Modeling perennial bioenergy crops in the E3SM land model

Abstract

Abstract. Perennial bioenergy crops are increasingly important for the production of ethanol and other renewable fuels, and as part of an agricultural system that alters the climate through its impact on biogeophysical and biogeochemical properties of the terrestrial ecosystem. The Energy Exascale Earth System Model (E3SM) Land Model (ELM) does not represent perennial bioenergy crops, however. In this study, we expand ELM’s crop model to include perennial bioenergy crops whose production increases in modeled socioeconomic pathways owing to their potential for mitigating climate change. We focus on high-productivity miscanthus and switchgrass, estimating various parameters associated with their different growth stages and performing a global sensitivity analysis to identify and optimize these parameters. The sensitivity analysis identifies eight parameters associated with phenology, carbon/nitrogen allocation, and photosynthetic capacity as the most sensitive parameters for carbon and energy fluxes. We calibrated the model against observations collected at the University of Illinois Energy Farm for carbon and energy fluxes, and found that the model closely captures the observed seasonality and the magnitude of carbon fluxes. The model accurately represents the seasonality of energy fluxes, but their magnitude is not well captured. This work provides a foundation for future analyses of the interactions between perennial bioenergy crops and carbon, water, and energy dynamics in the larger earth system and can also be used for studying the impact of future biofuel expansion on climate and terrestrial systems.

Artificial Intelligence for Earth System Predictability (AI4ESP) (2021 Workshop Report)

“…Using surrogates therefore decreases the time required to generate the much larger number of ensembles needed for GSA and calibration. There are several examples where surrogate models have led to improved model predictions at individual sites or collections of sites (Sinha et al 2021;.…”

“…Surrogate modeling can help speed up sensitivity analysis and allow for more efficient model parameter calibration. Surrogate modeling therefore has gained traction in the land surface community in recent years (Sinha et al 2021;. Further development of ML algorithms and more efficient surrogate modeling will be critical for future land model parameterization and uncertainty quantification.…”