STRUCTURE AND BIOMASS ANALYSIS OF URBAN VEGETATION IN SQUARES OF SANTA CECÍLIA DISTRICT, SÃO PAULO, SP

Urban trees play an important role in urban planning and are directly linked to urban spaces that promote ecosystem services such as biodiversity conservation and carbon stock. A useful methodology to quantify the above-ground (ABG) biomass, and consequently, the carbon stocks, in this green infrastructure is the allometric models made for tropical areas. This work aimed to evaluate the carbon stock in public squares within the central region of São Paulo city, thus contributing to the comprehension of São Paulo’s green infrastructure. To test the models, tree density and ABG biomass of approximately 7 ha of urban green areas in the center of São Paulo city were evaluated. The activities involved measuring the diameter at breast height (DBH) of the trees and the total height of all individuals with a DBH greater than 5 cm. The results showed that the public squares varied in size and number of trees. The average height of trees also varied along the squares, probably infl uencing the ABG biomass allocation. Also, the results showed that there was a considerable variation between the ABG biomass estimated along the models. Some squares stood out in absolute terms of ABG biomass accumulation. The size of the green area correlated only with the total number of individuals found (R=0.44). The total ABG biomass, tree density, and ABG biomass density did not present a signifi cant relationship with the size of the squares. This work demonstrates, for the fi rst time, the attributes of tree communities in public squares ofSão Paulo city and provides technical information for the management of public policies related to the protection and maintenance of urban green areas.


INTRODUCTION
The uncontrolled growth of urban centers, together with intense deforestation, has altered the landscape with serious ecological consequences. The Atlantic Forest encompasses most of the large cities in Brazil and features diff erent forest formations with great species richness and fl oristic diversity, overcoming in some areas the tree diversity of the Amazon Forest (Tabarelli and Mantovani, 1990;Tanus et al., 2012;Eisenlohr et al., 2013). However, the biome underwent intense deforestation, as a result of the changes in land use and the expansion of the cities (Lapola et al., 2014). Thus, the urbanization process has brought serious consequences to the ecosystems, generating losses of biodiversity, biological homogenization and damages to ecosystem services in urban areas (McDonald et al., 2013;Ferreira et al., 2017a).
A very promising proposal that has been currently discussed to solve urban infrastructure problems involves the use of Nature-based solutions (Kabisch et al., 2017;Keesstra et al., 2018). Among several existing methodologies, the use of green infrastructure has been highlighted in scientifi c journals (Lafortezza et al., 2013;Sanesi et al., 2017) and in european government reports (EEA, 2011;EC, 2012).
Both biosphere and atmosphere are major carbon sinks (Zhang et al., 2015), and the sequestration or emission of CO 2 plays an important role in the ecosystem dynamics (Saleska et al., 2003). Thus, the reduction of atmospheric carbon promoted by photosynthesis can imply climatic eff ects at local and regional scales (Malhi and Grace, 2000). The overall levels of carbon stored in trees diff er among species, biome, geographical location and phytosociology (Scolforo et al., 2015). It is possible to measure carbon contents in vegetation by direct (destructive) methods or by allometric (nondestructive) modeling, which is a method widely used to assess above-ground biomass in tropical forests (Vieira et al., 2004;Higuchi et al., 2008;Vieira et al., 2008;Alves et al., 2010). These models are usually specifi c to communities of a particular region or to a special state of forest regeneration, e.g., primary or secondary forests (Burger and Delitti, 2008).
A problem intrinsic to above-ground biomass studies in urban green areas is the choice of a model that best represents or estimates the biomass value contained in the tree community in question. Since no specifi c allometric model is available for the vegetation of the São Paulo City, it is necessary to analyze the variability among the estimates provided by diff erent existing equations, before proposing a method to evaluate the green areas of the city in the context of its green infrastructure.
In addition, Brazilian urban green areas have not received much attention when it comes to studies involving the quantifi cation of biomass (Ferreira and Uchiyama, 2015;Ferreira et al., 2017b). As well as large expanses of native forest, urban green areas also provide signifi cant environmental services to cities, which is a key tool to creating sustainable and healthy places.
An important environmental service provided by public green areas in urban centers is their capacity to store carbon (Díaz-Porras et al., 2014). As a consequence, CO 2 emissions from automobile exhaust and industrial processes are compensated (Weissert et al., 2014). In turn, the urban fl ora, due to the interference of anthropic activities, has become as vulnerable as the natural forests. It so happens because these areas are modifi ed by local regulars, which might infl uence the composition and the fl oristic structure by the direct planting of exotic species (Moro et al., 2014). Therefore, taking into account the importance and functions of urban green areas, it is necessary to understand the ecological characteristics inherent to them. An open question is whether tree density and biomass vary according to the size of the green area. Contrastingly to natural forests, such relationships are still unknown when it comes to urban green areas (Oertli et al., 2002;Villard and Metzger, 2014). Finally, as studies involving the capacity of urban green areas to stock carbon lack in Brazil, , this work also assesses the carbon stockage in public squares of the central region of São Paulo city, thus contributing to the comprehension of São Paulo's green infrastructure. It is important to highlight that the collection of these data may become the basis for the development of public policies aimed at the maintenance and expansion of such urban green areas.

Study area
The study was conducted in ten public squares in the Santa Cecília District (Figure I), central region of São Paulo City, including the districts of Santa Cecília, Campos Elísios, Vila Buarque, Várzea da Barra Funda and Bom Retiro. According to the prefecture of São Paulo (Sao Paulo, 2020), the central region of the city totalizes 26,20 km 2 , and the district of Santa Cecilia occupies approximately 14% of this area (3,9 km 2 ). The human population is estimated at about 86,132 inhabitants, with a population density of 22,726 Inh/km² (SEAD, 2014).
São Paulo City is located 780 meters above sea level on the State of São Paulo plateau and presents a historical annual average rainfall of 1,500 mm and amean temperature of 25 °C (IAG/USP, 2014).
The study area encompasses 7.1 hectares of urban green areas distributed in ten public squares. The squares of Santa Cecilia district were randomly selected according to the total area, including both green and paved areas.
Each square presents its own characteristics with respect to physical and structural attributes. The Santa Cecilia Square (LSC) includes a Catholic church that occupies most of the square area and a cemented courtyard sporadically used for festivals and other cultural and social activities. Its vegetation is limited to trees surrounding the church and a few gardens. The Princesa Isabel Square (PI) is densely vegetated with old trees. It is connected to an urban bus terminal with the same name. Large and medium trees are dispersed around the Marechal Deodoro Square (MD), where nearby hospitals and health units are reference points. PI and MD both include playgrounds and historical monuments. The Julio Prestes Square (JP) is located within a train station (São Paulo Metropolitan Trains Company). The State Symphony Orchestra House occupies part of this area, and the population is transient. Chão de Giz (SD), as described in this study, has been recently acknowledged by the São Paulo City Hall (PMSP) as a green area. It includes a few medium-sized trees and is located between two large avenues with heavy traffi c. Antonio Cândido de Camargo (ACC), Olavo Bilac (OB), Padre Luis Alves de Siqueira Castro (PLASC), and Vicente Celestino (VC) squares include vegetation, cement benches and bus stops, and are typically frequented by local residents and merchants. The Torquato Tasso Neto Square (TTN) is not paved. Taxi drivers help maintain the square by planting fruit trees, thus contributing to the preservation and diversity of the site ( Figure I).

Structure and carbon stock in urban green areas
One of the parameters obtained in the fi eld for the quantifi cation of the above-ground (ABG) biomass of each square of the study area was the Diameter at Breast Height (DBH). The Perimeters at Breast Height (PBH) of all trees greater than 14.8 cm were measured using a fl exible tape graduated in millimeters. Trees along sidewalks surrounding the squares were also considered. Should aerial roots, calluses or obstruction appear at 1.30 meters above the ground, PBH measurements were taken 50 cm above these obstacles (Clark et al., 2001).
PBH data were then converted to DBH by dividing the measured perimeter by pi (π). In case of branched trees, the PBH of each branch was measured.
Because there is no specifi c model for urban green areas for São Paulo City, diff erent equations were applied to estimate a probable ABG biomass range, considering also those that led to overestimated values of biomass (McHale et al., 2009). Therefore, fi ve diff erent allometric models applied to tropical forest trees from the Atlantic Forest biome, in which the São Paulo City is located, were used to quantify biomass, namely: Arevalo et al. (2002) [1] and Tiepollo et al. (2002) [2], based only on DBH as the independent variable; the allometric model adopted by Scatena et al. (1993) [3], which used DBH and height as independent variables; the allometric model proposed by Chave et al. (2005) [4], which used tree wood density and DBH as independent variables, and Chave et al. (2014) [5], which adopted DBH, height and tree wood density as independent variables. It is important to highlight that Chave's model was recently elaborated with data from diff erent pan-tropical forests (Chave et al., 2014). The equations of the used models are presented below (equations 1, 2, 3, 4 and 5): It is also important to stress that we used an estimated average value of wood density (0.603 g/cm 3 ) for the Atlantic Forest trees in the allometric equations of Chave et al. (2005) and Chave et al. (2014), as proposed by Vieira et al. (2008).
Carbon stock was considered as 50% of the biomass, since this procedure has been widely used

Eq2
Eq3 Eq4 Eq5 Eq1 and reported in the scientifi c literature (Lewis et al., 2009;Saatchi et al., 2011). It should be noted that some authors have shown a possible error (approximately 5%) associated with this estimate (Martin and Thomas, 2011;Melson et al., 2011).
In order to test whether the square size correlates with number of trees and biomass, a linear regression using the statistical PAST package was applied (Hammer et al., 2001).

RESULTS
The size of the squares, the absolute number of individuals, tree density and average height of the trees varied among the squares as follows. TTN was the largest square (24,817.1 m 2 ) while ACC was the smallest (421.9 m 2 ). The largest number of individuals (203) was reported at PI, followed by TTN (128) and MD (61). SD presented the highest tree density (380.9 individuals ha -1 ), and JP the smallest (19.3 individuals ha -1 ). The average height of the trees ranged from 4.5 m to 11.9 m, and the tallest individuals were found at MD. Considering the branched trees, TTN presented 42 individuals, a number that represents 33% of the total of its trees., Analogously: PLASC 9%, JP 14%, and VC 54% ( Table I).
The structure of the communities varied among the squares. Figure 2 shows the tree distribution in diff erent classes of DBH. Most of the VC and TTN individuals fall within the 5-10 cm DBH range (64% and 38%, respectively). ACC, JP, MD, PLASC, TTN and PI trees fall in all DBH classes. OB and JP have the largest trees, with about 63% and 41% of individuals with more than 30 cm DBH.  The use of a single allometric model may result in a skewed fi nal value. Thus, we used fi ve diff erent models, which allowed to verify the range of the data by comparing the trend of each model.
Regarding the absolute contribution of biomass, PI showed the highest value of all squares, followed by MD and TTN. ACC, SD and LSC presented the lowest amount of total ABG biomass (Figure 3). In some cases, the variability of data provided by the diff erent allometric models was quite high. As an example, Arevaldo's model yielded 74% more biomass than Scatena's model for PI. This was also the trend for the squares with low biomass, such as ACC, which showed 76% more biomass, as calculated by Arevaldo's model, than that computed with Scatena's model.
Although TTN is larger in area than MD and its tree density is also higher, MD yielded the second greatest value of biomass, a result of the diametric structure of the squares since for 38% of the MD trees DBH is greater than 30 cm, whereas for 38% of the TTN treesDBH ranges from 5 to 10 cm.
Based on the diff erent models, OB presented the greatest value of ABG biomass in relation to the total area of the square, even though this represents a comparatively small area. PI yielded the second highest biomass density. ACC, LSC and TTN yielded the lowest biomass densities (Figure 4). In general, Chave's and Tiepollo's models produced intermediate values of ABG in the diff erent squares, while Scatena's and Arevaldo's models produced the lowest and the highest values, respectively.
Such diff erences in ABG biomass between squares with diff erent structures may be attributed to a weak relationship between square area and the absolute number of trees. The increase in the number of trees was poorly explained by square area, as shown in Figure 4. Other attributes such as ABG biomass, tree density and ABG biomass density showed no signifi cant relationships with square area.

DISCUSSION
Antônio Cândido de Camargo square showed the lowest number of trees, but its green area had a high tree density, indicating that it is a comparatively wellwooded square. The results do not show a clear trend correlating square size with the density of branches, as represented by PI, which is one of the largest squares in the study area with the greatest number of individuals. In terms of density, JP and LSC presented the lowest values, and both squares have a predominance of buildings and paved areas of historical and religious importance to local residents.
The tallest trees are on average found in MD and OB. Generally, compared to trees in dense forests, urban trees have variations in their architecture, which favor the formation of multiple branches and greater lateral development of the canopy. Proportionally, VC and TTN showed the greatest number of individuals with this morphology, which is aesthetically pleasing because of the shade, as well as important ecosystem services, such as thermal control (Lindberg and Grimmond, 2011;Nowak et al., 2013).
Regarding the proportion of branches in diff erent DBH classes, the values found in this study point out that VC, SD and TTN present new or newly managed trees since they showed the highest percentage of trees in the smallest size class. Other authors who have studied the vegetation in urban areas have found similar results, such as Troian et al. (2011), who found a predominance of trees belonging to the smaller diameter classes in a fragment of urban green area in the metropolitan region of Porto Alegre in southern Brazil. Naturally, this type of trees in the green infrastructure are better for some ecosystem services, such as carbon sequestration, but are worse for others, such as thermal control.
This structural feature of public squares may highlight the need for proper management by public authorities, given the rapid city growth and the specifi c management of the green infrastructure in São Paulo, which occurs in a decentralized way (Benchimol et al., 2017). The expansion of present suburbs must be monitored and planned in relation to existing forest remnants so that construction activities respect the maintenance of biodiversity, the provision of ecosystem services, and the well-being of those who reside in such localities (Nowak et al., 2013). Breuste (2012) reported the participation of local residents in the preservation of tree biodiversity in urban streets of Mendonza, Argentina, with special attention to residential areas. Similar public participation may alter the tree structure of town squares, refl ecting the involvement of the local community in the conservation of green areas. This could ultimately result in varying the distribution of diameter size classes in urban areas, as well as control the amount of exotic species planted in such places (Zisenis, 2015;Lossová et al., 2016). It is important to highlight that although fl oristic composition is a key tool for treescape assessment, this paper focuses on the structure and biomass of the communities located in squares of the district of Santa Cecilia, in São Paulo city.
Hence, it is likely that the structure of the tree community in public green areas may diff er from the center to suburbs.
Interestingly, we found a high percentage of large trees in almost all squares, accounting for total ABG biomass and contributing to carbon stocks. This can be explained by the historical age of the city center, although São Paulo has experienced an urban sprawl during the last century. The central part of the city is the oldest area, currently consisting of offi ces and a few houses. This fact makes these squares places frequented by transients.
Several authors have reported that total tree carbon is approximately 50% of the biomass (Chave et al., 2008;Lewis et al., 2009;Saatchi et al., 2011). Other researchers have shown a possible error (approximately 5%) associated with this estimation (Martin and Thomas, 2011;Melson et al., 2011;Saner et al., 2012). In this sense, the carbon stock in urban trees is an extremely important ecosystem service once it works as a mechanism that compensates for local greenhouse gas emissions (Sharma et al., 2010).
Corroborating this statement, Zhao et al. (2010) showed a signifi cant absorption of CO 2 emitted by industrial sources in Hangzhou, China, by urban trees. The authors explained that approximately 18% of total CO 2 emitted annually was uptaken by urban vegetation alone.
In the same way, Russo et al. (2015) pointed out that public squares in Bolzano, Italy, stocked approximately 0.61 kg CO 2 ha -1 year -1 , playing an important role as a carbon sink. The authors noted the important role of treescapes in the C sink. Timilsina et al. (2014) also showed that urban trees were responsible for about 5% uptake of anthropogenic emissions in the US State of Florida. The eff ectiveness of urban green areas in the USA prevented 850 casualties and 670,000 cases of acute respiratory symptoms in 2010 (Nowak et al., 2014).
Based on United Nations recommendations, Martins and Araujo (2014) mentioned that each city must have 12 m 2 of green area per inhabitant. By 2030, some estimates predict that more than fi ve million people will inhabit urban centers (UNFPA, 2015). Thus, it very important to quantify the biomass stocked in urban trees and their potential for sequestering carbon from the atmosphere.
Such information may be used to develop specifi c public policies for the protection and maintenance of urban green areas, such as the PLANPAVEL program for São Paulo City, which integrates the Municipal System of Protected Areas, the Green Areas and Free Spaces program (SAPAVEL) and the Municipal Plan of Urban Aff orestation. Some public squares in this study showed high values of biomass per hectare compared to some areas of native Atlantic Forest, the biome of the local study. Rolim et al. (2005) found above-ground biomass of 334.5 Mg ha -1 in Linhares, southeastern Brazil, and Alves et al. (2010) found a total above-ground biomass of 239.3 Mg ha -1 in the same phytophysiognomy (submontane) in Ubatuba.
Proportionally, OB had the largest above-ground biomass per hectare, followed by PI. This demonstrates that despite the importance of tree density in a green area for carbon stock, other attributes are equally important, for example, the vertical and horizontal structure of the community. This is reinforced by MD, which has the third highest density of trees, and, on average, the tallest trees. Brown and Iverson (1992) showed that a DBH of 100 cm represents a biomass equivalent to 610 trees with 10 cm of DBH. Although this result was obtained for an area of Ombrophilus Dense Forest in the Amazon region, it can be inferred that diameter and biomass have no direct and proportional relationship and an individual's biometric values can be a decisive factor in higher or lower carbon stocks.
Special attention has been given to the best allometric models that fi t these tree communities. The results also showed that PI and ACC presented completely diff erent biomass profi les, highlighting the heterogeneity of treescape ABG biomass in Santa Cecília District in the central region of São Paulo City. A large range of ABG values yielded by the diff erent allometric models was also found, mainly in PLASC, with approximately 81% of variation in ABG biomass, when Arevaldo's and Scatena's models were used. McHale et al. (2009) tested diff erent allometric equations in an urban green area in the U.S. State of Colorado and found more than 300% diff erence among the equations, reinforcing the need for allometric models specifi c to urban green areas.
Interestingly, it was observed that the TTN total area is almost four times that of MD and that it presents more than twice the absolute number of trees. Yet it still yields a lower amount of ABG biomass. It should be noted that the average TTN tree heights is almost three times shorter than those of MD, resulting in the smallest amount of stored biomass, which emphasizes the importance of the community vertical structure to the stock of ABG biomass.
Another important issue related to the use of public green areas is the maintenance of biodiversity. Several squares play an important role in leisure, while others are important for their carbon uptake (Mazzei et al., 2007). An important consideration for public policy makers in São Paulo is to determine if trees can be planted in any empty space in urban green areas, since they promote shade and thermal comfort, as well as shelter for the biodiversity of urban arthropods and birds. Our results show that the increase in green area responds for only 42% of total number of individuals, which demonstrates the potential of these squares for planting more trees.

CONCLUSION
The present work allowed to verify that the biometric characteristics, related to the DBH classes and the height of the trees, varied throughout the squares of the study area. The frequency distribution of branch diameters in the squares varied according to regions. Other structural features also varied among the areas, such as average height of trees.
The study draws attention to the antagonistic results observed when analyzing the total biomass per square and the density of biomass per square, which shows that larger areas are responsible for the greater amount of absolute biomass stored in trees; however, when analyzing in terms of biomass density, it is observed that smaller squares are more vegetated or have trees with larger diameters, promoting an important ecosystem service (carbon stock).
This information may assist the Secretariat for Green and Environment in the planning of future green corridor, as well as improving existing programs that require ecological information on public green spaces, such as PLANPAVEL and the Municipal Urban Aff orestation Plan. So, as a positive point of the present study, it is possible to highlight that it is a contribution to public policies aimed at the management of the green infrastructure of São Paulo City. This is because the results showed that the higher biomass densities, verifi ed in the smaller squares, indicate that the larger squares have potential for the planting of more trees, promoting thus more ecosystem services to the city. In addition, it is possible to highlight the pioneerism of the present study in the eff ort to estimate biomass values in the present green infrastructure of São Paulo city. Lastly, we highlight the importance of the expansion of studies that seek the development of allometric models specifi c to the vegetation of the green areas of São Paulo City. This is especially important because the results indicate a high variability of data among biomass values calculated by diff erent allometric models, for example, the one obtained using the Arevalo's and Scatena's models.