Dynamic Responses of Landscape Pattern and Vegetation Coverage to Urban Expansion and Greening: A Case Study of the Severe Cold Region, China

Abstract

Urbanization is the natural trend of human social development, which leads to various changes in vegetation conditions. Analyzing the dynamics of landscape patterns and vegetation coverage in response to urban expansion is important for understanding the ecological influence of urban expansion and guiding sustainable urban development. However, existing studies on the effects of urbanization on vegetation conditions in severe cold regions are limited. Here, taking Harbin, China as an example, the study explored the evolution of the urban expansion process and adopted landscape metrics to derive landscape pattern changes from 2005 to 2020. Based on the fraction of vegetation coverage (FVC) derived from Landsat satellite observations during 2004–2020, we investigated the spatiotemporal change characteristics of FVC. By separating the direct and indirect effects of urbanization on vegetation growth, we quantified the impacts of urbanization on vegetation. The results show that the urban area increased by 70.37%, and urban expansion exhibited a compact sprawl pattern. Cropland and water were the major resources converted for urban expansion. The distribution of FVC exhibits a pattern that the urban fringe region is significantly higher than the central region. The FVC trend is decreased overall, but the changes are not significant with regional variation. Moreover, the average observed FVC decreased with increasing urban intensity. By contrast, the indirect impact is increased along the intensity gradient, with growth enhancement offsetting about 2.26%~2.71% of direct vegetation loss. The study further shows that vegetation growth responses to urbanization vary according to urbanization levels. Our findings provide detailed information and reveal the relationship between urban intensity and vegetation coverage, which could help to manage urban vegetation for planners and stakeholders.

1. Introduction

Globally, urban expansion has progressed rapidly over recent decades, and with population growth and the steady transition of human residence from rural to urban areas, urban areas are expected to absorb approximately 68% of the world’s population by 2050 [1], most of them in developing regions. China, one of the biggest developing nations, has experienced urban expansion at an unparalleled rate and scale against the background of social transformation during the past 60 years [2]. This trend will continue, and China’s urbanization level is estimated to reach nearly 75%, with an urban population of 2.55 billion by 2050 [1,3]. Urbanization has triggered the soaring national economy, and plays an important role in poverty reduction, safety and health, and employment, which is an inevitable requirement for the development of China.

However, urbanization is often accompanied by the expansion of built-up land and the appropriation of physical landscapes (e.g., forest, grassland, cropland), which leads to land use/cover change and landscape fragmentation [4]. Additionally, the continuing population growth of cities has accelerated the consumption and exploitation of natural resources [5] and induced many serious environmental issues, such as increased carbon emissions, air and water quality pollution [6], total climate change [7], and thus, ecosystem service degradation and loss threatening urban sustainable development [8,9]. These issues have had complicated impacts on the modernization development of China, have attracted the attention of experts and scholars, and generated demands for urban greenery to abate environmental problems and improve the quality of urban life.

Extensive studies have shown that vegetation in cities is critically important because it provides a wide diversity of ecosystem services that benefit people, for example, mitigating the impact of urban heat islands [10], reducing noise [11], increasing terrestrial carbon sequestration and storage [12], reducing air pollution [13] and stormwater runoff-induced surface erosion [14], and strengthening resistance to natural disasters (e.g., flood control) [15]. Moreover, urban green spaces provide significant well-being benefits to local urban residents, including reducing stress, improving cognitive ability and physical and mental health, and increasing property value [16]. The enhancement of city green space to improve environmental quality has received increased attention from urban management and planning departments, and a series of policies have been implemented, which resulted in highly dynamic and patchy urban green spaces [17]. In the face of increasing and rapid environmental change in urban areas, land managers and policymakers urgently need a comprehensive understanding of urban green space change and urban expansion processes in order to make informed decisions. Therefore, quantifying the process and extent of urbanization and greening has been a vital task for urban management, which is critical for China to implement the 2030 Agenda for Sustainable Development [18], and has been elevated to the forefront of research and practical applications to support an effective urbanization process in China.

Remote sensing typically serves as a key source in determining the status and dynamics of urban information attributes due to its extensive spatiotemporal coverage and data availability. Currently, remote sensing monitoring of urban expansion and greening mainly focuses on area changes, land cover type conversions, and range changes at different periods. Deng et al. [19] used the principal component analysis and a hybrid classifier based on SPOT and Sentinel-2A images to monitor the land use change in Hangzhou City, China from 1996 to 2016. Liu et al. [20] extracted information on urban expansion in China using the Defense Meteorological Satellite Program’s Operational Line-scan System (DMSP-OLS) nighttime light data from 1993 to 2008. However, area change analysis only captures the total space of urban coverage and cannot reveal interior changes of urban environments [21]. Landscape patterns, known as spatial metrics, can objectively measure the pattern and structure of an urban environment. Landscape indices were usually used to detect and describe the change in urban landscape patterns, which can quantify and classify complicated landscapes into recognizable patterns and reveal some ecosystem properties that are not readily visible. Moreover, urban landscape pattern describes the spatial interaction among landscape factors in urban areas, which has the potential to provide a more detailed analysis of the spatio-temporal patterns of the urbanization process. Combining the application of urban landscape indices and remote sensing, though still under-used, can map the landscape patterns of urban green spaces and explore urban vegetation fragmentation. Such a combination will help better illustrate and understand the heterogeneous characteristics of urban regions and the effects of urbanization on the surrounding environment.

It is worth mentioning that urbanization has effected water, heat, soil quality, and carbon cycle, and has gradually become an important factor affecting vegetation growth [22]. Apart from the necessity to depict and comprehend the spatial diversity in land use patterns, it is also vital to assess the impact of urbanization on plant growth. To investigate the response of vegetation to urbanization, a variety of vegetation indexes have been utilized. Zhao et al. [23] explored the impacts of urbanization on vegetation growth using an enhanced vegetation index (EVI). Previous studies found that vegetation production is decreased in urban areas using net primary production (NPP) data derived from process-based models and data-driven models [24,25,26]. However, as a key quantitative factor of vegetation growth, the spatiotemporal dynamics of the fractional vegetation coverage (FVC) under rapid urbanization have been studied rarely and, hence, are not yet fully understood. Here, FVC was generally defined as the ratio of the vertical projection of above-ground vegetation organs on the ground to the entire calculated area, it can be independent of the noise of NDVI or EVI extremes, and better reflects the vegetation change process than vegetation indices [27,28]. Wang et al. detected the dynamic change of urban FVC using multi-temporal remote sensing images and revealed the influence intensity and direction of the urban sprawl on vegetation coverage [29]. Moreover, because of the high rates of urbanization and obvious ecological impacts, metropolitan areas often serve as the ideal places to study the ecological effects of urbanization. Most previous studies on the effects of urbanization have focused on mega-temperate cities in southern regions [25,30]. However, few studies have been published specifically regarding how urbanization affects vegetation in relatively cold regions in China. As the largest city in China by area, Harbin is the capital of China’s northeast province Heilongjiang, it is a typical industrial city with recent rapid urbanization in the cold temperate of northeastern China. Harbin experienced a significant loss of green space as a result of implementation of the revitalization of the old industrial base of Northeast China’s urban development strategy and the rapid urbanization process [31]. Since 2014, the Harbin government has formulated a series of development strategies including the Harbin Urban Master Plan (HUMP, 2004-2020), which had direct guidance for the development of urban green space and strived to be an ecological garden city [32]. Therefore, Harbin is a typical and unique area to study urban expansion and greening.

Although remote sensing technology has made significant progress in recent years, leading to a notable improvement in the spatial resolution of satellite imagery (0.5–10m) such as Sentinel-2, Worldview-3, SPOT, and IKONOS, their limitation in high acquisition cost and time scale present major challenges for monitoring long-term urban changes [33]. Landsat TM provides a suitable spatial resolution, long time series, and cost-free availability, which makes it an ideal choice for studying long-term trends and changes in landscape, vegetation, and other features [34]. In the study, Harbin is used as a case study area. Here, we use time-series Landsat satellite imagery from 2004–2020 to investigate the following four aspects under the two seemingly contradicting forces of rapid urbanization development and urban greening strategies: (1) the process and patterns of urban expansion, (2) the evolution of landscape patterns under the rapid urbanization process, (3) the spatiotemporal dynamics and trends of vegetation coverage, and (4) the impacts of urbanization on vegetation coverage. This study will contribute to improving the understanding of urban greening trends and the impact pattern of urbanization on vegetation growth (FVC) changes in China. They will also provide scientific information for identifying the potential environmental effects and appropriate decision-making toward urban sustainable development in the future.

2. Materials and Methods

2.1. Study Area

Harbin city (125°42′–130°10′ E, 44°04 N–46°40′ N) is the political, economic, and cultural center of the northeast of China. The mean annual temperature is 5.2 °C, which is characteristic of a continental monsoon climate. The surface soil is mainly chernozem soil, also referred to as “black earth”, which has rich nutrients and provides excellent conditions for both forests and crops. The urbanization of Harbin city began in 1978 and it had a gross domestic product (GDP) of RMB 535 billion in 2021, an increase of 367 billion over 2004. With the greatest geographical area among all provincial cities in China, Harbin has a total population of over 10 million people, representing approximately 31% of the population of Heilongjiang Province [35]. Our study area includes six districts comprising the urban area of Harbin city, distributed as shown in Figure 1.

2.2. Data Source and Pre-Processing

The satellite remote sensing data of the study include two main parts: land use/cover data and Landsat TM/OLI images. First, the land use/cover data in 2005, 2010, 2015, and 2020 were collected from the Land Use and Land Cover in China (CN-LUCC) dataset with a spatial resolution of 30 m. CN-LUCC was provided by the National Resources and Environment Database of the Chinese Academy of Science, which is based on the manual visual interpretation of multitemporal Landsat images and is widely used in published research [36]. The overall accuracy of the classified CN-LUCC dataset was over 92% [37], hence the land use/cover data used in the study was fairly credible. CN-LUCC includes six main land cover types: forestland, grassland, waterbody, cropland, built-up land, and unused land.

Landsat images (path 118, row 28) in 2004–2020 were freely available from the United States Geological Survey (USGS, https://earthexplorer.usgs.gov, accessed date September 2022) of the Earth Resources Observation and Science Center (EROS). Due to the desirable properties of moderate spatial resolution and shorter revisit time, Landsat images have been widely used for the research of urban planning, urban heat islands, and impervious surfaces [38,39,40]. The dates of images were selected and acquired during the growing season to reduce the effects of differences in phenological status. In the study, we assembled a set of cloud-free Landsat images of similar dates (September 2004–2020). Before downloading the data (L1T level of systematic geometric accuracy), georeferencing was performed at the USGS and no additional refinement was conducted in this study. Atmospheric corrections were undertaken to the Landsat datasets using the FLAASH model to remove the effect of atmospheric disturbance on the pixel values. A geometric precision correction was conducted based on the 1:50,000 topographic maps, and correction error was controlled within 1 pixel using ENVI 5.3 software. The parameters in the metadata file were used to perform radiometric calibration, which involved converting the pixels values to reflectance values. Finally, all satellite images were clipped according to the boundary of the study area using “extracted by mask” tools in ArcGIS Pro 2.5 software. Additionally, we calculated the greenness-related NDVI for each Landsat image. NDVI images were produced by calculating the ratio between the red (R, 630–690 nm) and near-infrared (NIR, 760–900 nm) values of the satellite image using Equation (1):

$$\mathrm{NDVI}=(NIR-R)/(NIR+R)$$

(1)

2.3. Spatio-Temporal Dynamics Features of Urban Expansion

To explore urban expansion, superimposed analysis of multi-temporal remote sensing images, as a crucial tool for the study of land use change, can reflect detailed information of spatial and temporal expansion. Here, built-up lands from four temporal phases overlapped one another for quantifying and assessing the urban expansion in Harbin city. Additionally, the transfer matrix was used to reflect the structure and status of land use type transitions between every two periods, which is the application of the Markov model for land use. With a pixel-by-pixel comparison of the land-use patterns of two distinct time periods, we can identify the direction and quantity transition of each land-use type. Specifically, the transfer matrix formula is as follows:

$$A_{ij}=\left(\begin{array}{ccc}{a}_{11}& \dots & a_{1n}\\ ⋮& \ddots & ⋮\\ a_{m1}& \cdots & a_{mn}\end{array}\right)$$

(2)

where i and j represent land-use types at the beginning and end of the research. A_ij represents the area of land-use type i (before) transformed into land class j (after). The transfer matrix and the spatio-temporal urban dynamics were produced using the spatial analysis tool “Overlay” in ArcGIS Pro 2.5 between 2005 and 2020.

2.4. Landscape Patterns Analysis

Spatial characteristics such as location, direction, and distance are the bases to analyze the structure of landscape patterns. Landscape metrics, which incorporate characteristics such as patch shape, size, number, and spatial combination, are frequently used to detect and quantify the spatiotemporal variations in landscape patterns. These metrics indicated complexity in the type and arrangement of landscape patches. However, most of the metrics are correlated, which can lead to redundant information and make interpretation difficult. Appropriate landscape indices should be selected based on their properties, purposes, and contents to achieve reliable and quantitative descriptions of the research objects [41].

In the study, eight frequently-used landscape metrics by referencing relevant work were selected to evaluate the fragmentation, connectivity and diversity of landscape patterns (

Table 1. Description of landscape metrics used in the study for urban landscape analysis.

Landscape Metrics	Abbreviation	Equations	Description
Percent of landscape	PLAND/%	$PLAND=({\displaystyle \sum _{j=1}^n{{\displaystyle a}}_{ij}})/A$	aj = area (m2) of the patch ij
Number of patches	NP	NP = n	n = the number of patches of a class
Patch density	PD/(m/hm2)	$PD=\frac{n}{A}$	A = total landscape area
Largest shape index	LSI/%	$LSI=\frac{0.25{\displaystyle {\sum }_{k=1}^m{e}_{ik}^{*}}}{\sqrt{A}}$	$e_{ik}^{*}$=total length (m) of edge in the landscape between patch classes i and k, including landscape boundary and background segments of urban patch
Edge density	ED/(m/hm2)	$ED=\left(10000\right){\displaystyle \sum _{j=1}^n{e}_{ik}^{*}}/A$
Landscape division index	DIVISION	$DIVISION=1-{\displaystyle \sum _{i=1}^m{\displaystyle \sum _{j=1}^n(\frac{a_{ij}}{A})}}$	m = the number of patch types
Shannon’s diversity index	SHDI	$SHDI=-{\displaystyle \sum _{i=1}^m(p_i\ln }{p}_i)$	pi = proportion of the landscape occupied by patch class i
Shannon’s evenness index	SHEI	$SHEI=\frac{-{\displaystyle \sum _{i=1}^m(p_i\ln }{p}_i)}{\ln m}$

), including percent of landscape (PLAND), number of urban patches (NP), patch density (PD), landscape shape index (LSI), edge density (ED), landscape division index (DIVISION), Shannon’s diversity index (SHDI), and Shannon’s evenness index (SHEI). Subsequently, eight landscape metrics were divided into four classes based on their definitions and features, including expansion, shape complexity, compactness, and landscape heterogeneity. PLAND depicts the relative abundance of each land-use type. NP is used to evaluate the discrete urban areas in the landscape, as urban areas expand and merge into a continuous urban fabric, NP values may decline. PD is an effective indicator to characterize aggregation or fragmentation of the urban landscape. LSI and ED reflect the connectivity in patches and shape complexity of urban landscape patches [42]. DIVISION reflects the fragmentation of the landscape, and SHDI and SHEI are landscape-level metrics that show the landscape diversity and heterogeneity in the landscape. More detailed descriptions and specific mathematical equations of the seven metrics employed in the study are shown in Table 1. We calculated these landscape metrics using the public domain software Fragstats version 4.2.

2.5. Characterizing Vegetation Coverage and Trends Analysis

2.5.1. Dimidiate Pixel Modeling and Statistical Analysis

Dimidiate pixel modeling (DPM) is a mature method that is widely utilized in the practical remote sensing monitoring of FVC. DPM also can help reduce the impact of vegetation type, atmospheric conditions, and soil background on remote sensing images, and can preserve more information about vegetation cover [43]. Thus, DPM was chosen to calculate the fraction of vegetation coverage (f_vc) with the following Equation (3) [44]:

$$f_{vc}=\frac{(NDVI-NDV{I}_{soil})}{(NDV{I}_{veg}-NDV{I}_{soil})}$$

(3)

where NDVI_soil and NDVI_veg represent the NDVI values of bare soil and pure vegetation cover pixels, respectively. We selected NDVI values of the 5th and 95th percentile of the histogram as the NDVI of bare soil and pure vegetation, respectively. To quantitative investigate the spatial distribution and dynamic evolution traits of FVC, the FVC values are divided into five grades using the equal spacing grading method, including very high coverage (0.8 ≤ f_vc < 1), high coverage (0.6 ≤ f_vc < 0.8), medium coverage (0.4 ≤ f_vc < 0.6), low coverage (0.2 ≤ f_vc < 0.4), and very low coverage (0 ≤ f_vc < 0.2). In addition, the coefficient of variation (Cv) is also a widely used statistical measure of the dispersion or variability of observations around the mean, which can eliminate the impact of varying observational units or methods from the data to enable objective comparison of results. The Cv of FVC from 2004 to 2020 was calculated to analyze the temporal variation.

2.5.2. FVC Trend Analysis

To further study FVC change, a simple linear regression analysis was utilized to detect the trend in vegetation dynamics. Here, time t was set as the independent variable, and the FVC value of each pixel served as the dependent variable. The slope of linear regression is an indicator to quantify the vegetation dynamics trends over the study period, which was determined using the following Equation (4),

$$Slope=\frac{n{\displaystyle {\sum }_{i=1}^{n}i\times f_{vc}{}_i-({\displaystyle {\sum }_{i=1}^{n}i})({\displaystyle {\sum }_{i=1}^n{f}_{vc}{}_i})}}{n{\displaystyle {\sum }_{i=1}^n{i}^2}-{({\displaystyle {\sum }_{i=1}^{n}i})}^2}$$

(4)

where n is the number of years in the study period, i is the serial number of the year, and f_vci is the FVC value in the year i. Slope > 0 indicates a positive trend of vegetation dynamics, which implies increasing the activity or coverage of vegetation. Slope < 0 refers to a decreasing trend of vegetation dynamics or a weakening of vegetation activity. The F test was used to determine whether the vegetation dynamic trend (Slope) is statistically significant, the significance level was set to 0.01 and 0.05. If the slope passes the significance test, it indicates a significant ascending or descending trend, and the trend grade is shown in

Table 2. Gradation table of vegetation coverage trends.

Change	Slope	Significance
Very significant decrease	<0	p ≤ 0.01
Significant decrease	<0	0.01 < p ≤ 0.05
Insignificant decrease	<0	p > 0.05
No change	=0	-
Very significant increase	>0	p ≤ 0.01
Significant increase	>0	0.01 < p ≤ 0.05
Insignificant increase	>0	p > 0.05

.

2.6. Analyzing the Impacts of Urbanization on FVC

A conceptual framework proposed by [23] was generated to quantify the direct and indirect impacts of urbanization on vegetation growth. Here, the direct impacts refer to the impacts of LUCC on vegetation growth in a city during urban expansion. Theoretically, the FVC is inversely proportional to the urban ratio (β, also called “urban intensity”), and we assumed that there is a linear relationship between FVC and β when only direct effects are considered. Hence, the theoretical zero-impact straight line was determined by two FVC values of fully vegetated and fully urbanized pixels, as shown in Equation (5),

$$FV{C}_{zi}=(1-\beta )\cdot FV{C}_v+\beta \cdot FV{C}_{nv}$$

(5)

where β is urban intensity (i.e., the fractions of built-up land). FVC_zi was the theoretical FVC value of no impact, FVC_v was the FVC value of fully vegetated pixels (β=0), the FVC_nv was the FVC value of fully built-up surfaces (β = 1). To calculate urban intensity, we use the “fishing net” tool based on the GIS platform to generate a 300 m × 300 m grid as the evaluation unit (27,520 evaluation units in total). Coupling with the LUCC dataset, the urban intensity of each grid was determined as the fraction of falling into the built-up surfaces within the grid, with values ranging from 0 (full vegetation) to 1 (full built-up) with a 300 m spatial resolution. To match the grid size of urban intensity, FVC with a spatial resolution of 30 m was spatially resampled into a 300 m pixel.

The distribution of observed FVC points (FVC_obs) cannot completely be compatible with the zero-impact lines, which indicated the existence of indirect effects. Any observed points above or below the zero-impact line indicate positive or negative indirect impacts of urbanization on vegetation growth, respectively (Figure 2). The relative indirect impact (w_i) of urbanization on vegetation growth can be calculated as:

$$w_i=\frac{FV{C}_{obs}-FV{C}_{zi}}{FV{C}_{zi}}\times 100\%$$

(6)

To quantify the total growth change impacted by urbanization, coupling the direct impact and indirect impact, the growth offset (γ) was defined as Equation (7). The positive γ represents enhanced vegetation growth (offset) and, conversely, a negative γ suggests exacerbate vegetation loss by surface replacement.

$$\gamma =\frac{FV{C}_{obs}-FV{C}_{zi}}{FV{C}_v-FV{C}_{zi}}\times 100\%$$

(7)

To characterize the FVC~β relationship, we derived the average value of FVC with each urban intensity (β) bin at an interval of 0.01. We used a well-proved cubic regression model to fit the FVC~β response curve [23,25,45]: y = a₀ + a₁x + a₂x² + a₃x³, where y was the observed FVC and x was urban intensity. Additionally, the intercepts (a₀) of regressions were used to determine the FVC value of full vegetation (FVC_v), which depended on the FVC changing trend and were less affected by FVC outliers [23,45]. To determine the FVC value of no vegetation (FVC_nv), we manually selected 50 pixels of fully built-up land in the years 2005 and 2020, followed by [25].

The steps and sequences of data collection and analysis used in the study are shown in Figure 3. The flowchart has five parts, including pre-processing, urban expansion analysis, urban landscape pattern, FVC change analysis, and impacts of urbanization on vegetation coverage. The FVC change analysis and impacts of urbanization on vegetation coverage were performed using Python 3.6 with the numpy, pandas, gdal, and statsmodels packages.

3. Results

3.1. Characteristics of Spatiotemporal Change of Urbanization at 30 m Spatial Resolution

According to the dataset of land use in 2005, 2010, 2015, and 2020, the urban land in Harbin was mostly distributed on the south side of the Songhua River (Figure 4). The change in urban expansion of Harbin city over the past 15 years showed that the built-up area of Harbin expanded relatively quickly between 2005 and 2010, and even faster between 2015 and 2020, which demonstrates the rapid economic growth.

When the Sankey diagram (Figure 5) is coupled with the spatial distribution of LUCC (Figure 4), the built-up area expands by 66.96 km² at an average rate of 13.39 km² in 2005–2010. The cropland (106.71 km²) and waterbody (11.96 km²) have the highest fraction for land conversion and are the major land resources that urbanization has encroached upon. The urban converted from farmland areas and waters constituted about 84.71% and 9.5% of the total urban gain, respectively. The main expansion was in the form of westward expansion from the center, expanding in a radiating way, which mainly occurred in the Songbei and Dongli Districts (33.71 km², 30.02 km²). From 2010 to 2015, the pace of urban expansion keeps gradually stable, and the area of Harbin city expands by 12.39 km² at an average rate of 2.49 km². Only a total of 22.85 km² of land experienced changes, an increase of 15.46km² in urban areas accounting for 67.64% of the total change in land use, of which, 70.69% and 19.60% resulted from cropland and waterbody, respectively. In 2015–2020, the urban experience remarkably increased on a rapid scale. A total of 192.74 km² of land was encroached upon by built-up land; the result from cropland was 156.22 km² (81.05%) and grassland was 16.27 km² (8.44%).

3.2. Analysis of Landscape Pattern

The accelerating urbanization in Harbin resulted in a fundamental change of landscape patterns between 2005 and 2020. As shown in Figure 6, the cropland remained dominant according to PLAND, but the PLAND of cropland decreased gradually from 68.61% to 61.47%. In contrast, the process of urban expansion accelerated, and the PLAND of urban areas increased greatly from 14.21% to 24.19%. In terms of urban compactness, the NP and PD of built-up land are higher than other land use types, which indicates the relatively high fragmentation of built-up land. The NP, PD and DIVISION of urban areas conformed to decreasing trends during 2015–2020, which demonstrated that the city was becoming more compact. Therefore, the urban landscape of Harbin expanded by merging with existing diffuse sprawled and fragmented urban areas, which were connected to a more continuous urban patch.

Apart from the urban landscape, the NPs, PDs and DIVISIONs of forest and grassland exhibit decreasing trends and decreasing PLAND, which indicated the areas of forest and grassland have decreased in size, although they were likely not fragmented under urbanization pressure from 2005 to 2020. Alternatively, the decreasing PLAND and the increasing NP and PD for agricultural land indicated the fragmentation of farmland. Additionally, LSI and ED, which measured the shape complexity of the landscape, increased over the period 2005–2015. The finding indicates that the urban landscape adopted a more irregular and convoluted pattern. Subsequently, the value of LSI in urban and forests sharply increased, reaching peak values in 2015, then, the LSI metric abruptly declined until 2020 (Figure 6d), when the NP and PD also dropped abruptly. This trend may reflect the fragmented development cores expanded together to the densely built-up area, and the higher proportion of single-shaped artificial ecological patches, which simplifies the landscape shape.

In terms of the metrics of the entire landscape level, SHDI and SHEI were increasing between 2005 and 2015, as shown in

Table 3. Changes in the landscape-level metrics in Harbin between 2005–2020.

	2005	2010	2015	2020
SHDI	1.0285	1.0915	1.0927	1.0743
SHEI	0.574	0.6092	0.6098	0.5996

, which indicated the landscape was becoming more heterogeneous and equally distributed. By contrast, the slight reduction in SHDI and SHEI during 2015–2020 reflected the continued urban expansion, shrinking the heterogeneity and re-establishing unevenness in the landscape.

3.3. Spatial-Temporal Change of Vegetation Coverage

The annual average FVC in Harbin ranges between 35.57%~47.41%, and vegetation coverage varies regionally (Figure 7a). The overall spatial distribution of FVC in Harbin exhibits that the urban fringe region is obviously higher than the central region. The average FVC in the Nangang district is 0.39, consisting mainly of extremely high grade and low grade, occupying 40.14% and 22.68% of its area, respectively. The highest average FVC in the Pingfang district is 0.81, which is mainly extremely high grade and high grade, taking up 49.46% and 23.56% of its area, respectively. Overall, the very high grade accounts for the largest area proportion, which fills about 50.98% of Harbin’s area, mainly distributed in the Songbei and Daowai districts (424.22 km² and 322.98 km²), while it has the least distribution in the Pingfang district (48.85 km²).

The coefficient of variation reflects the discrete degree and temporal variability through time in 2004–2020 (Figure 7b), which was divided into four classes using the geometrical interval classification method, including very stable (0~0.2), stable (0.2~0.4), unstable (0.6~0.8), and very unstable (0.8~1.0). The spatial pattern of the coefficient of variation exhibits that the vegetation in the study area is stable overall, but fluctuates in the local area. The under stable class (Cv < 0.4) accounts for 86.54% of Harbin’s area, which is mainly distributed in the Songbei and Daowai districts. The unstable class (0.6 < Cv < 0.8) is distributed mainly in the Daoli and Songbei districts, accounting for 9.46% of Harbin’s area. The area with a high fluctuation class (Cv > 0.8) is concentrated in the urban region in the Daowai and Songbei districts, accounting for 4.0% of Harbin’s area.

Combining the linear regression slope method and F test, the spatial pattern and variability of vegetation coverage can be described to reflect the FVC trends. From 2004 to 2020, the variation trends of FVC ranged from −0.05~0.06. A decreasing trend in vegetation cover was noted in most areas of Harbin (Figure 8), with an extremely significant decrease in about 142.78 km², accounting for 5.85% of Harbin’s land (p < 0.01). Insignificantly decreased vegetation coverage (p > 0.05) accounted for the largest area (1281.26 km², approximately 52.47%), and was more evenly distributed in each district, taking up about half of each district’s land. In the contract, the regions with increased vegetation coverage only accounted for 31.53%. The regions where FVC significantly increased comprise only 7.48% of Harbin’s area (p < 0.05), which was mainly found in the southern part of the Songbei and northern Daowai districts, and least in the Nangang district. Additionally, the stable vegetation coverage change area of 3.01% was mainly distributed in the main urban area, with an area of about 73.44 km². In general, most regions exhibited insignificant FVC changes, accounting for 76.51%. Significant changes were mainly found in the interior of the Pingfang district and along the Songhua river in the Daowai district (Figure S1).

3.4. The Impacts of Urbanization on FVC in 300 m Grid-Cells

The relationship between the FVC and urban intensity (β) was fitted for Harbin in 2005 and 2020, the FVC~β curves and their related zero-impact straight lines are shown in Figure S2. Results show the observed FVC decreased with increasing urban intensity, and the cubic regressions of the FVC~β curves had statistical significance (p < 0.01). The values of FVC in 2005 and 2020 were both above the zero-impact line, which can be attributed to the positive indirect impacts of urbanization on vegetation growth. Additionally, the overlapping relationship between the two years (2005 and 2020) shows that vegetation growth exhibited a slightly stronger positive indirect in response to the urban environment over time.

The indirect impact of urbanization in 2005 and 2020 is shown in Figure S3; the negative impacts (w_i < 0) were almost presented over the urban intensity of 0~0.5, whereas the positive impact (w_i > 0) significantly increased non-linearly with urbanization intensity (slope > 0, p < 0.05), and could be as high as over 19%. The growth enhancement offset (γ) exhibited a trend of first increasing and later decreasing, and gradually stabilizing with the increasing urban intensity, with values of about 2.26% and 2.71% of the growth offset on average in 2005 and 2020, respectively (Figure S3). We additionally found that the relatively higher offset mainly occurred in low and medium urban intensity areas. Subsequently, we further classified urban intensity into low urbanization (β from 0 to 0.3), medium urbanization (β from 0.3 to 0.6), and high urbanization (β from 0.6 to 1), and then compared the magnitudes of the indirect impact on vegetation growth at different urbanization levels in 2005 and 2020 (Figure 9a). The results showed that the average indirect impact (w_i) in areas with high urbanization levels was highest (~11.87%), obviously higher than the medium and low urbanization levels. The indirect impact was near zero or even negative in low β levels, though positive impacts also exist. By contrast, there was the highest offset growth (~3.47%) in medium urbanization level compared to urbanized and rural areas (Figure 9b). Both indirect impact and offset growth on FVC became gradually stronger from 2015 to 2020, especially w_i in the high β level, but the offsetting growth in high β areas was slightly weaker in 2020 than in 2005.

4. Discussion

4.1. The Response of LUCC and Landscape Pattern to Urbanization

Urbanization is the natural trend of human social development and an inevitable result of economic and technological progress. As a major metropolis in northeast China, the land use pattern of Harbin’s main city zone has changed greatly and the urbanization development is obvious when examining the period of 2004–2020. The dominance, fragmentation, shape, and diversity of landscape features were examined using eight landscape metrics (Table 1) in the study. The dynamics analysis of LUCC and landscape patterns conducted in the study demonstrated that the rapid urbanization process occurred at the cost of the loss of green space in the urban fringe and the suburban area, which is a similar pattern to how most cities in many developing countries are urbanizing.

Due to the in-depth implementation of the revitalization of the old industrial areas in Harbin, there has been a continuous expansion of cities, development of industrialization, and population growth. To meet demand, the green space in Harbin has been sacrificed, and PLANDs and areas for cropland, grassland, and forestland have been reduced, most of which have been replaced by built-up land. Among the land use types for green spaces, the significantly increased PD and NP for farmland indicated it was largely encroached upon and dispersed by urban expansion. This result is the same as other research [46]. Both the area and edge density of woodland exhibit an increasing and then decreasing trend, which indicated the forest was also impacted, and were mainly artificial ecological patches (e.g., plantation forest) with a single shape, although the intensity was much less than that of agricultural land. Unfortunately, the superimposed spatial analysis revealed that the built-up land also increased in the side of the Songhua River, where many small-area wetlands and marshes exist, which may threaten these important ecological reserves. This is also evidenced by the decrease of PLAND and the increase of NP, LSI, and ED for the waterbody. Apart from the class-level metrics of fragmentation and connectivity for each type, the landscape-level metrics for diversity also were evaluated in this study. The urban landscape has become more diverse, homogeneous, fragmented, and complicated in its configuration in Harbin due to urbanization, as has been the case in numerous other cities in China, as recorded in previous studies [47,48,49].

Notably, the absolute values of landscape patterns are limited by the level of detail in landscape classification of land use/land cover data, and scale and spatial resolution of imagery. In the study, LUCC datasets with 30 m resolution and an interval of 5 years were used to analyze the LUCC and landscape patterns in Harbin. A few small areas of land use patches in the urban zones were not well recognized in a timely manner to be included in the study due to the constraints of the spatial data and temporal resolution and the classification method, and this also could affect the accuracy and interpretation of the results [50]. Finer spatial and temporal resolution satellite images could provide more information for mapping urban features. Moreover, although the study quantitatively evaluated the landscape patterns of Harbin and analyzed its dynamic change, we selected eight landscape indices of relative importance based on previous research, and there may be missing indicators.

Many factors drive the LUCC and landscape pattern changes, and human disturbance is a major factor in urban landscapes. As a crucial means of governmental involvement, the policies of greening and land use introduced by the local government control and guide a city’s development to some extent, which can limit the detrimental effects of urbanization on green space and balance the relationship between urbanization and greening. Since the enforcement of the “Harbin City Master Plan (2004–2020)” and the “Harbin City Urban Master Plan (2011–2020)” (revised draft in 2017), Harbin has formed a “one river, two rivers, three ditches, and four lakes” ecological planning layout. Many new urban garden parks and wetland parks have been constructed, such as Changqing park, Binjiang wetland park, and Shangzhi Park. Meanwhile, some urban parks are more concentrated in their distribution, and ecological corridors are better protected. The parks are linked by corridors to create complete ecological patches, thereby improving the connectivity and aggregation of ecological land. Additionally, the local government supported initiatives to expand suburban forests and increase urban public green space to improve the rural and urban ecological environments. “Harbin Land-use Master Plan (2006–2020)” also repeatedly emphasized the strict protection of basic farmland and reasonable use of agricultural land to protect national food security, which may explain the mitigation of the loss of farmland in 2010–2015 (Figure 5). Additionally, the spatial pattern of urban dynamics exhibited the expansion trend in southern Harbin (Figure 4), which is closely related to the development strategy of “North Jump, South Extension, Central Prosperity and Strong County” launched at the end of 2009.

Although the improved green space condition and urban area has expanded, the demand for construction is still high, and the distribution of ecological and non-ecological land continues to be contradictory. Therefore, there is still a need to develop planning strategies to optimize the allocation of land resources, and to coordinate land use and ecological construction. It is important to reserve a certain amount of farmland as buffer zones, particularly along both banks of the Songhua River, where most of the farmland is located and built-up land has increased (Figure 4). Ecological corridors should be preserved and further enhanced to improve ecological network connectivity and create a stronger green space network [32].

4.2. Impacts of Urbanization on Vegetation Growth

Vegetation plays an important role in the urban environment, however, the patterns and mechanisms of vegetation dynamics caused by urbanization are still poorly understood due to the limited availability of traditional ground-based observations [51]. Here, the FVC derived from remote sensing data with 30 m resolution, in combination with the LUCC dataset, were used to quantify vegetation activity and the impact of urbanization on vegetation growth. From 2004 to 2020, the vegetation coverage in Harbin city exhibited an overall insignificant decreasing and stable trend, but fluctuated on the local scale, which was consistent with the findings of other studies [52,53]. The FVC, as with other vegetation indicators (EVI, GPP, NDVI, LAI), was reduced significantly with increasing urban intensity [53,54]. By separating the direct and indirect impacts of urbanization on vegetation growth along with urban intensity, we further confirmed that urbanization could exert both positive and negative impacts on vegetation growth. The direct impact, which generally refers to the replacement of natural land cover by impervious surfaces and the reduction in vegetation cover and growth, is usually negative [55]. By contrast, the indirect impact is usually caused by the unique natural environments of urban areas (e.g., urban “heat island” warming, lower air humidity, higher CO₂ concentration, water shortage) and human management of urban vegetation, which may enhance or inhibit vegetation growth [56]. It has long and widely been believed that there is more impervious surface, dense buildings, and less open space in areas of high urbanization, which led to plants getting limited light, thereby increasing the relative allocation of plant biomass to shoots [57]. Yet, a previous study revealed that despite poor solar conditions and high air pollutant concentrations in central urban areas, aerosol particles can increase diffuse radiation, which can enhance canopy light use efficiency [58]. More surface sealing, soil compaction, and pollution caused by urbanization may affect root health and litterfall decomposition, and further affect the urban carbon cycle [59,60]. Additionally, urban vegetation growth also benefits from human management practices (e.g., irrigation, fertilization). The higher temperature caused by “urban heat island” warming and atmospheric CO₂ concentrations can effectively lengthen the growing season in cold regions [61] and may enhance the autotrophic respiration and photosynthetic processes of vegetation. This may be one of the drivers of the enhanced indirect impact (w_i) that accompanies increasing urbanization intensity. It is because of the unique conditions of urban spaces that they are often seen as “harbingers” of future global change [62]. It can be said that the impact of urbanization on vegetation growth can provide valuable information on how other non-urban ecosystems may react to future climate and environmental changes.

The driving factors of an FVC change study [52] found that the vegetation change in Harbin is influenced by various factors (e.g., temperature, precipitation, radiation, and CO₂ concentrations). Other environmental factors, such as plant species, O₃, and nitrogen deposition, may also have an impact on the growth of vegetation in urban areas [62], which may be an important reason for the complex changes of indirect impact, that merit further investigation. Moreover, we found, as in previous studies [23,25,63], enhanced growths (FVC) caused by urbanization can offset to some extent the direct loss of vegetation caused by land conversion, especially in medium urban intensity areas, which might be explained by more resources of ecological governance invested in peri-urban areas. Urbanization-induced growth enhancement has also been reported, showing that the average growth offsets of EVI and GPP were 24.63% and 16.97% in Shanghai [25], respectively, which is obviously higher than our results in Harbin (2.26%~2.71%). This may be partially explained by vegetation growth sensitivity varying for different vegetation variables. Additionally, in relatively cold environments, the restricting effect of low temperatures on vegetation growth is more prominent than in temperate cities (e.g., Shanghai). The indirect impact of urbanization on vegetation growth tended to vary across cities at different economic development levels, and the development of Harbin city is relatively backward compared to southern China. A study of 32 major Chinese cities also found a similar result; EVI along the urban intensity gradient demonstrated a linear form in Harbin, which indicates little growth offset generated by urbanization [53]. Additionally, there was a slight increase in the indirect impact from 2005–2020, which may reflect urban environmental improvements in Harbin. As urbanization in Harbin is likely to continue for decades to come, it is still necessary to adopt effective protection, management practices, and design plans to maximize the benefits of urbanization-induced environmental change to promote vegetation growth and minimize vegetation losses. This detailed study on the spatiotemporal trends of vegetation growth with urbanization in Harbin expands understanding of vegetation response to urbanization, which provides several valuable pieces of information for future urban planning.

5. Conclusions

Based on the Landsat TM/ OLI satellite images and LUCC dataset in 2004–2020, the study explored the spatial-temporal dynamic of landscape patterns and FVC, and evaluated the positive and negative impacts of urbanization on vegetation growth across the severe region, Harbin. In the past 17 years, 660.48 km² or 26.67% of the total land has undergone change, with the conversion of non-urban to urban land accounting for 43.12%. The rate of expansion varied in different periods, and dynamic changes were most significant in the period of 2015–2020. The two main resources that were transformed for urban growth were crops and water. Urbanization and related policies exerted significant impacts on the spatiotemporal dynamic of LUCC. Additionally, urban expansion exhibited a compact sprawl pattern, and the diversity and homogeneity of the landscape composition increased. Spatially, the distribution of FVC in Harbin exhibited that the urban fringe region was higher than the central region, and vegetation coverage was dominated by very high and high coverage. Temporally, the FVC in Harbin was stable but fluctuated on the local scale. The FVC continued to exhibit an overall decrease trend, but that trend was not significant. The FVC was observed to decrease with the increase of urban intensity, and the analysis of urbanization showed the positive impact of urbanization on vegetation growth in Harbin. Additionally, the growth enhancement of FVC resulting from the urban environment and human management was about 2.26%~2.71%, which was higher at the medium urbanization level, but lower in the low- and high-urbanized areas.

This study demonstrates the potential of our research approach as the first successful attempt to use the fraction of vegetation coverage rather than normal vegetation indices (e.g., EVI, NDVI, and production) to explore the impacts of urban expansion on vegetation growth in the severe cold region, China. Although the findings of this study are limited to Harbin, the methodology presented here could be easily implemented in many other cities to investigate whether urban vegetation growth is linked with distinct geographical settings and urban development, which can help to understand the factors that contribute to the growth and preservation of urban green spaces. Furthermore, quantifying the spatial variation of urban areas, landscape patterns, and vegetation coverage over time are crucial for guiding future urban planning and policy decisions toward sustainable urban development.