Assessment of UAS Photogrammetry and Planet Imagery for Monitoring Water Levels around Railway Tracks

Figure 1.
Relative location of test sites along the railway track in eastern Ontario for assessing change in surface water levels. Railway layer source: Ontario Ministry of Natural Resources and Forestry—Provincial Mapping Unit.

Figure 2.
Example UAS photographs showing the variability of water near the railway track right-of-way. (A) track crossing a wetland, June 2021; (B) track crossing a creek, October 2021.

Figure 3.
(A) Photograph of the M300 with P1 camera, circular scanning radar and D-RTK2 base station. (B) Photograph of the M600P with the X5 camera and integrated D-RTK unit.

Figure 4.
Flowchart of the SfM-MVS workflow for generating the final products from the UAS RGB photographs. Final products are shown in blue. Processes are coloured based on the software used where green = Pix4D Mapper, orange = CloudCompare and red = Quick Terrain Modeler.

Figure 5.
PlanetScope Dove classic (PS) satellite imagery for sites 1A, 1B, and 2 used for water surface classification at the landscape level. Images were clipped to show the sites.

Figure 6.
Examples of aquatic vegetation that are important to consider within the water class when mapping water surface in the study area. (A) Pondweed (Potamogeton nodosus), (B) water lily (Nymphea odorata), (C) duckweed (Lemna minor), (D) water lettuce (Pistia stratiotes), (E) UAS photograph illustrating the aquatic plants in the context of open water near the track.

Figure 7.
(A) Example 4-band Planet Dove spectrum from July illustrating the similarity of the wetland class to the tree class; (B) Example multi-temporal spectra from the 20-layer stacked image illustrating wetland, tree, and an agricultural field; (C) 20-band multi-temporal image stack; band combination shown is R: September band 3, G: July band 2, B: June band 2.

Figure 8.
Example of a final 3D point cloud (Site 3A—August) showing (A) RGB true color representation and (B) height above ellipsoid. The distribution along the vertical axis of the color bar shows the spread of the height values.

Figure 9.
Example orthomosaic (resampled to 3 cm pixel size) for Site 1A from 2 June 2021. Inset shows open water surface next to the culvert.

Figure 10.
Examples of the point density of the final 3D point clouds for (A) Site 3A—August and (B) Site 2—August.

Figure 11.
Comparison of difference in position of points (distributions) representing the railway track and ties, the positions of which are considered to have negligible change throughout the sampling period June–October. The values in cm represent the absolute different between the June–September point clouds in comparison to October for each site. The two July datasets (Sites 1A and 1B) were acquired with the M600P and the X5 camera, all other datasets were acquired with the M300 and the P1 camera. The distributions along the vertical axes of the color bars illustrate the spread of the values.

Figure 12.
Surface water (including aquatic vegetation) in percent cover across sites and periods based on a GEOBIA classification of the UAS orthomosaics (3 cm pixel size).

Figure 13.
Vertical distance between the rails crossing a culvert or bridge and the water level for each site across sampling dates. The error bars represent the uncertainty calculated as the variability in the rail and water surface heights.

Figure 14.
(A) Changes in exposed surface water area and the delineation of the wetlands from multi-temporal satellite image stack; (B) Comparison of wetland baseline classification (wetland symbology) to the Ontario Ministry of Natural Resources and Forestry wetland database (red outline). The scale is the same for panels A and B.

Figure 15.
Comparison of PS imagery from (A) July and (B) October of the railway track crossing the wetland in Site 2. In the July imagery the aquatic vegetation can be seen as bright green, obscuring the water surface below.