Rain on terrain.
Water through time.

Watch a storm scenario

Choose a completed simulation, then play or move through time. Play from start restarts the animation on every click; Pause holds the current frame. Rain falls uniformly for the first 30 minutes; the next 30 minutes show drainage with no further rain. Selecting a storm loads a separate solver run—not a recoloured copy of another storm.

Depth below 0.02 m is hidden for readability. No colour means below that threshold or outside the simulated interior—not proof of no water. North is up. Colours use the same depth thresholds for every storm and frame.

Click or tap a model cell to inspect its depth through time. The selected ground location stays fixed when you compare scenarios.

Both maps share time, background, depth colours and zoom. Scrolling is synchronised. On narrow screens the maps stack vertically. Click either map to compare the same ground location. The detailed chart below follows Scenario A.

What this experiment assumes

Rainfall: hypothetical uniform 25 mm or 50 mm events over 30 minutes (mean intensity 50 or 100 mm/h). Neither has a local return period or an observed-event attribution. The no-infiltration case sends all rain into surface water; it is a sensitivity case, not a certified worst case. The optional 5 mm/h loss is an assumed constant infiltration capacity, limited by available water, not a measured soil property.

Terrain and boundaries: the supplied terrain is averaged to 10 m or 5 m. Cells with less than 95% source coverage are excluded. The valid perimeter is fixed at a near-zero water depth; only interior cells receive rainfall. Missing-data cells are closed. These artificial boundary conditions omit external catchments and constrain drainage near the survey edge. Roads, small channels and barriers may be smoothed; buried structures are not resolved.

Surface resistance: uniform Manning n = 0.05, uncalibrated. Initial numerical film = 0.00001 m; the maintained perimeter film can introduce tiny numerical inflows recorded in the budget. No culvert, pipe, bridge opening or blockage has been invented. Deep simulated concentrations require terrain and drainage verification before any site interpretation.

Potential releases: scope for a next study

The construction survey shows the terrain at that stage. It does not establish a future pond level, tailings inventory or credible failure location. No release or breach has been simulated in this preview.

Water-only release: requires a defined source location, water level and storage, and opening geometry or discharge through time, plus downstream terrain and drainage. The rainfall scenarios above do not represent a release.

Tailings slurry release: additionally requires mobilisable tailings volume, solids concentration, density and flow properties. A suitable non-Newtonian model and credible release mechanism must be selected with the site engineers; water depths cannot be relabelled as slurry results.

Local terrain can support a review of possible downhill routes once a source is defined. Survey edges are data limits, not spill limits. No spill extent, arrival time or impact assessment is claimed. GISTM requirement 2.3 provides the reference for a documented breach analysis and independent review.

Methodology & validation

GISTM reference framework: requirement 5.3 addresses facility water balance and water management; requirement 5.5 addresses construction stages. This surface-runoff prototype is not a facility water balance, compliance assessment, dam-safety assessment, seepage model or tailings-flow model. Drainage drawings and Engineer of Record review are required for engineering use. GISTM reference

This module uses Landlab 2.10.1 OverlandFlow, an established local-inertial approximation to shallow-water flow following de Almeida et al. It calculates changing depth and exchange between grid cells with adaptive time steps and its steep-slope stability option. It is a reduced-physics research solver, not HEC-RAS, and steep-slope flow limitations still apply. AI assisted the model setup, code and presentation; depths are numerical solver outputs, not AI-generated images or predictions.

Checks include a closed flat-ground rainfall test, a no-rain site control, water-budget accounting, nonnegative finite depths, a smaller time-step run and 5 m versus 10 m resolution comparison. Numerical checks do not establish field accuracy. This setup has not been calibrated, independently reviewed or validated against observed flooding.

Landlab method documentation · Published OverlandFlow paper

The maps replay precomputed runs saved every two minutes. New rainfall depths, durations or engineering assumptions require another solver run. No hydraulic calculation is performed in the browser.