<p>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; Mapping bedrock outcrops is useful across disciplines, but is challenging in environments where ground surface visibility is obscured. The presence of soil or bedrock affects sediment production and transport, local ecology, and runoff generation. The distribution of bedrock outcrops in an area reflects the interplay between regolith production and sediment removal. Outcrop classification methods from Terrestrial-lidar produce millimeter or centimeter resolution DEMs that are highly successful because lidar penetrates through vegetation to the ground surface. However, data availability at such high resolution is limited, and the associated computational complexity required for identifying outcrop, or other surface features, is often impractical for landscape-scale analysis. Aerial lidar datasets at ~1-m resolution (e.g., moderate resolution) are more widely available and less computationally expensive than higher resolution datasets. With increasing accessibility of moderate resolution surface data, there is a need to develop outcrop classification methods and understand the efficacy of these methods across diverse environments. Our objectives are to present a simplified technique that builds on existing methods, and to examine the success of current outcrop identification methods in a variety of landscapes.</p><p>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; At moderate resolution, the two most cited metrics to differentiate bedrock from soil-mantled surfaces are based on gradient (e.g., DiBiase et al., 2012) or on surface roughness (e.g., Milodowski et al., 2015). We developed a method that simplifies and combines both metrics, and that improves overall accuracy. We applied all three methods to six landscapes in the USA. For each site, we delineated ground truth from high-resolution orthoimagery for 7-10 test patches with visible ground surface, that evenly spanned 0-100% exposed outcrop. Overall accuracy, true positive rate, and false positive rate for each patch were calculated by comparing the ground truth grids to each lidar-derived outcrop grids on a cell-by-cell basis. Metric success was evaluated for each landscape by assessing the mean and distribution of performance measures across patches. Our combined metric had the highest overall accuracy in an arid, horst and graben landscape (Canyonlands National Park, Utah). It also performed well in a vegetated, high sediment load, active volcano (Mount Rainier, Washington), a canyon carved by channel incision (Boulder Canyon, Colorado), and a chaparral mixed bedrock canyon environment (Mission Trails, San Diego, California). All three methods systematically failed for portions of the landscape in glacially carved canyons (Southern Wind River Range, Wyoming) and on terraced sea cliffs (Santa Cruz County, California). These environments have significant outcrop that is both smooth and low gradient, and therefore cannot be identified using a slope or roughness-based algorithm.</p><p>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; Our work highlights the importance of tailoring DEM-based bedrock mapping algorithms to its geomorphic context, and of the need for ground truth. Such data provides the basis for developing more robust methods for error evaluation. In addition, new methods are needed to identify bedrock outcrop from surface DEMs in smooth and low gradient, yet rocky landscapes.</p>
Comparative Analysis of Parallel Brain Activity Mapping Algorithms for High Resolution Brain Models
