Urban Tree DBH Response to Fast Urbanization— A Case from Coastal City Zhanjiang, China

Trees perform various ecosystem functions within urban green space, yet little is known about the magnitude of change in urban tree DBH, and its potential response to urbanization. Field investigation was used to determine current tree DBH within Urban Function Units (UFUs) in the coastal city Zhanjiang in China. The cover of each UFU was determined via visual interpretation of satellite images. We recorded 12,434 individuals within Zhanjiang green space belonging to 185 species, 137 genera, and 51 families. The dominant DBH range was 5-15 cm, which accounted for 43.72% of the total stems. The DBHs of 33 individuals were larger than 90 cm - 20 of these individuals were Ficus species. The average tree DBH within commercial areas was (32.29 cm ±1.74 cm), which was the highest among all UFU types, and lowest within woodland areas (7.11 cm ± 0.56 cm). Tree DBH was signicantly positively correlated with imperious surface rate, and signicantly negatively correlated with green space surface rate. Variation partitioning analysis showed that impervious surface rate had the highest explanatory power, followed by construction age, then patch density. These three prediction variables, however, only explained 20% of the total observed variation - this suggests that DBH was strongly inuenced by several additional factors.


Introduction
The bene ts of access to local green space are well documented, including, among many others, those related to physical health (Maas et al., 2006), stress recovery (Van den Berg etal., 2007), mental well-being (Fuller et al., 2007), social cohesion (Coley et al., 1997), provision of ecosystem services (Bolund & Hunhammar, 1999), and biodiversity conservation. Urban green spaces (UGSs) are becoming increasingly important in developing countries (Thaiutsa et al., 2008). Trees, especially those of maturity, are considered "key elements" within urban areas Stagoll et al., 2012); trees contribute to various ecosystem functions, such as habitat function (maintenance of biological and genetic diversity) and information function (spiritual enrichment, mental development and leisure) (McKinney, 2002;Dobbs et al., 2011).). Urban trees can also have a stronger effect on carbon budgets than non-urban trees (Dearborn & Kark, 2010), in addition to playing a high impact role in reducing the intensity of urban heat island effects (UHI) (Chow & Roth, 2006). Few studies, however, have assessed urban tree DBH and its in uencing factors -in fact, little is known about the magnitude and dynamics of urban tree DBH. This measure is considered to re ect the age structure of an urban forest, and also infers urban vegetation dynamics. The DBH structure is the most basic stand structure for a forest community, and could re ect the relationship between trees and their habitat. Typically, the greater the DBH, the greater the ecological bene ts associated with trees (e.g., carbon storage, carbon sequestration, and air puri cation), but the maintenance costs for trees with large diameters is often higher than for smaller trees. Both the size and the age of a tree affect characteristics such as the tree mortality. Typically, tree-level data are needed for maintaining the urban tree reserve. To maximize the ecological bene ts of urban green space, the DBH structure of trees should be reasonably con gured. Therefore, we ask two key questions: (1) What is the magnitude of tree DBH in Zhanjiang UGS? (2) Which factors could contribute to changes in urban tree DBH? We expect that trees with smaller DBHs would be more abundant than trees with larger DBHs, due to rapid urbanization and increased attention towards urban tree planting in recent decades.

Study area
Zhanjiang has a subtropical oceanic monsoon climate; the summer season is hot and wet, with the mean daily maximum air temperature often exceeding 34°C in July and August. Thunderstorms and heavy rain occur frequently from May to October, during which typhoons (tropical cyclones) may take place. The relatively cool and dry winter runs from December to February, with temperature occasionally dropping below 10°C.
Our study area was located on the Eastern side of Zhanjiang (Fig. 1).

Field investigation
We divided a Google Earth image of Zhanjiang's main urban area into 273 grids with an edge length of 650 m. In each grid, we choose one UFU within which to carry out our eld work. The eld work was carried out from June to August, 2017. The de nition of UFU used in this study was based on Wang & López-Pujol, (2015). A UFU has distinct boundaries, and the management and maintenance of green space within them was typically consistent. In each UFU, we set up three 20 m 2 plots.
All woody plants with a DBH ≥ 2 cm were inventoried within each plot. DBH was measured at 1.30 m from the ground using calipers or a diameter measuring tape. Woody plants with forks below 1.3 m were considered multi-stemmed; for trees with multiple stems, the aggregate DBH of individual stems was obtained from the square root of the sum of the squared DBHs of individual stems (Mauro et al., 2019).
All tree species present and the number of individuals of each were recorded. All recorded trees were identi ed into species level in the eld with the help of regional oras ( Urban vegetation is subject to biophysics and economic-social factors; with reference to previous work, several in uential factors were selected for study: land cover rate, landscape indices, soil qualities, green space maintenance, UFU type, and the construction age of each UFU (Cilliers & Bredenkamp, 1999 The landscape cover was divided into four types including green surface, wasteland, water body, impervious surface based on visual interpretation of the Systeme Probatoire d'Observation de la Terre (SPOT) image. The landscape indices were conducted using Fragstats 4.2. The 0-10 cm of surface soil samples were collected at each plot by foil sampler. The litters and stones were on the soil surface were removed before we collected soil sample. And the soil qualities were later measured in the lab. Green space maintenance information was acquired through interviews with associates of each UFU. UFU types and the construction age of each UFU were determined via the o cial web site of each UFU, during interviews, and/or through assessments of historical Google Earth image.

Data analysis
A linear regression model was adopted to choose prediction variables for analysis (Table 1), while nonsigni cant (p > 0.05) prediction variables were removed. The response variable was the average DBH of each plot. Greening rate and impervious rate were signi cantly correlated, which means there is multicollinearity between greening rate and impervious rate. Then we removed greening rate and kept impervious rate according to the high p value of impervious rate. Five prediction variables were chosen for further analysis according to linear regression model (Table 1). We implemented an all-subsets regression with the ve prediction variables to select the best subset of predictor variables. We used the regsubsets() function from the leaps package (Lumley, 2017) in R, the model with the highest predictive power was chosen according to adjusted R 2 (Schwarz, 1978). After the best subset of prediction variables was identi ed, variation partitioning analysis was used to identify the relative variations in average DBH in relation to these prediction variables. The average tree DBH in commercial areas was 32.29 cm (± 1.74 cm), which was the highest average among all UFUs, whereas the lowest average DBH recorded (7.11 cm ± 0.56 cm) was in woodland areas (Fig. 3). The average tree DBH within commercial areas was signi cantly higher than within vacant land, farmland, and woodland.
Tree DBH was signi cantly positively associated with the construction age of each UFU. Urban villages typically had older construction ages, whereas commercial apartments, commercial areas, vacant areas, and woodlands typically had younger construction ages (Fig. 6).
According to the results of an all-subset regression, adjusting R 2 values for construction age, UFUs, and patch density resulted in the highest predictive power for tree DBH (Fig. 7).

The different effects of greening surface and impervious surface on urban tree DBH
Contrary to expectation, urban tree DBH was signi cantly positively associated with impervious surface rate and signi cantly negatively associated with greening area rate. Our results indicate that large trees were usually found growing where there was high impervious surface, such as within urban villages and commercial areas. The largest tree in our investigation was located in urban village. Long-settled regions support more old trees mainly due to the positive effects of human management activities . Small trees, on the other hand, were usually found growing within parks and woodlands, which are some of the largest green space surfaces in Zhanjiang. As mentioned previously, during urbanization, some older, larger trees were protected by the government, so they were maintained over time rather than being removed. Newer, younger trees were increasingly planted in parks as people recognized the important ecological services of urban trees.
Impervious surfaces are generally poor areas for tree growth; trees surrounded by impervious surfaces have much shorter life expectancies than trees grown in parks or natural areas (Watson et al., 2014), and they tend to be more water-stressed due to higher temperatures, less soil moisture availability, and increased transpiration -these conditions and their combined effects can reduce overall tree growth and health (Cregg & Dix, 2001). Furthermore, impervious surface reduces the potential for future natural regeneration (Nowak & Green eld, 2020). Therefore, maintaining an appropriate impervious surface relative to green space surface is important for preserving urban trees.

Construction age and patch density
Typically, the DBH of urban trees is signi cantly affected by the construction age of green space, as humans allocate more resources towards greening measures, such as tree planting. Our results, however, indicate that the construction age of a UFU did not strongly affect the DBH. While new trees may be planted regularly as construction age increases (e.g., for aesthetic value or other ecosystem services), trees do not reach the full capacity of their natural lifespan due to environmental stress. Similar to urban tree cover, urban tree DBH is constantly changing as a consequence of many natural (e.g., regeneration, growth, storms) and anthropogenic (e.g., development, planting, tree removal) forces (Nowak & Green eld, 2020). Therefore, tree planting may have no lasting impact on the average DBH observed within a UFU when there are other natural and anthropogenic forces at play.
PD is considered an indicator of landscape con guration (Chen et al., 2020) and fragmentation. In our research, we found that greater PD was associated with greater average DBH. Large trees do not necessarily need large areas to grow, as suitable habitats such as abundant water and nutrients, lower pollution are more important. Older UFUs within Zhanjiang registered higher values of PD, and tended to have more green space patches overall -although these were usually small and highly fragmented . Habitat fragmentation resulting from urbanization separates and reduces the available space for tree species, thus, the survival of species that require larger growing spaces becomes less likely (Jian-feng et al., 2005). In the future, cities should consider allocating green spaces of various sizes to better aid the growth of natural species, such as native species Ficus species always needed large space as they have a lot of aerial roots to support the big thick crown.

Conclusion
The construction age, impervious surface rate, and patch density of UFUs are important habitat elements for urban trees, and are signi cantly correlated with urban tree DBH. The adjusted R 2 values of all three factors were low, however, indicating that DBH may be strongly in uenced by several additional factors, both biotic and abiotic, at multiple scales in the urban environment. Further research is needed to identify these factors. The compact development mode may not always militate against the nurturing and preservation of large urban trees; as long as some suitable niches are left by design in the built-up matrix, whether meritorious trees can ll such sites is contingent upon human activities. The results of this study can aid management and policies related to urban trees, speci cally the appropriate abundance of DBG classes, in addition to facilitating discussions regarding improvement to urban forest structure.

Declarations
Funding: This study was supported by the Natural Science Foundation of Guangdong Province, China (Grant No. 2018A030307059).
Con icts of interest: The authors declare no con icts of interest.
Availability of data and material: All raw data will be provided upon request.
Code availability: All codes will be provided upon request. All of the data used in this paper are included as Supplementary Data. The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
The distribution of individual trees sampled within Zhanjiang with DBHs ranging between 2 cm and exceeding 90 cm.  Results of a linear regression model comparing DBH to a) greening rate and b) impervious surface rate.

Figure 6
A scatterplot comparing the construction age and average DBH according to UFU type. All possible subset regression models for tree DBH, the best was the one with the highest adjusted R2 (adjusted R2 = 0.2). CA: construction age; IAR: impervious surface rate; UFUs: Urban function units; MF: maintenance frequency each year; PD: patch density. Results of a variation partitioning analysis. ISR: Impervious surface rate; CA: Construction age; PD: Patch density.