A regulatory requirement of a number of infrastructure projects, such as road and rail developments, is that there is no negative impact on natural drainage features due to altered surface water flows. Airborne lidar has traditionally been used to capture 3D digital elevation models (DEMs) to within a specified accuracy. The assessment of landform change is then a comparison of DEMs captured at two different points in time.
Astron was commissioned to provide a statistically valid threshold in classifying change detection maps into areas of real and spurious (or no) change; previously, thresholds were arbitrarily chosen. Utilising the fact that given multiple flight paths flown during a lidar mission provides information on DEM variability, spatially varying (or heterogeneous) confidence limits were built that reflected variations in lidar point density.
A key confounding factor was the presence of vegetation, particularly on steeper slopes. An interim solution of spatial aggregation over 16 m2 cells of the statistical filter was introduced to dramatically reduce noise in the threshold change detection map, though at some small loss in threshold sensitivity. A long term solution of supporting the lidar ground point classification with a spectrally based screening of the lidar point cloud was also trialled and recommended.
