Performing Department
Department of Chemical & Biological Engineering
Non Technical Summary
Concerns about land-use change (LUC) impacts currently drive the biofuel sustainability debate and hinder biofuel expansion. Many biofuel stakeholders are concerned that as biofuel feedstock demand grows, agricultural land will expand into natural lands. This potential agricultural expansion would deplete forests, grasslands, and wetlands that have great value for their storage of carbon and other ecosystem services. In the past few years, researchers have begun to look at LUC since the growth of the biofuel industry began in earnest (about 2007) using remote sensing tools. Remote sensing-based LUC analyses incur classification errors which tend to be higher in the textured landscapes where expansion of renewable feedstock agriculture could take place. Recent practices such as stover removal for animal feed, stover grazing, and cover cropping have been difficult to include in these LUC analyses, which may therefore overestimate LUC. In short, current analysis methods do not accurately characterize LUC. Undeserved faith in these methods could lead to misguided bioenergy land use policies and possibly delays in the realization of the bioeconomy. The biofuels community needs an improved, stakeholder-endorsed method to characterize LUC that takes explicit advantage of very-high resolution aerial imagery as a tool to improve what can be understood about LUC from farmer surveys and remote sensing products. Only when a method incorporates all of these tools and offers a robust and defensible estimate of LUC can the debate over LUC move forward, opening the gates to applying renewable biomass to near-term opportunities. Unlike today's LUC analyses that are either error-prone, time-intensive and/or expensive, and unresponsive to key stakeholders' concerns, the new approach will be low-cost, have high accuracy, and will address diverse stakeholder concerns. With this new method, analysts will be able to quantify with a transparent error estimate how many hectares of grassland, forested land, and wetlands have been converted to agriculture on gross and on net since a given year either nationally or regionally.The method we will adopt in this project is to evaluate and develop methods for best practices in working with three data sources that can inform LUC estimates. The first data source is farmer surveys conducted by the U.S. Department of Agriculture. We will conduct a gap analysis of these surveys. Some recognized gaps in farmer surveys include limitations on reporting to harvested acres at the exclusion of seeded and unseeded grasslands. Furthermore, we will propose defensible routes to incorporating these surveys into LUC estimates. The second data source is very-high resolution satellite imagery from the National Agricultural Imagery Program (NAIP). We will develop new machine-learning-based approaches to characterizing LUC among land in agriculture, forests, wetlands, and grasslands and use data fusion techniques to incorporate information from remote sensing data products, which constitute the third data set, including the Cropland Data Layer and the National Land Cover Data Set. Harnessing the information in these two data sources and the advantages of machine learning will produce high quality estimates of LUC that will be further informed and verified through incorporating insights from farmer surveys. Based on our work with these three data sources and development of analysis techniques, we will develop a new method that uses all three to estimate LUC transparently and with error estimates that could be used in biofuel sustainability analyses. We will develop a method to incorporate results from LUC analyses into biofuel life cycle analysis, which generates life-cycle GHG emissions estimates of biofuels and is used to establish their eligibility for programs including the Renewable Fuel Standard and California's Low Carbon Fuel Standard. Given the importance of stakeholder buy-in for this new method, we will hold stakeholder workshops at project initiation, midpoint, and conclusion to gather and incorporate feedback and disseminate results.The ultimate goal of this project is to develop a stakeholder-approved LUC quantification method that incorporates farmer survey data, remote sensing data, and satellite imagery to evaluate land-use change and a method to incorporate results from this analysis into biofuel sustainability analysis. Through this project, agricultural land will be used more sustainably. Moreover, concerns about LUC are a key stumbling block in the path towards demonstrating the sustainability of biofuels to the public, non-governmental organizations, and policy makers. With this stumbling block removed through demonstrating low LUC impacts and/or how to mitigate potential or actual LUC effects, the road to increased biofuel production may be smoother and rural communities may benefit from an increased market for biomass.
Animal Health Component
0%
Research Effort Categories
Basic
30%
Applied
70%
Developmental
(N/A)
Goals / Objectives
The major goal of this project is to providestrategic guidance through an improved approach to understanding historic and potential future land-use change (LUC) to improve the sustainability and environmental quality of the application of renewable biomass technologies. This project will develop an approach that will be multi-faceted, combining farmer-based reporting, remote sensing data, and aerial images while accounting for error will accurately, efficiently, and inexpensively quantify and characterize LUC at high spatial resolution.Objectives:Develop best practices for use of farmer survey data in LUC analyses based on a gap analysis of existing farmer survey data products (e.g., Census of Agriculture, June Area Survey, others),Use machine learning techniques including convolution neural networks to interpret LUC detectable in satellite imageryDevelop techniques for the fusion of the Cropland Data Layer/National Land Cover Data Set data and farmer-based surveys with National Agricultural Imagery Program imagery to yield a high resolution classification map.Hold three stakeholder workshops to introduce the project, gather feedback in the middle of the project, and disseminate information regarding results at the end of the project.Through venues including the project website, peer-reviewed publications, and conferences, disseminate research results and the final proposed approach to quantifying LUC.
Project Methods
The project will use the following methods:Gap analysis of farmer surveys in consultation with the USDADesign of convolution neural networks to provide an estimate of LUC from NAIP imageryDevelopment of an alternative deep network architecture to estimate directly the LUC areas from two consecutive-in-time input NAIP images of the same locationFusion of the CDL/NLCD data and farmer-based surveys with NAIP imagery to yield a high resolution classification mapWorkshop-based stakeholder feedback incorporation into the methodology for LUC estimationEfforts:Development of continuing education curriculum will be developed. The new approach and insights gained in developing the method for estimation of LUC will be taught in sustainability courses at all three universities and through the University of Illinois Extension Service to farmers as an element of continuing education.We will hold three workshops at the initiation, mid-point, and conclusion of the project to gain stakeholder feedback and dissemintate results.EvaluationCompletion of all activities, events, and other products listed in "products"The number of stakeholders who use all or part of the methodology in their assessment of LUC. Project team members will follow up with workshop participants annually for three years following the project.Submission for publication in a peer-reviewed journal article of two papers per yearPresentations at two conferences per year during the project period.Number of visits to the project website.