Uncertainty in Estimates of Bioenergy‐Induced Land Use Change: The Impact of Inconsistent Land Cover Data Sets and Land Class Definitions

While many data sets are increasingly available to describe land cover characteristics, these data require careful analysis and supplemental research before conclusions can be drawn about the scope, magnitude, and drivers of change. Land cover data, typically derived from remote sensing, are frequently analyzed to estimate land use change (LUC) that may be attributable to policies such as the U.S. Renewable Fuel Standard, which encourages bioenergy. However, land cover classifications such as grassland do not clearly differentiate among multiple land uses (pasture, fodder, crop, yard, wildlife corridor, decorative cover, and erosion control strip), many of which may occur on a single parcel simultaneously. A land cover class in one data set rarely corresponds exactly to the same land cover class in another data set. Further, persistent improvements in remote-sensing systems result in changes in spatial and temporal resolutions that lead to changes in classification methodology, which limit the ability to use those data sets to accurately measure changes occurring on the ground. Most LUC estimates are derived by comparing land cover across a few points in time and using aggregate land cover classes such as “forest,” “cropland” or “urban.” Analyses that cover short time spans or rely on just a few points in time are likely to generate spurious results if dynamic interacting classes such as U.S. cropland and grassland are considered. We review land cover dynamics in the Western Corn Belt (WCB) region, which comprises the U.S. states of Iowa, Minnesota, Nebraska, North Dakota, and South Dakota, to illustrate how the selection and manipulation of data can result in estimates of change associated with the cropland-grassland transition that vary by over 100%.