Selection of Water-Saving Plants and Annual Water Consumption Estimation for Garden Green Spaces in Beijing

: Currently, the number of urban garden green spaces (GGSs) being constructed in Beijing is increasing, and their high water resource demands contradict the severe water shortage situation in Beijing that is restricting urban construction and economic development. This has created an urgent need to build water-saving GGSs. This study analyzed and compared the daily/annual water consumption of 79 common plants in Beijing, and low-water-consumption (LWC) trees, shrubs and herbs were selected; additionally, the total annual water consumption (TAWC) of all plants in the built-up areas of all 16 districts in Beijing was calculated according to the result of the eighth general survey of landscaping resources in Beijing. The results are as follows: (1) ﬁfteen LWC tree species were selected from among 25 species, and the average daily water consumption (DWC) was <1.09 kg · m − 2 ; (2) nineteen LWC shrubs were selected from among 35 shrubs, and the average DWC was <1.17 kg · m − 2 ; (3) eleven LWC herbs were selected from among 19 herbs, and the AWC was <460.3 kg · m − 2 ; (4) the TAWC of all trees, shrubs and herb plants in the Beijing GGSs was 1.104 × 10 9 , 0.139 × 10 9 , and 0.16 × 10 9 m 3 , respectively. Based on the above results, it was estimated that the TAWC of all plants in the built-up areas of all 16 districts in Beijing is approximately 1.403 × 10 9 m 3 . These ﬁndings provide a better understanding of the water consumption of GGS plants in cities in semiarid and semihumid climates and can be used to help select LWC greening plants that can reduce water consumption when expanding green areas in cities.


Introduction
Water is recognized as the resource that is most critical for human survival, and it is also an important strategic resource for social and economic development and ecological environmental protection [1]. With intensifying global climate change and explosive population growth, industrialization and urbanization, two-thirds of the world's population (approximately 4.0 × 10 9 people) faces a serious shortage of water resources, nearly half of whom live in China and India [2][3][4]. The total amount of water resources in China is 2796.26 × 10 9 m 3 , accounting for only 5.1% of the total global water resources. Therefore, the per capita water resources are small, at 2039.25 m 3 , accounting for only 25% of the global per capita water resources [5]. The water resource shortages in many Chinese cities and Makkink equations [22]. The research objects of this method are mostly uniform grassland and farmland crops [23], and a large deviation may occur if it is adopted for the urban GGSs studied in this paper. Remote sensing is the most economical and effective estimation method for performing a large-scale study of transpiration [24,25]. However, urban GGSs have small and fragmented landscape patterns. The classification results from low-and medium-resolution remote sensing images cannot reflect the heterogeneity of plants inside these GGSs, as high-resolution remote sensing images are not accurate at the point and quadrat scales [25]. The data in this study come from a general survey that not only includes an urban green garden space resource survey and determines the number of urban GGS plants in Beijing, but also measures the detailed growth indicators of each plant; thus, the estimates have higher discrimination and greater accuracy. Many studies have used general survey data to estimate the water consumption of GGS plants. In 1998, Chen et al. studied the green level of the 37 most commonly used and representative garden plants in the eight districts of Beijing and estimated that the combined annual water consumption of the green areas of the eight districts of Beijing was 439 million tons by measuring the annual water demand of the leaf area [26]. In 2004, Zhang et al. used the green amount method to calculate the annual water consumption of common trees and shrubs in Beijing with a breast diameter of 15 cm [27]. Che expanded the time and space scales according to the water consumption rules of 35 common garden plants in Beijing and estimated that the annual water consumption of the main vegetation in eight districts of Beijing was 389 million tons [28]. In 2015, a census was again conducted in Beijing. Compared to the previous 10 years, i.e., since 2005, the urban garden green space construction scale had rapidly increased. At the same time, the built-up area of Beijing surveyed also expanded from the previous eight districts to 16 districts. Therefore, calculating the current water consumption of urban greening tree species is crucial for the management and full utilization of Beijing's urban water resources.
In summary, this study will compile relevant research results, analyze and compare the water consumption amounts of common green plants in Beijing, and identify excellent water-saving plants such as native and perennial plants that are suitable for the Beijing urban scape. Moreover, this study will estimate the TAWC of trees, shrubs and herbs of the GGSs in the built-up areas of 16 districts in Beijing based on the results of the eighth landscaping resources survey. The findings will also provide a better understanding of the amount of water GGS plants in semiarid and semihumid climate cities consume and help to select LWC greening plants that can reduce water consumption when expanding green areas in cities.

Selecting the Water-Saving Plants to Use in the Green Areas of Beijing
By consulting the relevant literature (in Tables A1 and A2 and Chapters 3.1.3), the water consumption levels of the main plants in Beijing, including trees, shrubs and herbs, are compared and summarized. In the literature, different plants' water consumption amounts are calculated via direct reference or according to equations, such as Equation (1) [15]. Although the studies may differ in their water consumption amounts for the same plant, the average value was obtained from the different research results, and the DWC of a plant was then obtained. Finally, the water consumption amounts of 79 plants (25 trees, 35 shrubs and 19 herbs) were determined, where the water consumption for trees and shrubs was the DWC per unit of leaf area, and the water consumption for herbs was the AWC per area [15,28]. Sorting plants according to the water consumption amount revealed differences in the water consumption capacities of various plants to identify LWC plants. where E is the transpiration water consumption of the plant throughout the day, g; 18 is the molar mass of water; e i is the instantaneous transpiration rate of the initial measurement point, mmol·m −2 ·s −1 ; e i+1 is the instantaneous transpiration rate of the next measurement point, mmol·m −2 ·s −1 ; t i is the measurement time of the initial measurement point, h; t i+1 is the measurement time of the next measurement point, h; j is the number of measurements; 3600 converts h to s; and 1000 converts mol to mmol. The water consumption results of certain plants were measured as the AWC per unit of leaf area, and these data were converted into the DWC using Equation (2) [28].
where W d (g·m −2 ) is the DWC per unit of leaf area, W a (g·m −2 ) is the AWC per unit of leaf area, and the AWC is calculated over 190 days from the beginning of April to the end of October.

Estimation of the AWC of the Trees, Shrubs and Herbs in the GGSs in Beijing
The estimated water consumption data were based on the results of the eighth general survey of landscaping resources in Beijing (2013-2015) (which is referred to as the general survey hereafter) [29]. The general survey included two items, i.e., the forest management inventory and the urban GGS resources survey, that were organized by the Beijing Municipal Landscaping Bureau. This general survey was prepared in 2013, and the field survey began in May 2014, and ended in December 2014. A total of 381 investigation teams were formed in built-up areas of 16 districts and counties in the city, with 2403 technicians and nearly 10,000 nonprofessionals directly or indirectly involved in the investigation. These research data mainly came from the survey of urban GGS resources, which is the basic work used to comprehensively and accurately grasp the development status of urban GGSs in Beijing. The method uses sub-compartments of GGSs as the survey units, including green park management units, such as parks and nurseries, or district/county-level administrative areas. In this study, 1,448,321 pieces of data from the urban GGS resource survey were obtained, and calculations and analyses were conducted according to different tree, shrub and herb species.
According to a statistical analysis of the general survey, the spatial scale of plant water consumption was expanded. Through diameter at breast height (DBH) daily water consumption and DBH-leaf area models, the water consumption of 13 tree species was extended to different diameter classes. For coniferous trees, the water consumption of the individual trees in this class of trees was the average water consumption for the individual trees of P. tabulaeformis and P. orientalis, and the TAWC was estimated in the same manner. The leaf area of individual trees and the AWC per of unit leaf area of broad-leaved trees were calculated as the average for 8 tree species, such as Ailanthus altissima, and the TAWC of broad-leaved trees was estimated in the same way. Shrubs were considered according to their leaf area index and green surface area and were mainly divided into the following four categories: deciduous shrubs, hedges, shrub balls, and evergreen coniferous shrubs. The water consumption of herbs was calculated based on the herb area and AWC per area. Finally, the total water consumption of the trees, shrubs and herbs in the GGSs in Beijing was determined. The GGSs in Beijing were mainly found in the built-up areas of 16 districts, including Haidian, Chaoyang, Xicheng, Dongcheng, Fengtai, Shijingshan, Mentougou, Fangshan, Shunyi, Changping, Tongzhou, Daxing, Huairou, Pinggu, Miyun and Yanqing Districts, with a total green area of 83,501.3 hm 2 , accounting for 5.09% of the total Beijing city area [10].

Tree Water Consumption Scale Expansion Method
The water consumption amounts of individual trees of different diameters of the five species of P. tabulaeformis, P. orientalis, Acer truncatum, Ginkgo biloba and Robinia pseudoacacia was calculated using diameter-water consumption models depending on the different diameter classes, as listed in Table 1 [15]. The model in Table 1 directly measures the sap flow rate of the plant sapwood through thermal dissipation sapwood sap flow velocity probe (TDP) technology, and then obtains the transpiration water consumption of the tested single plant. The experimental sites of P. tabulaeformis, P. orientalis and A. truncatum are in the Beijing Botanical Garden. The G. biloba and R. pseudoacacia experimental sites are on the campus of Beijing Forestry University. The above tree species were continuously observed for two years. Both experimental sites are within the scope of the Beijing urban garden green space (GGS) studied in this paper and are representative. The resulting model is suitable for estimating the water consumption amounts of these trees in the Beijing urban GGS in this study. Table 1. Diameter-water consumption models of trees.

Diameter-Water Consumption Models Correlation Coefficient
Pinus tabulaeformis W = A·V·T/1000 = 0.594D − 0. The trunk sap flow is generally measured from the beginning of April to the end of October. P. orientalis experiences normal sap flow throughout April, and Sophora japonica experiences normal sap flow only until 1 May, while the sap flow in the other tree species occurs around 20 April. Therefore, when calculating the AWC, that of P. orientalis was calculated over one month, i.e., April, while that of S. japonica was not determined in April, and that of the other tree species was calculated over 10 days in April [15].
Scale expansion of the water consumption of broad-leaved trees refers to the regression model of the DBH and the leaf area proposed by Chen et al. [26], as summarized in Table 2. In this model, the leaf area of different individual trees was calculated based on the DBH, and the water consumption of the entire tree was then computed based on its water consumption per unit of leaf area and its leaf area. The models in Tables 2 and 3 are derived from Chen Zixin's (1998) research [26]. In 1998, Chen Zixin conducted a large number of field measurements on the green quantity of the 37 most commonly used and representative plants in urban garden green spaces in Beijing's eight districts (the number of plants accounted for 81% of the city's total, including 15 trees, 17 shrubs and 5 herbs), for a total of 870,000 plants. According to the correlation between the leaf area of different plants and the diameter at breast height, crown height or crown width, a regression model for calculating the individual leaf area of different plants was established. Therefore, the research area of the relevant model is within the GGSs in the built-up areas of the 16 districts in Beijing in this study, and it is also applicable for estimating the water consumption of related plants in the Beijing urban GGS estimated in this study. According to Equation (3) [28], the water consumption amounts for different diameter classes of each tree species were calculated, and the total water consumption amounts of all diameter classes of each tree species were summed according to Equation (4) [28].
where Q is the AWC of the individual trees in diameter class I, Q i is the AWC of all trees in diameter class i, and N is the number of all the trees in this diameter class.
where Qn is the total water consumption of all diameter classes of a certain tree species, n is the number of diameter classes of a certain tree species, and Q i is the AWC of all trees in diameter class i.

Shrub Water Consumption Scale Expansion Method
To estimate the AWC of the shrubs in the GGSs in Beijing, the shrub water consumption per unit of leaf area was spatially expanded, and the key point was to calculate the leaf area [30]. In urban GGSs, the main shrubs were single plants, hedges, shrub balls, etc., and they were also divided into the categories of deciduous and evergreen shrubs. Therefore, this paper mainly divided shrubs into deciduous shrubs, shrub hedges, shrub balls, and evergreen coniferous shrubs and then estimated the water consumption of all shrubs.
(1) Deciduous shrubs The water consumption expansion method of deciduous shrubs was similar to that of certain broad-leaved trees, and the water consumption scale was also expanded by the leaf area model of Chen et al. (1998) [26], as summarized in Table 3. According to the model, the leaf area of an individual shrub was calculated, and the annual shrub water consumption per-area was then calculated by using Equation (5) [28], while the TAWC of all deciduous shrubs was determined with Equation (6) [28].
where Q (kg·m −2 ) is the AWC per unit of leaf area, and Q d (kg·m −2 ) is the DWC per unit of leaf area. The number of days in the AWC calculation is 190 days based on the DWC time scale expansion. where Q t (m 3 ) is the TAWC of all shrubs, Q (kg·m −2 ) is the AWC per unit of leaf area of the shrub, S d (m 2 ) is the plant leaf area of the shrub, calculated with the model in Table 3, and N is the actual number of deciduous shrubs, obtained from the general survey in Beijing.
(2) Hedges The method for calculating the green surface area of shrub hedges usually follows the cuboid surface area calculation equation, and the green surface area is calculated according to the leaf area index [31]. The product of the green area and leaf area index of the green surface is the total hedge leaf area. The green surface area is the sum of the upper surface area of the hedge and the two side surface areas. The leaf area index is based on the results of Ma et al. [15]. The green surface area and total leaf area of the hedge were calculated with Equations (7) and (8), respectively [28]. Then, the TAWC of the hedge was calculated with Equation (9) [28].
where S s (m 2 ) is the green surface area of the hedge, a is the hedge height, b is the hedge width, and S (m 2 ) is the area covered by the hedge. The hedge height is generally 0.6 m, and the width is 0.8 m.
where S t (m 2 ) is the total leaf area of the shrub hedge, S s (m 2 ) is the area of the actual green surface of the hedge, and K is the leaf area index of the hedge.
where W (m 3 ) is the TAWC of each hedge, S s (m 2 ) is the green surface area of the hedge, K is the leaf area index of the hedge, and W a is the AWC per leaf area of the hedge. (

3) Shrub balls
The water consumption of the shrub green ball was calculated according to its water consumption per unit of leaf area, leaf area index and surface area, as expressed in Equation (10) [28].
where W (kg) is the water consumption of a single shrub green ball, W s (g·m −2 ) is the water consumption per unit of leaf area of the shrub green ball, S (m −2 ) is the surface area of the shrub green ball, calculated as a hemispherical area, and K is the leaf area index of the green ball, based on the results of Ma et al. (2009) [12] and Che (2008) [16]. The water consumption Q of each green ball was calculated according to Equation (11) [28]. The shrub green ball diameter is 1 m.
where Q (m 3 ) is the total water consumption of all shrub balls, W (kg) is the water consumption of a shrub green ball, and n is the number of shrub balls.

(4) Evergreen coniferous shrubs
The coniferous shrubs in the GGSs in Beijing are mainly Sabina procumbens and Sabina vulgaris. The method for calculating the water consumption of these shrubs entails obtaining the total leaf area from the green surface area and leaf area index. The leaf area index was obtained from the results of Ma et al. (2009) [15], and the total water consumption Q was calculated using Equation (12) [28].
where Q (m 3 ) is the total water consumption of each shrub, S (m 2 ) is the actual shrub area, K is the leaf area index of a shrub, and W (kg·m −2 ) is the shrub water consumption per unit of leaf area.

Herb Water Consumption Scale Expansion Method
The AWC of herbs was calculated from the herb area and water consumption per-area, as expressed in Equation (13) [28].
where Q t (m 3 ) is the TAWC of each herb, Q d (kg·m −2 ) is the AWC per unit of herb, and S t (m 2 ) is the herb area. The above research structure diagram is shown in Figure 1.
was obtained from the results of Ma et al. (2009) [15], and the total water consumption Q was calculated using Equation (12) [28].
where Q (m 3 ) is the total water consumption of each shrub, S (m 2 ) is the actual shrub area, K is the leaf area index of a shrub, and W (kg·m −2 ) is the shrub water consumption per unit of leaf area.

Herb Water Consumption Scale Expansion Method
The AWC of herbs was calculated from the herb area and water consumption perarea, as expressed in Equation (13) [28].
where Qt (m 3 ) is the TAWC of each herb, Qd (kg·m −2 ) is the AWC per unit of herb, and St (m 2 ) is the herb area. The above research structure diagram is shown in Figure 1.

Selecting the Water-Saving Plants
Through a comprehensive comparative analysis of the literature, the water consumption levels of 79 plant species were calculated, where those of trees and shrubs are the DWC per unit of leaf area (kg·m −2 ·d −1 ) and those of herbs are the AWC per area (kg·m −2 ·a −1 ).

Selecting LWC Trees
Twenty-five types of common trees in the GGSs in Beijing were selected as research objects for the DWC unit leaf area comparison (Appendix A, Table A1).

Selecting the LWC Shrubs
According to the literature, the DWCs of 35 shrubs common to Beijing GGSs were selected (Appendix A, Table A2).
The DWC range of the 35 shrubs was 0.27-2.42 kg·m −2 , and the average DWC was 1.17 kg·m −2 . Nineteen shrub species had a lower than average DWC and were as follows in descending order:  [15,19,[32][33][34], the AWC per unit of green space area of 19 common herbs in Beijing are summarized (Appendix A: Table A3).

Diameter Class Distribution of the Trees
According to the general survey in Beijing, there are 98 main tree species, including 13 coniferous species, and 85 broad-leaved species, and the total number of trees is 46,948,967. Based on the model, this paper estimates the water consumption levels of the two coniferous species (Table 4) and 11 broad-leaved species (Table 5), and estimates the water consumption levels of other coniferous and broad-leaved trees through the above tree species  According to the data analysis, the diameter class distribution of coniferous trees is summarized in Table 4. There were 8,896,082 coniferous trees, accounting for 18.95% of the total trees, and P. tabulaeformis accounted for 26.80% of the total coniferous trees. The most widely distributed diameter class is 6-12 cm, accounting for 57.57% of the total P. tabulaeformis, and the number of trees with diameters above 38 cm is relatively small, accounting for only 0.40%. P. orientalis accounted for 37.27% of the total coniferous trees, which was 10.47% more than the number of P. tabulaeformis. The most widely distributed diameter class is 6-12 cm, accounting for 62.96% of the total P. orientalis, and the number of trees with diameters exceeding 38 cm is relatively small, accounting for only 0.39%. The other coniferous trees accounted for 35.93% of the total number of coniferous trees.
The diameter class distribution of the broad-leaved trees is summarized in Table 5, indicating that the total number of broad-leaved trees in Beijing is 38,052,885, accounting for 81.05% of the total number of trees. The diameter class distribution of the different broad-leaved tree species varies; 63.11% of the total broad-leaved trees are mainly in the 6 to 24 cm diameter class. The numbers of broad-leaved trees in the 26 to 36 cm and larger than 38 cm diameter classes were relatively small, accounting for 6.72% and 2.00%, respectively, of all trees. The number of S. japonica trees was the largest, accounting for 9.22% of the total broad-leaved trees, followed by P. tomentosa (5.38%), R. pseudoacacia (4.76%), G. biloba (4.45%), F. chinensis (4.39%), A. truncatum (2.40%), K. paniculata (1.33%), P. acerifolia (1.01%), A. altissima (0.86%), S. xaureo-pendula (0.85%) and Paulownia (0.19%). These 11 broad-leaved trees accounted for 34.85% of the total broad-leaved trees, while the other broad-leaved trees accounted for 65.15%.

AWC of the Trees of the Different Diameter Classes
According to the DBH-leaf area regression model in Table 2, the individual tree leaf areas of eight broad-leaved trees, such as K. paniculata, A. altissima, and F. chinensis, were calculated for the different diameter classes (Table 6). Then, the AWC unit leaf area (Table 7) was multiplied by the individual tree leaf area to obtain the AWC of the broadleaf trees of the different diameter classes (Table 8).   According to the DBH daily water consumption model in Table 1, the individual tree AWCs of the different diameter classes of P. tabulaeformis, G. biloba, A. truncatum, P. orientalis and R. pseudoacacia were calculated ( Table 8).
The general survey divided the tree DBH into five diameter classes. The models in Tables 1 and 2 are not suitable for trees with DBHs smaller than 5 cm, but in the general survey, there were 1,528,078 coniferous trees with DBHs smaller than 5 cm, accounting for 17.17% of the total coniferous trees, and 10,719,849 broad-leaved trees with DBHs smaller than 5 cm, accounting for 28.17% of the total broad-leaved trees. Therefore, the AWC of a single tree with a DBH smaller than 5 cm is calculated as 60% of the individual tree AWC of the 10 cm diameter class (Table 8). Table 6 indicates that the individual tree leaf areas of the different diameter classes vary greatly. In the same diameter class, the individual tree leaf area of P. acerifolia was much larger than that of the other broad-leaved trees. Among them, the individual tree leaf area of P. acerifolia was 1.36 times that of the other broad-leaved trees in the 10 cm diameter class. The leaf area of the different broad-leaved trees was positively correlated with the diameter class.
In Table 8, the individual tree AWC is positively correlated with the tree diameter class. For the coniferous trees, the individual tree average AWC of P. orientalis was 34.54% higher than that of P. tabulaeformis, but the DWC per unit of leaf area of P. orientalis was 34.76% lower than that of P. tabulaeformis (Appendix A, Table A1). This is due to the sap flow fluctuation of P. orientalis that occurred in April, while that of P. tabulaeformis occurred over only 10 days in April [15]. Moreover, in the same diameter class, the leaf area of P. orientalis was 1.4-1.9 times that of P. tabulaeformis [35].
For the broad-leaved trees, the individual tree average AWC of F. chinensis in the average-diameter class was the highest (103,203.6 kg), which was 61.29 times that of A. truncatum, with the lowest individual tree average AWC in the average-diameter class. The individual tree average AWC of broad-leaved trees is 79.57 times that of coniferous trees and that of broad-leaved trees is considerably higher than that of coniferous trees [17].

AWC of the Trees in the GGSs in Beijing
Based on the above calculations, the AWC of all trees in the Beijing GGSs includes the AWC of the 13 tree species and other coniferous and broad-leaved trees, as listed in Table 9. The AWC of the coniferous trees in Beijing is 4,187,204.86 m 3 (0.004 × 10 9 m 3 ), and the AWC of broad-leaved trees is 10,954,391.29 m 3 (1.100 × 10 9 m 3 ). The TAWC of all trees in the Beijing GGSs is 1,103,736,321.14 m 3 (1.104 × 10 9 m 3 ). The TAWC of P. orientalis among the coniferous trees was 79.53% higher than that of P. tabulaeformis, and the TAWC of the broad-leaved trees was 262.60 times that of the coniferous trees. The AWC of the other broad-leaved and coniferous trees accounted for 72.58% of the AWC of all trees.

Estimated AWC of the Shrubs in the GGSs in Beijing
According to the general survey, the total number of shrubs in the Beijing GGSs is 99,324,466, and the hedge area covers 14,876,207.5 m 2 . There were 48 main species of deciduous shrubs, with a total number of 59,222,441, accounting for 59.63% of all shrubs in the GGSs. There were 22,150,424 shrub balls, accounting for 22.30% of all shrubs, mainly including E. japonicus and B. microphylla, and 17,951,601 coniferous shrubs, accounting for 18.07% of all shrubs, mainly including S. procumbens and S. vulgaris. The AWC of the shrubs in the Beijing GGSs was estimated based on the regression model of the deciduous shrub leaf area in Table 3 and Equations (5)-(12).

AWC of the Deciduous Shrubs in the GGSs in Beijing
The AWC of the 11 common deciduous shrubs (Table 10) was estimated according to the regression model of the deciduous shrub leaf area in Table 3 and Equations (5) and (6). The individual tree AWC and leaf area of the other deciduous shrubs were estimated based on the averages of these 11 shrubs. The height and crown width of most Rosa shrubs are both 0.6-0.7 m, and the height and crown width of the other 10 deciduous shrub species are 1-2 m and 1.2-1.8 m, respectively [28]. Therefore, the height and crown width of the Rosa cultivar Floribunda were both calculated as 0.6 m, while the height of the other 10 deciduous shrubs was set to 2 m, and the crown width was regarded as 1.5 m. The specific calculation results are listed in Table 10.
The calculation shows that the total number of the 11 types of deciduous shrubs is 19,984,453, accounting for 33.74% of the total deciduous shrubs. The most abundant type was R. cultivar Floribunda, which accounted for 13.41% of the total deciduous shrubs, the least abundant type was C. chinensis, accounting for only 0.34% of the total deciduous shrubs, and the other deciduous shrubs accounted for 66.26%.
The TAWC of all deciduous shrubs in the Beijing GGSs is 60,930,420.93 m 3 (0.061 × 10 9 m 3 ). The AWC of the 11 deciduous shrubs was 11,992,841.54 m 3 (0.012 × 10 9 m 3 ), account-ing for 19.68% of the TAWC of all deciduous shrubs. The AWC of the other deciduous shrubs was 48,937,579.40 m 3 (0.049 × 10 9 m 3 ), accounting for 80.32% of the TAWC of all deciduous shrubs.

AWC of the Hedges in the GGSs in Beijing
According to the general survey, the shrub hedges in Beijing mainly include P. orientalis, S. chinensis, J. rigida, E. japonicus, B. microphylla, L. vicaryi and B. thunbergii, with a total area of 14,876,207.50 m 2 . The total area of E. japonicus, B. microphylla, L. lucidum and B. purpurea was 12,585,410 m 2 , accounting for 84.60% of the total hedge area. Therefore, the hedge green surface area and total leaf area are calculated with Equations (7) and (8), respectively, and the shrub hedge AWC per unit of leaf area follows the result of Ma et al. [15]. Finally, with the hedge AWC per unit of leaf area and total leaf area, the shrub hedge TAWC can be calculated with Equation (9), as summarized in Table 11.  Table 11 reveals that the TAWC of the four major hedges in the Beijing GGSs, E. japonicus, B. microphylla, L. lucidum and B. purpurea, is 30,960,653.55 m 3 (0.031 × 10 9 m 3 ). Since the total area of these four hedges accounts for 84.60% of the total hedge area, the water consumption of these four hedge shrubs is adopted to estimate the water consumption of all hedges in proportion to their area. Finally, the AWC of all shrub hedges in the Beijing GGSs is 36,580,216.95 m 3 (0.037 × 10 9 m 3 ).

AWC of the Shrub Balls in the GGSs in Beijing
There were 22,150,424 shrub balls in the Beijing GGSs, which mainly consisted of 16,548,928 E. japonicus shrub balls and 3,375,475 B. microphylla shrub balls, accounting for 89.95% of the total shrub balls. Hence, according to Equations (10) and (11), the total water consumption for E. japonicus and B. microphylla in the Beijing GGSs was 33,902,133.90 and 15,479.64 m 3 , respectively. Finally, based on the ratio of the number of these two shrub balls to the total shrub balls in Beijing, the total water consumption of all shrub balls in Beijing was estimated, and the green ball AWC in Beijing is 39,766,926.14 m 3 (0.0398 × 10 9 m 3 ), as indicated in Table 12.

AWC of the Evergreen Coniferous Shrubs in the GGSs in Beijing
The evergreen coniferous shrubs in Beijing mainly include S. procumbens, S. vulgaris, S. chinensis, and C. macrolepis. According to the general survey data (Table 13), S. procumbens and Sabina vulgaris account for 95.84% of the total evergreen coniferous shrubs. Therefore, the AWC of the evergreen coniferous shrubs in the Beijing GGSs is primarily estimated based on these two evergreen coniferous shrubs. Through a large number of field investigations, it was found that evergreen coniferous shrubs are mostly planted in clusters in GGSs, and the average planting density of these two shrubs is three plants·m −2 [16]. Hence, the actual areas of S. procumbens and S. vulgaris were calculated to be 674,951.7 and 5,060,127 m 2 , respectively. Moreover, the AWC per unit of leaf area was calculated according to the DWC of shrubs (Appendix A, Table A1), and the AWC period was 190 days. According to Equation (12), the AWC of S. procumbens was 230,972.74 m 3 , and the AWC of S. vulgaris was 932,485.20 m 3 . Finally, according to the percentages of S. procumbens and S. vulgaris in the total number of evergreen coniferous shrubs, the total water consumption of the Beijing GGS evergreen coniferous shrubs was 1,199,346.75 m 3 (0.0012 × 10 9 m 3 ).

Estimated AWC of the Herbs in the GGSs in Beijing
Herbs are divided into ground cover plants and lawns. The common ground cover plants in Beijing include I. tectorum, I. lacteal var. chinensis, Sedum, and H. fulva. The lawns include warm-and cold-season lawns. The warm-season lawns include B. dactyloides and Z. japonica, and the cold-season lawns include F. elata, P. pratensis and L. perenne.
According to Equation 13, the herb TAWC is provided in Table 14. The water consumption per area of the other warm-season lawns is the average value of B. dactyloides and Z. japonica, and the water consumption per area of the other cold-season lawns is the average value of F. elata, P. pratensis and L. perenne. The water consumption per area of the other ground cover plants is the average value of I. tectorum, I. lacteal var. chinensis, Sedum, and H. fulva. According to calculations, the lawn TAWC in the Beijing GGSs was 146,851,689.20 (0.147 × 10 9 ) m 3 , and the ground cover plant TAWC was 13,484,675.65 (0.013 × 10 9 ) m 3 . Therefore, the herb TAWC in Beijing was 160,336,364.85 (0.16 × 10 9 ) m 3 .

TAWC of the Trees, Shrubs and Herbs in the Beijing GGSs
According to the calculation of the AWC of the Beijing GGS trees, shrubs and herbs based on Tables 9-14, the TAWC of all plants was 1,402,549,596.76 (1.403 × 10 9 ) m 3 . The tree TAWC was 1,103,736,321.14 (1.104 × 10 9 ) m 3 , accounting for 78.69%. The shrub TAWC was 138,476,910.77 (0.138 × 10 9 ) m 3 , accounting for 9.83%. The herb TAWC was 160,336,364.85 (0.16 × 10 9 ) m 3 , accounting for 11.40%. The tree total water consumption in Beijing was 3.70 times the total water consumption of the shrubs and herbs, making trees the main water-consuming species.
Among the GGSs in the built-up areas of 16 districts of Beijing (Table 15), the AWC of the trees, shrubs and herbs in Chaoyang District is the highest, at 335,789,127.76 m 3 , or approximately 0.336 × 10 9 m 3 , accounting for 23.94% of the plant TAWC in the Beijing GGSs, and the lowest occurs in Miyun District, at 0.016 × 10 9 m 3 , accounting for only 1.14% of TAWC. The TAWC in the remaining districts are shown in Table 15.

Discussion
In this study, according to the transpiration water consumption law of common GGS plants in Beijing, the water consumption capacities of plants were evaluated, which objectively reflect the different water consumption capacities of plants. In the process of selecting water-saving plants, by comparing the water consumption levels of the various plants, 15 species of water-saving trees, 19 species of water-saving shrubs and 11 species of water-saving herbs were finally selected, which can guide the construction of water-saving GGSs in Beijing.
GGS plants can not only beautify the environment but also provide multiple ecological benefits, such as storm runoff regulation, pollutant absorption, solar energy consumption, cool ambient environments and alleviation of the urban heat island effect [36][37][38]. For example, P. orientalis, R. pseudoacacia, U. pumila and A. altissima have strong anti-pollution abilities and have strong abilities to absorb urban car exhaust emissions [39]. L. vicaryi, S. japonicus, G. biloba, etc. have a strong ability to absorb smoke and dust in the urban environment [40]. R. pseudoacacia, P. tabulaeformis, P. tomentosa, etc. have strong resistance to or capacities to absorb heavy metals in urban areas [41]. Additionally, C. chinensis, M. denudate, S. oblata, B. thunbergii var. atropurpurea, C. coggygria, etc. have high esthetic value [39]. Not all of the above tree species are water-saving plants. Furthermore, it should be noted that different plant configuration patterns will have a greater impact on plant water consumption. For example, for the same shrub, the water consumption of the green ball configuration is higher than that of the block configuration, and it is better to arrange shrubs in blocks in garden green spaces [42]. Under the same conditions, compared to composite and single configurations of shrubs, the transpiration rate would be reduced by 30-40% [43]. In a configuration with many types of plants, matching plants with different seasonal water consumption levels can balance the water consumption [44]. Therefore, in the subsequent construction of water-saving gardens and green spaces in Beijing, we must not simply consider the characteristics of the plants themselves, but also the social and ecological benefits of these plants. In particular, we should choose plants that are water-saving, meet people's esthetic requirements and improve the environment. It is also necessary to make a reasonable selection according to the characteristics of the water-saving plants selected, which is of great significance for improving the water use efficiency of Beijing, improving the urban ecological environment, and improving the quality of life of residents.
This study finally concludes that the total annual transpiration water consumption of all trees, shrubs and herbs in the built-up areas of all 16 districts in Beijing GGSs was 1.403 × 10 9 m 3 , accounting for 52.4% of Beijing's total water resources (the total amount of surface and groundwater formed by precipitation in Beijing is 2.676 × 10 9 m 3 ) in 2015 and 36.7% of the total water consumption (3.82 × 10 9 m 3 ) (BWRB, 2015). The water consumption of plants in Beijing's urban GGSs was very high. In this study, the total water consumption of trees, shrubs and herbs in Haidian District, Dongcheng District (including Xuanwu District), Xicheng District (including Chongwen District), Chaoyang District, Fengtai District and Shijingshan District equals 0.644 × 10 9 m 3 , which is 65.56% higher than the level of 0.389 × 10 9 m 3 calculated by Che [28] through the fifth general survey of landscaping resources in Beijing. In addition, the highest level of water consumption in 2005 was from 11,361,219 trees, with a water consumption of 0.33 × 10 9 m 3 , accounting for 84.8% of the total water consumption. By 2015, the number of trees was 18,960,002 in the same districts, an increase of 66.88% compared to that in 2005, and the water consumption was 0.53 × 10 9 m 3 . Compared to 2005, the water consumption increased 55.45% and accounted for 79.66% of the total water consumption in 2015. The above results show that the conclusion of this study is reliable and that the choice of garden-greening plant species in Beijing in recent years tends to be water-saving plant species, and the relative increase in water consumption has slowed. Sun et al. calculated the actual AWC of 16 districts in Beijing in 2012 to be 0.809 × 10 9 m 3 , which is 42.33% lower than that in this study. This is mainly due to the rapid economic development of Beijing in recent years [45]. An enormous number of trees have been newly planted, and the water consumption of the Beijing GGS coniferous trees and ground cover plants was not determined at that time, while the shrubs were not divided into shrub balls and hedges. Therefore, there are differences between previous results and the conclusions of this study.
In current estimation methods of plant transpiration water consumption, the difficulties mainly concern scaling up from a single plant to a stand. Normally, the transpiration water consumption of a stand can be obtained by scaling up the relationships between the sap flow and DBH, stem cross-sectional area, sapwood area or leaf area coefficient [46], vegetation density [47], and single-tree area [48]. Relevant research shows that the accuracy of transpiration water consumption scale conversion using the leaf area and DBH is the highest [49]. In this study, tree transpiration water consumption was expanded by the DBH daily water consumption and DBH-leaf area models. In urban gardens, most trees are planted individually, in rows or in pieces, the spacing between the rows is generally relatively large, and most of them are in the form of sparse forest. Therefore, when the scale of the water consumption model is extended, it is not necessary to extend the water consumption per plant to the stand, but only to extend the water consumption per plant to individual plants of different diameter classes [35]. Additionally, the shrub transpiration water consumption was mainly expanded by the leaf area index and green surface area, while the herb transpiration water consumption was accounted for by the herb area and AWC per area. The plant transpiration water consumption models were all reliably expanded and accurately estimated, and this study attains a higher estimation accuracy.
Although the model in this paper does not include related environmental factors such as climate and soil, the relevant plants are commonly used plants in the Beijing urban GGSs, and they are all measured in the Beijing urban GGSs over a long time span. This is equivalent to considering the water consumption of plants under changes in Beijing's urban environment, so the results of this model can be considered to already include the impact of Beijing's environmental factors. Meanwhile, a large number of experiments have shown that when the soil water content is sufficient, the water consumption per plant is linearly related to the leaf area [50], and the leaf area is a group of scalar quantities with extremely high correlations with the trunk diameter and sapwood area [51]. In addition, there is a self-balance between the leaf area and sapwood area. The trunk can provide sufficient water supply to the leaf area, which in turn affects the cross-sectional area of the trunk. This long-term interrelated balance effect leads to mutual adjustment and adaptation between leaf area and sapwood area, so that the two can maintain a similar water potential gradient in the tree. This relationship is relatively stable for the same tree species and does not change with changes in climate and site conditions [52][53][54]. Since the trunk diameter directly reflects the size of the tree and is also the easiest to observe and there are also many models to calculate leaf area, this study uses the relationship model between the trunk diameter or leaf area and water consumption to predict the water consumption of plants of different diameter classes.
Due to the complex structure of the tree species in urban gardens and the changing environment, it is difficult to estimate the water consumption of GGSs. In this paper, the results of previous water consumption studies are adopted to compare the water consumption of different tree species and to estimate the total water consumption of the trees, shrubs and herbs in Beijing. At present, the research on the scale expansion of plant water consumption is not comprehensive. For the water consumption of many other species, the average value of the same type of species (coniferous trees, broad-leaved trees, deciduous shrubs, herbs, etc.) was adopted, and there was no detailed calculation for each tree species because there were no related models. In addition, errors occur when expanding the water consumption of plants temporally and spatially. Although these errors will not affect the evaluation results of the individual and group water consumption levels of plants, they will affect the design of precision irrigation systems. All these deficiencies need to be addressed and resolved in future research.

Conclusions
In this study, by consulting the relevant literature, the water consumption of 79 common plants in Beijing was assessed, and plants with a low water consumption level were identified. Moreover, the AWC of the trees, shrubs and herbs in the Beijing GGSs were estimated.