Testing the hypothesis of hierarchical predictability in ecological restoration and succession.

To advance predictive ecology, the hypothesis of hierarchical predictability proposes that community measures for which species are interchangeable (e.g., structure and species richness) are more predictable than measures for which species identity matters (e.g., community composition). Predictability is hypothesized to decrease for response measures in order of the following categories: structure, species richness, function, and species composition. We tested this hypothesis using a 14-year, oak savanna-prairie restoration experiment that removed non-native pine plantations at 24 sites in northwestern Ohio, USA. Based on 24 response measures, the data showed minimal support for the hypothesis, because response measures varied in predictability within categories. Half of response measures had over half their variability modeled using fixed (restoration treatment and year) and random plot effects, and these "predictable" measures occurred in all four categories. Pine basal area, environment (e.g., soil texture), and antecedent vegetation accounted for over half the variation in change within the first three post-restoration years for 77% of response measures. Change between the 3rd and 14th years was less predictable, but most restoration measures increased favorably via sites achieving them in unique ways. We propose that variation will not conform with the hypothesis of hierarchical predictability in ecosystems with vegetation dynamics driven by stochastic processes such as seed dispersal, or where vegetation structure and species richness are influenced by species composition. The ability to predict a community measure may be more driven by the number of combinations of casual factors affecting a measure than by the number of values it can have.

A multi-scale spatial analysis of native and exotic plant species richness within a mixed-disturbance oak savanna landscape.

Impacts of human land use pose an increasing threat to global biodiversity. Resource managers must respond rapidly to this threat by assessing existing natural areas and prioritizing conservation actions across multiple spatial scales. Plant species richness is a useful measure of biodiversity but typically can only be evaluated on small portions of a given landscape. Modeling relationships between spatial heterogeneity and species richness may allow conservation planners to make predictions of species richness patterns within unsampled areas. We utilized a combination of field data, remotely sensed data, and landscape pattern metrics to develop models of native and exotic plant species richness at two spatial extents (60- and 120-m windows) and at four ecological levels for northwestern Ohio's Oak Openings region. Multiple regression models explained 37-77 % of the variation in plant species richness. These models consistently explained more variation in exotic richness than in native richness. Exotic richness was better explained at the 120-m extent while native richness was better explained at the 60-m extent. Land cover composition of the surrounding landscape was an important component of all models. We found that percentage of human-modified land cover (negatively correlated with native richness and positively correlated with exotic richness) was a particularly useful predictor of plant species richness and that human-caused disturbances exert a strong influence on species richness patterns within a mixed-disturbance oak savanna landscape. Our results emphasize the importance of using a multi-scale approach to examine the complex relationships between spatial heterogeneity and plant species richness.