(471a) Spatially Explicit Optimization of Cellulosic Biofuel Supply Chains Considering Landscape Design and Biomass Yield Uncertainty

Planting dedicated bioenergy crops on lands less suitable for food production has the potential to lower the competition between food and fuel. Additionally, fuels made from second generation cellulosic feedstocks provide an environmentally attractive sustainable alternative to fossil fuels and first-generation grain-based biofuels. Because cellulosic feedstocks have yet to be planted in large quantities and biorefineries have yet to be constructed, there is an opportunity to use mathematical programming techniques to better understand the major environmental and economic drivers, and ultimately optimize the biofuel supply chain network design (SCND) from field to product.

When considering supply chains covering large study areas, there is considerable variation in local weather and soil quality which leads to uncertainty in biomass yields that, if not taken into account, can lead to sub-optimal facility siting and biomass establishment decisions. The varying soil quality provides an opportunity to optimize the landscape design (i.e. deciding where in the landscape to plant which bioenergy crop and how to manage that land) along with supply chain network design decisions simultaneously while exploring their tradeoffs. The optimization of landscape design and SCND together has the potential to protect against uncertainty and exploit the spatially explicit nature of land for biomass to achieve better solutions both economically and environmentally.

While the area of biofuel supply chain optimization is relatively mature, a flexible high resolution spatially explicit model that considers feedstock yield uncertainty, environmental outputs, and landscape design simultaneously has yet to be proposed. We present an integrated two-stage stochastic mixed integer linear model that, given (i) the spatially explicit available land, (ii) biomass yield potential at each site, (iii) potential biorefinery and preprocessing depot locations, (iv) potential preprocessing and conversion technologies, (v) cost parameters and environmental parameters finds the optimal (i) biorefinery and depot locations, (ii) technology, (iii) capacity, (iv) transportation, production, and inventory planning, (v) crop establishment locations, (vi) and land management; the combination of which minimizes the total annualized cost of the supply chain.

We integrate the yield and environmental outputs of SALUS biogeochemical crop model simulations over multiple years of weather data with the spatially explicit supply chain network design model and present a case study in southern Michigan, USA. Our results show that fertilization closer to the biorefinery leads to a decrease in costs and greenhouse gas emissions related to transportation. Additionally, the two-stage stochastic formulation leads to solutions which perform more favorably in expectation with respect to the uncertain biomass yields. By adjusting the cost of carbon ($/MgCO2e), decision makers can determine how much to value reductions in GHG emissions with respect to SC costs.

Model complexity is sensitive to the number of uncertainty scenarios, the size of the study area, and the number of supply chain nodes considered. As model complexity increases, approximation methods are introduced to maintain computational tractability and solve larger problems. Because large, realistic, problems are difficult to solve with an extensive form of the two-stage stochastic model, we present and compare solution strategies and decomposition methods to solve detailed models more efficiently.