• Benning, T., 1993: Fire frequency and topoedaphic controls of net primary productivity in the tallgrass prairie: Development and tests of remote sensing indices. Ph.D. dissertation, University of Colorado, 169 pp.

  • Blad, B. L., and D. S. Schimel, 1992: An overview of surface radiance and biology studies in FIFE. J. Geophys. Res.,97, 18829–18835.

  • DeAngelis, D. L., W. M. Post, and C. C. Travis, 1986: Positive Feedback in Natural Systems. Springer–Verlag, 290 pp.

  • de Mazancourt, C., M. Loreau, and L Abbadie, 1997: Grazing optimization and nutrient cycling: A model of potential impact of large herbivores in a savanna ecosystem. Ecology, in press.

  • Detling, J. K., and M. I. Dyer, 1981: Evidence for potential plant growth regulators in grasshoppers. Ecology,62, 485–488.

  • Dyer, M. I., 1980: Mammalian epidermal growth factor promotes plant growth. Proc. Natl. Acad. Sci. USA,77, 4836–4837.

  • ——, and L. L. Wallace, 1994: Control of plant productivity variation by herbivores Bull. Ecol. Soc. Amer. Suppl.,75, 59.

  • ——, J. K. Detling, D. C. Coleman, and D. W. Hilbert, 1982: The role of herbivores in grasslands. Grasses and Grasslands, J. R. Estes, R. J. Tyrl, and J. N. Brunken, Eds., University of Oklahoma Press, 255–295.

  • ——, D. L. DeAngelis, and W. M. Post, 1986: A model of herbivore feedback on plant productivity. Math. Biosci.,79, 171–184.

  • ——, C. L. Turner, and T. R. Seastedt, 1991a: Remote sensing measurements of production processes in grazinglands: The need for new methodologies. Agric. Ecosyst. and Environ.,34, 495–505.

  • ——, ——, and ——, 1991b: Mowing and fertilization effects on productivity and spectral reflectance in Bromus inermis plots. Ecol. Appl.,1, 443–452.

  • ——, M. A. Acra, G. M. Wang, D. C. Coleman, D. W. Freckman, S. J. McNaughton, and B. R. Strain. 1991c: Source-sink carbon relations in two Panicum coloratum ecotypes in response to herbivory. Ecology,72, 1472–1483.

  • ——, C. L. Turner, and T. R. Seastedt, 1993a: Herbivory and its consequences. Ecol. Appl.,3, 10–16.

  • ——, D. C. Coleman, D. W. Freckman, and S. J. McNaughton, 1993b:Measuring heterotrophic-induced source-sink relationships inPanicum coloratum with the 11C technology. Ecol. Appl.,3, 654–665.

  • ——, A. M. Moon, M. R. Brown, and D. A. Crossley Jr., 1995a: Effects of animal biochemical messengers on plant growth and development: A review of the evidence. Bull. Ecol. Soc. Amer. Suppl.,76, 71.

  • ——, ——, ——, and ——, 1995b: Grasshopper crop and midgut extract effects on plants: An example of reward feedback. Proc. Natl. Acad. Sci. USA,92, 5475–5478.

  • Herms, D. A., and W. J. Mattson, 1992: The dilemma of plants: To grow or defend. Quart. Rev. Bio.,67, 283–335.

  • Hik, D. S., and R. L. Jeffries, 1990: Increases in the net above-ground primary production of a salt-marsh forage grass: A test of predictions of the herbivore-optimization model. J. Ecol.,78, 180–195.

  • Hilbert, D. W., D. M. Swift, J. K. Detling, and M. I. Dyer, 1981: Relative growth rates and the grazing optimization hypothesis. Oecologia,51, 14–18.

  • Holland, J. N., 1995: Effects of above-ground herbivory on soil microbial biomass in conventional and no-tillage agroecosystems. Appl. Soil Ecol.,2, 275–279.

  • ——, W. Cheng, and D. A. Crossley Jr, 1996: Herbivore-induced changes in plant carbon allocation: Assessment of below-ground C fluxes using carbon-14. Oecologia,107, 97–94.

  • Kanemasu, E. T., and Coauthors, 1992: Surface flux measurements in FIFE: An overview. J. Geophys. Res.,97, 18547–18555.

  • Kato, R., K. Itagaki, K. Uchida, T. H. Shinomura, and Y. Harada, 1991: Effects of an epidermal growth factor on the growth of Zea coleoptiles. Plant Cell Physiol.,32, 917–919.

  • ——, E. Nagayama, T. Suzuki, K. Uchida, T. H. Shinomura, and Y. Harada, 1993: Promotion of plant cell division by an epidermal growth factor. Plant Cell Physiol.,34, 789–793.

  • ——, T. Shimoyama, T. Suzuki, K. Uchida, T. H. Shinomura, and Y. Harada, 1995: Effects of an epidermal growth factor on growth of Zea primary roots and mesocotyls. Plant Cell Physiol.,36, 197–200.

  • Loreau, M., 1995: Consumers as maximizers of matter and energy flow in ecosystems. Amer. Nat.,145, 22–42.

  • McNaughton, S. J., 1979: Grazing as an optimization process: Grass-ungulate relationships in the Serengeti. Amer. Nat.,113, 691–703.

  • ——, 1985a: Ecology of a grazing system: the Serengeti. Ecol. Monogr.,55, 259–294.

  • ——, 1985b: Interactive regulation of grass yield and chemical properties by defoliation, a salivary chemical, and inorganic nutrition. Oecologia,65, 478–486.

  • ——, M. B. Coughenour, and L. L. Wallace, 1982: Interactive processes in grassland ecosystems. Grasses and Grasslands, J. R. Estes, R. J. Tyrl, and J. N. Brunken, Eds., University of Oklahoma Press, 167–193.

  • ——, D. G. Milchunas, and D. A. Frank, 1996: How can net primary productivity be measured in grazing ecosystems? Ecology,77, 974–977.

  • Milchunas, D. G., and W. K. Lauenroth, 1993: Quantitative effects of grazing on vegetation and soils over a global range of environments. Ecol. Monogr.,63, 327–366.

  • Odum, W. E., E. P. Odum, and H. T. Odum, 1995: Nature’s pulsing paradigm. Estuaries,18, 547–555.

  • Oesterheld, M., 1992: Effect of defoliation intensity on aboveground and belowground relative growth rates. Oecologia,92, 313–316.

  • ——, and S. J. McNaughton, 1988: Intraspecific variation in the response of Themeda triandra to defoliation: The effect of time of recovery and growth rates on compensatory growth. Oecologia,77, 181–186.

  • Painter, E. L., and A. J. Belsky, 1993: Application of herbivore optimization theory to rangelands of the western United States. Ecol. Appl.,3, 2–9.

  • Paruelo, J. M., and W. K. Lauenroth, 1995: Regional patterns of normalized difference vegetation index in North American shrublands and grasslands. Ecology,76, 1888–1898.

  • Peterman, R. M., 1990: The importance of reporting statistical power. The forest decline and acidic deposition example. Ecology,71, 2024–2027.

  • SAS, 1988: SAS/STAT user’s guide. Release 6.03. SAS Institute, 1028 pp.

  • Schimel, D. S., T. G. F. Kittel, A. K. Knapp, T. R. Seastedt, W. J. Parton, and V. B Brown, 1991: Physiological interactions along resource gradients in a tallgrass prairie. Ecology,72, 672–684.

  • Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1992: An overview of the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE). J. Geophys. Res.,97, 18345–18371.

  • Smith, E. A., W. L. Crosson, and B. D. Tanner, 1992a: Estimation of surface heat and moisture fluxes over a prairie grassland. 1: In situ energy budget measurements incorporating a cooled mirror dew point hygrometer. J. Geophys. Res.,97, 18557–18582.

  • ——, and Coauthors, 1992b: Area-averaged surface fluxes and their time-space variability of the FIFE experimental domain. J. Geophys. Res.,97, 18599–18622.

  • Sokal, R. R., and F. J. Rohlf, 1981: Biometry: The Principles and Practices of Statistics in Biological Research. W. H. Freeman, 776 pp.

  • Steel, R. G. D., and J. H. Torrie, 1960: Principles and Procedures of Statistics. McGraw–Hill, 481 pp.

  • Tieszen, L. L., B. C. Reed, N. B. Bliss, B. K. Wylie, and D. D. DeJong, 1997: NDVI, C3 and C4 production, and distributions in Great Plains grassland cover classes. Ecol. Appl., 7, 59–78.

  • Turner, C. L., T. R. Seastedt, M. I. Dyer, T. G. F. Kittel, and D. S.Schimel, 1992: Effects of management and topography on the radiometric response of a tallgrass prairie. J. Geophys. Res.,97, 18855–18866.

  • ——, ——, and ——, 1993: Maximization of aboveground grassland production: The role of defoliation frequency, intensity, and history. Ecol. Appl.,3, 175–186.

  • Vinton, M. A., D. C. Hartnett, E. J. Finck, and J. M. Briggs, 1993: Interactive effects of fire, bison (Bison bison) grazing and plant community composition in tallgrass prairie. Amer. Midl. Nat.,129, 10–18.

  • Wallace, L. L., 1987: Effects of clipping and soil compaction on growth, morphology and mycorrhizal colonization of Schizachyrium scoparium. Oecologia,72, 423–428.

  • ——, 1990: Comparative photosynthetic responses of Andropogon gerardii to clipping versus grazing. J. Range Manage.,43, 58–61.

  • ——, and M. I. Dyer, 1995: Grassland management: Ecosystem maintenance and grazing. The Changing Prairie: North American Grasslands, A. Joern and K. H. Keeler, Eds., Oxford University Press, 177–197 pp.

  • ——, and ——, 1996: Grazing effects on grassland ecosystems. Ecosystem Disturbance and Wildlife Conservation in Western Grasslands, D. M. Finch, Ed., USDA General Tech. Rep. RM–GTR–285, 13–19., [Available from Rocky Mountain Experiment Station, Fort Collins, CO 80526.].

  • Wessman, C. A., C. A. Bateson, and T. L. Benning, 1997: Detecting fire and grazing patterns in tallgrass prairie using spectral mixture analysis. Ecol. Appl.,7, 493–511.

  • View in gallery

    Location map of FIFE project grazing areas for 1987 and 1988, and 1987 mowing experiments conducted on Konza Prairie Research Natural Area (KPRNA). The study area near site 2 focused on simulating grazing intensity by mowing at different heights in pastures planted to an introduced brome grass; the study area near state road K-177 focused on both mowing intensity and frequency of grasses in the native prairie.

  • View in gallery

    Notched box-whisker graph showing distribution statistics of the average proportion of plant grazed for four grazed pastures used in the FIFE study during 1987 and 1988. Graphics for each site show notched box (indication of 25th to 75th percentile of data with notch surrounding the median giving a 95% confidence interval), whiskers (vertical error bar extending from the edge of the box to the 5th and 95th percentile for the data), and mean (horizontal line across box).

  • View in gallery

    Notched box-whisker graph for growing season-long averages of grass standing crop (StCp) and total production (Prod) in each of the grazed pastures (see Fig. 1 for locations) studied during the 1987 FIFE project.

  • View in gallery

    Notched box-whisker graph for grass standing crop (StCp) and total production (Prod) in each of the grazed pastures studied during the 1988 FIFE project.

  • View in gallery

    Estimates of Δ relative growth rates of grasses in grazed pastures (RGRg) relative to plants in ungrazed exclosures (RGRug), both measured in terms of average growth rates (g) of new plant tissue in a standard unit of space (m2) computed on a daily basis between successive sampling dates [see text, Eq. (2)] (g m−2 day−1). Regression with heavy dashed line = 1987 data; dotted regression = 1988; solid regression line = both years together.

  • View in gallery

    Distribution of Δ (in %) for total productivity and relative variance (vCV: Sokal and Rohlf 1981) for mowed and grazed pastures when compared to controls in the 1987 and 1988 FIFE projects. We show both sets of variables plotted against grazing intensity, (1 − G), for four grazing classes grouped by quartiles (grazing classes = 0.25) following Dyer and Wallace (1994, 1997 unpublished manuscript) [because the boundaries of no grazing (=0, or control conditions) and complete vegetation removal (=1.0) define the HO hypothesis, we use a quadratic curve to force as expression of relationships for each variable through 0 and 1.0 intercepts].

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 13 13 5
PDF Downloads 5 5 0

Biotic Interactivity between Grazers and Plants: Relationships Contributing to Atmospheric Boundary Layer Dynamics

View More View Less
  • 1 Institute of Ecology, University of Georgia, Athens, and Biosphere Research, Inc. Lenoir City, Tennessee
  • 2 Minnesota Department of Natural Resources, St. Paul, Minnesota
  • 3 Institute of Alpine and Arctic Research, University of Colorado, Boulder, Colorado
© Get Permissions
Full access

Abstract

During 1987 and 1988 First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE) studies conducted in the tallgrass prairie of central Kansas, variations in ungulate grazing intensity produced a patchy spatial and temporal distribution of remaining vegetation. Equally variable plant regrowth patterns contributed further to a broad array of total primary production that resulted in a pronounced mosaic of grazing impacts. This regrowth potential, derived from a relative growth rate (RGR) equation comparing ungrazed and grazed plants, determines much of the ecosystem dynamics within and among the grazed pastures and between years. Rates of change in new plant growth (ΔRGRg) ranged from −100% to +40%; however, 78% of the time in 1987 and 71% in 1988, productivity increased as a function of grazing intensity. Since plant growth potential in ungrazed (RGRug) and grazed systems (RGRg) have inherently different attributes, interactions with the abiotic environment may develop many uncertainties. Thus, changes in growth rates in grazed areas compared to ungrazed areas (ΔRGRg) may impose major controls over system productivity and associated biological processes currently not accounted for in ecosystem models.

Because FIFE microsite atmospheric boundary layer (ABL) studies did not directly incorporate grazing intensity into their design, Type I and Type II statistical errors may introduce significant uncertainties for understanding cause and effect in surface flux dynamics. As a consequence these uncertainties compromise the ability to extrapolate microsite ABL biophysical findings to other spatial and temporal scales.

Corresponding author address: Dr. M. I. Dyer, Institute of Ecology, University of Georgia, Athens, GA 30602-2202.

Abstract

During 1987 and 1988 First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE) studies conducted in the tallgrass prairie of central Kansas, variations in ungulate grazing intensity produced a patchy spatial and temporal distribution of remaining vegetation. Equally variable plant regrowth patterns contributed further to a broad array of total primary production that resulted in a pronounced mosaic of grazing impacts. This regrowth potential, derived from a relative growth rate (RGR) equation comparing ungrazed and grazed plants, determines much of the ecosystem dynamics within and among the grazed pastures and between years. Rates of change in new plant growth (ΔRGRg) ranged from −100% to +40%; however, 78% of the time in 1987 and 71% in 1988, productivity increased as a function of grazing intensity. Since plant growth potential in ungrazed (RGRug) and grazed systems (RGRg) have inherently different attributes, interactions with the abiotic environment may develop many uncertainties. Thus, changes in growth rates in grazed areas compared to ungrazed areas (ΔRGRg) may impose major controls over system productivity and associated biological processes currently not accounted for in ecosystem models.

Because FIFE microsite atmospheric boundary layer (ABL) studies did not directly incorporate grazing intensity into their design, Type I and Type II statistical errors may introduce significant uncertainties for understanding cause and effect in surface flux dynamics. As a consequence these uncertainties compromise the ability to extrapolate microsite ABL biophysical findings to other spatial and temporal scales.

Corresponding author address: Dr. M. I. Dyer, Institute of Ecology, University of Georgia, Athens, GA 30602-2202.

1. The atmospheric boundary layer and herbivore dynamics

Grazers—including large ungulates such as cattle and native North American fauna, and invertebrates, such as chewing and sucking insects—have direct and indirect influences on the plant community that have arisen in many different grassland systems on all continents (Milchunas and Lauenroth 1993). Removal of green-leaf biomass above ground and root biomass below ground and the physical effects of trampling control primary production (Wallace and Dyer 1995). However, indirect effects associated with plant response to feeding may have even greater total effects on system production(Dyer et al. 1982; McNaughton 1979, 1985a; McNaughton et al. 1982). This response, one that results from a series of positive and negative nonlinear feedbacks from the grazer or herbivore to the plant, may either decrease or increase plant community regrowth and overall productivity, which results in defined optimal use patterns and ecosystem productivity in space and time (Dyer et al. 1993a; Loreau 1995; de Mazancourt et al. 1997). This concept has its critics (Painter and Belsky 1993). However, several recent studies, which include major reviews (Herms and Mattson 1992; Milchunas and Laurenroth 1993); designs to test specific grazer influences (Hik and Jeffries 1990); and theoretical studies (Loreau 1995) coupled with long-term field studies (de Mazancourt et al. 1997); and our data in this report on the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE) database, give strong support to the multiple feedback hypothesis.

a. The FIFE database—Spectral reflectance as a measure of productivity

As part of the FIFE surface radiances and biology design (see Sellers et al. 1992), three studies conducted in 1987 and one in 1988 examined results from experimental simulated grazing and grazed-pasture treatments to evaluate and interpret spectral reflectance characteristics throughout the growing season (Dyer et al. 1991a,b, 1993a; Turner et al. 1992, 1993). In each of the studies we found unpredictable results involving grazing or mowing intensity (Dyer et al. 1991a,b; Turner et al. 1992, 1993), frequency (Turner et al. 1992, 1993), grazing history in managed pastures (Turner et al. 1992, 1993), system N availability and plant N content (Dyer et al. 1991a,b; Turner et al. 1992, 1993). These reports show considerable uncertainty related to total greenness indexes and normalized difference vegetation index (NDVI) computations based on leaf area index (LAI) or standing crop per se, as well as to overall season-long productivity (Blad and Schimel 1992; Turner et al. 1992).

Almost no grassland studies have sampled herbivore/grazer effects sufficiently to give accurate descriptions of NPP (e.g., primary production minus the effects of grazers) throughout a growing season (McNaughton et al. 1996). This fundamental problem results in confounding of statistical results with respect to Type I error (rejection of the null hypothesis when it is true, with the acceptance of an incorrect alternative hypothesis) and Type II error (acceptance of null hypothesis when it is false, with the rejection of an alternative hypothesis when it is true) (Steel and Torrie 1960; cf. Peterman 1990).

Standing crop and total production over a growing season in grasslands vary largely as a function of the proportion of biomass removed by grazers. These estimates derive simply from measures of vegetation mass in areas not protected from grazing (g) compared to measures of vegetation in moveable exclosures nearby (ug). The measure of grazing intensity (G) thus becomes
i1520-0469-55-7-1247-e1
(McNaughton 1979, 1985a).

Most estimates of G come from comparing grazed pastures with a nearby long-term exclosure, most often over an entire growing season. Such an approach cannot provide adequate measures because of the inherent changes in vegetation type and physiology that develop in exclosures, and differences in the way vegetation responds to grazing during the growing season. As McNaughton (1979, 1985b) pointed out, because grazed plants respond differently through their physiology and the alteration of growth rates compared to plants inside an exclosure, an investigator can make appropriate estimates only through sampling a network of moveable exclosures. By not employing this approach an investigator may 1) miss important seasonal growth dynamics necessary for making correlations with other events, such as surface dynamics, or 2) obtain an incorrect measure of grazing intensity, one often associated with Type II error. These result in part from inadequate estimates of plant regrowth (McNaughton et al. 1996).

b. The herbivore optimization curve hypothesis

The herbivore optimization (HO) hypothesis suggests that stresses associated with herbivory induce more or less instantaneous physiological changes in several plant production processes that involve pulses with medium frequency and amplitude lasting from hours to days. Depending upon the level of grazing stress, these shifts in leaf and whole plant physiology may result in either increases or decreases in overall plant function (cf. McNaughton 1979, 693–694). Subsequent reports describe alterations in 1) photosynthetic rates (Dyer et al. 1991c; Dyer et al. 1995a); 2) individual leaf dynamics, such as stomatal conductance and transpiration (Dyer et al. 1995a); 3) translocation of labile C and N compounds from leaf sources to stem and root sinks (Dyer et al. 1991c, 1993b); and 4) describe several physiological effects of putative growth promoting substrates found in herbivore regurgitants and saliva (Dyer 1980; Dyer et al. 1995a,b; Kato et al. 1991, 1993, 1995; McNaughton 1985b).

As herbivory stresses increase, the plant comes to a maximal level of increased response (Dyer et al. 1986; Loreau 1995) [defined as the optimization point (Hilbert et al. 1981)], following which production potential drops through a “break even” point and then becomes negative. Regrowth of tissue lost during herbivory exhibits a short lag effect. The extent of its total regrowth comes about as a function of a shift in plant source-sink dynamics. Now instead of a downward translocation of labile C and N compounds induced by the herbivore, the plant reverses the flow of compounds from the rhizosphere and plant storage organs to help organize new leaf formation and new rates of C fixation.

A major problem exists in measuring these growth responses in sufficient detail to give a full and accurate description of productivity potential following a grazing episode. The problem lies in simultaneously determining during a time interval (t) 1) any changes in shoot biomass (S) and 2) any concomitant changes inherent in the instantaneous growth rate of S2 at t2 compared to S1 at t1 for plants not stressed by grazing (nominally the controls) and those subjected to grazing [cf. Hilbert et al. (1981) for basic theoretical arguments]. Hilbert et al. (1981) presented an argument designed to compute and compare growth rates for grazed and ungrazed plants found together at any given place or time. Oesterheld and McNaughton (1988) and Oesterheld (1992) conducted a series of validation experiments on this method, showing how it should work for laboratory andfield experiments and ultimately for management purposes.

With the input of four variables [1) mean shoot relative growth rate, 2) change in relative growth rate after grazing, 3) grazing intensity, and 4) recovery time following grazing] an equation comparing relative growth rates (RGR), as originally stated by Hilbert et al. (1981), provides the basis for defining a production isoline along which productivity of grazed plants equals that in ungrazed controls [cf. Fig. 2, Hilbert et al. (1981) for additional background], such that
i1520-0469-55-7-1247-e2
where
i1520-0469-55-7-1247-eq1

Thus, the Hilbert et al. (1981) growth equation presents a way to determine nonlinear responses of plants to any grazing stress over a relatively short regrowth period. For longer regrowth periods, the overall regrowth model with its combined positive and negative feedback loops may require additional functions derived from the Pearl–Verhulst logistic equation (DeAngelis et al. 1986; Dyer et al. 1986) or a Michaelis–Menton-like model (Dyer et al. 1986), or Lotka–Volterra modeling approaches used by Loreau (1995) and de Mazancourt et al. (1997). Provision of the necessary information for equation model, in particular estimating G and obtaining statistically adequate sample numbers for RGRΔt, has proven difficult in the field (McNaughton 1985a; McNaughton et al. 1996). For that reason few studies give sufficient information for computing RGR [both the ungrazed rates (RGRug) and new rates for grazed plants (RGRg)].

The most critical test of the hypothesis has come fromlaboratory studies (Oesterheld 1992). As a consequence, almost all reports in ecological and range science literature give significant underestimates of short-term (those ranging from hours or days to season-long estimates) production responses in grazed systems, resulting in a morass of interpretations throughout the entire grazing literature where both Type I and Type II errors emerge sporadically, and without the statistical evidence to sort out their inherent error terms. Because of such problems it becomes nearly impossible to determine the level of uncertainty in any given report. These biological problems and logistical constraints have important implications for the study of atmospheric boundary layer (ABL).

c. Atmospheric science models associated with the boundary layer and grazing effects

To couple atmospheric science with ecological investigations scientists have focused primarily on use of 1) the Bowen ratio (β) and 2) the eddy correlation (EC) models. FIFE investigators established a network of 20 stations, including fixed and mobile instrument packages, to obtain data required by these models to determine flux dynamics. The sites represented a statistical design that stratified topographic features and managed burn studies at the Konza Prairie Research Natural Area (KPRNA) (Kanemasu et al. 1992). The primary FIFE study design did not include grazing impacts, although some of the flux station locations existed in locales grazed at differing intensities (Smith et al. 1992a, b).

Although basically common knowledge to meteorologists and biophysicists involved routinely with planetary surface fluxes, few ecologists encounter the standard model for one-dimensional surface energy balance:
i1520-0469-55-7-1247-e3
where
i1520-0469-55-7-1247-eq3
ignoring energy taken up by photosynthesis (Kanemasu et al. 1992; Smith et al. 1992a,b).

Grazing stresses inherently will have differing effects on each of the variables, ranging from short-term, direct effects on plant structure and physiology that in turn tend to control Rn, H, and λE to long-term indirect effects that regulate soil parameters affecting Gs (Table 1). As we suggest in Table 1, grazers may affect elements of surface energy fluxes at varying system pulse rates, ranging from pulses with short response times of hours to days to those with time lags lasting a full growing season and contributing to equlibria over years to decades. Variation in these flux rates will come aboutthrough altered plant growth, changes in nutrient cycling processes, accumulation of litter, and soil compaction, all as a function of the grazing intensity.

As far as we know, no one has investigated either the direct or indirect impacts of grazing on the models used to determine ABL dynamics. Two major limitations exist: 1) the practical inability involving potential damage to expensive instruments or logistical limitations in constantly moving hard-to-calibrate instruments to measure simultaneously grazed and ungrazed sites in areas impacted by ungulates, and even more important, the measurement of a gradient of spatially and temporally distributed stresses; and 2) the assumption that measurements reflect values of conditions unstressed by herbivores when the investigators have not accounted for impacts of arthropods aboveground or below ground microarthropods and nematodes in their study sites, which gives a major potential for random error effects. These limitations as well provide for a complex array of Type I and Type II statistical errors noted previously for which insufficient data exist to give adequate assessment in the FIFE study.

Thus, in this paper we address the hypothesis that plant–heterotroph feedbacks—namely, those arising from grazing stresses—result in a principal control of plant regrowth potentials, and processes associated with RGRg govern many rates that describe surface energy flux dynamics from area to area within a growing season.

2. Materials and methods

a. Field studies and data sources

For this study we have utilized the 1987 and 1988 grazed pasture data collected for the FIFE database, much of it reported in Turner et al. (1992, 1993). The data reflect standing crop and productivity (g m−2) from four grazed pastures associated with nearby surface flux stations (Kanemasu et al. 1992). Reading from north to south, in 1987 we obtained measurements from a moderately grazed pasture at site 29, a lightly grazed pasture near site 42, a heavily grazed pasture near site 32, and a moderately grazed pasture near site 21. In 1988 sampling of the lightly grazed pasture changed to the nearby site 40 (Fig. 1).

Samples from 1) within permanent cattle exclosures;2) inside replicated (10) 1-m2 wire cages moved at intervals [intervals that depended upon grazing and vegetation growth rates in each pasture as applied by McNaughton (1985a) throughout the growing season]; and 3) from grazed plots adjacent to the wire cages provided the entire dataset. For a more complete discussion of the study sites and related findings we refer the reader to Kanemasu et al. (1992) and Turner et al. (1992, 1993).

b. Analytical procedures

We utilized the FIFE database for each of the four sampling areas (Fig. 1) from 1987 and 1988 growingseasons to 1) determine standing crop biomass (g m−2) for each sampling period (t0+n, where t0 = initial sampling period for each year and n = subsequent sampling dates); 2) compute separate production (Σ g m−2 t−10+n); 3) compute estimates of RGR [Eq. (2)], the calculated instantaneous growth rate (g m−2 day−1) of grasses in each grazed treatment; and 4) the proportion (G) of grasses removed by grazers during any given sampling period.

We compared total productivity with standing crop values at each sampling period by determining Δproductivity [(productivity − standing crop)/standing crop)] at each site. We transformed these proportions to natural logarithms and conducted a nested analysis of variance (ANOVA). For the ANOVA design we used sampling dates (n = 4 to 9) within classes of grazingsites (n = 4) within years (n = 2) using SAS GLM (general linear model) procedures (SAS 1988).

For the analysis of RGR variation we first conducted an ANOVA comparing RGRug values for ungrazed exclosures among sites within each year. Then we conducted an ANCOVA for the statistical design noted above (using RGRug as the covariate) to compare ΔRGRg to grazed controls, using a ln transformation of (RGRg − RGRug)/RGRug.

Lastly, we used the Dyer and Wallace (1994) database, which, in addition to several hundred studies reported in the herbivore and grazing ecology literature they have assembled to assess mean productivity and relative variance, includes FIFE data from all experiments conducted in 1987 and 1988 [including brome field (Dyer et al. 1991b) and native prairie simulated grazing frequency and intensity plots (Turner et al. 1992, 1993)] to determine the relationship between these two statistical measures at KPRNA. For this analysis we used ANCOVA (SAS 1988) to determine differences between grazed and ungrazed controls for ln-transformed values of mean productivity and their associated, sample-corrected coefficient of variation (vCV: Sokal and Rohlf 1981). For covariates we used ungrazed plot mean productivity (or vCV), photosynthetic pathway of plant (e.g., C3 or C4 pathway), scale of study (experimental plots and estimates of whole pastures), and level of fertilization (fertilizer added compared to no fertilizer additions).

3. Results

Because we conducted three different sets of experiments during the main FIFE study in 1987 and different weather patterns between 1987 and 1988, our workload did not permit us to sample grazed pastures at the same intervals each year. In 1987 intervals ranged from 6 to 48 days [mean = 31 ± 3.4 day (s.e.)] and in 1988 from 11 to 26 days (mean = 20.5 ± 1.7 day). These intervals fall roughly within the guidelines set by McNaughton (1985a), and we have no evidence these different intervals affect statistical results that we present for standing crop, productivity, or RGR. Furthermore, the methods for computing the mean relative growth rate tend to normalize data where Δt varies (Hilbert et al. 1981).

a. Vegetation removed: Measures of grazing intensity

Figure 2 shows averages of grazing intensity for each site and each year. Even though overall grazing intensity showed slight but nonsignificant differences (p = 0.797) between the relatively wet growing season of 1987 (grazing intensity = 0.181) and a dry growing season of 1988 (grazing intensity = 0.146), local differences developed across the FIFE study area. Sites 32 and 21 showed the highest grazing intensities for both years, with sites 29 and 40/42 having lower grazing impact. At sites 29 and 40/42 no differences emergedbetween years (P = 0.478 for site 29; P = 0.591 for sites 40/42). Sites 21 and 32 showed considerable scatter for grazing intensity: site 32 showed between year differences (P = 0.009), whereas site 21 did not (P = 0.426).

b. Standing crop and production values

Turner et al. (1993) provided an initial assessment of the standing crop, estimated consumption, and cumulative production data for four FIFE grazing sites studied during 1987 and 1988. We present a more detailed analysis of standing crops and productivity for each site during each of the two years following grazing at each location. In all instances standing crop values represent averages of biomass of grass measured throughout the growing season inside exclosures, essentially the vegetation associated with instruments at these sites that measured energy fluxes. Average total aboveground productivity provides estimates of grass biomass over the growing season in the grazed areas. This includes estimates of consumption by livestock and new biomass attributed to new regrowth rates of grass following grazing.

In Figs. 3 (1987) and 4 (1988) results from the four study sites arranged from left to right present a generalized north to south gradient in the FIFE study area (see Fig. 1 for site locations). Mean standing crop across all sampling dates in 1988 [108.1 ± 9.03 (s.e.) g m−2, Fig. 4] dropped significantly (Δ% = 24.8, P = 0.0093) from 1987 values (143.8 ± 17.67 g m−2, Fig. 3). Mean total productivity showed a similar significant (P ⩽ 0.05) change between the two years (1987 = 211.4 ± 23.18 g m−2; 1988 = 158.1 ± 11.56 g m−2).

For 1987 standing crop means over the growing season ranged from 80.2 ± 10.01 g m−2 at site 32 to 235.9 ± 34.5 g m−2 near site 42 (a total difference of 192.2%). At the same time total production means across these sites ranged from 162.4 ± 36.17 g m−2 at site 21 to 309 ± 44 g m−2 (a total difference of 90.3%). In 1987, a year with a relatively high amount of precipitation during the growing season, site 29 production values averaged 46.3% higher than standing crop; for site 42, 31%; for site 32, 122.7%; and for site 21, 49.9%. For all grazing data a sampling period-weighted difference showed that total production increased 46.5% over standing crop during the growing season (Fig 3).

For 1988 the data show less of a difference between standing crop and productivity over the growing season (Fig. 4). Site 29 production values averaged 24.9% higher than standing crop; for site 40, 41.1%; for site 32, 42%; and for site 21, 111.1%. However, even with these lower values possibly induced by differences in weather, sampling period-weighted production values exceededstanding crop by an amount almost identical to that in 1987 (45.5%).

The main ANOVA model for Δproductivity showed significant effects (F12,38 = 2.09, P = 0.033), giving differences between years (P = 0.003) and between sites within years (P = 0.047). In 1987 mean productivity = 211.4 ± 23.18 g m−2, while in 1988 mean productivity = 158.1 ± 11.56 g m−2. At the same time Δproductivity related to standing crop showed significantly different levels between years (1987 overall Δ = 17.3 ± 5.75%; for 1988 Δ = 12.2 ± 3.72%).

c. Relative growth rates for four FIFE experimental sites

Considerable variation developed for ΔRGRg within and between each site for the two years (Fig. 5), and overall showed a nonlinear association between RGRug and RGRg. In 1987 RGRg values ≥ RGRug values 78% of the time (18 of 23 sampling periods); for 1988 RGRg ≥ RGRug 71% of the time (22 of 31 sampling periods). By segregating the 54 sampling periods into their respective quadrants (Fig. 5), we found the relationships between ΔRGRg as a function of RGRug yielded 1) 36 (67%) ΔRGRg values ≥ RGRug when RGRug was positive (I); 2) four (7%) ΔRGRg values ≥ RGRug when RGRug was negative (II); 3) seven (13%) ΔRGRg values ⩽ RGRug when RGRug was negative (III); and 4) seven (13%) ΔRGRg values ⩽ RGRug when RGRug was positive (IV) (Fig. 5). Linear representations between change in ΔRGRg show slight differences between years in the ANOVA, Fig. 5 (P = 0.076), suggesting a possible (weather–grazing) interaction.

d. Overall productivity and variance

From data assembled for all FIFE actual and simulated grazing studies reported in 1987 (Dyer et al. 1991a,b; Turner et al. 1992, 1993), we show Δproduction relative to controls and its associated ΔvCV (Sokal and Rohlf 1981) as a function of grazing (1 − G) defined by McNaughton (1979, 1985b) (Fig. 6).

Here, Δproduction increased significantly to as much as 50% over controls (for ANCOVA F7,39 = 264.75, P ⩽ 0.0001) at what most ecologists and range scientists consider moderately heavy grazing levels (midpoint for 1 − G = 62.5% of vegetation removed—see Fig. 6), and showed lesser values for both smaller and larger grazing levels. Moderate differences developed between grazing intensity classes (F3,39 = 2.24, P = 0.099). SNK and least-squares mean separation tests show that Δ% for grazing class one (1 − G = 0.125), two (1 − G = 0.375), and four (1 − G = 0.875) did not differ (P ≥ 0.05), but grazing class three ≥ grazing classes one, two, and four (P ⩽ 0.034). Since the complete HO model includes the boundaries of both ungrazed (1 − G = 0) and conditions with all vegetation removed (1 − G = 1.0), we have used a quadratic to express the entire relationship (values for mass, Fig. 6). This relationship gives a highly significant fit to all the data (P ⩽ 0.01).

At the same time relative variance, vCV, showed a significant decrease almost inversely proportional to the nonlinear biomass changes (F7,39 = 8.30, P ⩽ 0.0001), reaching its minimal value (circa −45% relative to variance of ungrazed controls) at grazing class two (37.5% of vegetation removed) (Fig. 6). Significant differencesdeveloped among grazing classes (F3,39 = 4.2, P = 0.011). For ΔvCV, grazing class two = grazing class three (P = 0.464) and these ⩽ grazing classes one and four (P ⩽ 0.042). The application of a quadratic, applied as discussed above for biomass values, gave a highly significant (P ⩽ 0.01) fit through all of the ΔvCV data.

For the ANCOVA of production means two covariates showed highly significant influences on variation of production, including background variance of ungrazed plot production values (P = 0.0001) and C3 or C4 photosynthetic pathway (P = 0.0001). Fertilization had little influence on variance (P = 0.0952), and too few data existed to determine the influence of scale in the studies. The ANCOVA for vCV showed that none of the covariates (vCV for ungrazed plots, C3 or C4 photosynthetic pathway, and fertilization) had any effect on overall vCV variance.

4. Discussion

The two clipping experiments simulating grazing conducted in 1987 and the grazed pasture experiments conducted in 1987 and 1988 demonstrate clearly a distinction between apparency—for example, standing crop—and its less perceptible counterpart, function—for example, productivity—collectively the processes driving dynamics in this grassland ecosystem. Biologists and ecologists have a well-developed understanding of tallgrass prairie concerning state variables, such as species composition and estimates of mass and how they vary temporally and spatially; however, these represent only a surrogate or approximation of production. Many processes, such as short-term C and N distribution throughout the plant as a function of grazing, that drive grassland productivity and associated energy fluxes have much less known about them. These processes become the most important for those interested in ABL models and dynamics.

a. Grazing intensity, RGR, and ecosystem productivity

In our introductory comments we gave a short review of the RGR concept related to plant regrowth following the stresses of herbivory. Hilbert et al. (1981) outlined several constraints pertinent to the use of the concept such as 1) plant age, 2) species-specific growth potentials (RGRmax), 3) interval between successive episodes of feeding, 4) grazing intensity, and 5) how these parameters may interact. Indeed, because of the pervasive nature of herbivory and complexity of such interactions they concluded that “The type and degree of these responses in relative growth rate determine the effect of herbivory on primary production and hence ecosystem trophic dynamics” (italics added). If so, as Oesterheld (1992) showed in his validation studies where he included root dynamics as well in the HO equation, and as Holland (1995) and Holland et al. (1996) have shownfor aboveground-herbivore induction of soil microbial growth and rhizosphere dynamics, then logic suggests that such controls over primary production in ecosystem function may also result in controls over ABL dynamics that depend upon input from primary production processes.

By applying this logic to the FIFE studies, we get some sense of how a gradient of grazing stresses in a tallgrass prairie ecosystem might affect surface energy fluxes. First, we have shown that G, the measure of green plant tissue removed, varied considerably across the FIFE study area. It ranged from high values in heavily grazed pastures east and south of KPRNA to very low levels in moderately grazed pastures adjacent to KPRNA along the northeast corner (we have not included values for the KPRNA, which ranged from zero in large ungrazed areas to those unmeasured in sites moderately grazed by bison, nor have we included herbivory above ground from insects, or from insects and a large variety of microarthropods and nematodes below ground).

Across the FIFE study area a noisy pattern emerged for RGRug during both years. Nonetheless, we found no statistically significant evidence in the ANOVA for mean differences in RGRug between years, between sites within years, or between sampling dates within sites. In short, the FIFE dataset indicates that average grass RGR in exclosures used as experimental controls did not vary;however, we point out that with additional sampling, including both added temporal and spatial sampling, this interpretation might have changed.

At the same time the FIFE data show that different stocking rates in the four study sites over the two years resulted in differing levels of G that translated to a highly variable landscape picture for RGRg, one that showed no covariance with RGRug. Thus, we conclude that the highly significant changes noted for ΔRGRg result from a highly significant association with G. Furthermore, this variance in G shows an association with initial estimates of production. Therefore, this initial estimate of production multiplied by RGRg gives even a greater sense of variation due to grazer effects. Thus, we conclude that, as predicted by the Hilbert et al. (1981), Dyer et al. (1986), Loreau (1995), and de Mazancourt et al. (1997) models, herbivory has highly significant effects over vegetation dynamics in this ecosystem. By not accounting for RGRg, ABL scientists addressing ecosystem-related processes may have missed important cause-and-effect relationships in surface energy flux dynamics. This potential deficiency involves a complex set of interactions in the primary producers that display neither linear nor monotonic patterns relative to heterotrophic processes.

The HO hypothesis invoking ΔRGR as a function of grazing intensity predicts that low to moderate grazing intensity results in short-term nonlinear increases in productivity and associated processes, such as rates of photosynthesis; CO2 and H2O fluxes; and translocation oflabile C and N from internal plant sources [such as leaves and stems, Dyer et al. (1991c, 1993b)] to sinks [such as roots, and ultimately the rhizosphere, Dyer et al. (1991c, 1993b), Holland (1995), Holland et al. (1996)]; stomatal conductance (Dyer et al. 1995a); transpiration (Dyer et al. 1995a); and perhaps plant water use efficiency (Wallace 1990). Operating together, these processes often produce increases in RGRg following light to moderate grazing. The extremes of heavy grazing more often may result in a negative RGRg, and hence results in lowered regrowth potentials, thus adding even more adversely to the impact of lost plant tissue.

We do not know the exact range of differences to expect in surface energy fluxes as a function of ΔRGRg. Should ΔRGRg display a linear relationship with other biotic and abiotic variables (which it may not), it could range from decreases of more than 100% to increases of circa 20% in heavily grazed areas to fluctuations in lightly to moderately grazed areas that range from Δ = ca. −30% to 40% (see Fig. 5). Obviously, such a range in biotic conditions governing ΔRGRg across the landscape will introduce variance that those constructing and applying ABL models must take into consideration.

b. Influences of grazed plant RGR on measurements of spectral reflectance

Turner et al. (1992, 1993) reported results of reflectance measurements at the sites we cover here. Those reports, plus the reports of simulated grazing (mowing) (Dyer et al. 1991a,b; Turner et al. 1992, 1993), suggest that ΔRGRg seriously constrains the ability to pick up predictable remotely sensed signals within and between pastures containing a patchy distribution of grazing intensity and 1 − G. Herbivory removes both leaves and stems, which, during the growing season, initially reduces both LAI and biomass, and hence affects reflectance, R, directly. By having neither a static nor a linear response to G, plant RGRg quickly alters the vegetation canopy by inducing leaf growth, thus R once again will change at some proportion to the addition of new plant tissue. At the same time this new growth rate also changes spectral reflectance properties by altering C/N ratios in the plant. The HO hypothesis states that both LAI and biomass may change to levels above, equal to, or below ungrazed controls, and experience shows that this change has a fast pulse rate, occurring within minutes to hours and lasting for days (Dyer et al. 1991c). Thus, RGRg, as a function of G and 1 − G, has considerable importance for estimating LAI or biomass with remote sensing indexes. At the same time it affects estimates of net radiation, an important component in determining surface energy fluxes.

Schimel et al. (1991) and Wessman et al. (1997) showed that NDVI values correlate more strongly with intercepted photosynthetic radiation (IPAR) than with either foliate biomass or leaf area. Furthermore, enhanced nitrogen content of foliage also shows a strongcorrelation with IPAR (Schimel et al. 1991), and, as we pointed out above, since grazing tends to induce increases in N content of regrowth foliage (Dyer et al. 1991b), grazed vegetation consequently gives higher NDVI values than predicted by biomass or LAI alone. Indeed, Wessman et al. (1997) could not detect differences in grazed and ungrazed watersheds on Konza Prairie Research Natural Area in 1990 from NDVI values, in spite of known, large differences in leaf area biomass.

NDVI shows strong regional patterns in the Great Plains that correlates with RGR and potential productivity (e.g., Paruelo and Laurenroth 1995; Tieszen et al. 1997). Yet, no data exist to suggest that actual leaf area or biomass shows a similar function. Management of grazing intensity in remaining grasslands of the Great Plains results in relatively small differences in aboveground biomass or LAI across this entire region. Instead, as we noted above, NDVI variables appear to sense IPAR (Schimel et al. 1991, Wessman et al. 1997). Thus, while Tieszen et al. (1997) found that tallgrass prairie regions of the Great Plains produced both highest maximum and integrated NDVI values, Benning (1993) demonstrated that among years, maximum NDVI at a single site correlated negatively with ungrazed plant biomass.

Instruments onboard satellites or high-flying aircraft sense only a part of the ungrazed canopy, and in productive years N content and chlorophyll density of the uppermost portion becomes diluted below values in less productive years. These relationships, plus induced increases in N mobilization throughout the plant and concentration in the leaf as a function of grazing, all add to the overall ambiguity in interpreting remote sensing signals. Thus, because of the inherent nonlinear relationship between NDVI and LAI with respect to instantaneous energy transformations, this remote sensing concept cannot provide clear, unambiguous information for vegetation–atmosphere models.

Wessman et al. (1997) used a high-resolution imaging spectrometer (AVIRIS) and linear spectral mixing analysis in comparison with standard NDVI analyses to evaluate relationships among plant biomass, various management treatments, and the spectral signature of the landscape. They could not identify an endmember (a unique spectral signature of some specific surface component) that represented graminoid biomass and one that showed any sensitivity to the presence of grazing. Therefore, rather than concentrate on the state variable components within the system, we suggest these indexes, either alone or in conjunction with other remotely derived information, might more correctly identify RGR of a site, thereby allowing an estimation of energy or water flux, and in so doing provide a metric far superior to LAI alone.

We have demonstrated that photosynthetic activity, as measured by ΔRGRg in the FIFE study, changes as a function of grazing intensity. IPAR, photosynthetic activity, and RGR undoubtedly have strong relationshipsto energy, C, and H2O fluxes within the landscape. As a consequence, we reiterate a stance taken earlier (Dyer et al. 1991a) that to correctly assess system-level functions using remote sensing tools we must find new approaches or techniques not dependent solely upon LAI or standing crop biomass that correctly identify the intensity of vegetation–atmosphere interactions.

c. Influences of grazed plant RGR on measurements of boundary layer dynamics

In Table 1 we present a simplified summary of the potential effects that grazing dynamics may have on variables describing surface energy balance. Our results and discussion to this point highlight an extremely dynamic grassland ecosystem as a function of 1) differences in grazing intensity giving a measure of a continuum of plant tissue removed; 2) 1 − G, or the remaining vegetation following grazing throughout the season; and 3) RGRg, the measure of regrowth of 1 − G following grazing. Furthermore, these variables make grassland ecosystem dynamics complex by operating at differing pulse rates [e.g., those environmental, ecological, and physiological oscillations regulating system behavior, including grazing systems as defined by McNaughton (1985a), Odum et al. (1995), and added to by our inclusion of a new category involving rapid rates involving definitive processes lasting from hours to days in duration: see Table 1].

First, we have argued that light to heavy grazing levels will result in rapid-pulse changes of the grassland canopy, reducing biomass and LAI by some measurable fraction and leaving a vegetation component equal to 1 − G (e.g., ΔLAI at Tn to LAI1−G at Tn+1) that affect both temporal and spatial estimates of R composed of visible and infrared irradiances (Smith et al. 1992a). These estimates of R may dramatically alter the basic component of flux calculations by affecting Rn. Furthermore, these estimates of events in the vegetation component, LAI1−G, will change rapidly over the course of time as the plants regrow, and become subject to new grazing stresses throughout the growing season, all potentially different as a function of grazing intensities that will vary according to grazer population levels and their interactions with any changes in the physical environment.

Second, ΔLAI1−G may affect the range of soil heat fluxes, but in ways somewhat different than the rapid-pulse changes potentially affecting Rn. Indirectly, ΔLAI1−G may have a rapid-pulse linkage to Rn, largely through the regulation of the amount of energy reaching the actual soil surface. Perhaps of more importance to soil heat fluxes, pulse differences in soil compaction and the availability and incorporation of litter may affect evapotranspiration dynamics. These differences arise through changes in ungulate density or total herbivore feeding rates from area to area through time. In a practical sense, such differences associated with longer pulses may have less significant overall effect on uncertainties related to the surface flux equation because investigators can more easily detect these changes in soil properties and stratify their sampling designs accordingly.

Finally, ΔLAI1−G may have significant consequences for determining sensible heat and latent heat fluxes and Bowen ratios. Such uncertainties in the main determinants could arise in both rapid- and slow-pulse events as a function of grazing intensity by altering irradiance fluxes and evapotranspirational cooling functions in the surface layer where sensors pick up changes in temperature and moisture gradients throughout the day and from day to day. Here we would expect a complex series of spatial and temporal interactions over a growing season governed by stochastic abiotic controls dictated by changes in weather patterns, as well as grazing frequency, and biotic controls as a result of the feedbacks to grazing. To our knowledge, no one has given a sufficient accounting of these interactions to permit those interested in ABL dynamics to move ahead with development of more dynamic models than those that currently exist.

d. Apparent relationships between the HO hypothesis and measures of variance

The literature review on over 800 studies on the responses of plants to herbivory initiated by Dyer and Wallace (Dyer and Wallace 1994; Wallace and Dyer 1995, 1996; Dyer and Wallace 1997, unpublished manuscript) gives strong support to the HO hypothesis. These studies cover over 55 plant species in grasslands, agriculture, and other ecosystems, and a large variety of herbivores, including ungulates, small mammals, birds, and many plant-feeding insects. In that study Dyer and Wallace have included the data from FIFE studies that we report here.

Fundamentally, the relationships in our FIFE work give much the same responses as monocots reported from the broader literature. Biomass changes across a landscape as a function of 1 − G in the way we report in Fig. 6 give a slightly more conservative peak than the broader literature (Δmax for mass = ∼40% vs ∼90%). For relative variance, as measured by vCV, the FIFE data give a slightly stronger response (Δmin ∼−45% vs −30%) of 1 − G compared to that of the broader study. Furthermore Dyer and Wallace (1997, manuscript unpublished) found significant influences imposed by a large variety of covariates, some of which we report. For example, C3 and C4 photosynthetic pathways affect both Δbiomass and ΔvCV across a large number of studies. Fertilization level, as a covariate, showed no effect for Δbiomass, but a large effect for ΔvCV. Several other covariates, such as plant part eaten, type of herbivore, scale of study, community type, and grazing intensity interactions with fire history, almost no data of which emerged during the FIFE studies, show a complex of control mechanisms (Dyer and Wallace 1997, unpublished manuscript). While the FIFE database has too few data to provide an adequate test for these assumptions, the lessons from a broader literature just now under study may have major implications for future work in describing biotic influences over ABL dynamics.

Dyer and Wallace (1997, unpublished manuscript) have suggested this biomass/relative variance relationship might contribute to questions of production and stability in ecosystems dominated by graminoids. If so, it raises the question of how large regional programs such as FIFE should view the consequences of grazing. Thus, we have argued that the biotic effects of grazers in grasslands have unknown and little understood influences over primary productivity. Many of these influences may affect our understanding of ABL dynamics.

5. Conclusions

Throughout this paper we have focused on ecological and biological processes induced directly or indirectly by grazers and the uncertainties they give to understanding ABL dynamics. For FIFE data, these uncertainties center mainly on large ungulates, although we cannot ignore the potential effects of invertebrate grazers. Our data suggest a high potential impact that biological processes may have on ABL model determinations, and perhaps ultimately on algorithms developed for global change programs. By ignoring the controls induced by heterotrophs we argue that Type II statistical error may preclude interpretations of both local and regional cause-and-effect mechanisms related to climate, weather, and biotic systems.

How has the FIFE project fared in measuring and assessing such basic ecological concepts as trophic ecology and its associated processes? Overall we suggest a mixed record. Almost certainly lesser known above- and below-ground processes associated with rapid and fast pulsing owing to differences in grazing stresses have significant uncertainties. However, it seems logical to assume that at some level of integration fluxes within the ABL microsite models merge to provide adequate ABL macroscale coverage. For instance, flux-measuring aircraft and the Lidar studies ought to have picked up the integrated signals due to grazing stress; however, such flux dynamics probably do not have an adequate basis to determine relevant cause-and-effect relationships with biological surface dynamics. Thus, as noted previously, the surface flux measurements probably contain a complex mix of Type I and Type II errors.

New studies should focus on at least some of the major uncertainties that we point out, principally those associated with rapid pulses in ecosystem function that most investigators have either found troublesome to measure or have ignored altogether. In some way new study designs need to incorporate the actual stress of herbivory—surrogates such as simple clipping of individual plants or mowing of large plots simply do not give the same feedback to the plant that actual animal feeding provides (Detling and Dyer 1981; Vinton et al. 1993; Wallace 1987, 1990).

Biotic processes do not operate in an isolated environment—they all have important rapid- and slow-pulse interactions with changes in weather. Thus, a new, sophisticated series of process studies must address this complex of interactions before scientists and managers alike can describe cause and effect for processes involved in positive feedbacks between herbivores and their plant hosts in pulsed ecosystems such as grasslands.

Acknowledgments

We thank D. A. Crossley Jr., S. J. McNaughton, E. P. Odum, L. L. Wallace, and three anonymous reviewers for valuable discussions that have helped us present our arguments. NASA Grants NAG 5-897 and NAG-5-910 along with NSF long-term ecological research Grant BSR-8514327 to Kansas State University and BSR-8904632 to M. I. Dyer helped obtain the database from which we have derived our analyses for this study.

REFERENCES

  • Benning, T., 1993: Fire frequency and topoedaphic controls of net primary productivity in the tallgrass prairie: Development and tests of remote sensing indices. Ph.D. dissertation, University of Colorado, 169 pp.

  • Blad, B. L., and D. S. Schimel, 1992: An overview of surface radiance and biology studies in FIFE. J. Geophys. Res.,97, 18829–18835.

  • DeAngelis, D. L., W. M. Post, and C. C. Travis, 1986: Positive Feedback in Natural Systems. Springer–Verlag, 290 pp.

  • de Mazancourt, C., M. Loreau, and L Abbadie, 1997: Grazing optimization and nutrient cycling: A model of potential impact of large herbivores in a savanna ecosystem. Ecology, in press.

  • Detling, J. K., and M. I. Dyer, 1981: Evidence for potential plant growth regulators in grasshoppers. Ecology,62, 485–488.

  • Dyer, M. I., 1980: Mammalian epidermal growth factor promotes plant growth. Proc. Natl. Acad. Sci. USA,77, 4836–4837.

  • ——, and L. L. Wallace, 1994: Control of plant productivity variation by herbivores Bull. Ecol. Soc. Amer. Suppl.,75, 59.

  • ——, J. K. Detling, D. C. Coleman, and D. W. Hilbert, 1982: The role of herbivores in grasslands. Grasses and Grasslands, J. R. Estes, R. J. Tyrl, and J. N. Brunken, Eds., University of Oklahoma Press, 255–295.

  • ——, D. L. DeAngelis, and W. M. Post, 1986: A model of herbivore feedback on plant productivity. Math. Biosci.,79, 171–184.

  • ——, C. L. Turner, and T. R. Seastedt, 1991a: Remote sensing measurements of production processes in grazinglands: The need for new methodologies. Agric. Ecosyst. and Environ.,34, 495–505.

  • ——, ——, and ——, 1991b: Mowing and fertilization effects on productivity and spectral reflectance in Bromus inermis plots. Ecol. Appl.,1, 443–452.

  • ——, M. A. Acra, G. M. Wang, D. C. Coleman, D. W. Freckman, S. J. McNaughton, and B. R. Strain. 1991c: Source-sink carbon relations in two Panicum coloratum ecotypes in response to herbivory. Ecology,72, 1472–1483.

  • ——, C. L. Turner, and T. R. Seastedt, 1993a: Herbivory and its consequences. Ecol. Appl.,3, 10–16.

  • ——, D. C. Coleman, D. W. Freckman, and S. J. McNaughton, 1993b:Measuring heterotrophic-induced source-sink relationships inPanicum coloratum with the 11C technology. Ecol. Appl.,3, 654–665.

  • ——, A. M. Moon, M. R. Brown, and D. A. Crossley Jr., 1995a: Effects of animal biochemical messengers on plant growth and development: A review of the evidence. Bull. Ecol. Soc. Amer. Suppl.,76, 71.

  • ——, ——, ——, and ——, 1995b: Grasshopper crop and midgut extract effects on plants: An example of reward feedback. Proc. Natl. Acad. Sci. USA,92, 5475–5478.

  • Herms, D. A., and W. J. Mattson, 1992: The dilemma of plants: To grow or defend. Quart. Rev. Bio.,67, 283–335.

  • Hik, D. S., and R. L. Jeffries, 1990: Increases in the net above-ground primary production of a salt-marsh forage grass: A test of predictions of the herbivore-optimization model. J. Ecol.,78, 180–195.

  • Hilbert, D. W., D. M. Swift, J. K. Detling, and M. I. Dyer, 1981: Relative growth rates and the grazing optimization hypothesis. Oecologia,51, 14–18.

  • Holland, J. N., 1995: Effects of above-ground herbivory on soil microbial biomass in conventional and no-tillage agroecosystems. Appl. Soil Ecol.,2, 275–279.

  • ——, W. Cheng, and D. A. Crossley Jr, 1996: Herbivore-induced changes in plant carbon allocation: Assessment of below-ground C fluxes using carbon-14. Oecologia,107, 97–94.

  • Kanemasu, E. T., and Coauthors, 1992: Surface flux measurements in FIFE: An overview. J. Geophys. Res.,97, 18547–18555.

  • Kato, R., K. Itagaki, K. Uchida, T. H. Shinomura, and Y. Harada, 1991: Effects of an epidermal growth factor on the growth of Zea coleoptiles. Plant Cell Physiol.,32, 917–919.

  • ——, E. Nagayama, T. Suzuki, K. Uchida, T. H. Shinomura, and Y. Harada, 1993: Promotion of plant cell division by an epidermal growth factor. Plant Cell Physiol.,34, 789–793.

  • ——, T. Shimoyama, T. Suzuki, K. Uchida, T. H. Shinomura, and Y. Harada, 1995: Effects of an epidermal growth factor on growth of Zea primary roots and mesocotyls. Plant Cell Physiol.,36, 197–200.

  • Loreau, M., 1995: Consumers as maximizers of matter and energy flow in ecosystems. Amer. Nat.,145, 22–42.

  • McNaughton, S. J., 1979: Grazing as an optimization process: Grass-ungulate relationships in the Serengeti. Amer. Nat.,113, 691–703.

  • ——, 1985a: Ecology of a grazing system: the Serengeti. Ecol. Monogr.,55, 259–294.

  • ——, 1985b: Interactive regulation of grass yield and chemical properties by defoliation, a salivary chemical, and inorganic nutrition. Oecologia,65, 478–486.

  • ——, M. B. Coughenour, and L. L. Wallace, 1982: Interactive processes in grassland ecosystems. Grasses and Grasslands, J. R. Estes, R. J. Tyrl, and J. N. Brunken, Eds., University of Oklahoma Press, 167–193.

  • ——, D. G. Milchunas, and D. A. Frank, 1996: How can net primary productivity be measured in grazing ecosystems? Ecology,77, 974–977.

  • Milchunas, D. G., and W. K. Lauenroth, 1993: Quantitative effects of grazing on vegetation and soils over a global range of environments. Ecol. Monogr.,63, 327–366.

  • Odum, W. E., E. P. Odum, and H. T. Odum, 1995: Nature’s pulsing paradigm. Estuaries,18, 547–555.

  • Oesterheld, M., 1992: Effect of defoliation intensity on aboveground and belowground relative growth rates. Oecologia,92, 313–316.

  • ——, and S. J. McNaughton, 1988: Intraspecific variation in the response of Themeda triandra to defoliation: The effect of time of recovery and growth rates on compensatory growth. Oecologia,77, 181–186.

  • Painter, E. L., and A. J. Belsky, 1993: Application of herbivore optimization theory to rangelands of the western United States. Ecol. Appl.,3, 2–9.

  • Paruelo, J. M., and W. K. Lauenroth, 1995: Regional patterns of normalized difference vegetation index in North American shrublands and grasslands. Ecology,76, 1888–1898.

  • Peterman, R. M., 1990: The importance of reporting statistical power. The forest decline and acidic deposition example. Ecology,71, 2024–2027.

  • SAS, 1988: SAS/STAT user’s guide. Release 6.03. SAS Institute, 1028 pp.

  • Schimel, D. S., T. G. F. Kittel, A. K. Knapp, T. R. Seastedt, W. J. Parton, and V. B Brown, 1991: Physiological interactions along resource gradients in a tallgrass prairie. Ecology,72, 672–684.

  • Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1992: An overview of the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE). J. Geophys. Res.,97, 18345–18371.

  • Smith, E. A., W. L. Crosson, and B. D. Tanner, 1992a: Estimation of surface heat and moisture fluxes over a prairie grassland. 1: In situ energy budget measurements incorporating a cooled mirror dew point hygrometer. J. Geophys. Res.,97, 18557–18582.

  • ——, and Coauthors, 1992b: Area-averaged surface fluxes and their time-space variability of the FIFE experimental domain. J. Geophys. Res.,97, 18599–18622.

  • Sokal, R. R., and F. J. Rohlf, 1981: Biometry: The Principles and Practices of Statistics in Biological Research. W. H. Freeman, 776 pp.

  • Steel, R. G. D., and J. H. Torrie, 1960: Principles and Procedures of Statistics. McGraw–Hill, 481 pp.

  • Tieszen, L. L., B. C. Reed, N. B. Bliss, B. K. Wylie, and D. D. DeJong, 1997: NDVI, C3 and C4 production, and distributions in Great Plains grassland cover classes. Ecol. Appl., 7, 59–78.

  • Turner, C. L., T. R. Seastedt, M. I. Dyer, T. G. F. Kittel, and D. S.Schimel, 1992: Effects of management and topography on the radiometric response of a tallgrass prairie. J. Geophys. Res.,97, 18855–18866.

  • ——, ——, and ——, 1993: Maximization of aboveground grassland production: The role of defoliation frequency, intensity, and history. Ecol. Appl.,3, 175–186.

  • Vinton, M. A., D. C. Hartnett, E. J. Finck, and J. M. Briggs, 1993: Interactive effects of fire, bison (Bison bison) grazing and plant community composition in tallgrass prairie. Amer. Midl. Nat.,129, 10–18.

  • Wallace, L. L., 1987: Effects of clipping and soil compaction on growth, morphology and mycorrhizal colonization of Schizachyrium scoparium. Oecologia,72, 423–428.

  • ——, 1990: Comparative photosynthetic responses of Andropogon gerardii to clipping versus grazing. J. Range Manage.,43, 58–61.

  • ——, and M. I. Dyer, 1995: Grassland management: Ecosystem maintenance and grazing. The Changing Prairie: North American Grasslands, A. Joern and K. H. Keeler, Eds., Oxford University Press, 177–197 pp.

  • ——, and ——, 1996: Grazing effects on grassland ecosystems. Ecosystem Disturbance and Wildlife Conservation in Western Grasslands, D. M. Finch, Ed., USDA General Tech. Rep. RM–GTR–285, 13–19., [Available from Rocky Mountain Experiment Station, Fort Collins, CO 80526.].

  • Wessman, C. A., C. A. Bateson, and T. L. Benning, 1997: Detecting fire and grazing patterns in tallgrass prairie using spectral mixture analysis. Ecol. Appl.,7, 493–511.

Fig. 1.
Fig. 1.

Location map of FIFE project grazing areas for 1987 and 1988, and 1987 mowing experiments conducted on Konza Prairie Research Natural Area (KPRNA). The study area near site 2 focused on simulating grazing intensity by mowing at different heights in pastures planted to an introduced brome grass; the study area near state road K-177 focused on both mowing intensity and frequency of grasses in the native prairie.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Fig. 2.
Fig. 2.

Notched box-whisker graph showing distribution statistics of the average proportion of plant grazed for four grazed pastures used in the FIFE study during 1987 and 1988. Graphics for each site show notched box (indication of 25th to 75th percentile of data with notch surrounding the median giving a 95% confidence interval), whiskers (vertical error bar extending from the edge of the box to the 5th and 95th percentile for the data), and mean (horizontal line across box).

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Fig. 3.
Fig. 3.

Notched box-whisker graph for growing season-long averages of grass standing crop (StCp) and total production (Prod) in each of the grazed pastures (see Fig. 1 for locations) studied during the 1987 FIFE project.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Fig. 4.
Fig. 4.

Notched box-whisker graph for grass standing crop (StCp) and total production (Prod) in each of the grazed pastures studied during the 1988 FIFE project.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Fig. 5.
Fig. 5.

Estimates of Δ relative growth rates of grasses in grazed pastures (RGRg) relative to plants in ungrazed exclosures (RGRug), both measured in terms of average growth rates (g) of new plant tissue in a standard unit of space (m2) computed on a daily basis between successive sampling dates [see text, Eq. (2)] (g m−2 day−1). Regression with heavy dashed line = 1987 data; dotted regression = 1988; solid regression line = both years together.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Fig. 6.
Fig. 6.

Distribution of Δ (in %) for total productivity and relative variance (vCV: Sokal and Rohlf 1981) for mowed and grazed pastures when compared to controls in the 1987 and 1988 FIFE projects. We show both sets of variables plotted against grazing intensity, (1 − G), for four grazing classes grouped by quartiles (grazing classes = 0.25) following Dyer and Wallace (1994, 1997 unpublished manuscript) [because the boundaries of no grazing (=0, or control conditions) and complete vegetation removal (=1.0) define the HO hypothesis, we use a quadratic curve to force as expression of relationships for each variable through 0 and 1.0 intercepts].

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1247:BIBGAP>2.0.CO;2

Table 1.

Comparison of elements in surface energy balance model and the effects of grazing intensity and trampling on ecosystem behavior across temporal and spatial scales.

Table 1.
1

A confusing usage of terms appears in the literature, one that we wish to resolve. In Eq. (1) we noted that McNaughton (1979, 1985a) defined “G” as (1 − g/ug), whereas here we point out that Hilbert et al. (1981) use “G” as the proportion of S1 reduced by grazing so that at t1, the time grazing occurs, shoot weight = S1(1 − G). Throughout this paper we retain the original expressions (e.g., we have not attempted to reconcile the conflict in terminology by redesignating either of the two terms); thus, we use them in their proper form for computation of grazing intensity where needed and similarly for determining ΔRGR values.

Save