Results – Wrights Road Storage Ponds dam-breach model (shakedown 15 Sep 2026; east and north runs extended to Diversion Road 16 Sep 2026)

Screening-level results. Method, assumptions and caveats: SPEC.md §3 and §7.3; sources: docs/source-notes.md. Time zero for all 2D results is the moment the external (downstream) breach opens; for cascade scenarios that is 2.2 h after the Pond 1 failure begins.

1. Breach outflow (all four scenarios)

Damwatch (design report s3): "approximately 2,500 m³/s" for the type of breach analysed.

2. 2D flood routing (east and west, 4 h, 20 m grid)

Road arrival times – east breach vs Damwatch 2016 (WIL Evacuation Plan Table 1)

The model reaches each road ~45 min earlier and generally deeper than the 2016 study – the expected direction for uniform roughness (n = 0.045), a 20 m grid that smooths road/race embankments, and a 1.1 h breach formation. These are the phase-1 calibration levers.

Roads reached – west breach

2b. East breach routed to Diversion Road (extended domain, 12 h, 20 m grid)

The east cascade breach re-run on the full 31.5 km × 16 km study domain (site to Diversion Road and the Waimakariri River; the east edge sits 1 km beyond Diversion Road so the open boundary is on the river side of the road). Same 20 m LiDAR grid, mesh densities, roughness and hydrograph as the shakedown; 272,000 triangles, 12 h simulated (12 h wall-clock on one loaded Apple-silicon core). Outputs: outputs/east/*_extended.*; animation: docs/east-extended.

Road arrival times – east breach vs Damwatch 2016 (WIL Evacuation Plan Table 1)

Reading: the full-length wave stays in the Damwatch corridor (between South Eyre Road and the Waimakariri terrace, along the WIL Main Race embankment) all the way to Diversion Road, and the inundated area (69 km²) matches the Damwatch flood zone (73 km²) closely. Arrival is ~1 h earlier at Browns and Two Chain Roads and ~1.5 h earlier at Diversion Road, the same "faster and deeper" bias as the near-field roads (uniform n = 0.045, 20 m grid smoothing embankments and drains, 1.1 h breach formation). The largest difference is at Diversion Road itself: 0.43 m here versus "<0.1 m" in 2016. At that point the model is at the far end of a 26 km routing on a 20 m grid with no drains, culverts or stopbank detail, and the water then leaves through the open east boundary; treat the Diversion Road depth as an upper-bound screening value and the arrival time as ±1 h. The near-field rows change by ≤ 0.06 h / ≤ 0.2 m compared with the shakedown domain, confirming the smaller domain did not bias the earlier results. All 12 roads west of Diversion Road that Damwatch lists are now reached; Thongcaster, Domain and Barrett Roads (south-west of the site) stay dry in the east scenario.

Maximum depth

Arrival time

Hazard class

Maximum flow speed

Depth × velocity

Time of peak depth

2c. North breach (Dixon Road side): 4 h shakedown and 12 h routing to Diversion Road (20 m grid)

The north cascade breach (Pond 1 → Pond 2 → NORTH embankment, 4.7 Mm³ released, peak 817 m³/s, invert 216.7 m) routed in 2D in two steps: the 18 × 11.5 km shakedown domain for 4 h (95,000 triangles, current repaired mesh) and the full 31.5 × 16 km domain to Diversion Road for 12 h (201,000 triangles). Same 20 m LiDAR grid, mesh densities, roughness (n = 0.045) and hydrograph basis as the east runs. Outputs: outputs/north/*_shakedown.* and outputs/north/*_extended.*; animations: docs/north (4 h, 60 m cells) and docs/north-extended (12 h, 80 m cells, 10-min frames). Wall-clock on one core: ~2 min (shakedown) and ~10 min (extended) – far less than the east runs because the lower peak keeps the wet area and velocities small.

Road arrival times – north breach vs Damwatch 2016 (WIL Evacuation Plan Table 1)

Times are hours after the Pond 2 breach opens (2.2 h after the Pond 1 failure begins). Shakedown and extended values agree within 0.04 h / 0.05 m where both reach a road; the extended values are listed.

Reading: the outflow crosses Dixon Road within 20 min, turns east with the fall of the plain and runs in a band 1–3 km wide north of the east-breach corridor (between the Main Race and South Eyre Road), crosses Downs Road near Pashbys Road at 4.5 h, splits into two arms either side of Hetherton Road, and reaches Diversion Road at 9.6 h before draining to the Waimakariri River. Thongcaster and Barrett Roads stay dry. The inundated area (49 km²) is three-quarters of the Damwatch north flood zone (66 km²) and the at-risk building count (30) close to the 35 households at risk, so the extent is broadly consistent.

The arrival times are the opposite of the east case: the model is ~0.6 h later than Damwatch at Carleton Road and the gap grows to ~4 h at Diversion Road, whereas the east run is ~0.7–1.5 h earlier than Damwatch everywhere. Damwatch's 2016 north-breach times are the fastest in their table (Carleton Road 0:50 vs 1:30 for the east breach) even though the north toe is 6 m higher than the east toe and the north embankment is furthest from the corridor. In this model the north breach is a Pond 2 overtopping cascade with a 217 m invert, which limits the head to 6.9 m and the peak to 817 m³/s (40 % of the east peak) on a gentle 1:250 plain. The 2016 study evidently assumed a much larger or faster north outflow – e.g. a breach that also releases Pond 1 directly (FSL 226.5 m, 3.7 m higher head), a lower breach invert scoured into the foundation, or a shorter formation time. Which pond the north embankment retains along its length and the 2016 north breach parameters are therefore the first things to reconcile with Appendix H (assumption register A1, A4, A15 and A16 in docs/clg-notes.md; sensitivities below); a --scour 1 --time-factor 0.5 sensitivity on the north hydrograph is the quick model-side check. Depths at the far roads (Browns, Two Chain, Diversion) match the 2016 values within 0.1 m.

Reconciling the north arrival times – breach sensitivities (4 h domain)

Six variants of the north hydrograph were routed on the shakedown domain to see which breach assumption would reproduce the 2016 table (outputs/north/*_shakedown_<tag>.*, outputs/north_p1/). Times are hours after the Pond 2 (or Pond 1) breach opens; depths are maxima at the road.

The full-depth case was then routed on the 12 h full domain (outputs/north/*_extended_full.*):

What this shows:

1. Breach size, not routing, is the difference. Scour, formation time and width within the usual Froehlich sensitivity range move Carleton Road by at most 0.5 h and leave Downs Road beyond 3.5 h. Only a breach with the east-size outflow (~2,100 m³/s) gets within 0.1–0.9 h of the 2016 north times, and it also reproduces the 2016 finding that the north breach is about as severe as the east at Carleton Road (2016: 0.71 m north vs 0.68 m east; model full-depth north 0.88 m vs east 1.00 m).
2. The reconciling assumption is the breach invert. This model stops the breach at natural ground at the embankment toe (216.7 m on the north side), which leaves the 6 m of Pond 2 that sit below natural ground (floor 210.8 m) in the pond. The 2016 study evidently let the north breach cut down to (or near) the pond floor, releasing the full 7.2 Mm³ at the same peak as the east breach; that is also the only reading under which "approximately 2,500 m³/s" applies to every side. Physically it means the outflow headcuts through ~6 m of the natural gravel between the pond floor and the toe, which is a judgement for the reviewing engineer (erodibility of the in-situ gravels, duration of flow, whether the liner/floor detail resists it).
3. A direct Pond 1 breach through the north embankment (north_p1) is not the explanation: the Pond 1 toe on that side is even higher (219.4 m) and the outflow smaller (473 m³/s), though it starts 2.2 h earlier because there is no cascade.
4. A residual remains. With the full-depth breach the model is still 0.2–0.9 h later than 2016 near the site and 1.0–2.1 h later at Browns, Two Chain and Diversion Roads, whereas the east run with the same hydrograph is 0.7–1.5 h earlier than the 2016 east times. The 2016 north wave therefore travelled faster than the 2016 east wave over the same plain, which breach size cannot explain; it points to a different flow path or terrain representation in the 2016 north model (e.g. a route along the Eyre-side terraces or the Main Race with lower effective roughness) and can only be settled from the Appendix H maps. The 2016 depths at the far roads (0.3–0.5 m) sit between the two model cases, and the 2016 north flood zone (66 km²) lies between the base (49 km²) and full-depth (74 km²) extents, so the 2016 north breach was probably somewhat smaller than the east-size hydrograph but much larger than the toe-level breach.
5. The full-depth PAR screening (~500) is far above the 2012 survey figure (95) because OSM footprints count every building, including sheds, and the north band crosses the West Eyreton lifestyle blocks; the east extended run shows the same inflation (225 vs 107). Use the building counts for ranking only.

Recommendation for the certified assessment: treat the north-side breach invert (toe level vs pond floor) as a bounding pair until drawing WIL1125/30/2 (floor levels, cut/fill on the north side) and Appendix H (2016 breach parameters) are available. For the PIC the full-depth case governs on the north side: 203 buildings ≥ 0.5 m on the 4 h domain against 17 for the toe-level breach. Assumption register: A15/A16 in docs/clg-notes.md.

Maximum depth – full-depth north breach, 12 h full domain

Maximum depth – full-depth north breach, 4 h domain

Maximum depth (12 h, full domain)

Arrival time

Hazard class

Maximum flow speed

Depth × velocity

Time of peak depth

Shakedown domain (4 h): maximum depth, arrival time, hazard class

3. Maps

East breach

Maximum depth

Maximum flow speed

Depth × velocity

Hazard class H1–H6

Arrival time

Time of peak depth

West breach

Maximum depth

Maximum flow speed

Depth × velocity

Hazard class H1–H6

Arrival time

Time of peak depth

3b. Earthquake scenario – all embankments fail at once (quake)

Postulated seismic failure at full supply (e.g. Alpine Fault Mw ~8 or a Darfield-type Mw 7 event): the east, south, north and west embankments and the dividing embankment all start to breach at t = 0, with Froehlich formation times halved for erosion through cracked, slumped fill, and no warning time. Assumption list in config/dam.yaml: scenarios.quake; method in SPEC §3.1 item 7. Time zero is the earthquake.

Road arrival (h after the initiating event, water > 0.1 m) and maximum depth:

Reading: the east breach still carries 80 % of the water because its invert is 5–6 m lower than the others, so the downstream picture along Carleton–Downs Road resembles the east cascade, arriving 10–20 min sooner and, measured from the initiating event, about 2.3 h sooner because there is no Pond 1 → Pond 2 stage. The extra 4 km² of inundation and the additional buildings reached are on the Dixon Road (north), MR4 (west) and R2 race (south) sides, which the single-breach scenarios do not touch. The number of buildings ≥ 0.5 m is slightly lower than the east case because the same volume is spread over more directions.

Earthquake maps

3c. Storm scenario – breach during heavy rain with the Eyre River in flood (storm)

NZSOLD "rainy day" combination: the east cascade breach (identical hydrograph, peak 2,098 m³/s) opens after 2 h of steady 10 mm/h rain on saturated ground (no infiltration) with the Eyre River already carrying an assumed 150 m³/s where it enters the domain. Breach-added figures are differenced against an identical storm run without the breach, so rain and river water are not credited to the dam. Assumptions in config/dam.yaml: scenarios.storm and docs/clg-notes.md (A9–A12); method SPEC §3.1 item 8. Time zero is the breach opening.

Road arrival (h after the breach opens, first time the breach raises the depth by > 0.1 m above the storm-only run) and maximum total depth:

Reading: the breach wave itself is barely faster on a wet plain (0–12 min earlier at Poyntzs–Downs Road), but it spreads over 10 % more ground and is 0.05–0.35 m deeper where it meets the Eyre River and the ponded runoff east of Poyntzs Road, because the river channel and the drains are already full. What changes most is the total picture people would experience: 608 buildings with water against 439, and 113 rather than 95 with 0.5 m or more, since the storm alone already wets 40 km² of the plain in this (very heavy, saturated-ground) rainfall assumption. The storm-only baseline run is outputs/storm/storm_shakedown_nobreach.sww.

Mesh note: the storm runs use a repaired mesh (95,140 triangles, no sliver triangles – see SPEC §7.3). Re-running the east scenario on the repaired mesh changes its inundated area from 39.5 to 39.2 km², road arrivals by ≤ 2 min and buildings ≥ 0.5 m from 95 to 92, so the published east results were left as they are.

Storm maps (total depth, including rain and river)

4. Key finding on the cascade

With the design report's volumes and areas, Pond 1 draining into Pond 2 statically equalises at ~223.8 m RL, 0.5 m below the Pond 2 crest (224.3 m). The overtopping cascade Damwatch describes therefore requires dynamic surge or a different stage–storage shape. The model initiates the Pond 2 breach at 223.6 m RL (config/dam.yaml: cascade.pond2_trigger_mRL); reconcile with Appendix H before any reportable use.

5. Interactive

• East: https://thorstenbux.github.io/wrights-road-dam-breach/east/
• East routed to Diversion Road: https://thorstenbux.github.io/wrights-road-dam-breach/east-extended/
• North: https://thorstenbux.github.io/wrights-road-dam-breach/north/
• North routed to Diversion Road: https://thorstenbux.github.io/wrights-road-dam-breach/north-extended/
• West: https://thorstenbux.github.io/wrights-road-dam-breach/west/
• Earthquake (all embankments): https://thorstenbux.github.io/wrights-road-dam-breach/quake/
• Storm (east breach + rain + Eyre River in flood): https://thorstenbux.github.io/wrights-road-dam-breach/storm/