Hydrological Recovery and Gas‑Phase Memory Across Green Roof Substrates: Evidence from Auckland, New Zealand

Living roofs are commonly evaluated using event-scale runoff metrics, while gas-phase dynamics are rarely considered in relation to rainfall timing. This study investigates how storm sequencing and hydrological memory jointly influence runoff response and near-surface CO₂ concentration in living roof systems.

Rainfall, runoff, and near‑surface CO₂ concentration were monitored on five experimental roof trays in Auckland, New Zealand, representing three substrate configurations of equal depth: an unvegetated stone ballast reference and two vegetated substrates (Daltons living roof mix and eco‑pillows). We analysed a six‑month winter‑to‑spring period (1 June–30 November 2025) with variable inter‑event dry durations. Rainfall events were classified by inter‑event dry duration to distinguish closely spaced and isolated storms. Runoff response was quantified using runoff coefficients and peak discharge metrics normalized by rainfall forcing, while CO₂ dynamics were assessed during rainfall and inter‑event periods and expressed as anomalies relative to the stone reference (ΔCO₂).

Closely spaced storms generally produced higher runoff coefficients and reduced peak attenuation compared with isolated events, consistent with incomplete hydrological recovery. However, isolated events associated with exceptionally large or intense rainfall like the one in July 2025, with a depth of 82.8 mm and an intensity of 4.17 mm/hr, can produce high peak discharges, indicating that storm characteristics may override memory effects under extreme conditions. CO₂ concentrations increased during rainfall and remained elevated between closely spaced events, indicating a gas‑phase “memory” associated with rainfall‑driven state changes. Substrate type strongly modulated the CO₂ signal: Daltons showed persistent CO₂ drawdown relative to stone (mean ΔCO₂ ≈ −8.9 ppm), whereas eco‑pillows exhibited net enrichment (mean ΔCO₂ ≈ +11.8 ppm, increasing in spring). These results highlight non‑stationary coupled hydrological and gas‑phase behaviour in living roofs, while noting that concentration‑based metrics capture near‑surface signals rather than CO₂ fluxes.