Significant decadal channel change 58-67 years post-dam accounting for uncertainty in topographic change detection between contour maps and point cloud models

摘要

Construction of digital elevation models (DEMs) and the subtraction of DEMs between different points in time as a method to determine temporal patterns of scour and fill is a highly valuable procedure emerging in geomorphology. These DEMs of Differences (DoDs) must be assessed for error in order to distinguish actual topographic change from uncertainty and surface error. Current methods include: (1) uniformly excluding all values that fall below a minimum threshold; (2) using a spatially variable approach such as the construction of minimum Level of Detection (LoD) grids; or (3) the creation of a fuzzy inference system. Although spatially variable methods for determining error have been more accurate in excluding noise without discarding large amounts of meaningful data, a challenge remains in performing DoDs against preexisting contour-based maps for which no original point data are available. The goals of this study were to (1) develop a method that overcomes the unknown point density of contour (and other historical) data sets and allows for some assessment of DoD uncertainty on the basis of information on topographic variability, (2) perform comprehensive uncertainty analysis testing to understand the opportunities and constraints associated with this new method, and (3) report and interpret the overall pattern and volume of decadal topographic change for a regulated river 67 years post-dam in light of alternate conjectured mechanisms of post-dam longitudinal profile adjustment. The key feature of the new approach is the introduction of a high-density artificial point grid that samples the topographic variability evident in the available historical data set. The testbed used to develop and assess this new DoD method was the ~37.5-km lower Yuba River, California. Historical data consisted of 0.6-m contours from a 1999 survey, while a more detailed point cloud was available for the most recent survey in 2006-2008. To evaluate uncertainty in the method, this study applied seven different uncertainty metrics of varying strictness (t= 1, t = 1.96, uniform 0.3-m exclusion, t = 1 plus uniform 0.3 m exclusion, t = 1 with LoD minimum at 0.3, t= 1.96 plus uniform 0.3-m exclusion, and t= 1.96 with LoD minimum at 0.3) and two different DoD adjustment methods (exclusion and subtraction). A stringent approach involving joint use of the contour half-interval and the spatially distributed LoD grid at a 95% confidence limit excluded 44.3% of the study area from spatial assessment of channel change and volumetric change computation. This preferred approach yielded an estimated 2.518 million m~3 of total scour and 2.455 million m~3 of total fill in 7-9 years. After considering different mapping epochs, the net annual average export was 17,000 m~3 (～32,500 tons) to the Feather River and a river-valley sediment yield of 2205 tons/km~2/year. This amount is 36% of the post-dam annual yield of gravel and cobble to the upstream reservoir that blocks sediment conveyance into the study domain, suggesting that the lowland system is still highly dynamic 67 years after the dam was built. The scientific significance of this is that the response of rivers to dams can be far more long-lasting and complex, depending on the suite of cumulative societal impacts to rivers. The ability to account for spatially explicit DoD uncertainty (i.e., data retention when uncertainty is low and data removal when uncertainty is high) when comparing historic contour-based and modern point cloud-based DEMs will allow for more detailed and reliable DoDs and sediment budgets in these cases.