Biotic-Abiotic Interaction Experiment
Ecosystem structure and function are controlled by a variety of biotic and abiotic factors acting singly and in combinations. The role of multiple factors in controlling ecosystem function has been variously described over the years to characterize the complexity of these relationships (e.g., Jenny 1941, Billings 1952, Chapin et al. 1996), but these conceptual descriptions have not been evaluated experimentally or quantified. Comparative approaches to understanding ecosystem structure and function (IBP, e.g., Reichle 1981) have provided great insights into understanding the diversity of system structures and their relationship to function (Cole and Rapp 1981). Continued research over the years has elucidated many important ecosystem-level mechanisms with respect to material cycling and energy flow. However, because replicated whole-ecosystems are difficult to establish, we have learned a great deal about many factors taken singly, but much less about how these factors interact to control ecosystem-level processes.
The term "interaction" is used by ecologist in many different contexts, and clarification is required before proceeding. Almost all definitions and discussions of ecosystems refer, in some manner, to communities of organisms interacting with each other and with biogeochemical factors that collectively represent the environment. Van Dyne (1969), for example, emphasizes that ecosystems are represented by the "totality of interactions" and further argues that only a model that depicts interactions can assess the structure and function of the ecosystem as a whole. This totality of interactions includes species-species interactions (competition, predation, etc.); species-abiotic factor interactions (resource limitations and responses, physiological stress, etc.), and non-additive relationships among factors (recognizable/quantifiable as statistical interactions). It is this latter array of interactions that we explore in this study.
When the combined effect of two factors on an ecosystem process (e.g., growth, control of nutrient cycling) is different than the sum of the two effects taken alone, then the interaction of the two factors result in a non-additive situation often recognized as an antagonistic or synergistic relationship. From a quantitative perspective, this situation results in statistical interactions among "treatment" variables (independent variables) that influence the outcome or level of a response variable (dependent variable). The statistical aspect of this factor interaction can be tested and quantified by analyzing components of variation (F statistic and expected mean squares). Interactions of this type are important, if not defining, attributes of many ecological systems. For the purpose of this discussion, we refer to this type of interaction as "contextual" interaction, to differentiate it from other interactions such as direct pairwise interactions (e.g., competitive interactions between two species, or the additive responses of an organism to individual environmental factors). The notion here is that the response of a process (e.g., growth) to a driving variable (e.g., nitrogen level) depends on other variables -- i.e. the "context" in which the plant is growing (e.g., levels of light, soil moisture, other nutrients, symbionts, the associated plant and animal community, etc.). This context effect tends to be non-additive, non-linear, and difficult to predict. We believe that this contextual interaction explains a substantial and significant component of ecological responses to environment factors, and this interaction is critical during ecosystem development. If the response to an enhanced level of light and a higher level of nitrogen is not equivalent to the sum of the two factors taken alone, then a "contextual" interaction (also a statistical interaction) exists indicating an antagonism or synergy between light and nitrogen. For example, if enhanced light in the presence of higher nitrogen had no additional growth effect (but enhanced light under low nitrogen did have an effect), then a factor antagonism exists.
This well-recognized but perhaps little-appreciated type of interaction in agronomy studies and probably most factorial treatment experiments may be much more important than a statistical nuisance requiring alternative statistical models. At the organism and population levels, these contextual interactions often play a significant role (Chapin 1991). In studies of population genetics contextual interactions must be interpreted in order to understand the role of gene combinations in controlling performance of trees across their range of distribution. The use of statistical tools in analyzing genetic provenance trials allows quantification of the interaction, and often these analyses demonstrate the importance of this factor in understanding the fundamental relationship of organisms to their environment (Zobel and Talbert 1984; Namkoong 1979; Wright 1973,1974). Previous work in Vermont has shown that yellow birch (Betula alleghaniensis) growing under different light intensities and temperature treatments can have between 45 and 75% difference in photosynthetic efficiency explained by the biotic X abiotic interaction (Shane 1992, 1989). Another one of our studies demonstrated that over 90% of the variation in balsam fir (Abies balsamea) seedling growth was attributed to the population X fertilizer concentration interaction (Marshall 1989) Thus, the concept of "performance" in making genetic selections is heavily dependent on the outcome of the contextual interactions among abiotic and biotic factors that control plant fitness and growth rates.
Although described with a different terminology, this type of interaction has also received considerable attention with respect to multi-species interactions such as competition and predation. In forestry, this phenomenon translates to important differences in productivity depending on the species context (Kelty 1992). The determination, interpretation, and importance of "higher order interactions" have been discussed by several authors (Case and Bender 1981, Abrams 1983, Worthen and Moore 1991, Woottoon 1994, Billick and Case 1994), mostly with respect to animal species interactions. In this context, the effect of a third species on the direct, pairwise interaction of two other species is considered a higher order interaction. Adler and Morris (1994) use the term "interaction modification" to more clearly distinguish this phenomenon from indirect effects and non-additive interactions driven by multiplicative terms and provide a quantitative approach to evaluating it. They suggest that any realistic model of an ecological interaction should include interaction modifications because it is so prevalent in nature.
At the community level, with respect to studies of succession, this contextual interaction is at the foundation for a proposed predictive theory of competition and succession (Tilman 1990). "A given plant species may have different constraints in different habitats, and different species living in the same environment may be limited by different factors" (Tilman 1990, 1977). The interaction of these varying constraints and factors are critical determinants of the success of species over successional time. For example, the nutrient:light ratio hypothesis suggests that succession is controlled by variations in this ratio and the differential ability of species to respond to these variations (Tilman 1993). Typically these multifactor responses exhibit strong contextual interactions which are interpreted, for example, as "different constraints in different habitats" (Tilman 1990).
If contextual interaction is important at the organism, population, and community levels, what is its role at the ecosystem level? What kind of experimental designs will permit this type of interaction to be quantified and how do we interpret and find ecological meaning for the phenomenon? Chapin et al. (1996) proposed a framework in which ecosystems are characterized by the four interactive controls of 1) local climate, 2) soil resource supply, 3) major functional groups of organisms, and 4) disturbance regime. Their model clearly recognizes the "interactive" nature of these factors, and they and others (DeAngelis and Post 1991) suggest that contextual interactions within and among interactive controls generate feedbacks (e.g., mutualisms) that are key to providing ecosystem resistance and resilience to natural and anthropogenic modification of the four factors controlling ecosystem structure and function. Thus, their theoretical perspective suggests that contextual interactions at the ecosystem level are fundamental to understanding how nature works to create stable or meta-stable conditions. Empirical, comparative studies of ecosystem biomass and net primary production at different sites suggest that strong/large contextual interactions exist between species and site fertility (e.g., Binkley 1983).
Because field sites cannot be chosen with all other factors being constant and true replication does not exist, the statistical analysis of contextual interaction can not be conducted/pursued in these types of studies. Thus we chose to use artificially created mesocosm ecosystems.