Rainfall. Terrain. Perspective.

Flow trails use velocity derived from Landlab link discharge divided by the solver hydraulic depth, averaged from opposing links at each cell and rotated into east/north directions. The two-minute snapshots are reconstructed for visualization. Color shows derived speed with the same scale on both maps; animated trails are illustrative, not tracked water parcels or travel-time predictions. Only interior cells with four eligible links and at least 2 mm hydraulic and adjacent water depth are retained. Masked areas do not prove zero flow. Rising/receding colors show the difference between successive saved depths, including rainfall and drainage effects; they do not identify lateral displacement.

Every scenario is calculated separately with Landlab 2.10.1 OverlandFlow, using a local-inertial approximation to shallow-water flow. Mapbox displays the actual survey orthomosaic and computed depth frames; it does not calculate water movement. Depth cells are reprojected with nearest-neighbor sampling for display. This is a plan view, without terrain exaggeration. Colors represent fixed depth intervals, not hazard categories; depths below 2 cm are transparent.

Uniform rainfall of 25, 50 or 100 mm lasts 30, 60 or 120 minutes, followed by 60 minutes without rain. Frames are saved every two minutes. The 100 mm options are hypothetical stress tests, not site-specific design storms or assigned return periods. All options use a 10 m grid, assumed Manning n = 0.05 and no infiltration. Real storms vary through time and space.

An artificial drained perimeter is used. Upstream inflows, culverts, buried drainage, pumps and tailings releases are not represented. Elevations are in meters. Survey quality documentation has been reviewed. The 10 m computational grid can omit smaller terrain and drainage features. Survey quality does not establish the accuracy of these hypothetical runoff scenarios. Apparent ponding can reflect missing drainage or terrain errors. Water remaining at the end is not assumed to have drained.

Both maps use the same elapsed time and fixed depth legend. If one run ends sooner, its water layer and metrics are hidden beyond its final frame; later conditions are not inferred. Pan and zoom are synchronized.

Numerical consistency does not establish field accuracy. The results are not calibrated or independently reviewed and do not replace engineering analysis. The two sites are not a controlled comparison of risk. AI assisted the software and presentation; it did not generate the water-depth values.

Original detailed viewers, including point inspection and grid/infiltration comparisons: Mining · Infrastructure. These retain the earlier scenario set.