Understanding the effect of plant richness, patch configuration and landscape composition on biodiversity in bioenergy production systems

Consistent and low yielding areas of agricultural fields are prevalent across the US Midwest and present the opportunity to enhance regional biodiversity through grass land conservation and land transfers from annual production to perennial bioenergy crops. Medium resolution, regional scale assessment of biodiversity using remote sensing is needed to inform land use policy at the regional scale. From the literature, direct observation of plant richness at the plot scale and compositional assessment at the landscape scale may be integrated to assess biodiversity regionally. This assessment requires translating plot diversity in both flora and fauna richness to transferable models using medium resolution remote sensing imagery. Our objectives are 1) to model plant, bird, butterfly, and bee richness at the field scale and 2) isolate the effects of plant biodiversity on non-plant group's richness from other site features. Using data from an extensive field campaign of bioenergy plots at the Kellogg Biological Station in SE Michigan, we developed a random forest model to assess each group's richness from monthly features constructed from visible, near infrared, and thermal imagery and elevation parameters. Additionally, plant cover was classified into three categories: annual monocultures, perennial monocultures, and diverse perennials. Nested predictions, where inclusion of the plant cover was used in the prediction of non-plant group richness, improved model performance. The trained and validated models were further evaluated at the field scale with sites across SE Michigan. Sites were selected along a gradient of low to high diversity perennials. Plant cover classification was used to calculate landscape composition and configuration metrics and to investigate the separation of plant cover, composition, and configuration as predictors of each group's richness. High level biodiversity assessments from remote sensing may extend the findings of costly and time-consuming field investigations and inform in-field and regional conservation policy towards increased biodiversity.