When Does Calibrating an Urban Heat or Water-Sensitive Design Model Become R&D?

Quick answer: Running a stormwater or microclimate model to a published method is professional practice — the outcome is determinable in advance, and tuning parameters until the output looks right does not change that. A core R&D activity may exist in the narrower case where a model cannot reproduce a behaviour the decision depends on, where the reason is a technically uncertain question about the physics of that site, and where it can only be settled by a systematic progression of work from hypothesis to experiment, observation and evaluation conducted to generate new knowledge. Missing a performance target is not itself the unknown. The company must assess each activity against the statutory tests and self-assess its own claim.

Regulatory and planning boundary

A limit on scope first. This article assesses nothing against a planning control, a development standard or a discharge requirement, and takes no position on whether any product, media specification or software package performs. Assessment against planning and drainage requirements, flood risk determination and heat-health advice are matters for the accredited professionals and consenting authorities you engage. What follows is about one question only: whether a technical outcome was genuinely not knowable in advance.

This piece sits alongside our earlier interview article on greening, rooftop photovoltaics and buildings as a coupled system, which covers the research-trend picture across integrated greening, energy and building systems. This one is narrower: the eligibility test for calibrating and validating urban-heat and water-sensitive-design models against site-specific conditions.

Paragraph (f) — Statutory requirements and soils/atmospheres testing

Section 355-25(2)(f) sits close to this work. It excludes "activities associated with complying with statutory requirements or standards", expressly including "routine testing and analysis of materials, components, products, processes, soils, atmospheres and other things". This may be relevant to routine filter-media testing or modelling activities undertaken specifically to satisfy a consent condition, depending on the purpose and circumstances of the activity. The same paragraph names "calibrating secondary standards": metrological calibration of an instrument against a reference, a different operation from fitting a hydrological model's parameters, though the shared word causes confusion.

Calibration vs parameter selection: The first is between calibration against observed data and parameter selection against a convention: fitting parameters to measurements you took can carry a hypothesis, while choosing parameters so the output reproduces a published coefficient cannot, because the answer was fixed before the work started. Both get called calibration.

Fit vs validation: The second is between a fit and a validation — a model tuned on a dataset and then reported against that same dataset has demonstrated nothing about its predictive adequacy, however good the residuals look.

1. The stormwater model is usually not calibrated against anything measured

The Water Sensitive Urban Design – Greater Adelaide Region Technical Manual is blunt about the normal position: "Ideally models should be calibrated against local flow data, however in most cases information is not available to achieve this." Where it is not, the manual directs that "the selection and calibration of model input parameters should be based on replicating 'accepted' annual volumetric runoff coefficients" — worked through with an example urban catchment at an AVRC of 0.4 (Stormwater Management Authority SA).

That is a sound convention, but it constrains only annual volume — not the drying trajectory between storms, which determines media moisture when a heat event peaks. The manual notes that effective impervious area shifts with rainfall depth/intensity. Furthermore, continuous series from the Bureau of Meteorology's IFD design rainfalls and Australian Rainfall and Runoff tell you when it rained, but not the soil state through dry spells (antecedent moisture, evapotranspiration, root uptake).

2. The thermal map is not the quantity the model predicts

South Australia's thermal imagery provides 2 m surface temperature data (DEW Urban Heat FAQs; Green Adelaide), but it is land surface temperature, not air temperature or mean radiant temperature.

A microclimate model can be tuned to match a surface thermal raster via compensating errors (e.g., overestimating evaporative flux while underestimating thermal admittance), producing the right skin temperature while predicting air temperature and radiant load 1 m above the pavement completely incorrectly.

3. The state both models depend on and neither resolves

The joining variable is soil moisture in the filter media. Stormwater models lump moisture as a store to be drained; microclimate models typically treat vegetation water status as a static "well-watered" setting. In winter-dominant climates, winter water must sustain summer cooling. On extreme days, transpiration can be limited atmospherically by vapour pressure deficit (stomatal closure) rather than by root-zone water supply, a threshold that no published table provides for local species and media configurations.

Baseline & targets (recorded October 2025): Baseline: uncalibrated AVRC 0.4 stormwater model (84% TSS, 46% TN reduction) and microclimate model predicting 2.4 K air-temp reduction at 15:00 on a 41 °C day (measured on a pilot street at 0.7 K). Target: one parameter set driving both models, predicting 15:00 pedestrian-level air-temp depression within ±0.4 K on ≥ 8 of 10 held-out hot days (> 38 °C), and annual TN load reduction within ±5 percentage points of flow-weighted measurements.

Held and varied: Held: Ulmus parvifolia, pit geometry (1.8 × 1.2 × 1.0 m), sandy loam media (180 mm/h Ksat), kerb inlets, meteorological forcing mast. Varied: submerged zone depth (0, 150, 300, 450 mm; 2 pits each), rainfall–runoff parameters, and transpiration scheme.

Trials & results:
• Trial 1 (Calibrate microclimate to 2 m thermal raster, Failed): Model matched raster (1.1 K RMSE), but predicted 15:00 air depression remained 2.1 K vs 0.7 K measured. Compensating errors in sensible vs latent fluxes masked pedestrian air temperatures.
• Trial 2 (Refit runoff parameters to 6-min inlet flow, Partial): Annual volume matched within 6%, but model predicted media moisture at 0.19 m³/m³ when probes read 0.08 m³/m³ because it lacked summer tree root-uptake.
• Trial 3 (Test supply-limited cooling hypothesis, Failed): 450 mm submerged pits held moisture > 0.22 m³/m³ all summer, but air-temperature depression differed by only 0.1 K (within ±0.2 K measurement uncertainty). Above 38 °C, sap flow dropped in all pits as VPD exceeded ~4.6 kPa, disproving the supply-limited hypothesis on extreme days.
• Trial 4 (Coupled model with VPD-limited transpiration & root uptake, Target Met): Calibrated on 4 pits (Dec 1–Jan 14); evaluated on 4 held-out pits (Jan 15–Mar 17). Predicted media moisture within 0.03 m³/m³ across all 62 days; 15:00 depression predicted within ±0.3 K on 9 of 10 held-out hot days. Annual TN reduction was 41% (vs 38% measured, within ±5% target).

Engineering consequence: On extreme days, cooling was driven by tree shading (mean radiant temp 17 K lower), so capital was directed to canopy area with a modest 150–200 mm submerged zone.

Candidate boundary: Spans recorded hypothesis/criteria through to held-out evaluation in Trial 4. Streetscape civil works, consent-mandated drainage modelling, routine media testing, and rollout to future streets sit outside, requiring separate supporting-activity analysis under s 355-30.

Talk to Ignition Research before the pits are built, while the instrumentation can still answer a question rather than record an outcome. As a Registered Research Service Provider at Lot Fourteen in Adelaide, we design and conduct the experimental programme and produce the technical record while the work is happening. We are not a registered tax agent, and we are not your planning, drainage or heat-health assessor: your company self-assesses and remains responsible for its own claim. Get in touch.

This article is general information from a Registered Research Service Provider about the R&D Tax Incentive. It is not tax, legal or financial advice; eligibility depends on your circumstances and you should self-assess and seek your own advice.