Carbon sequestration by plant species used in green roofs across different periods 1

: This study evaluated the influence of three plant species and two periods on the amount of carbon stored in the substrate in an extensive green roof. The study was carried out in March and September 2020 at Charles Darwin Business Garage Building, Recife, Pernambuco state, Brazil. A completely randomized design with a split-plot scheme was adopted, with plant species (zoysia grass [ Zoysia japonica ], inch plant [ Tradescantia zebrina ]


Introduction
The urbanization process has drastically intensified in recent decades, interfering with the landscape and local weather patterns. As a result, one of the most noticeable changes is the large removal of vegetation cover, which in turn ends up directly interfering with the impermeability of soils, increase in local temperature, formation of heat islands, in addition to accelerating the greenhouse effect through the emission of greenhouse gases (GHG), in particular carbon dioxide (CO 2 ) (Lima, 2013).
Among GHGs, CO 2 is the most abundant and primary anthropogenic contributor to global warming (Rei et al., 2017). In this sense, green roof application is an effective alternative for reducing the effect of the urban heat islands, habitat restoration, rainwater management, energy efficiency improvements (Talebi et al., 2019), and GHG reduction, especially CO 2 (Agra et al., 2017).
Green roofs are constituted by a vegetal covering implanted in conventional roofs or slabs (Souza et al., 2015). These are classified into three types: extensive green roofs, semi-intensive and intensive, whose main difference is in the vegetation kind and the substrate thickness used (IGRA, 2017).
Plant species such as Sedum floriferum, Sedum album, Allium schoenoprasum, Cynodon dactylon, Paspalum vaginatum and Zoysia japonica have been used in studies involving the theme of green roofs (Ismail et al., 2012;Heusinger & Weber, 2017;Ntoulas et al., 2017). However, research involving the use of different species of plants is required as they can contribute to a better performance of green roofs.
In this way, the current research aimed to evaluate the amount of carbon stored in the substrate influenced by three different plant species found in an extensive green roof in two different periods.

Material and Methods
The study was conducted in March and September 2020 at the Charles Darwin Business Garage Building, Rio Ave Empreendimentos Construction Company, Recife, Pernambuco state, Brazil (8º 04' 03" S, 34º 55' 00" W, at an altitude of 4 m). The building is eight floors high. According to the Köppen climate classification, the climate of the region is an As type, tropical wet (Pereira et al., 2002), with a rainy season from May to August and an annual accumulated rainfall of 2263.4 mm. The annual average air temperature is 25.9 °C, with maximum and minimum average temperatures of 29.5 and 22.3 °C, respectively. The relative air humidity is 78.3% (INMET, 2020).
To enhance the characterization of the two periods, data from the Meteorological Database for Teaching and Research (BDMEP), managed by the National Institute of Meteorology, were obtained from the Recife (Curado) meteorological station in Pernambuco state (8º 04' 03'' S, 34º 55' 00" W, and altitude of 4 m) and used for analysis. Monthly data related to maximum and minimum air temperatures, and relative air humidity for the two periods were selected ( Table 1).
The installed green roof was extensive with an area of 2.800 m² and three plant species, including zoysia grass (Zoysia japonica), inch plant (Tradescantia zebrina), and fountain grass (Pennisetum setaceum). In addition, the roof had a 7.5cm thick substrate layer of treated sewage sludge from the Charles Darwin Business Garage Building, a drainage system formed by a geotextile drainage fabric (bidim), a 2.5-cm thick stone layer (made from recycled concrete debris), as well as a bidim fabric and asphalt layers for slab waterproofing. In addition, the green roof had a retractable sprinkler irrigation system that could supply 9.2 mm of water per day, and a reservoir with a storage capacity of 380 m 3 .
The substrate used was derived from treated sewage sludge (C/N ratio: 15.2; pH: 7.0; total nitrogen: 21.5 g kg -1 ; phosphorus: 16.75 g kg -1 ; potassium: 2.27 g kg -1 ), with a dry density of 781.35 kg m -3 , a density of 1130 kg m -3 at the maximum water retention level, and a maximum volumetric retention of 49.6%.
The experimental design adopted was a completely randomized design in a split-plot scheme, with plant species (FG: fountain grass [Pennisetum setaceum]; ZG: zoysia grass [Zoysia japonica]; IP: inch plant [Tradescantia zebrina]) allocated to the plots and the material collection periods (March and September) to the subplots, with five replicates for each treatment. The FG, ZG, and IP flower beds had areas of 39.00 m² (13.00 × 3.00 m), each corresponding to an experimental area of 117.00 m².
The plant species used in the present study are recommended for green roofs due to their aesthetic, functional, and recreational properties (Beard & Green, 1994), in addition to their air pollution reduction capacity (Lu et al., 2016). Furthermore, the plants have diminished root systems so that they are suitable for application as extensive green hedges.
The variables analyzed were organic carbon concentration and content of CO 2 evolution from the substrate. The analyses were performed in March and September 2020.
A square of known area (0.5 × 0.5 m) was used to select the location where the substrate samples were obtained. The square was randomly selected within each replicate, with five replicates in each treatment, yielding 15 substrate samples during each period of evaluation, and 30 samples per period for each variable analyzed.
The substrate samples on the green roof were collected from a depth of 0 to 5 cm to determine CO 2 evolution. The samples were packed in plastic bags and subjected to chemical characterization analyses. The chemical method described by Grisi (1978) was used to determine the variable based on  (2020) the CO 2 amounts absorbed from the soil. The substrate was incubated in a closed container with 10 mL of potassium hydroxide (0.5 N KOH).
Sample collection for organic carbon and CO 2 evolution determination in both periods were carried out on the same day. The substrate samples were collected using a garden shovel from a depth of 0 to 5 cm. Afterward, the samples were transferred to previously labeled plastic bags and sent to the laboratory. Organic carbon concentration was determined using the standard method established by Yeomans & Bremner (1988) based on dichromate (Cr 2 O 7 2-) reduction by organic carbon compounds and subsequent determination of the remaining Cr 2 O 7 2by titration of excess chromium with ferrous ammonium sulfate.
The percentages of organic matter in the substrate associated with each plant species and period were also determined ( Table  2). The value obtained for each treatment was multiplied by a factor of 1.724 because generally, carbon accounts for 58% of the average humus composition (Claessen, 1997).
The data obtained were subjected to Analysis of Variance, and if the F-test detected significance, the average values were subjected to Tukey's test (p ≤ 0.05). Statistical analyses were performed using SISVAR 5.0 (Ferreira, 2011).

Results and Discussion
No significant interactions (p > 0.05) were observed between plant species and collection periods. In addition, no significant difference was observed in CO 2 content among the species based on microbial respiration. However, CO 2 content across the different periods differed significantly, with average values of 14.05 and 96.97 mg in March and September, respectively (Table 3).
Precipitation was experienced in Recife, Pernambuco state, Brazil, on the week preceding sample collection and on the collection day in the second period, which could explain the relatively high CO 2 evolution values observed. According to data acquired from BDMEP (BDMEP, 2020; INMET, 2020), a total precipitation of 11 mm was observed in Recife, Pernambuco state, hours before sample collection in September. In contrast, no rainfall was observed on the collection day in March. Such rainfall increased water supply to the green roof, which justifies the high CO 2 production. Soil moisture is one of the main factors influencing microbial activity, and enhances CO 2 content.
Similar results have been observed by Santos et al. (2016), who investigated soil microbial respiration within a research area located in Olho D' Água do Casado municipality in Alagoas State, Brazil. In addition to the results of Azevedo et al. (2019), analysis of various sources of organic fertilizers in an area under agroforestry systems in the Brazilian semiarid region revealed that CO 2 evolution was higher in the rainy season than in the dry season.
Soil moisture regulates CO 2 release (Macleod et al., 2008;Correia et al., 2009). Seasonal fluctuations in microbial development are more relevant to the soil surface, where humidity and temperature fluctuations are the most notable (Cattelan & Vidor, 1990). Therefore, varying soil conditions and their environments can generate fluctuations in carbon content (Araújo et al., 2011).
A significant difference was observed in organic carbon concentrations among species and between the research periods (Table 3). Inch plant and fountain grass organic carbon concentrations differed significantly, with mean average values of 129.22 and 95.06 g kg -1 , respectively. Furthermore, the mean organic carbon concentrations in March and September were significantly different, with average values of 131.15 and 100.81 g kg -1 in March and September, respectively.
Soil organic carbon concentrations indicate the amount of organic matter accumulated in soil (Monteiro, 2020). Similarly, the organic carbon concentration (Table 3) was higher in the period when substrate organic matter was high, in March, than in September (Table 2).
A similar result was observed by Brown & Coton (2011) who investigated total organic carbon deposition, available nutrients, soil nitrogen, and microbial activity, among other trends, in agricultural soils in California. The researchers observed that as the soil organic matter concentration increased with the addition of organic waste, the organic carbon concentration also increased. In addition, the authors observed that microbial activity in soils treated with organic residues was approximately 2.3-fold higher than that in the control soil (p ≤ 0.009). In the present study, microbial activity increased significantly following the addition of organic matter to the soil, which provided nutrients to the microorganisms.
Means followed by different lowercase letters in the line are statistically different by the Tukey test at p ≤ 0.05. 1 ZG -Zoysia Grass (Zoysia japonica); 2 FG -Fountain Grass (Pennisetum setaceum); 3 IP -Inch Plant (Tradescantia zebrina); 4 SEM -Standard error of the mean; 5 MAR -March; 6 SEP -September; 7 CO 2 -Content of CO 2 ; 8 OC -Concentration of organic carbon Table 3. Average content of CO 2 and concentration of organic carbon in the substrate, for the species zoysia grass (Zoysia japonica), fountain grass (Pennisetum setaceum), and inch plant (Tradescantia zebrina), and periods of March and September Table 2. Average concentration of organic matter in the substrate for the species ZG (Zoysia Grass -Zoysia japonica), FG (Fountain Grass -Pennisetum setaceum), and IP (Inch Plant -Tradescantia zebrina), and for the periods of March and September 2020 Good management practices can increase soil organic matter concentration and enhance soil carbon, in addition to reducing CO 2 content in the atmosphere and potentially mitigating global climate change (Wells et al., 2019).
Over the years, it has increasingly become evident that greenhouse gas emissions have considerable detrimental effects on populations and cause climate change, which adversely affects plant growth and development, and in turn global agriculture production (Wargent & Jordan, 2013). Therefore, carbon sequestration in soils as CO 2 considerably reduces the greenhouse effect. In addition, organic matter concentrations in soil sequestered with high carbon levels would also be high (Trombetta et al., 2020).
Notably, carbon dynamics have been studied extensively in natural, agricultural, urban, and community forests. However, understanding of the the carbon sequestration potential of green roofs remains poor.

Conclusions
1. The content of CO 2 in the substrate does not differ between plant species, but it differs between the periods, with September showing the greatest potential for CO 2 storage in the substrate.
2. The organic carbon concentration does differ between species and periods, with higher values for the species Inch Plant and Zoysia Grass, and for March, which was the period with higher water content in the substrate.