The GPR system involves computational models and statistical models of data at three levels: single plant, community, and plot. Specifically, it includes three parts. The first part follows the two-dimensional indicators (green space rate and green space coverage); the second part is an indicator to measure the ecological benefits of green space and greening level (green volume and green volume rate); the third part is the GPR and greening construction that combines green space construction with urban planning and construction index (Zhao & Rao, 2005) .

Through the guidance of GPR system, the scientific plant selection and plant configuration of low ecological efficiency residential green areas can form a closely related residential microenvironmental ecosystem, improve the function of plants in carbon sequestration and oxygen release, dust and bacteria reduction, wind and noise prevention, regulate local microclimate and optimize ecological habitat.

The measurement of GPR is related to the two key indicators of plant leaf area (LA) and LAI, which can mainly reflect the richness of plant community structure, GPR is calculated as formula (1), and the green space GPR can be calculated after knowing either LA or LAI and plant projection area (Li et al., 2019) . In addition, the Chongqing region has introduced the Chongqing Green Building Evaluation Standard (DBJ50/T-066-2020) based on the Chinese Green Building Evaluation Standard, and gives the estimation formula (2) of GPR, and hopes to apply it to all types of urban green space ecological construction evaluation (HAURDC, 2020) .

GPR = LA S = ∑ ( LAI 1 × s 1 + ⋯ + LAI n × s n ) S (1)

In formula (1), LA is the total leaf area of a site, and S is the total area occupied by the site.

GPR = ∑ ( LAI t × S t × n t ) + ∑ ( 3 × S s × n s ) + S g S (2)

In formula (2), LAI_t is the leaf area index of trees, S_t is the canopy projection area of trees, S_s is the canopy projection area of shrubs, n_t is the number of trees, n_s is the number of shrubs, S_g is the footprint of the lawn, S is the total area of the site. In the formula, the leaf area index of canopy sparse class of trees is taken as 2. The leaf area index of canopy dense class of trees is taken as 4. The projected area of trees is calculated according to the seedling table data, and the three-dimensional greening in the site can be included in the calculation. However, the variability of tree LAI is large, and judging tree LAI only by dense or sparse canopy may cause large errors to the calculation results.

LAI was introduced by Watson in the 1940s, and the leaf area index is half of the total green leaf area per unit area of the ground surface (He et al., 2022) , and is often used as an important parameter of vegetation canopy as a state variable in vegetation monitoring and ecosystem cycle models, and is important in quantitative assessment of physio and physical processes such as photosynthetic respiration, transpiration, and precipitation retention in vegetation (Ong, 2003) .

The measurement methods of LAI are mainly divided into direct and indirect methods. The direct measurement method has the most accurate results, but the disadvantages of high workload and time consumption make it extremely unlikely to monitor vegetation LAI spatially and temporally; while the indirect measurement method mainly uses the relationship between LAI and leaf variables (green matter, infrared reflectance) inferred, thus allowing more spatial and temporal LAI values to be obtained, but the LAI accuracy of the indirect method needs to be calibrated by the direct method (Chason et al., 1991) . The hemispheric photography method is the most accurate of the indirect methods and has the advantage of being easy to measure, fast to calculate, and highly accurate, allowing to reach an average error of only 11%. Therefore, this method will be adopted to complete this study.

Chongqing (29.51'02" - 30.22'24"N, 105.58'37" - 106.40'37"E) is in the southwestern China and has a humid subtropical monsoonal climate, which contains rich regional plant resources. A total of 45 families and 87 genera were involved. According to the Urban Green Space Classification Standard (MOC, 2017) and the current status of native plant distribution, the study sampling sites were determined to be residential green spaces (RG) in 22 residential areas within 3 - 10 years of completion in the central urban area, and 15 park green spaces were added to the study area range considering the actual plant data collection for the study (Table 1 and Figure 1).

Native garden plant research sampling was conducted using Digital Hemispherical Photography (DHP), and the sampling time was used during the summer season when the leaf area is the largest. Hemispherical images were acquired using a Nikon CoolPix 4500 digital camera with an external 183˚ wide-angle Nikon FC-E8 fisheye lens. The camera’s built-in shooting parameters were set before image acquisition, and the camera’s built-in “Fisheye 1” mode was selected to save the photos in JPEG image format with 2272 × 1704 resolution and low compression ratio (1:4) to ensure image clarity while reducing image file size and increasing the speed of subsequent software processing speed (Kay et al., 2021) . To set the external parameters of the camera, the camera needs to be fixed on a tripod after a more horizontal alignment, with the camera 1.6 m above the ground, the lens facing upward, and the horizontal direction of the photo is the magnetic north pole (Ke et al., 2011) , and the sampled images are shown in Figure 2. The sampling point was 1/2 of the line from the projected edge line of the tree’s crown to the center of the trunk (Figure 3), and the tree species name, tree height, diameter at breast height, crown width, and sampling point data were recorded. The shooting time was chosen in the early morning or evening when there was a poor contrast between the sky and the branches and leaves caused by direct sunlight.

According to the urban residential area planning and design standards (GB50180-2018) on the residential area green space type division and residential area as-built drawings to determine the type of green space within the residential area, measuring the green space area and taking photos to record the surrounding environment, recording the planting points within the green space, and analyzing the plant configuration method.

The captured hemispherical photos were screened to exclude blurred and overlapping canopy photos, and a total of 2253 valid summer data were obtained. Based on the fact that manual determination of thresholds can affect the calculation results by about 10%, the photo data were processed using the Hemispherical plug-in for edge detection (Figure 4), an automatic thresholding algorithm to separate canopy and sky to improve the accuracy of the calculation results (Montero et al., 2008) , and the GLA (GapLight Analyzer Version 2.0) software for hemispheric photos (Figure 5). The software processing yielded LAI4R (effective leaf area index extracted by integrating the zenith angle between 0˚ and 60˚), LAI5R (effective leaf area index extracted by integrating the zenith angle between 0˚ and 75˚), canopy openness, direct light on the forest, and scattered light on the forest floor to provide a basis for the GPR calculation (Nölke et al., 2015) .

The mean analysis (95% confidence interval) of LAI of regional plants was performed in SPSS 26.0 to form a table of LAI indicators of common native garden plants in Chongqing (Table 2).

To further investigate the relationship between the frequency of use of summer native garden plants and LAI, the regional plants were clustered and analyzed according to LAI1. The LAI was divided into three clusters based on the final clustering center, and the LAI of 113 native garden plants was divided into three clusters based on the cluster members (Table 3), with high LAI plants (LAI ≥ 3.77), medium LAI plants (2.92 ≤ LAI < 3.77), and low LAI plants (LAI < 2.92).

The optimization was carried out on the basis of the following principles:

1) Green development, focus on ecology. First of all, on the basis of respecting the topography of the site, understanding the ecological habits of native and introduced plants, reasonably configuring plants to achieve optimal GPR, maximizing the ecological benefits of residential green areas, increasing the total annual average carbon sink, and enriching biodiversity.

2) Aesthetic perception and seasonal configuration. The planting of each layer of the complex community is sparse and dense, and the space is created freely. The planting of seasonal plants is increased moderately, so that leaves, flowers and fruits can be seen throughout the year, and each scene is different.

3) People-oriented, sustainable development. First of all, we should respect people-oriented needs and choose planting according to the functional attributes of different green areas. For example, green areas next to houses need to ensure privacy for low-rise residents, shade and shelter for road green areas, ventilation and light for centralized green areas, and open sight lines for green areas attached to public service facilities. In addition, the selection of plant species for each type of green space should be arranged according to the ecological habits of plants to reduce plant loss, replanting and later management and maintenance, and to create ecological green spaces close to nature in residential communities.

(Note * indicates that the sample size of the species was too small for 95% confidence interval description; LAI1 indicates the mean of leaf area index with 5% clipping; LAI2 indicates the mean of leaf area index at 95% confidence interval).