Developmental morphology, biomass yield and compositional differences among upland and lowland switchgrass (Panicum virgatum L.) ecotypes grown as a bioenergy feedstock cro

Sustainable and successful development of the bioenergy industry strongly depends upon the amount and quality of bioenergy feedstock produced. Switchgrass (Panicum virgatum L.) has been identified as a model lignocellulosic bioenergy crop in the U.S. Information regarding its growth and development is considered critical for making management decisions, production of high quality feedstock and to support simulation model calibration and subsequent analysis.;In the first study (Chapter 2), we evaluated upland ('Cave-in-Rock', 'Trailblazer' and 'Blackwell') and lowland ('Kanlow' and 'Alamo') ecotypes of switchgrass for harvest management, morphological (phenology and leaf area index) and biomass yield differences. A field study was conducted near Ames, IA during 2012 and 2013. The experiment was laid out as randomized complete block design. Six in-season destructive biomass harvests were collected at approximately 2-week intervals each year. In both years, delaying harvest to later maturity increased biomass yield in all cultivars. Averaged over years lowland cultivars 'Kanlow' and 'Alamo' produced higher dry matter yield (6.31 and 5.98 tons ha-1, respectively) than upland ecotypes 'Cave-in-Rock', 'Trailblazer' and 'Blackwell' (5.89, 4.78 and 4.53 tons ha-1, respectively). Lowland cultivars had delayed reproductive development compared with upland ecotypes. The widely recommended cultivar in Iowa 'Cave-in-Rock' had the highest mean stage count and leaf area index at the end of both growing seasons, but did not produce the greatest biomass. There was a nonlinear relationship between MSC and biomass yield. However, the magnitude and form of the response was different between cultivars and from year to year.;In the next study (Chapter 3), our objective was to quantify the chemical composition of switchgrass varieties as influenced by harvest management, and to determine if ecotypic differences exist among them. We found that delaying harvest increased cellulose, hemicellulose and lignin concentrations while decreasing nitrogen and ash concentrations in all varieties. On average, Kanlow had the highest cellulose and hemicellulose concentration (354 and 321 g kg -1 DM respectively), and Cave-in-Rock had the highest lignin concentration (33 g kg-1 DM). The lowest nitrogen and ash concentrations were observed in Kanlow (14 and 95 g kg-1 DM respectively). In general, our results indicate that delaying harvest until fall improves feedstock quality, and ecotypic differences do exist between varieties for important feedstock quality traits.;The objective of the third and final study (Chapter 4), was to developed a new mechanistic model to describe switchgrass phenology with the objective to assist agronomists and on-going breeding programs. Switchgrass is sensitive to photoperiod. However, existing switchgrass phenology models rely on thermal time coupled with ad-hoc empirical modifications to account for the effects of photoperiod when the model is to be applied across a wide range of environments. Our model simulates four phases of switchgrass development (emergence to juvenile, juvenile to elongation, elongation to flowering and flowering to maturity). It uses daily temperature and site latitude as driving variables, contains five cultivar specific biological meaningful parameters, and two model constants (base and optimum temperature of 10°C and 30°C, respectively). Three of the cultivar specific parameters (a1, a2, a 3) define the thermal optimum time needed to complete a phase and the other two parameters describe the critical photoperiod (pcrit) and the photoperiod sensitivity (psen). The model matched Iowa's observations with a RMSE (relative mean square error) of 2.6 days for each cultivar or with 5.1 days error when average parameters by ecotype were used. The next step in our analysis was to generalize the model by providing estimates of the photoperiod effect, and for that we used diverse literature database. We found that pcrit was 13.5 h and 12.7 h for upland and lowland cultivars. By using these parameters the overall prediction ability of the model across representative sites in the U.S. had an 8.4 days error. This model can be a helpful tool for improvement and development of calibration protocols for other models and it can also serve as a template for the development of phenology models for other perennial grasses.