RESPECTING UNCERTAINTY IN GEOMORPHIC CHANGE DETECTION USING REPEAT AERIAL LIDAR OVER A MASSIVE AREA: BLUE EARTH COUNTY, MN

High-resolution topography data (lidar) are being collected over increasingly larger geographic areas. These data contain an immense amount of information regarding the topography of bare-earth and vegetated surfaces. Repeat lidar data enables extraction of an unprecedented level of information about landscape form and function and provides an opportunity to quantify geomorphic change. However, significant technological and scientific challenges remain in the analysis of repeat lidar data over massive areas (>10³ km²), not the least of which is a robust quantification of uncertainty.

Excessive sedimentation has been documented in the Minnesota River. The Blue Earth River and its tributaries have been identified as one of the main sources of sediment to the Minnesota River. Much of the Blue Earth watershed is located in Blue Earth County (1,982 km²) where airborne lidar data were collected in both 2005 and 2012. One of the largest floods on record (100-year recurrence interval) occurred in September 2010. A sediment budget for the watershed is being developed to inform strategies to reduce current sediment loads and better predict how the basin may respond to changing climate and management practices.

Here we evaluate the geomorphic changes that occurred between 2005 and 2012 to identify and quantify hotspots of erosion and deposition. To make meaningful interpretations of the differences between the 2005 and 2012 lidar digital elevation models (DEMs), total uncertainty must be accounted for between and within DEMs. We identified, quantified, and corrected for a systematic vertical bias by comparing elevation differences between control and additional digitized points. To evaluate horizontal bias, we digitized points at the corners of structures and compared magnitude and direction of the differences. There was no detectable horizontal bias. A new uncertainty model was developed and applied to each DEM that accounts for uncertainty due to vegetation interference and topographic complexity relative to sampling. Uncertainty models were propagated into the change detection estimates to establish a probabilistically thresholded estimate of elevation change. Results will be presented of the spatial patterns of net erosion and deposition and their contribution to the overall sediment budget being developed.