Native grasslands have been reduced to a fraction of their original extent, with estimated total loss prior to the 1990s of 70% for prairie grassland (Federal Provincial and Territorial Governments of Canada 2010). Conversion of native grassland to cropland and tame hayfields or pasture has been one of the leading drivers of native grassland loss in North America. Degradation of native grasslands also continues in some areas due to changes in natural disturbance regimes such as fire suppression and intensive prolonged cattle grazing, threats from invasive non-native species, fragmentation, intensification of agriculture, and economic development associated with population growth(Federal Provincial and Territorial Governments of Canada 2010). With increasing pressure and stressors on remnant native grassland, land managers and land use decision makers need to be able to provide clear guidance on locations of priority landscapes (i.e., remnant native grassland) and allow opportunities for industry to conduct coarse-level, pre-development planning that can identify native grasslands as sensitive habitat. Furthermore, grassland-associated species, many of which are species of conservation concern, prefer to settle in native grasslands compared to tame grasslands (Davis et al. 2013). Identification of important habitat for these species requires reliable information on the types of grassland over the landscape (Davis et al. 2013). In order to accomplish this objective, it is imperative that remnant native grassland be identified and mapped. However, the spatial location of these remaining patches have yet to be accurately identified, as it has proven difficult to distinguish between parcels of land that are native grass-dominated and those dominated by tame grasses. An effective and efficient process to differentiate native prairie from surrounding land covers will be fundamental for grassland conservation in the Northern Great Plains region.
Other studies have attempted to distinguish native and tame grassland using a variety of methods. The most relevant study (McInnes et al. 2015) compared the classification accuracy of air-photo interpretation to a time-series analysis of MODIS Normalized Difference Vegetation Index (NDVI). McInnes et al. (2015) examined the NDVI time-series because of potential differences in rates of green-up between the two landcover types. McInnes et al. (2015) concluded that the time-series approach performed better than a single-date approach and the aerial-photo interpretation approach.
Our objective was to test three methods for distinguishing parcels of land dominated by native grasses from those dominated by tame grasses. The first method involved air-photo interpretation using high resolution color infrared digital stereo models. The second method used a time-series of vegetation indices (NDVI, EVI, SAVI, and MSAVI) derived from Landsat. The third method attempted classification of different grassland types using a high-resolution LiDAR product.
