On 14 February 2023, Cyclone Gabrielle made landfall on the North Island's east coast with a central pressure of 958 hPa. What followed was the most significant flood event in Hawke's Bay in living memory. Rainfall totals across the region were equivalent to 1:50-year to 1:200-year ARI events depending on location and duration interval. The Tukituki River recorded peak flows among the highest since the gauge network was established. Seven stopbanks on the Tukituki system breached upstream of Waipukurau and Waipawa, inundating productive farmland and rural properties across a wide area.
SAE Ltd had completed a flood hazard assessment for a proposed subdivision at Mt Herbert Road, Waipukurau, before the cyclone arrived. The design had been finalised. There was no ability to adjust parameters once the storm began. The design performed as intended.
This post documents the engineering decisions embedded in that assessment and what Cyclone Gabrielle revealed about those decisions.
The Tukituki River Flood Assessment
The project was a flood assessment for a proposed subdivision at Mt Herbert Road, Waipukurau, on a site adjacent to the Tukituki River, on the edge of HBRC's 2008 regional flood study boundary.
The primary engineering challenge was hydrological: which variant of the SCS (Soil Conservation Service) rainfall-runoff method should be used to calculate design peak flows for the Tukituki catchment? This is not a trivial question. Three credible variants produce materially different results:
- Original SCS (initial abstraction ratio λ = 0.2): Q100 = 894 m³/s
- Brazilian SCS (λ = 0.05): Q100 = 1,129 m³/s - an increase of 21% over the original
- Auckland TP108 (Ia = max 5 mm): Q100 = 1,186 m³/s - an increase of 25% over the original
The range between the lowest and highest estimate is 292 m³/s - equivalent to roughly 33% of the original estimate. This is not a rounding error. It is a fundamental methodological choice that directly affects finish floor levels and, ultimately, whether the site is safe under flood conditions.
The project site was at the edge of the existing 2008 HBRC regional flood study boundary, which meant the available modelled flood levels did not extend to the full site area. Extrapolation was required, using HEC-RAS cross-section geometry derived from a LiDAR dataset. The raw LiDAR file was a 6 GB point cloud of approximately 120 million points, processed in Cloud Compare to generate Civil 3D-compatible cross-sections at the required intervals.
The methodological decision was straightforward in principle: where multiple credible methods produce materially different results, and the consequences of underestimating are irreversible (a building constructed too low cannot easily be raised), adopt the most conservative result as the design baseline. Not the midpoint. Not the average. Not the "most commonly used" variant. The most conservative.
For this project, that was the TP108 method: Q100 = 1,186 m³/s. All design flood levels - and therefore all finish floor level recommendations - were derived from this flow.
What Cyclone Gabrielle Did to the Tukituki
The Tukituki River behaved as a large catchment under extreme rainfall. Seven stopbanks on the Tukituki system breached upstream of the project site. The river flooded surrounding rural and residential properties across the Waipukurau and Waipawa areas, causing significant damage and disruption.
The project site at Mt Herbert Road was examined in post-event aerial photography and by ground inspection. Lots 1 through 5 on the assessed area showed no visible flooding. The surrounding properties were affected; the site was not.
The 1:100-year ARI event is, by definition, a flow that is equalled or exceeded on average once every 100 years. Cyclone Gabrielle's flows in the Tukituki catchment were broadly consistent with this return period, though the precise annual exceedance probability of the specific peak flow at this gauge location would require detailed hydrological analysis to confirm. What is clear is that this was a major flood event - and the finish floor levels derived from the TP108 Q100 of 1,186 m³/s proved adequate.
Had the design been based on the original SCS Q100 of 894 m³/s - a method that is widely used and broadly accepted - the recommended finish floor levels would have been materially lower. Whether those lower levels would have resulted in inundation during Gabrielle cannot be stated with certainty, but the margin of safety would have been significantly reduced.
What This Means for Engineering Methodology
Post-event analysis is the most honest test of an engineering design. There is no opportunity to adjust model parameters, revise assumptions, or recalibrate after the flood arrives. The design either worked or it did not.
Conservative methodology is not the same as over-engineering. Over-engineering means adding unnecessary cost or redundancy without a clear engineering basis. Conservative methodology means choosing the correct analytical approach when multiple approaches exist and produce different results, and choosing it at the more cautious end of the defensible range. The TP108 variant is not exotic or unpublished; it is a recognised SCS implementation used across New Zealand. Choosing it over the original SCS was not a departure from standard practice; it was a deliberate, documented methodological preference with a clear engineering rationale.
Cyclone Gabrielle provided a rare and unusually direct confirmation that this principle, applied carefully before the event, produced finish floor levels that performed as intended when the event arrived.
One flood assessment. One cyclone. The Tukituki River lots remained dry while surrounding properties flooded. Conservative methodology, applied consistently at the design stage, is what made the difference.