Our first objective was to determine how organic C storage and organic N turnover, both key ecosystem processes, were affected by fire at two frequencies in this historically N-limited ecosystem (Aber et al., 1989). The N mineralization rates of soils of our annually and periodically burned plots were approximately 70% and 20% greater, respectively, than those of the unburned controls. It is common to such increases in N mineralization rate and/or TIN in response to fire in many ecosystem types. For example, fire-induced increases in TIN and/or N mineralization have been reported in Pinus ponderosa forests in western North America (Wagle and Kitchen, 1972), mixed pine (P. echinata and P. taeda) forests in east Texas (Webb et al., 1971) and California chaparral (Debano et al., 1979).
The soils of the frequently burned units exhibited net nitrification while those from the periodic fire units and controls immobilized NO3
−. Neither soil organic C content nor total inorganic N (TIN) in the soil solution was affected by fire at either frequency. In contrast to N mineralization, the nitrification and TIN results were not consistent with the majority of existing fire research. For example, the meta-analysis of the effects of fire on N cycling done by Wan et al. (2001) concluded that post-fire increases in NO3
− are common, and the [NO3
−] peaks approximately one year after fire.
Caution must be exercised in extrapolating the effects of a single fire to those of multiple fires over time. Vance and Henderson (1984) found that annual burning over 30 years reduced N mineralization and TIN to a greater extent than did periodic burns over the same period. We also found that the initial effects of the first burn in these two study sites (measured one month postfire) differed considerably from what we observed three years later, even in the treatment units burned only once (Boerner and Brinkman, 2004a).
Our second objective was to determine the degree to which these ecosystem properties and processes were spatially autocorrelated and at what maximum distance such spatial autocorrelation existed. The maximum range of spatial autocorrelation essentially establishes the size of non-random patches of resources on the forest floor. Our 50 m2 macroplots permitted estimation of spatial autocorrelation at ranges from approximately 1.0 m to 7.8 m, and only in approximately 1/3 of the site-by-fire treatment combinations were we able to demonstrate significant spatial autocorrelation in N turnover within this range of spatial scales; of those we could resolve, almost all were in recently burned plots.
Our 1m2 microplots permitted estimation at ranges from 20 cm to 80 cm, and once again we observed significant spatial autocorrelation in N turnover in only approximately 1/3 of the possible cases. Of those, five of seven (71%) were from the Young’s Branch study site, and four of seven (57%) were from unburned control units. Thus, in contrast to what we observed for the larger, 1–8 m patches, most of the discernable patches <1m were in unburned plots.
Spatial dependency in soil biological, chemical, and biochemical properties at a range of nested scales seems to be a common observation in the literature (review by Ettema and Wardle, 2002). However, in few cases have the range of scales and their interrelations been empirically estimated. In sites similar to ours, spatial structure at ranges < 1m have been documented for fungal and bacterial biomass (Morris, 1999), mycorrhizal fungal infectivity (Boerner et al., 1996), and N mineralization (Boerner and Koslowsky, 1989, Bruckner et al., 1999). Other studies have demonstrated spatial structure at ranges between 1 and 10 m, though most have employed sampling designs that precluded quantification at scales <1 m (e.g. Robertson, 1987; Boerner et al., 1998). Other studies yet demonstrate spatial structure at ranges >10 m, including many which clearly document spatial structure in soil properties that correlate with landscape geomorphic features (e.g. Rahman et al., 1996; Boerner and Brinkman, 2004b).
The results of this study suggest that relatively little spatial structure in soil organic C and N turnover exists in our study sites over the range of 0.2 m to 7.8 m. As strong spatial structure at ranges of <0.2 m (e.g. Morris 1999) and > 7.8 m (e.g. Boerner and Brinkman, 2004b) have already been established in these and nearby study sites, we believe we have now with some certainty established quantitatively the range of nested, hierarchic scales at which these and similar ecosystem processes are structured. Our next challenge is to link this hierarchic pattern of patches in soil properties with vegetation properties that are important to longer term management, especially the diversity of the understory flora, the population dynamics of threatened and endangered plant species, and the abundance and survivorship of tree seedlings.
Our third objective addressed the relationship between sample plot size (in this case 1 m2 vs. 50 m2) and judgments about the magnitude, significance, and possible importance of fire effects. In the analysis of variance of the full experiment, plot size was not a significant source of variation for any of the four soil response variables we measured. Pairwise t-tests of responses between plots sizes within individual combinations of fire treatment and response variable also demonstrated no significant effect of plot size. Overall, this suggests that estimates in these study sites derived from intensive sampling of small plots (25 samples within 1 m2) could be scaled up to 50 m2 with confidence. It does not necessarily follow from this, however, that such extrapolation would still be reliable if the sampling intensity within the 1 m2 plot were reduced greatly.
Several authors (e.g. Bonnicksen and Stone, 1982) have suggested that fire suppression tends reduce spatial heterogeneity in western mixed conifer forests, and Raison (1979) concludes from an exhaustive review that fire tends to increase the spatial heterogeneity of soil resources, especially N. This led us to hypothesize suggest that fire suppression should lead to the loss of spatial structure dominated by small patches in the ecosystems we studied as well. In support of this hypothesis, at our macroplot scale significant patchiness at the 1.0 – 7.8 m scale could be quantified only in recently burned areas. However, to conclude from this study that this might be a general trend in hardwood forests would not be justifiable given the relatively small number of sites, treatment units, and sample plots analyzed.
Future research should address the mechanism(s) that underlay this change in spatial structure, if indeed such a relationship between fire and patchiness exists in other forested ecosystem types. In our region with fire suppression comes an increase in understory biomass and stem density. These changes can, in turn, could be expected to reduce the heterogeneity of the forest floor through an increase in fine root biomass (Dress and Boerner, 2001). Increases in understory woody stem density would also be expected to reduce patchiness through reductions in autumn and winter leaf litter redistribution, a major mechanism for generating spatial heterogeneity in these landscapes (Boerner and Kooser, 1989). Increasing fire frequency, in contrast, results in a more open understory, akin to that described by early settlers and surveyors (Whitney, 1994) and a much more spatially heterogeneous stem distribution. We suggest this fire-induced increase in above-ground spatial heterogeneity may translate to differences in C storage and N turnover in the forest floor and soil, and we urge others to test this hypothesis in a broader range of forested ecosystem.
Studies of presettlement forest structure suggest that frequent, low intensity fires tended to produce and maintain high diversity and high spatial heterogeneity. Thus, fire suppression may result in increasing homogeneity. Just as increasing patch size and coarse-grained structure may serve as an indicator of desertification in desert grasslands, fire-suppression induced changes in spatial structure may signal the onset of negative impacts on deciduous forest ecosystems.