed (i.e. management to the ‘middle’).
2A model of the fire–grazing interaction argues that grazing and fire interact through a series of positive and negative feedbacks to cause a shifting mosaic of vegetation pattern across the landscape. This interaction was important in the evolution of species in the North American Great Plains grasslands. This approach has the potential to serve as an ecological-based model for management of grasslands with a long evolutionary history of grazing.
3We compared a heterogeneity-based approach, in which fire is applied to discrete patches, with typical homogeneity-based land management in the North American Great Plains, to determine if patch burning followed by focal grazing creates a shifting mosaic pattern of vegetation structure and composition.
4Our data suggest that spatially discrete fires promote focal grazing, where grazing animals devote 75% of grazing time within the one-third of the area that has been burned within the past year. These focal disturbances cause local changes in the plant community and increase patch-level heterogeneity across landscapes. As the focal disturbance is shifted to other patches over time, successional processes lead to changes in local plant communities and the patchwork landscape can be described as a shifting mosaic.
5A patch-dynamic approach can be accomplished in the tallgrass prairie through applying spatially discrete fires and allowing animals free access to a diversity of landscape elements that vary in time since focal disturbance. This increases heterogeneity across the landscape, a variable that has been shown to be critical to some wildlife species as well as the structure and function of grassland ecosystems.
6Synthesis and applications. Our study demonstrates that the fire-grazing model may be useful for generating heterogeneity in grassland management. Discrete fires are applied to patches, and patchy grazing by herbivores promotes a shifting vegetation mosaic across the landscape. Furthermore, application of the model has the potential of increasing the area of rangelands under management for conservation purposes, because livestock production is maintained at a level similar to traditional management. So, by managing transient focal patches that move through the landscape, heterogeneity has the potential to be a central paradigm for managing landscapes for multiple objectives, such as biodiversity and agricultural productivity.
Most approaches to managing native ecosystems are based on an equilibrium paradigm that rarely considers spatial or temporal variability (Fuhlendorf & Engle 2001; Briske, Fuhlendorf & Smeins 2003). Understanding the variability inherent within ecosystems or associated with variable patterns of disturbance can be critical in describing and managing the structure and function of ecosystems. Heterogeneity may actually be the root of biological diversity at all levels of ecological organization and should serve as the foundation for conservation and ecosystem management (Christensen 1997; Ostfeld et al. 1997; Wiens 1997). Therefore, it is important that management approaches apply state-of-the-art ecological theories that incorporate the role of spatial and temporal variability in the structure and function of ecosystems.
In grassland ecosystems, such as those in the Great Plains of North America, which evolved with fire and ungulate grazing, the frequency and intensity of disturbances are critical to ecological processes, biological diversity and heterogeneity across multiple spatial scales (Collins 1992; Fuhlendorf & Smeins 1999). However, most research of the effects of these processes has focused on the main effects of grazing and fire, with little or no attention given to their interaction in space and time. A model of the fire–grazing interaction (fire-grazing model) argues that grazing and fire, interacting through a series of positive and negative feedbacks to cause a shifting mosaic of vegetation pattern across the landscape, were operative in the evolution of Great Plains grasslands (Fig. 1) (Hobbs et al. 1991; Fuhlendorf & Engle 2001). According to the model, the probability of fire is greatest on areas with high biomass accumulation within a grazed grassland landscape. A positive feedback occurs when a recent fire event attracts grazing animals, which further disturbs the site. On tallgrass prairie landscapes grazed by bison Bos bison L., the most recently burned patches are preferentially selected from a diverse landscape that includes patches with variable fire histories (Coppedge et al. 1998; Coppedge & Shaw 1998). The model predicts that tall graminoid species decrease in dominance, and bare ground and forbs increase, on recently burned patches that are focally grazed. These changes in composition and productivity are associated with a negative feedback because focal grazing reduces biomass. This reduces the probability and intensity of fire, which in turn lowers the probability that the patch will be grazed. The grazing animal subsequently focuses on other patches that have been burned more recently (Coppedge, Leslie & Shaw 1998) and the tall graminoid species eventually recover dominance. So, the fire-grazing model predicts that grazing animals and fire interact through positive and negative feedbacks to cause a shifting mosaic. The landscape includes local patches that have been burned and heavily grazed, dispersed within a patchwork of areas in various states of recovery (Fuhlendorf & Engle 2001).
Heterogeneity associated with the fire-grazing model may be critical for conservation of many grassland species (e.g. grassland birds). However, management approaches that focus on increasing or maintaining heterogeneity have yet to be developed or empirically tested (Fuhlendorf & Engle 2001). Land management has long operated under the paradigm of minimizing spatially discrete disturbances, so reducing inherent heterogeneity. Grazing management of grassland ecosystems generally promotes uniform disturbance through uniform distribution of grazing animals within a year (Holecheck, Pieper & Herbel 2003). The fire-grazing model, and its associated land management regime known as patch burning (Fuhlendorf & Engle 2001), offers an alternative heterogeneity-based approach to traditional homogeneous land management.
A few observational studies have described unreplicated landscapes with discrete fires and free-roaming grazers, but no experimental studies have compared the effects of traditional homogeneous management to this or similar alternative approaches. Therefore, through the application of replicated treatments, we evaluated the fire–grazing interaction and explored its potential as a land-management approach for grassland ecosystems. We assessed whether patch burning followed by focal grazing of Great Plains grasslands creates a shifting mosaic pattern of vegetation structure and composition. In addition, we compared the effects of homogeneous and heterogeneous land-management approaches on cattle weight gains and grazing patch selection. We also evaluated the effects of treatments on the primary invasive plant species in the area, sericea lespedeza or chinese lespedeza Lespedeza cuneata (DuMont) G. Don. Our focus was not on the main effects of fire or grazing, but instead on the fire–grazing interaction and its potential as a conservation or land-management practice.
The study area was located approximately 21 km south-west of Stillwater, Oklahoma, USA (36°22′N; 99°04′W) and managed by Oklahoma State University. The climate is continental, with an average frost-free growing period of 204 days extending from April to October. Average annual precipitation for the area is 831 mm, with 65% falling as rain from May to October. The mean annual temperature is 15 °C, with an average daily minimum of −4·3 °C in January to an average daily maximum of 34 °C in August. Major soil and community types (ecological sites) in the study area are shallow prairie, loamy prairie, eroded prairie and sandy savanna. Most of this area would be classified as typical of tallgrass prairie, but some local communities are representative of cross timbers vegetation, Quercus stellata Wang. and Juniperus virginiana L. Dominant grasses include little bluestem Schizachyrium scoparium (Michx.) Nash, big bluestem Andropogon gerardii Vitman and indiangrass Sorghastrum nutans (L.) Nash. Secondary grasses include switchgrass Panicum virgatum L., tall dropseed Sporobolus asper (Michx.) Kunth, sideoats grama Bouteloua curtipendula (Michx.) Torr. and Scribner's dicanthelium Dicanthelium oligosanthes (Schult.) Gould. The dominant forb is western ragweed Ambrosia cumanensis Kunth. All units have been burned historically to minimize the encroachment of eastern redcedar Juniperus virginiana L. (McCollum et al. 1999). Lespedeza cuneata is an exotic species that has invaded all of the study pastures and is expected to continue to increase.
We used a completely randomized design to evaluate the fire-grazing model. Two treatments, replicated three times (a total of six pastures), were established at the Oklahoma State University Research Range. Treatment units varied in size from 45 to 65 ha. The treatments included (i) burning of spatially distinct patches within a treatment unit and free access by moderately stocked cattle (patch treatment), which is an approximation of the fire-grazing model (Fig. 2), and (ii) no burning with free access by moderately stocked cattle (traditional). Unlike the Flint Hills of Kansas and Oklahoma, in which most units are burned annually, this is the standard management practice for the regional grasslands. Experimental units (i.e. pastures) of both treatments were treated similarly except for the application of spatially discrete fires in the patch treatment. In both the traditional and patch treatments, the experimental unit consisted of a pasture divided into six distinct patches (Fig. 2). Each patch was delineated at the corners by permanent markers (metal posts) to facilitate ecological monitoring, but not to interfere with livestock or wildlife behaviour and distribution. Fences surrounded each experimental unit (six fenced pastures) but free movement of animals was allowed within each. In the patch treatment, one-sixth of the experimental unit was burned each spring (March to April) and one-sixth each summer (July to October). Application of fires within each patch treatment pasture was in sequentially contiguous fashion to assure a 3-year return interval (Fig. 2). No patches were burned in the traditional treatment. The first year of data (1999), recorded before initiation of treatments, represented pretreatment data.
Pastures in both treatments were moderately grazed with mixed-breed yearling cattle from about 1 December to 1 September, at 3 ha animal−1 based on long-term research projects to optimize sustainable production (Gillen et al. 1991; McCollum et al. 1999). Cattle were permanently identified and randomly assigned to each experimental unit. Cattle were weighed individually with 1-kg resolution on electronic scales at the beginning (1 December), middle (1 April) and end (1 September) of the grazing season each year. Gain per hectare and average daily gain (kg day−1; ADG) were analysed with one-way analysis of variance to identify differences between patch and traditional treatment pastures.
From 13 April to 1 September 2000, cattle were observed three times a day once each week (54 observations) to determine patch preference and grazing activity within the landscape. Observations were evenly spaced during the day to span all daylight hours. Patch preference, indexed by total number of cattle observed grazing in each patch for the 2000 grazing season for each treatment, was subjected to χ2 analyses to determine if patch preference differed between observed and expected (each patch grazed one-sixth of the time).
Vegetation composition was sampled in late August–early September of 1999, 2000, 2001 and 2002. For each patch within each pasture, we recorded canopy cover for plant functional groups, amount of bare ground and cover by litter within 30 0·1-m2 randomly located quadrats (180 pasture−1). Plant functional groups were tallgrasses, S. scoparium, other grasses, forbs, annual grasses and L. cuneata. Schizachyrium scoparium was given its own functional group because it is important to many wildlife species and is particularly sensitive to the combined effects of fire and grazing (Pfeiffer & Hartnett 1995). Lespedeza cuneata was isolated because it is an important exotic species invading throughout the region. Analyses of cover and species composition were by a univariate approach using repeated-measures analysis of variance conducted on pasture averages, to test for significance of the main effects and the interaction of year × treatment (P < 0·05). A multivariate approach was also applied using principal component analysis (PCA) to investigate the relationship of plant community composition among patches within treatments. We used PCA site scores (for patches) as an indication of patch-scale heterogeneity and vegetation change associated with patch-level responses to patch disturbances. Heterogeneity was also assessed using the standard deviation for bare ground, litter and forbs among patches within pastures. Finally, to assess resilience to patch burning, we regressed the difference in total grass canopy cover, forb canopy cover, litter cover and bare ground on time (months) since fire.
Quantifying structure is crucial to understanding biological diversity and wildlife habitat (Rotenberry & Wiens 1980; Schulte & Niemi 1998; Sutter & Brigham 1998). The angle of obstruction (AOB) was used to measure vegetation structure simultaneous to measures of vegetation composition in 1999, 2000 and 2001 (Harrell, Fuhlendorf & Bidwell 2001; Harrell & Fuhlendorf 2002). AOB was systematically sampled in eight cardinal directions from a pivot point 15 cm above the soil surface using a digital level, and measured by recording angles (0–90°) to the top of the nearest obstructing vegetation. Average AOB calculated from these right angles provided a structural measurement of the vegetation at each sampling point. Thirty randomly located AOB points were measured in each patch (180 in each pasture). Analyses included repeated-measures analysis of pasture mean AOB and the standard deviation of AOB among patches for each pasture as an indicator of heterogeneity.
vegetation composition and structure
Plant communities varied by treatment, year and the treatment–year interaction. Cover of forbs, litter and bare ground were best described through the treatment–year interaction of the repeated-measures analyses. In 1999, before treatment began, these three variables were similar for each treatment (Fig. 3). Cover of forbs and bare ground increased in the patch treatment in 2000 and remained higher than the traditional treatment for the duration of the study. Litter increased in the traditional treatment and remained more abundant than the patch treatment throughout the remainder of the study. Effect of treatment was significant for tallgrasses, but tallgrasses were more abundant in the traditional treatment throughout the study, including pretreatment (1999), indicating that treatment had little effect on this variable at the pasture scale. Annual grasses and S. scoparium were variable across years, and did not differ between treatments (data not shown).
The invasive species L. cuneata occurred in all treatment units and increased overall during the study (Fig. 4). The effect of treatment on cover of L. cuneata was significant, and cover was greatest in the traditional treatment throughout the 4 years of this study. The rate of increase over 4 years was also greatest in the traditional treatment. In fact, within all traditional treatment units, L. cuneata increased every year of the study (highest in 2002). In the patch treatment, temporal change was best described as fluctuating or non-directional because, for all three replications, year of greatest cover was either 2000 or 2001 with a slight decline by 2002. So, L. cuneata fluctuated at low values in the patch treatment, the amount being dependent upon which patch served as the focal fire-grazing patch (data not shown), while in all traditional treatment units this invasive species increased over time.
The differences in structure of these grasslands were best described by the treatment–year interaction. Angle of obstruction of the traditional treatment was near 90° every year, indicating a dense, continuous herbaceous canopy resulting from uniform moderate to light disturbance (Fig. 5a). Initiation of patch treatment in 2000 reduced the overall angle of obstruction to about 80°, indicating a more open habitat structure in the patch treatment. This effect was due primarily to the effect of fire and focal grazing on the most recently burned patches.
heterogeneity among patches
Initiation of the patch treatment in 2000 led to increased patch-level heterogeneity resulting from increased variability among patches. Variability of litter, bare ground and forbs increased among patches (Figs 6 and 7). By 2001, heterogeneity of the patch treatment had increased by three- to fivefold over the pretreatment 1999 measurements. Structural heterogeneity, based on AOB, followed patterns similar to the cover variables. By 2001, heterogeneity of structure (AOB) had increased more than fourfold in the patch treatment (Fig. 5b).
A multivariate approach, using PCA ordination on patch-level data, provided additional indication of differences in patch-level heterogeneity between treatments. The first two axes of the PCA accounted for 51% of the variance in the data (Fig. 6). PCA axes 1 and 2 had eigenvalues of 2·33 and 1·73, respectively. The order of importance of functional group variables on PCA axis 1 was litter, forbs, bare ground and L. cuneata. On PCA axis 2, the most important variables were S. scoparium, bare ground, tallgrasses and other perennial grasses. Axis 1 separated patches based on treatment effects (i.e. years since focal disturbance), whereas axis 2 was a function of year (Fig. 6b,c).
Variability within the traditional treatment pastures, although minimal within the ordination space, was associated primarily with variation among years. Variability within years among patches was minimal and all changes among years were consistent among patches. Dynamics within the traditional treatment were associated primarily with an increase in litter cover and canopy cover of the invasive L. cuneata over time. Variation among patches within years in the traditional treatment pastures was less than the overall variation between 1999 and 2002.
The area occupied within the ordination space was much greater within the patch treatment, indicating greater heterogeneity than in the traditional treatment (Fig. 6b,c). In 1999, the pretreatment data indicated little difference between the patch treatment and the traditional treatment. In 2000, two patches within the patch treatment were burned and focally grazed (see below), causing them to shift down and to the right in ordination space, which reflected an increase in bare ground and forbs and a decrease in tallgrasses, litter and L. cuneata. The four unburned patches were similar to patches in the traditional treatment. In 2001, two additional patches were burned and focally grazed. They responded similarly to patches burned the first year. The first-year burn patches shifted left towards unburned patches, reflecting an increase in tallgrasses and litter but continuing to reflect a contribution by forbs. The two remaining unburned patches moved down and to the left, similar to the patches within the traditional treatment. By 2002, the remaining two patches were burned and focally grazed and they moved to the right and down in ordination space, similar to the other burned patches in previous years. The patches burned in 2001 began shifting back to the right and the original burn patches were similar to the patches in the traditional treatment. Patches burned in spring and summer differed little in ordination space, which indicated burn season had little influence on functional group composition in the patch treatment.
resilience under focal fire-grazing disturbance
Results of the PCA suggested that focal patches within the patch treatment differed little from the traditional treatment after 3 years since fire. This was confirmed by regression analysis of months since focal disturbance and differences in cover between patch and traditional treatments (Fig. 8). Cover of grass on focal patches matched that of traditional treatment pastures after about 2 years and generally exceeded it after 3 years. Rapid increase of grass cover was probably explained by limited production in the traditional treatment, because of the high accumulation of litter (Fig. 3). In the patch treatment, litter was reduced by the focal disturbance but was resilient, and 3 years after focal disturbance litter approximated the traditional treatment. The difference in bare ground between the patch treatment and the traditional treatment approached zero after 3 years. Forbs were highly variable from year to year, but forb cover was highest during the first year to 18 months after the focal disturbance and consistently more abundant in the patch treatment than the traditional treatment.
Grazing behaviour differed between treatments largely because cattle preferentially selected focal patches in the patch treatment pastures, while in the traditional treatment pastures grazing animals were randomly distributed among patches. Grazing animals were observed grazing fairly evenly among all patches in the traditional treatment, with only one patch grazed more than 20% of the time (26·0%) and one patch less than 10% of the time (8·6%). In the patch treatment pastures, 75% of grazing time was in the two patches that were most recently burned, and no more than 8% of grazing time was spent in any single patch not burned within the past year. The weight gain of grazing animals differed among years but did not differ between treatments (P > 0·10).
Grassland ecologists have long recognized that grazing and fire are important ecological processes that contribute to the development of many grassland ecosystems (Knapp et al. 1999; Collins 2000). Using fire to manage grasslands for grazing by large ungulates is a logical extension and suggests that grazing coupled with fire can be sustainable and important for conservation. Grazing can be sustainable on most grasslands, but numerous studies of the effects of grazing have demonstrated that creating permanently located reoccurring patches on the landscape that serve as focal grazing points (i.e. near water points) can lead to degradation of the entire ecosystem (Martin & Ward 1970; Foran & Bastin 1984; Fuls 1992; Watkinson & Ormerod 2001; Landsberg et al. 2003; Tobler, Cochard & Edwards 2003). These and many other studies in the past 100 years have led to management practices that minimize spatial variability by altering inherent livestock behaviour to promote uniform, moderate forage use. However, an ecosystem approach to rangeland management should focus not only on restoring the late successional composition of grasslands, a goal of traditional range management, but also on restoring the heterogeneity within the landscape (Hartnett et al. 1996; Coppedge et al. 1998; Fuhlendorf & Engle 2001). Research focused on fixed focal disturbance has resulted in conflicting paradigms between traditional rangeland management for agricultural productivity and land management that promotes heterogeneity. Our study demonstrates that the fire-grazing model may be a useful heterogeneity-based model for grassland management, in which discrete fires are applied to patches, and patch-selective grazing by herbivores promotes a shifting mosaic across the landscape while simultaneously maintaining livestock production. So, by managing transient focal patches that move through the landscape, heterogeneity has the potential to be a central paradigm for managing landscapes for multiple objectives, such as biodiversity and agricultural productivity.
Grasslands evolved with spatially variable disturbances in which grazing animals responded to heterogeneity across multiple spatial and temporal scales, promoting a shifting mosaic landscape that included severely disturbed habitats, relatively undisturbed habitats and a matrix of patches that varied in time since disturbance. The habitat requirements of the grassland species that evolved within these disturbance-driven landscapes are also an indication that these systems were highly heterogeneous in space and time. For example, the habitat requirements of co-occurring grassland bird species vary from plant communities described as heavily grazed to those described as ungrazed (Knopf 1996). Also, populations of co-existing bird species can increase, decrease or be minimally influenced by grazing and fire (Fuhlendorf & Engle 2001). Other species groups, such as insects and small mammals, follow similar patterns. So, diverse grassland faunal communities require heterogeneous landscapes that may be best described as a shifting mosaic, which suggests that heterogeneity is a critical feature to consider in conservation of grassland ecosystems (McIntyre, Heard & Martin 2003). In addition to playing a critical role in biodiversity, spatial and temporal variability can be important to ecosystem function. It is for these reasons that understanding heterogeneity has been argued as an important focus and potentially a basis for conservation biology (Patten & Ellis 1995; Christensen 1997; Ostfeld et al. 1997; Wiens 1997; Burnett et al. 1998; Benton, Vickery & Wilson 2003).
Most studies of disturbances such as grazing and fire have focused separately on their main effects (Fuhlendorf & Smeins 1997; Engle & Bidwell 2001). The fire-grazing model predicts that the interaction of these two factors is more important than the sum of their main effects because of spatially controlled feedback mechanisms. Our data indicate that spatially discrete fires promote focal grazing, in which grazing animals spend more than 70% of their grazing time within the one-third of the area that was burned within the past year. Native herbivores of the North American prairies, if given free access, focus on recently burned areas that provide relatively higher forage quality (Vinton et al. 1993; Coppedge & Shaw 1998). This study demonstrates that this behavioural response can be used in the management of domestic animals, which respond similarly to forage quality cues.
Focal disturbances of grazing and fire caused ephemeral change in plant community composition consistent with predictions of the fire-grazing model that the interaction would lead to a shifting mosaic (Fig. 1). The primary influence of focal fire and grazing disturbance was a temporary decrease in tallgrasses and an increase in forbs that persisted for 2 years. Without disturbance, tallgrasses gain dominance and litter accumulates, limiting plant-available light and reducing the abundance and diversity of forbs, which are critical to overall diversity and conservation of prairies (Collins 1992; Briggs & Knapp 1995). Fire and focal grazing increase forb abundance, forb diversity and structural complexity and decrease litter and tallgrass dominance. These temporary changes in plant community composition and structure make local sites less attractive for continued selection of grazing animals and reduce the probability of ignition or fire spread in the patch. Additionally, with the patch treatment, new patches are periodically burned so grazing animals can shift to more recently burned patches. Patches that have recently been burned and heavily grazed are the most diverse in terms of structure and composition but they are transitional in that the plant community returns to its pre-fire state after 2–3 years. Three years following the initial patch fire, grasses again dominate the plant community and litter accumulates to pre-fire levels, which increases the potential of fire ignition and spread.
Application of the fire-grazing model contributes to a shifting mosaic of vegetation patterns dispersed across the landscape that may be more similar to evolutionary disturbance patterns within Great Plains grasslands than typical homogeneous-based management approaches. However, the patterns of these fire-grazing focal disturbances would have been historically much more stochastic, with boundaries that were much less linear than in our study. An important reason for evaluating alternative approaches for land management is the conservation of endemic and native species, but habitat for many species may be scale dependent and historical patterns may have been at scales that are not practical for land management and conservation on private lands. It is not yet possible to reconstruct the exact pattern of disturbance that would have occurred historically, but these data represent an experimental evaluation of an approach that attempts to simulate more closely the spatial and temporal patterns that arise from the fire–grazing interaction than traditional management.
application of patch dynamics as a management model
The importance of spatial and temporal patterns in native ecosystems has long been recognized (Watt 1947; Wu & Loucks 1995) but patch dynamics have been discussed only recently as critical to conservation and land management (Christensen 1997; Wiens 1997). Spatial and temporal patterns that contribute to patchiness in native ecosystems have many causes (Wiens 1976; Roughgarden 1978; Forman & Godron 1986; Patten & Ellis 1995). These include (i) inherent patterns in resources and conditions due to site differences and species interactions (e.g. water-holding capacity or nutrient availability of soils) and (ii) spatially discrete disturbances such as fire and grazing that can create a shifting mosaic driven by out-of-phase succession. Patchiness that is inherent to native ecosystems and driven by spatially variable disturbance is critical to the function of rangeland ecosystems (Ludwig & Tongway 1995; Fuhlendorf & Engle 2001).
The management and conservation of rangelands have recently embraced a non-equilibrium paradigm, although many practices remain equilibrium based (Briske, Fuhlendorf & Smeins 2003). Incorporating non-equilibrium principles facilitates the integration of temporal variability into natural resource management and adds complexity in the sense that ecosystems are not considered static with single equilibrium points. By classifying range sites and by developing landscape-level plans, rangeland conservationists have long considered inherent spatial variability associated with site differences and resource availability (Stuth, Conner & Heitschmidt 1991). However, the primary outcome of identifying spatial variability has been to apply practices that minimize it and control many of the ecosystem-level responses coupled to it (e.g. animal movement patterns) (Martin & Ward 1970; Holechek, Pieper & Herbel 2003). No effort has been made to integrate spatially discrete disturbances into management of native grasslands. At the very best, traditional approaches to managing rangelands incompletely reduce spatial variation, but the remaining spatial variation is almost always spatially fixed.
A patch dynamic approach in the tallgrass prairie can be accomplished through application of spatially discrete fires and by allowing animals free access to a diversity of landscape elements that vary in time since disturbance. This approach creates a shifting mosaic landscape that is constantly changing but always includes heavily disturbed communities, undisturbed communities and a matrix that varies in time since disturbance while maintaining livestock production. Also, control of L. cuneata under patch burning indicates that a shifting mosaic pattern may be effective in managing this invasive species. Many management practices, other than the fire–grazing interaction, could be used to promote focal disturbances and spatial heterogeneity. Additionally, the interaction between fire and grazing is dependent on stocking rate and the relationship among grazing animal density, vegetation productivity and the area burned. These complex relationships merit additional study.
Traditionally, the management of natural resources has focused on single objectives leading to simplification of ecosystem structure and function. This has met some societal needs but is not sustainable (Liu & Taylor 2002). An approach based on certainty and predictability of simplified systems has given way in ecology to uncertainty and the adaptation of new ecological paradigms that highlight spatial and temporal variability (Wiens 1984; Westoby, Walker & Noy-Meir 1989; Pickett, Parker & Fiedler 1992; Wiens, Van Horne & Noon 2002). An understanding of heterogeneity may serve as the theoretical basis for conservation biology and natural resource conservation. This evaluation of the functionality of heterogeneity in tallgrass prairie indicates the potential of using heterogeneity to integrate conservation biology and production agriculture on native grasslands.
This research was funded by the Oklahoma Agricultural Experiment Station and USDA-NRI Managed Ecosystems Program (02-00777). We thank John Wier, Timothy Tunnell, Chris Stansberry and Chad Cummings for contributions to the application of treatments and data collection, and Terry Bidwell, Jon Marshall and two anonymous referees for review of the manuscript. This article is published with the approval of the director, Oklahoma Agricultural Experiment Station.