Energy production is expected to increasingly rely on dedicated bioenergy crops. Field experiments have demonstrated that low input high diversity mixtures of native grassland perennials can produce more usable energy, greater greenhouse gas reductions, higher carbon and nitrogen accumulation than monocultures. Furthermore, the mixtures have higher ecological restoration value relative to corn, Miscanthus or any other potential high-yield monoculture. As experiments progress, there is a growing need for adequate models and theories to provide robust generalizations, predictions, and interpretations of experimental results. To support this goal a database containing trait and yield information of possible biofuel crop species is needed to (1) evaluate the biofuel potential of the grasses and (2) parameterize models that will be used to predict how richness and composition of plant species/plant functional types change in polycultures and link community and ecosystem processes in large scale. In this analysis we focus on 28 tall grass prairie species that were planted in the tall-grass prairie restoration plots at the Energy Biosciences Institute (EBI) Energy Farm (Urbana, IL, USA). The mixed-species polycultures consist of C4 grasses, C3 grasses, sedges, warm season and cool season legumes and forbs. Species-level plant yield and trait data was collected from a combination of peer-reviewed literature and field experiments conducted during the summer of 2009.

Results/Conclusions

Based on this research a database with more than 5000 entry for the 28 potential biofuel plant species has been constructed as part of the EBI Biofuel crop database (http://ebi-forecast.igb.uiuc.edu/). Analysis of yield data shows that certain species like big bluestem (Andropogon gerardii) cup plant (Silphium perfoliatum) and wholeleaf rosinweed (Silphium integrifolium) have relatively higher productivity while yield of culver’s root (Veronicastrum virginicum), yellow coneflower (ratibida pinnata) is very low. Trait data was synthesized in a Bayesian meta-analytical model and posterior distributions of traits was used to parameterize the Ecosystem Demography Model. Ensemble runs were conducted to propagate parameter uncertainty through the model. These runs were used to construct error estimates around model forecasts, to compare modeled and observed yield, and to identify which processes and which species require further study.