Unveiling the driver behind China’s greening trend: urban vs. rural areas

Urban and rural areas play an important role in the greenness change in China, despite exhibiting divergent landscape ecologies. Although recent studies have revealed an overall greening pattern in China, the relative contribution of urban and rural vegetation to nationwide greening trend and their driving mechanisms behind these changes remain poorly understood. Here, we first utilized a high-resolution land use/cover dataset (GlobeLand30) to establish a framework for distinguishing between urban and rural areas. We then assessed and compared the greenness changes in both urban and rural areas using multiple vegetation indices from 2000 to 2020. By employing Random Forest model and generalized linear model regression, we further investigated drivers behind the changes in urban and rural vegetation trends. Our results demonstrated a significant greening trend in China, and the greenness increased 13.71% from 2000 to 2020. Vegetation changes in both urban (+4.96%, 0.0011 yr−1) and rural areas (+14.25%, 0.0026 yr−1) have contributed positively to China’s greening trend, with their contribution being 11.3% and 88.7%, respectively. Urban core areas exhibited the largest trend magnitudes (0.0043 ± 0.0035 yr−1) among all the urban–rural subregions. Increased tree cover was identified as the primary driver of greening trends in both urban and rural areas, explaining 36% and 29% of the greening, respectively. However, the pathways of tree cover increase differed between urban and rural areas, with urban areas focusing on green space construction and rural areas implementing afforestation programs. In contrast, climate change and the CO2 fertilization effect had a greater contribution to the greening trend in rural areas than in urban areas. Our study demonstrates the positive role played by both urban and rural areas in China’s greening trends and elucidates the underlying mechanisms driving these changes, highlighting the need for differentiated strategies in urban and rural areas for future vegetation restoration.


Introduction
Satellite data have revealed an unambiguous greening trend in China in recent decades (Yun et al 2020, Chen et al 2021, Zhang et al 2022a). Remarkably, with only 6.6% of the global vegetated area, China explains 25% of the global net increase in leaf area (Chen et al 2019a). This enhanced vegetation growth in China plays a crucial part in 'Greening Earth' phenomenon and has a significant impact on the terrestrial biogeochemical cycles and energy balance (Zhu et al 2016, Burrell et al 2020. Long-term changes in vegetation greenness in China are driven by multiple factors concerning climates and anthropogenic activities. Since the end of the 20th century, human activities, represented by rapid urbanization (Zhang et al 2022b) and large-scale afforestation activities (Piao et al 2015) have been accelerating (Chen et al 2021), leading to profound alterations in vegetation patterns in both urban and rural areas of China (Zhang et al 2022c). However, the specific pattern of gradual and abrupt greenness changes in urban and rural areas under increased disturbances is not yet clear. This affects our perception of the role of urban and rural areas in China's greening trends and hampers the development of corresponding ecological strategies to promote green development in urban and rural construction.
Urban and rural areas together form the fabric of China's landscape ecology while exhibiting distinct vegetation patterns due to different anthropogenic activities. In recent decades, China has undergone rapid urbanization, accompanied by a large number of people moving from rural to urban areas (Zhao et al 2022, Zhang et al 2022a. This process has been shown to cause a substantial loss in vegetation productivity in urban areas via the expansion of impervious surfaces (He et al 2014, Liu et al 2020b. Since 2012, when China began incorporating 'Ecological Civilization' into national plans and constitution (Xu et al 2017, Zhao et al 2022, Zhang et al 2022a, urban green spaces has been promoted on a large scale. However, the role of urban greening projects in improving vegetation coverage remains ambiguous. In contrast, rural areas, which hold great value in terms of ecological services and productivity, have also undergone conspicuous reshaping and transformation of the vegetation landscape (Hu et al 2019, Jia et al 2021. In the urbanization process, the migration of rural residents to cities has resulted in the phenomenon of 'rural hollowing' (Zhang et al 2022c), which entails abandonment of cultivated or settled land and the invasion of woody plants , Chen et al 2020. Large-scale afforestation activities in rural areas since the 1990s have also been remarkable. Several studies have shown that such afforestation efforts have made significant contributions to improving vegetation density and carbon storage in China (Ma et al 2019, Shi et al 2020, Yang et al 2021, Zhao et al 2022. However, vegetation degradation still persists in some rural areas. For instance, in the northwestern arid and semi-arid grassland areas, overgrazing and aridity have led to a decline in ecosystem function and biodiversity (Guo et al 2020b, Li et al 2021. The role that urban and rural areas play in China's vegetation greening trends during the urbanization process is controversial (Hu et al 2019, Yang et al 2021, Zhang 2022. This calls for a comprehensive quantification and assessment of urban and rural greenness changes to help advance national top-down initiatives for urban-rural green development (He et al 2014). Previous studies on urban and rural greenness change in China have generally concentrated on localized areas, delineating urban and rural gradients by creating buffer zones centered on cities , Jia et al 2021. However, due to the lack of a clear division between urban and rural areas from a national ecologicalspatial perspective, these studies have failed to provide a comprehensive picture of the differences between urban and rural vegetation changes at the national scale. This will affect our projections of future patterns of vegetation evolution and subsequent Chinese urban-rural carbon management (Wu and Zhang 2021). Besides, previous studies on the temporal trends of vegetation dynamics in China have typically focused on two types of change: phenological change (seasonal dynamic) driven by short-term climate fluctuations (Wang et al 2011, Nguyen et al 2020 and gradual changes (interannual dynamics) (Yao et al 2018, Wang andFriedl 2019). However, abrupt changes, which are often associated with landuse disturbances such as deforestation, afforestation, and extreme climate conditions in both urban and rural areas, have been largely overlooked (Chen et al 2014, Zhao et al 2022. In addition, the factors and mechanisms driving vegetation responses to climate variability and anthropogenic disturbance are still poorly understood across spatial gradients in urban and rural areas (Zhang et al 2022b).
Here, we advance current understanding of the differences in the role of urban and rural greenness variation in China's widespread greening trend at the pixel level. We first establish a framework for the delineation of urban and rural areas on a national scale using high-resolution (30 m) land-use data from an ecological-spatial perspective (Zhang et al 2022c). We use multiple vegetation indices to evaluate and compare urban and rural greenness change patterns and their contribution to greening trends in China. Random Forest (RF) model (Breiman 2001) and generalized linear model (GLM) regression (Tao et al 2015, Girardin et al 2016 were employed to quantify climatic and anthropogenic effects on urban and rural greenness changes. We aim to answer the following questions: (1) what are the spatiotemporal patterns of vegetation greenness in urban and rural areas across China and where have the trends shift occurred? (2) What are the main driving factors of the difference between urban and rural greenness change? (3) In particular, how will urban greenness trends vary depending on the level of urban development? Answering these questions is crucial for developing a comprehensive understanding of the dynamics of urban and rural greenness changes, and therefore, contributes to the formulation of effective urban-rural vegetation management strategies for promoting harmonious and sustainable urban-rural ecological development.

Generating annual vegetation greenness time series
To investigate the changes in vegetation in urban and rural China over the past 20 years, we first used Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI) datasets (Fensholt et al 2009) (Badgley et al 2017) and the enhanced vegetation index (EVI) (Huete et al 1997) datasets for the same period, were also used in this study (table S1). NDVI and EVI time series data were derived from MOD13Q1 (collection 6), while kNDVI and NIRv products were calculated based on MOD09Q1. All these four vegetation index datasets were resampled to a spatial resolution of 1 km. Further details can be found in the supplementary materials.

Detection of turning points and trend shifts
To detect if there was a turning point in vegetation time series as well as how the trend of NDVI shifted over China, a piecewise regression model with one turning point (Wang et al 2011, Chen et al 2014, Pan et al 2018 (equation (1)) was applied to vegetation greenness time series for each pixel across China from 2000 to 2020 where t is the year; y t is NDVI value in time-series; T 0 means the turning point of vegetation greenness trend. b 1 , b 2 and β 0 are the coefficients and intercept of regression; ε represents the residual. The greening/browning rates of g trends are b 1 and b 1 + b 2 , before and after the turning point, respectively. We also restricted T 0 within the period 2000-2002 to avoid linear regressions over a greenness period with scarce data. We used maximum likelihood to determine the turning point. At the P < 0.05 level, if the null hypothesis (i.e. b 2 = 0) tested by a two-tailed t-test is rejected, the difference in trends before and after the turning point is considered statistically significant (Yuan et al 2019).
We also analyzed the greenness trends before and after the turning point for each pixel to investigate the impact of external disturbances on vegetation dynamics. Four types of greenness trend shifts were identified based on changes around the turning point: greening-to-browning, greening-to-greening, browning-to-greening, and browning-to-browning trends.

A framework for distinguishing urban and rural areas
We proposed a conceptual framework to distinguish urban and rural areas based on the gradient of human pressure (Zhang et al 2022c) (figure 1). GlobeLand30 land cover maps (Chen et al 2015) for 2000 and 2020 with a spatial resolution of 30 m were used in this study. For a grid of 1 × 1 km 2 , we defined its urban or rural level according to the percentage of coverage by certain land cover classes, such as 'artificial surfaces' , 'cropland' and 'forest' , which reflect a gradient from urban to rural, indicating varying levels of human activity.
In this study, urban areas included urban cores, urbanization and suburban areas, reflecting different urban intensity levels. Rural areas were located outside urban areas, consisting of croplands and their surrounding areas (including rural residential areas). We classified rural areas into six types: towns, high agricultural pressure areas, low agricultural pressure areas, forests, shrubs and grassland, and others. Our classification framework adopts a top-to-bottom hierarchy for assigning pixels to the most applicable land cover type in cases where a pixel satisfies multiple land use criteria (table 1) , Zhang et al 2022c.

Driving factors of changes in vegetation greenness
To investigate the impact of potential drivers on the greenness change in urban and rural areas of China, we first used the RF model (Breiman 2001), a machine learning algorithm, to predict the greenness trend class (i.e. greening, browning, and non-significant trend) based on environmental and anthropogenic characteristics over the last two decades.
Time series predictors, including temperature, precipitation, solar radiation, CO 2 concentrations, tree cover and population change, were used as explanatory variables in the RF model. Three types of NDVI trends (greening, browning, and nonsignificant trend) were considered as the variables to  Table 1. Definition of urban and rural land cover classes. Urban core areas, urbanization areas and urbanized areas are categorized as urban areas. Towns, high agricultural pressure areas, low agricultural pressure areas, forests, shrubs and grassland and others are classified as rural areas.

Urban and rural land cover class Definition
Urban core areas Grids with artificial surfaces covering >50% in 2000.

Urbanization areas
Grids where >50% are covered by artificial surfaces in 2020 but not in 2000.

Suburban areas
Grids that 25%-50% are covered by artificial surfaces in 2020. Towns Grids that 12.5%-25% of their areas are covered by artificial surfaces in 2020. High agricultural pressure areas Grids where croplands dominate ⩾50% in 2020. Low agricultural pressure areas Grids that 12.5%-50% are covered by croplands in 2020.

Shrubs and grassland
Grids that >25% are covered by shrubs or grassland in 2020.

Others
Remaining grids that are not classified as above types, including deserts or permafrost areas.
be explained (Leroux et al 2017, Liu et al 2020a. We removed highly correlated variables (coefficients >0.75) with high average absolute correlation by computing pair-wise correlations to avoid variables with strong collinearity affecting their interpretability (Berner et al 2020). In this way, the driving factors, including temperature, precipitation, CO 2 concentration, tree cover and population change, were preserved to establish the attribution model. The time series of NDVI and its potential drivers at pixel level were randomly partitioned into the training set (70%) and validation set (30%). To quantify the contribution of each driving factor, the feature importance was computed based on the mean decrease accuracy in the attribution model . To reduce uncertainties of the RF model, we repeated the training 20 times and computed the mean value of the determination coefficient (R 2 ) and standard deviation (SD) as overall model performance. Our results demonstrated that over 83% of the vegetation greenness trend in China was explained by RF models. The RF model was constructed and evaluated using the RF software package in R.
For comparison, we further quantified the independent explanation ability of each selected variable to the variation of vegetation in urban and rural areas by developing GLM regressions (Tao et al 2015, Girardin et al 2016. Different from ordinary least squares, standard machine learning algorithms and partial least square-based models, GLM regression broadens the distribution range of dependent variables and employs a continuous function, offering a simple and flexible framework for nonlinear regression where X i are explanatory variables (i.e. temperature, precipitation, CO 2 concentration, tree cover and population change), representing driving factors of greenness change; Y is dependent variable; µ i represent n independent samples subject to exponential distribution family; η i refer to k linear combinations of explanatory variables; g (µ i ) is a link function that links µ i and η i . Through the link function, GLM can be applied to the data without the compliance of a specific normal distribution shape, such as normal, Poisson, gamma and binomial distributions (Lopatin et al 2016). Thus, the weight of each explanatory variable and their relative contributions were quantified through GLM. All analyses were performed in R Version 3.6.1 (Smith and Warren 2019, Adeleye et al 2021). Other statistical methods and results were described in the supplementary materials.

Greenness dynamics in urban and rural areas
Our results indicated that at the national scale, vegetation greenness in China increased 13.71% from 2000 to 2020, with an increase rate of 0.0025 ± 0.001 yr −1 (P < 0.05) (figure S1, table S3). Most (96%) of the significant changes were greening (significant positive), accounting for 65% of the total area. Due to divergent local environmental conditions, this widespread vegetation greening trend exhibited conspicuous spatial heterogeneity across China's urban and rural areas (figures 2 and S1). By dividing urban and rural areas from a national ecological-spatial perspective, we assessed the role they played in national greening trends. We found that both urban and rural vegetation exhibited a greening trend, which increased 4.96% (0.0011 yr −1 ) and 14.25% (0.0026 yr −1 ) from 2000 to 2020, respectively (figures 2, 3(a) and table S3). In urban areas, more than half (51.8%) of the area showed greening, accounting for approximately three-quarters (78.8%) of the significant trends, and browning (significant negative) trend occupied 13.9% of the urban areas ( figure 3(c)). By contrast in rural areas, the greening trend was more extensive, accounting for 68.8% of the rural area, and only 2.1% of the rural areas showed a browning trend ( figure 3(c)). Urban cores showed the largest magnitude of trends (0.0043 ± 0.0035 yr −1 ), which exceeded the greening rate of all rural subregions ( figure 3(b)).
The relative contribution of urban and rural vegetation change to China's greening trend was 11.3% and 88.7%, respectively (figure 4). High agricultural pressure areas and forest areas in rural areas were the main contributors to vegetation greening in China, explaining 29.5% and 28.1% of the greening trend, respectively. For the browning trend in China, urban areas (43.1%) contributed less to the browning trend than rural areas (56.9%). The largest source of China' browning trend was the urbanization areas, which explain 24.4% of the browning trends with the area occupying only 2% of the whole nation.
Our NDVI-based greenness change patterns in urban and rural areas show clear consistency with those based on kNDVI, NIRv, and EVI time series from 2000 to 2020 (figure 5). All these four vegetation indices revealed increasing vegetation trends in both urban and rural areas, although kNDVI exhibited more variable trend magnitudes (higher SD, 0.0032 ± 0.0013 yr −1 and 0.0033 ± 0.0011 yr −1 in urban and rural area, respectively) and NIRv less variable magnitude (0.0013 ± 0.0006 yr −1 in urban areas and 0.0017 ± 0.0004 yr −1 in rural areas). Urban areas showed a predominant greening trend in all four vegetation indices, with a median greening trend area accounting for a 55% of the urban areas (ranging from 54% for NIRv to 62% for kNDVI data). From the urban core to the urban periphery, the trend and magnitude of vegetation greening showed a consistent stepwise decrease in NDVI, kNDVI, EVI, and NIRv data. In rural areas, the proportion of greening trends in NDVI was also similar to the results of kNDVI (78%), NIRv (82%) and EVI (85%), and the areas with significant greening trends were consistently concentrated in the middle reaches of the Yangtze River and the Loess Plateau region (figures 2 and S3-S5).

Detection of turning point and trends shift with NDVI time series
In addition to spatiotemporal changes, we also detected the timing and trend shifts of greenness time series to reveal abrupt changes in vegetation during 2000-2020. Our results showed that the year when the turning point of greenness trend occurred spanned from 2005 to 2014 and exhibited heterogeneity across urban and rural areas (figure 6(a)), although previous studies tend to set the same year for the turning points of greenness trends for all regions of China (Ma et al 2019, Zhao et al 2022. Over the past 20 years, 63% of vegetation in China had experienced a shift in greenness trends, with a higher proportion in rural areas (62%) than in urban areas (54%) (figure 6). Before the turning points, greening trends in urban and rural areas accounted for a similar share, with proportions of 64% and 65%, respectively (figure 6(c)). After the turning points, however, greenness changes in urban and rural areas presented a conspicuous difference. Large-scale greening trends were observed in rural areas, accounting for 72% of the significant greenness change, which was higher than the proportion (52%) of greening in urban areas ( figure 6(d)).
There was a clear difference in the vegetation trend shifts between urban and rural areas. In urban areas, the most prominent trend shift was greening-to-browning reversals, accounting for 41% of the areas where vegetation trends have shifted (figure 6(e)). By contrast in rural areas, greening-togreening (interrupting greening) trend prevailed during the study period and accounted for the largest proportion (40%) among the four trend shifts types. Among all subregions in both urban and rural areas, the urban cores showed the largest proportion of browning-to-greening shifts (43%), indicating a widespread reversal of vegetation trends in this region. Different from urban cores, suburban and urbanization areas underwent large-scale shifts from greening to browning, occupying 48% and 43% of the total trend shifts, respectively. By comparison, the types of greenness trend shift in all six rural subregions were dominated by interrupting greening.

Drivers of greenness change in urban and rural areas
To explore the key drivers of greenness change in urban and rural areas in China, RF models were first established to associate the vegetation greenness trend category (i.e. greening, browning, or nonsignificant trend) with CO 2 concentrations, climate factors, tree cover, and population density change (figure 7). Our results suggested that tree cover was the largest contributor to greenness change in urban and rural vegetation. Tree cover explained more than one-third (36 ± 13%) of urban greenness change, higher than its contribution (29 ± 21%) to rural vegetation ( figure 7(g)). Except for tree cover, other drivers of vegetation greenness change between urban and rural areas showed clear inconsistency.
In urban areas, the population was the second largest driver of vegetation, with a contribution of 20 ± 9%, followed by CO 2 concentration (18 ± 4%). By contrast, in rural areas, CO 2 concentration was the second largest driver, explaining 26 ± 4% of the greenness change, and the population contributed the least to vegetation (6 ± 7%). The relative contribution of climate change to the greenness change in urban areas (14 ± 6% for precipitation and 12 ± 9% for temperature) was generally smaller than that in rural areas (23 ± 4% for precipitation and 16 ± 3% for temperature).
We use the GLM model to further complement the RF model results by identifying the positive and negative effects of the driving factors on vegetation change. The GLM model revealed the strongest positive correlation between greenness changes and tree cover in urban and rural areas (figure S7), which was consistent with the results from RF models. The GLM analysis also showed a negative correlation between population and greenness in urban areas, particularly in the urban core where the correlation coefficient was −0.23. By contrast, in rural areas, the population change showed an overall slightly positive correlation with greenness changes in all rural subregions.

Difference in greenness trend among cities with different development levels
To reveal the impact of urban development on urban vegetation change. We further compared the direction and amplitude of greenness change among cities with different development levels (figures 8 and S9-S11). Tier 1, Tier 2, and Tier 3 cities, which were defined by the National Bureau of Statistics of China according to city size and economic aggregate, were selected to represent cities from relatively high to low levels of development. We found that Tier 1 cities in the urban cores showed the fastest NDVI trend of 0.0034 ± 0.0061 yr −1 , which was 32.4% higher than that in Tier 2 (0.0023 ± 0.0042) and 52.9% higher than that in Tier 3 cities (0.0016 ± 0.0047). Tier 2 cities in urbanization area showed a browning trend (−0.0013 ± 0.0071) compared with Tier 1 (0.0013 ± 0.0064) and Tier 3 (0.0011 ± 0.0052) cities which exhibited a trend of greening, and the proportion of Tier 2 cities with vegetation browning reached 80% (tables S4-S7). This difference is related to the development stage of the cities, where the rapid urbanization expansion of Tier 2 cities exerted pressure on the vegetation surrounding the cities. In contrast, Tier 1 cities have entered a mature and stable stage, while Tier 3 cities have not achieved rapid urbanization expansion.

Discussion
Urbanization is generally assumed to lead to a decline in vegetation cover in urban areas through encroachment on agricultural land and natural vegetation. For example, Liu et al (2019) found that urban expansion reduced nationwide net primary production by 40.56 (Tg C) from 2000 to 2010 in China. Our results showed that although the vegetation trend in urban areas was dominated by browning from 2000 to 2010 ( figure S6(a)), there was a clear reversal of vegetation trends after 2010 ( figure S6(c)). Especially in the urban core, the proportion of vegetation showing shifts from browning to greening and greening to greening exceeded two-thirds (figure 6). This enhanced urban greening trend was related to the construction of green spaces in urban areas. Since 2012, when China implemented the 'ecological civilization' (Xu et al 2017), the development of parks and green spaces in urban areas has been notably prominent. Indeed, all Tier 1 cities showed a significant increase in tree cover (figures 8 and S12). Particularly for Beijing, 92% of the area exhibited an upward trend of tree cover (figure S12(b)). Another reason for greening urban areas was that urban expansion mainly occurs on croplands (figure S12(c)), a land use type with relatively low aboveground biomass relative to forests (Hwang et al 2022, Zhang et al 2022c. Our results suggest that rapid urbanization and enhanced vegetation greening are not mutually exclusive. As the level of urban construction and management improves, the vegetation in urban areas will play a positive role in promoting a nationwide greening trend.
Unlike urban areas, which focus on the construction of green spaces in residential agglomerations, the widespread greening trend in rural areas has been realized mainly through the large-scale afforestation projects carried out in China since the late 1990s. Afforestation projects aimed at protecting existing forests and reversing forest degradation, with the goal of mitigating ecological problems such as soil erosion and desertification (Qu et al 2018). We found that cropland and grassland accounted for 50.4% and 42.3%, respectively, of the new forests in rural areas over the past 20 years (figure S12). Regions with significant growth in tree cover were concentrated in areas where major afforestation and reforestation programs were implemented (figure S13). This supports previous research that afforestation and OL represent urban core areas, urbanization areas, suburban areas, towns, high agricultural pressure areas, low agricultural pressure areas, forests, shrubs and grassland, and others, respectively. and reforestation are primary drivers of the greening trend in China (Yu et al 2022) and demonstrates China's efforts to conserve and restore forests in response to global initiatives such as New York Declaration on Forests and the Bonn Challenge (Qu et al 2020, Zhao et al 2022. Climatic factors, especially precipitation, had a larger contribution to rural vegetation than to urban vegetation. Areas where precipitation positively affects vegetation were primarily located in the farming-pastoral mixed zone of northern arid and semi-arid regions ( figure S7). For widespread short-rooted grasslands and croplands in this zone, water availability was the dominant control for plant growth (Guo et al 2020a, Miao et al 2021. In contrast, temperature had a lesser impact on rural vegetation compared to precipitation. This is because as temperature increases, the constraints on vegetation growth in middle and high latitudes in China are reduced (Wang et al 2011. In contrast, climatic factors had a limited impact on urban vegetation. This is mainly due to the fact that vegetation in urban areas is strongly influenced by human management (e.g., irrigation, etc) and is therefore less constrained by precipitation, temperature, or nutrient availability than natural vegetation in rural areas and matches previous studies which found that urban management had a significant positive impact on vegetation in highly urbanized cities in cold and arid regions (Zhang et al 2022b). . Spatial patterns of NDVI trends for different city development levels (Tier 1, 2, or 3 cities) and different urban intensity levels (urban core areas, urbanization areas, or suburban areas). The spatial distribution of the NDVI trends for the cities with different development levels in (a) urban core areas, (b) urbanization areas and (c) suburban areas. (d) NDVI trends vary with different city development levels and different urban intensity levels. NDVI trends are estimated using the Mann-Kendall test. Box plots indicate the interquartile range (IQR, box), the median (central line), and whiskers extended to 1.5 (−1.5) times the IQR.
An overall stagnation in vegetation productivity growth was observed in recent years in China (Ma et al 2019, Zhang et al 2022c, which may be related to the rapid adoption of fast-growing tree species in rural areas following the implementation of afforestation policies in 2000. Although this stagnant trend is not yet evident in our vegetation index-based greening detection, the growth of greening trends in rural areas may also tend to slow down as environmental factors (e.g., CO 2 fertilization or warm temperature) reach a natural limit (Piao et al 2020). This requires the implementation of more effective strategies (e.g., enhanced forest management, site-specific tree species selection) to achieve vegetation restoration and carbon neutrality goals.
Although greenness changes in both urban and rural areas were assessed in this study, however, other factors driving urban and rural greenness changes, such as forest management, cropland irrigation and nitrogen deposition effects, need to be further incorporated into the attribution analysis , Xu et al 2019, Dara et al 2020. Furthermore, land use changes resulting from urbanization were primarily represented by the addition of artificial surfaces between 2000 and 2020 in our analysis. Further research is required to distinguish the impact of vegetation replacement by artificial surfaces specifically.

Conclusion
This study enhances our understanding of the roles of urban and rural areas in vegetation greenness change in China by leveraging multiple satellite-based vegetation indices. Our results revealed that both urban and rural areas contributed positively to the vegetation greening trend in China, with their contribution being 11.3% and 88.7%, respectively. We also found that the dominant driver of both urban and rural greening trends was tree cover, accounting for 36% and 29% of urban and rural greening, respectively. However, how this increase in vegetation coverage was achieved differed between urban and rural areas, with green space construction in urban areas and afforestation programs in rural areas being the main drivers. Climate factors, particularly precipitation, had a greater impact on rural areas than urban areas. Given the dissimilar ways urban and rural areas have contributed to China's greening, targeted strategies should be implemented in urban and rural areas for future vegetation restoration and carbonneutral efforts.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Code availability
All code used for preprocessing the data and creating the plot in this analysis is freely available on request from the author.