High-resolution broad-scale mapping of soil parent material using object-based image analysis (OBIA) of LiDAR elevation data

Precise mapping of subsurface properties, especially soil parent material (SPM), is critical to understanding most geomorphic and hydrologic processes. Because spatial variability of the subsurface properties is highly correlated to the morphology of the earth's surface, DEMs and aerial/satellite imagery have been used successfully for lithological mapping. However, most of the studies using this approach have used either coarse-resolution DEMs or covered limited areas. LiDAR-derived topographic attributes may provide an opportunity for finer-resolution mapping of subsurface properties at much larger spatial scales. Here, we present an approach that uses object-based image analysis (OBIA) - the process of clustering pixels into homogeneous objects - paired with a classification tree (CT) algorithm to map SPM at 1 m resolution and at a regional spatial scale using LiDAR-derived topographic attributes. Combining OBIA of topographic attributes with the CT method proved to be an efficient approach for mapping of SPM at fine resolution and broad scales. Using field-acquired SPM samples, we found that this approach was able to explain 73.5% of the variance of three SPM classes (i.e., bedrock, glacial till, sands) using only four common topographic metrics (i.e., slope, topographic wetness index, terrain ruggedness index, multi-resolution topographic positioning index). When comparing our approach to other classification methods and to other studies that have mapped SPM, we obtain the highest overall accuracy with a value of 85%. This approach allowed us to produce a fine-resolution 185 km(2) SPM map in a heterogeneous post-glaciation Precambrian Shield setting.