Although fire played an important role historically in the dynamics of many mixed tree-grass vegetation systems, the details are poorly understood, especially for oak savannas in central N. America prior to European settlement. The forest - grassland ecotone provides a number of puzzles about its stability, the role of disturbance, and internal vs. external controls on composition, diversity, productivity and nutrient cycling (Scholes and Archer 1997). Although we know an increasing amount about the influence of trees on grassy vegetation (e.g., Jackson et al. 1990, Belsky 1994, Haworth and McPherson 1994) via effects on light, water, and nutrient cycling, we know less about the dynamics of these ecosystems as a whole. In particular, although substantial effort has been directed at understanding controls on NPP and N cycling for forests (e.g., Pastor et al. 1984, Gower et al. 1992; Gholz et al. 1994) and grasslands (e.g. Risser et al. 1981; Knapp et al. 1993, 1998, Turner et al. 1997) generally and at Cedar Creek (Wedin & Tilman 1990, 1993; Reich et al. 1997a) this is not as true for savannas (Mitchell et al. 1999). Plant traits, disturbance, and climate exert joint control of ecosystem functioning (e.g., Pastor et al. 1984; Tilman 1988, Parton et al. 1987, 1988; Aerts 1990, 1992; Wedin & Tilman 1990), but studies of climate effects have usually been made at a single time across sites varying in climate, rather than examining across climate (i.e., internannually) at a single site.
Description of Experimental Methods
We are addressing the above issues as part of Theme 1, and within the context of fire and oak savannas, also addressing the other 3 themes. To do so we focus on a long-term prescribed burning experiment (E015, begun in 1964; expanded to E133, involving 29 landscape units (3 to 27 ha) which range widely in fire frequency. These semi-native oak woodlands may have experienced some selective logging and grazing, but were never plowed, and were protected from fire from 1938 to 1964. Fire treatments (prescribed spring ground fires) range from complete fire protection to near-annual burning (8 fires per decade), spanning the range of presettlement fire frequencies along the forest-prairie border. Permanent plots (50 x 75 m) were established in 12 burn unit plots in 1984, and in 17 more by 1995. Several other ongoing fire experiments will also be continued, including E012 (Effects of fire frequency on old field succession), E143 (Accelerated savanna restoration), and E002 (Effects of grassland fire and N).
Results to Date
- Periodic fires (at medium to high frequency) gradually kill mature trees (Fig. 23 [pdf]) and suppress sapling recruitment, but responses differ by species (Peterson & Reich 2000a). As a result, there is a gradient of tree composition, density (Fig. 24 [pdf]) and biomass associated with fire frequency (Peterson & Reich 2000a, Reich et al. 2000). These long-term changes enable us to characterize the fire regimes required to sustain a presettlement landscape mosaic with savannas, and the prescribed fire regimes needed to restore and maintain oak savannas (Peterson & Reich 2000a).
- Fire, directly and indirectly by altering tree canopy cover, has had major effects on breeding bird composition and diversity (Davis et al. 2000a, b) and on plant species composition (Fig. 25 [pdf]). Major plant functional groups diverged in presence and abundance due to within- and cross-plot variation in light, N and water availability (Fig. 26 [pdf]).
- Patterns of species diversity weakly supported the intermediate disturbance hypothesis (Peterson & Reich 2000b), with total vascular plant diversity highest at intermediate fire frequencies in communities dominated jointly by trees and grasses. However, individual functional groups did not follow that pattern. The data suggest that maintenance of spatial heterogeneity in woody cover and associated resources is likely the mechanism leading to higher plant diversity in savannas than grassland or forested plots.
- Fire has direct effects on ecosystem function and indirect effects via changes in the vegetation composition (Reich et al. 2000, Peterson et al. in review). Differences in traits between woody and herbaceous species influence C and N cycling because of how they change their acquisition and turnover rates at tissue-to-ecosystem scales. As a result, frequently burned, open savanna differs greatly from unburned, closed woodland in terms of NPP (Fig. 27 [pdf], Reich et al. 2000), C storage (Fig. 7A [pdf], Tilman et al. 2000), and N cycling (Fig. 28 [pdf], Reich et al. 2000). Belowground NPP as a proportion of total NPP is not higher in grass than tree-dominated communities, contrary to many hypotheses. Both fire and vegetation feedbacks play important roles. Net N mineralization rates beneath trees are greater than beneath adjacent grassy patches in the same fire treatment unit (Wrage et al., unpublished data) and even for a given fire regime, net N mineralization and ANPP are 3-fold greater in 80% tree-dominated than 80% grass-dominated communities (Reich et al. 2000).
Past, Ongoing, & Future Mechanistic & Theory Work
Are savannas merely additive mixtures of grassland and forest patches in terms of their ecosystem function or compositional dynamics? Although empirical data address this, we are also beginning to use modeling as a means of integration. This involves both biogeochemistry modeling (PnET and Century) and modification of vegetation dynamics models to examine tree-grass compositional patterns. A variety of manipulative experiments enables tests of specific mechanisms. An example is a field study of competitive interactions under multiple resource combinations, which found that greater N availability led to increased competition for water by increasing herbaceous biomass, thereby leading to greater oak seedling mortality (Davis et al. 1998, 2000b). In contrast, without competition, a precipitation regime that simulated a 100-yr drought had minimal effects on young oaks, regardless of N supply rates.
In 1999, we expanded our set of on-going measurements to more fully address all 4 Themes in E133. This sampling will continue, in some cases annually, for 2000-2006. We will examine C and N cycling to better understand the roles of disturbance (fire), plant traits (woody vs. herbaceous), resource supply (N) and climate (interannual variation). The annual studies will be made in 12 plots spanning the range of fire frequency. Composition will be characterized by annual censusing of vegetation. We will examine many aspects of C and N cycling, including aboveground litter N flux (Reich et al. 1997a, 2000), if possible fine root turnover rates (Eissenstat 1997; Eissenstat & Yanai 1997) and belowground litter N flux; decomposition rates of aboveground and belowground litter (Aber & Mellilo 1982, Melillo et al. 1982, Hobbie & Vitousek 2000); soil net N mineralization rate (Grigal & Homann 1994, Reich et al. 1997a, 2000), soil CO2 flux (Craine et al. 1999b), aboveground NPP (Reich et al. 1997a, 2000), and three methods to assess belowground NPP- (1) N budget, (2) in-growth root cores, and (3) standing fine root biomass combined with fine root birth and mortality rates stratified by root order and diameter (Fahey et al. 1985, 1999, Vogt et al. 1989, 1998, Reich et al. 2000). Decomposition studies will include manipulations to determine effects of vegetation change on C and N cycling across the fire frequency gradient. Experiments will include in situ and reciprocal decomposition studies of both above- and belowground biomass; soil incubations; and reciprocal soil transplants. In addition to field mineralization studies (see above), in one year, C and N mineralization will be studied in laboratory incubations and with stable isotope techniques (e.g., Wedin et al. 1995). We will quantify N losses associated with fires compared to N immobilized in litter in short-term decomposition experiments, in an attempt to develop a complete N budget (including inputs and outputs) for these stands. Other research activities will complement these core measurements. These include: (1) an ongoing annual acorn predation and seedling establishment experiment; (2) a 1-year study of spatial heterogeneity in composition and function in two 4 ha savanna plots; (3) detailed measurements of composition and resource availability in all 24 subplots of the 29 oak savanna burn unit plots: these are made every 5 to 6 years and provide a more detailed basis for long-term changes than the 12 plots (8 subplots) to be examined annually; (4) a 1-year study of foliage and fine root ecophysiology in the core savanna plots, (5) collaboration on mycorrhizal research, and (6) use of stable isotope techniques to evaluate uptake and utilization of C, N and water. In total, these activities will enable us to concurrently explore root and foliage linkages to turnover, productivity, soil properties and C and N cycling at tissue-to-ecosystem scales, as influenced by interannual variation in climate, variation in fire frequency, and N availability and vegetation composition gradients.