DNR-funded mapping project finds automated landform mapping detects sites but needs better boundaries

Dan Miller, co-principal investigator for the object-based landform (OBL) mapping project, told the Forest Practices Board that automated mapping using high-resolution lidar reliably identifies the presence of landforms such as bedrock hollows and inner gorges but often disagrees with experts on precise polygon boundaries.

Miller summarized a phased project that compared computer-derived polygons from two automated approaches (a virtual-watershed method and OBIA/GEOBIA) to polygons hand-drawn by four to five experienced practitioners across several study sites. The team used quantitative overlays and a voting exercise in which experts evaluated whether computer polygons represented real features and, if so, whether polygon boundaries were correct. The findings showed substantial variation among human mappers and between human and computer delineations.

"There's a lot of variation, even between people who have decades of experience doing this kind of work," Miller said, noting that while the computers consistently detect the existence of landforms, delineation depends heavily on input parameters. Experts flagged issues including insufficient topographic convergence in some polygons and incorrect stream adjacency (some inner gorges were aligned to road prisms instead of stream channels). The team therefore recommended tuning parameters (for example setting minimum steep-area proportions) and proceeding to a second phase focused on empirical evaluation using lidar-change detection to build a landslide inventory and test susceptibility and run-out metrics.

Board members asked whether the project would examine climate drivers (rain-on-snow and precipitation trends) and how results would be shared with counties and local planners. Miller said climate relationships would be examined in later phases and that the project is focused on susceptibility and hazard characterization to better guide field practitioners rather than producing absolute ‘‘truth’’ maps. He said teams plan to complete the susceptibility/run-out subproject and deliver a report to the board as the next milestone.

The board acknowledged the research’s potential to give practitioners consistent, scalable mapping tools while also noting that ground-based judgment will remain important for many decisions.