Towards best practices for assessing land use change with data and imagery

Recently, examinations of land use change (LUC) from agricultural expansion into native grasslands and forests have employed data such as the Cropland Data Layer (CDL) and the National Land Cover Data (NLCD). Some of these... [ view full abstract ]

Recently, examinations of land use change (LUC) from agricultural expansion into native grasslands and forests have employed data such as the Cropland Data Layer (CDL) and the National Land Cover Data (NLCD). Some of these studies have assessed large amounts of LUC, notably in environmentally sensitive regions such as the Prairie Pothole Region.  It is critical to understand losses in native ecosystems that can influence biodiversity and carbon stocks among other important aspects of native ecosystems.  The methodology that is used, however, to employ these datasets deserves examination if the results of such analyses are to be used to potentially inform stakeholders and policy makers.  

To examine how results of LUC assessments vary with methodology, we have undertaken an analysis in changes in land area in cropland, grassland, and forested land in twenty counties in the Prairie Pothole Region using the CDL and NLCD (at 30 m resolution) and compared our results to those we obtained comparing side-by-side images of land in these counties from the National Agricultural Imagery Program (at 1 to 2 m resolution).  We used two techniques to asses land use change with the Cropland Data Layer.  One approach used raw Cropland Data Layer data.  A second modified these data with an approach developed by Lark et al. to limit false identification of native land converted to agricultural land.   We estimated approximately 700 thousand hectares of land had been converted to cropland from forested lands and wetlands between 2006 and 2014 when we used unmodified CDL data.  With the modified CDL approach, this number dropped to 15 thousand hectares.  A NAIP-based approach resulted in an estimate of approximately 4 thousand hectares.  In this presentation we will discuss the underlying reasons behind these differences and suggest routes towards a discussion among the community of analysts undertaking these types of analyses regarding best practices.