This study was conducted on an unobstructed rooftop in Tochigi, Japan. New 1 m² thin-layer containers (soil thickness: 35 mm) filled with a medium primarily composed of perlite were set up, with three replicates. Zoysia (Zoysia matrella (L.) Merr.) sods were planted in these containers. Three months later, on November 25, 2021, gas sampling began. Hourly temperature data for this site during the experimental period are shown in Supplementary Figure S1. Fertilizer (N:P:K = 10:10:10) applications of 20 g/m² were conducted on April 14 and July 19, 2022. An automated bottom irrigation system supplied water to the containers. The turfgrass was clipped when exceeding a height of 5 cm, and the clippings were collected. The properties of the experimental medium, identified in previous studies (Kuronuma et al., 2012; Kuronuma et al., 2014), are listed in Supplementary Table S1.

Gas sampling was conducted from November 25, 2021, to November 10, 2022, at approximately fortnightly intervals, at 10:00 and 22:00 each time. On July 21, August 4, and August 22, 2022, sampling was conducted at 8:00 and 22:00 to prevent turf damage due to high temperatures in the closed chamber. The sampling frequency was set to be equal to or higher than the frequency of previous studies that have investigated long-term GHG fluxes from green roofs (Teemusk et al., 2019) and grass lands (Law et al., 2021). Sampling times were established throughout the preliminary study.

In each rooftop lawn (n = 3), gas samples were collected at 1, 7, 13, 19, and 25 min after a clear acrylic chamber was placed over the center of the turfs (Figure 1). Regardless of weather conditions on the designated sampling days, the automatic open/close chamber collected gases at the aforementioned interval times and sealed these in a vial (Kuronuma et al., 2023).

CO₂, CH₄, and N₂O concentrations in the gas samples were analyzed using a gas chromatograph equipped with flame ionization, thermal conductivity, and ⁶³Ni electron capture detectors (GC-2014, Shimadzu, Kyoto, Japan). The fluxes of the three gases in each sampling and the emissions during the experimental period were calculated according to the closed chamber guidelines (Minamikawa et al., 2015) as follows: F l u x G H G s = ∆ C ∆ t × V A × ρ × 273 273 + T where ΔC/Δt is the concentration change over time (ppm-CO₂/h, ppm-CH₄/h, or ppb- N₂O/h), V is chamber volume (0.00408 m³), A is chamber area (0.0295 m²), ρ is gas density (1.977 kg/m³ for CO₂, 0.717 kg/m³ for CH₄, and 1.977 kg/m³ for N₂O at 0°C), and T is the mean air temperature inside the chamber (°C). Cumulative GHG emissions over a 1-year period were calculated using a trapezoidal integration method. (i.e., linear interpolation was conducted for each GHG flux result (sampling time) and total GHG emissions were calculated based on numerical integration between sampling times) (Minamikawa et al., 2015). Trapezoidal integration method is the most commonly used in literature to calculate cumulative GHG fluxes (Matson et al., 1996; Levy et al., 2017).

To eliminate the influence of the sampling times, cumulative CO₂ emissions were determined from the difference between the cumulative values of the light and dark periods. For convenience, to calculate cumulative CO₂ emissions during light period, we assumed that the light period continues until sampling during the subsequent light period. Similarly, for the dark periods, calculations were performed assuming that the dark period would continue until sampling during the subsequent dark period.

Unlike other GHGs, CO₂ fluxes are strongly influenced by photosynthesis. Thus, cumulative CO₂ emission calculated by GHG flux (above-mentioned method) is likely to be over- or underestimated. Thus, we also calculated the cumulative emissions of CO₂ based on the difference between the total carbon contents in rooftop lawns (soil and turf) at the beginning (November 1^st, 2021) and end (November 10^th, 2022) of the experiment. Using an HSC-5 soil sampler (Fujiwara Scientific Co. Ltd., Tokyo, Japan), we collected 20 cm² samples of soil and turf from three sites per rooftop lawn unit (n = 9). These samples were dried and separated into turf and soil and weighed. The total carbon concentrations in the samples were determined using a 2,400 Series Ⅱ CHNS/O analyzer (PerkinElmer, USA), on the basis of which, we calculated the carbon contents per rooftop lawn area by multiplying the dry weight by the carbon concentrations.

To quantify the annual GHG emission from a thin-layer rooftop lawn as CO₂ equivalents (CO₂e), the global warming potential (GWP) of CH₄ and N₂O were set at 28 and 265 times greater than that of CO₂, respectively (IPCC, 2023). Total GHG emissions from rooftop lawns were calculated by integrating these results. For determining the cumulative emissions of CO₂, we used the result quantified based on two calculation methods (CO₂ flux and carbon sequestration) to present more objective information.

CO₂, CH₄, and N₂O fluxes during the experimental period are shown in Figure 2. Error bars indicated standard errors (n = 3). Since May 17, CO₂ absorption (photosynthesis) during the light period and CO₂ emission (respiration) during the dark period were detected (Figure 2A), signaling that the Zoysia turf had broken dormancy and was growing healthily. The mean values of CO₂ flux during the light and dark period were calculated at −6.35 and 3.08 mg-CO₂• m^-2• min^-1, respectively• Cumulative CO₂ emissions in light and dark period were calculated at −3,416 g-CO₂• m^-2• year^-1 and 1654 g-CO₂• m^-2• year^-1, respectively. Consequently, the annual CO₂ balance during the experiment was −1762 g-CO₂• m^-2• year^-1.

For CH₄, positive fluxes were observed during high temperature periods (between July 7 to September 22), while no notable fluxes were observed from the beginning of the experiment to June 16 (Figure 2B). No significant differences or trends were identified in the flux between light and dark periods. The mean value of CH₄ flux during the experiment period was 0.156 µg-CH₄• m^-2• min^-1 which is intermediate between those from perennial ryegrass and bermudagrass in the previous study (Law et al., 2021). The annual CH₄ emission from the thin-layer rooftop lawn was 92.33 mg-CH₄. m^-2• year^-1 (2.59 g-CO₂e•m^-2. year^-1).

N₂O fluxes were only observed on August 4 during the experimental period (Figure 2C). Additionally, the standard errors were very large because significant flux was detected only in certain samples (Figure 2C). No significant N₂O emissions were observed immediately after fertilization. The annual N₂O emission from the thin-layer rooftop lawn was 0.53 mg-N₂O.m^-2. year^-1 (0.14 g-CO₂e.m^-2. year^-1).

Total carbon contents in rooftop lawns (soil and turf) at the beginning and end of the experiment are shown Table 1. Over a 1-year period, we observed significant increments in the dry weight of turf and in the total carbon concentrations of soil, contributing to an increase in total carbon in rooftop lawn (soil and turf) of 716.1 g-C/m² from the beginning to the end of the experiment, thereby indicating that rooftop lawns sequestrated −2,626 g-CO₂•m^-2•year^-1.

From these results, the annual GHG emission from the thin-layer rooftop lawn was calculated at −1759 to −2,623 g-CO₂• m^-2• year^-1 (Table 2). These results varied greatly depending on the method used to calculate CO₂ emission (CO₂ flux or carbon sequestration). The cause of this is unknown, but continued research is needed to derive more robust data.

In this study, we found that the sequestration of CO₂ by rooftop lawns far exceeded the emissions of CH₄ and N₂O (Table 2). High sequestration of CO₂ by rooftop lawn during the first year after construction are similarly observed in our previous study [Z. matrella: 670 g-C•m^-2•year^-1 (Kuronuma and Watanabe, 2017)]. Likewise, Getter et al. (2009) have reported that the plants and soils of Sedum green roofs sequestrated 375 g-C•m^-2•year^-1 during the first 2 years after construction. In addition, even if the carbon sequestration capacity of rooftop lawns is 110 g-C• m^-2• year^-1 (403 g-CO₂e•m^-2• year^-1) for 15 years (Kuronuma et al., 2014; 2018a), CO₂ absorption far exceeds GHG emissions since the GHG (CH₄ and N₂O) flux was calculated at only 2.73 g-CO₂e•m^-2• year^-1. Therefore, this study suggests that rooftop lawns could serve as a GHG sink.

In contrast, previous studies have estimated that increasing soil organic matter reduces carbon sequestration and increases soil GHG fluxes (Bandaranayake et al., 2003; Zhang et al., 2013; Gu et al., 2015). Thus, the quantification of GHG emissions from rooftop lawns at different ages is a topic for future research. Additionally, there have been reports of environmental impacts associated with the use of thin-layer substrates, fertilizers, and other materials (Kuronuma et al., 2018b; Scolaro and Ghisi, 2022). A more comprehensive analysis that includes these environmental impacts will be required in the future to ensure that the overall environmental benefits of rooftop lawns are accurately assessed. The use of materials with lower environmental impact will be important to increase the GHG absorption capacity of rooftop lawns.

The CH₄ and N₂O fluxes in this experiment were approximately one-half and one-eighth, respectively, of the values reported in similar previous studies using Andosols (Kuronuma et al., 2023). Additionally, results from other studies of ground-level (not rooftop) lawns indicated that the CH₄ and N₂O fluxes observed in our experiment were very low (Dutt and Tanwar, 2020; Brandani et al., 2021; Law et al., 2021). This is likely due to the use of a perlite, which has low total carbon and nitrogen concentrations (Supplementary Table S1), as the main component of the medium. The extremely thin soil thickness in the rooftop lawn compared to ground-levels lawns may also be a reason for the lower GHG emissions. Moreover, microbial biomass C and microbial activity, measured by fluorescein diacetate (FDA) assay, were lower in the medium (main component: perlite) that in the organic layer (Kuronuma et al., 2018a). These findings suggest that green roofs, which require the use of light artificial soils (e.g., perlite), are likely to have lower GHG emissions than ground-level lawn.

In this study, N₂O fluxes were only observed on August 4 and no significant N₂O emissions were observed immediately after fertilization (Figure 2C). Teemusk et al. (2019) had reported N₂O fluxes in green roofs were very low, correlation with meteorological parameters was insignificant. To detect trace amounts of N₂O emissions, lengthening the interval between gas samplings (e.g., 0 min, 30 min, 60 min, and 90 min) might be an effective approach. For the CH₄ fluxes, the findings of some studies have indicated that urban turfgrass can function as a CH₄ sink throughout the year (van Delden et al., 2018; Teemusk et al., 2019). However, in the present study, we observed positive CH₄ fluxes during summer, which we suspect could be attributable to a high soil water content as a consequence of frequent irrigation. In any case, it is difficult to make precise reference in this study to the relationship between CH₄ and N₂O fluxes and various environmental conditions, etc. Accordingly, further investigations are essential to gain insights into these areas. In this study, we adopted a closed chamber method having issues such as the limited number of observation points and the inability to quantify the flux in all directions. In the future, we plan to investigate different gas flux measurement methods in order to improve the output.