As users decide on which model is appropriate for their purposes, they will need to consider the level confidence in the inputs required and the appropriate level of model complexity to accomplish their goals. Process — based simulation models are split into those comprised of: 1) detailed leaf photosynthesis components that are integrated up to the leaf canopy level or 2) canopy level models that contain the relationship between plant functions and leaf area index (LAI), light interception, and radiation use efficiency (RUE). There are some concerns with both approaches. Leaf level photosynthetic rates are often not directly related to productivity, as described with RUE or above-ground net primary productivity (ANPP). A good example is a study by Kiniry et al. (1999) reported that sideoats

grama (Bouteloua curtipendula (Michx.) Torr.) had higher photosynthetic rates throughout the range of light levels than Alamo switchgrass, but sideoats gramma is far less productive than switchgrass. Similarly, Aspinwall et al. (2013) looked at several switchgrass ecotypes and concluded that "leaf — level physiological traits are often uncorrelated with genotype ANPP due to confounding of development with physiology, covariation among leaf traits, feedbacks with sink capacity, and increased self-shading". However, they identified "a syndrome of leaf functional traits" which aligned with genotype ANPP revealing that more productive genotypes initiated growth earlier and flowered later.

On the other hand, parameters for whole canopy models such as ALMANAC, EPIC, and SWAT are derived by measuring leaf area and dry matter destructively during the active growing period and fraction of light interception of plants assumed to be grown under nonlimiting water and nutrients conditions. Ideally such models use plant parameters derived at one site, with adequate soil moisture and soil nutrients, which are then applied for simulations under a wide range of environmental conditions. However, problems can arise when applying the model parameters, outside the regions of adaptation of a particular switchgrass ecotype. Latitudinal differences include photoperiod, number of hot days during the growing season, and number of cold days during the winter. Realistic simulation of processes controlling location differences, especially with differences in latitude, requires realistic understanding of the factors affecting such adaptation. These adaptation processes still need to be identified and quantified to more accurately simulate switchgrass ecotypes across a wide range of locations.

Another concern when simulating switchgrass with these process-based plant models is that many of these models were developed for annual crops. The perennial growth process is quite different from that of annuals, and more work is needed to accurately incorporate these differences in models, such as rooting during the establishment year as compared to subsequent years. Especially important for switchgrass modeling is N, P, and carbohydrate storage in roots in autumn and translocation out to above­ground plant parts in the spring. In addition, there has not been sufficient research regarding the possible differences in base temperature, optimum temperature, and root:shoot partitioning for different ecotypes. These will be key to simulating greenup in the spring and growth and development in the hottest part of the growing season.

Production of bioenergy requires the protection of soil and water associated with emerging bioenergy landscapes (Graham et al. 1996). The widespread degradation of the soil resource base and water quality due to past and current agricultural practices is well documented (USEPA 2009). The limited set of sustainability criteria attached to the 2007 Renewable Fuel Standard, which include stipulations about what types of land feedstocks are grown on, and the GHG intensity of biofuel production, were a promising start but may need to be expanded to include additional sustainability dimensions such as soil and water quality.

These mechanistic models will be useful for comparing switchgrass production systems to more conventional agricultural crops. Meki et al. (2011) applied a version of the EPIC model, APEX, to assess the sustainability of corn stover removal from the Upper Mississippi River Basin, based on a set of ‘acceptable planning criteria’ used in the CEAP analysis (USDA-NRCS

2010) , to judge whether or not a farm field needed additional conservation treatment. The ‘acceptable criteria’ included; (a) N in surface runoff < 16.8 kg ha1 y-1 (15.0 lb ac-1 yr-1), (b) N in sub-surface runoff < 28.0 kg ha1 y-1 (25.0 lb ac-1 yr-1), (c) total P losses < 4.5 kg ha-1 y-1 (4.0 lb ac-1 yr-1), (d) Sediment loss < 4.5 Mg ha-1 y-1 (2.0 ton ac-1 yr-1), and (e) SOC with a more ‘stringent’ restriction that the annual rate of change be positive. Given the critical functions of SOC in maintaining soil quality and productivity, biomass removal can only be justified if it does not deplete the SOC pool. These ‘acceptable’ levels represent field-level losses that are feasible to attain using traditional conservation treatment (nutrient management and soil erosion control), are agronomically feasible, and can equally be adapted to switchgrass production systems. Scientific literature on field research and edge-of-field monitoring in the U. S. Midwest, coupled with model simulations of conservation practices effects, provided guidance for identifying these thresholds (USDA-NRCS 2010).

Conclusions

Switchgrass is the "model" bioenergy crop for a potential bioenergy industry throughout the southeastern and south central USA (Wright and Turhollow

2010) . Given the infancy and urgency of the fast-evolving bioenergy industry, crop simulation models can complement and extend the applicability of information collected in field research trials, and when combined with the appropriate climate, soil, crop, and management databases, can be applied effectively to assess the sustainability and long-term impacts of converting land to bioenergy crops in a timely and cost-effective manner.

Acknowledgements

We thank Philip Fay, Daren Harmel, and Lara Reichman for comments on the manuscript. USDA is an equal opportunity provider and employer.