Our study area covers the eastern coastal region of China (108°12′E∼135°05′E, 20°54′N∼53°33′N), including 13 provinces, which are distributed in three different zones (see Figure 1, zones II to IV). Zone II and zone III belong to the monsoon climate of medium latitudes, and zone IV belongs to the subtropical monsoon climate. They all have distinct seasons, with higher temperatures and more precipitation in summer. These three different zones belong to other vegetation types, including temperate coniferous and broadleaved mixed forest (zone II), warm temperate broadleaved deciduous forest (zone III), and subtropical evergreen broadleaved forest (zone IV). After data screening, we selected phenological records from 17 phenological observation sites (Figure 1), and these observation sites were located in parks or school gardens in cities. Zone II contains one site in Heilongjiang. Zone III contains one site in Liaoning, two in Beijing, three in Shandong, one in Shanxi, and one in Henan. Zone IV contains two sites in Jiangsu, one site in Zhejiang, one in Hunan, one in Chongqing, one in Guangxi, one in Guizhou, and one in Sichuan.

The phenological data come from the Chinese Phenological Observation Network (CPON) established in 1963 (http://www.geodata.cn), administered by the Institute of Geographical Sciences and Natural Resources Research (IGSNRR) and the Chinese Academy of Sciences (CAS). The weather data were collected by the Chinese Meteorological Agency (http://data.cma.cn/). Population and GDP figures come from the National Bureau of Statistics of China (https://data.stats.gov.cn/).

To discriminate the different effects of factors, linear fitting regressions of temperature, precipitation, population, and GDP were conducted with the GSL. The fitting results were tested for significance by the two-tailed method. The correlation coefficients were calculated by Pearson’s and Spearman’s rank correlation analyses. Significant differences in factor effects between the sites were obtained by one-way analysis of variance (ANOVA). In the correlation, a p-value less than 0.05 was considered significant, and a p value less than 0.01 was considered extremely significant. Principal component analysis (PCA) was used to determine the importance of influencing factors on the GSL in each sub-region. All data were statistically analyzed by IPM SPSS Statistics 22. The above analysis is presented by Origin pro 8 in the form of a figure. The spatial distribution of the correlation of linear analysis is a product made by ArcMap 10.5.

By calculating the annual mean temperature and the annual total precipitation over all study sites, we obtained their changes from 1963 to 1988 (Figure 2). The temperature changes in this period showed a trend in fluctuation and increase at the whole regional scale, with obvious inter-annual differences. The maximum annual mean temperature is 14.36°C, and the minimum value is 12.58°C. During this period, the precipitation showed a trend of fluctuation and decline at the scale of the whole area, and there was a great difference in precipitation between years. The maximum annual mean precipitation was 1,097.09 mm, with a minimum of 816.74 mm.

Due to the influence of geographical location and topography, the climate of subregions in the whole area varies greatly. The average temperature in the subregion was the highest in Guangxi (19.31°C) and the lowest in Heilongjiang (3.64°C). The annual total precipitation of subregions was the highest in Guangxi (1,869.8 mm) and the lowest in Heilongjiang (530.7 mm). The temperature of most regions is increasing, and the most significant is in Guangxi and Liaoning. Three subregions showed a decreasing trend in temperature, namely, Sichuan, Chongqing, and Henan. Among them, the temperature decrease in Sichuan is the most significant. There are five sub-regions with increased rainfall, none of which is extremely significant, and only Zhejiang is significant. There are seven subregions with reduced rainfall, but decreased rainfall is significant only in Shandong and Guizhou. Overall, temporal variations in temperature are more significant than precipitation variations.

The temporal and spatial dynamics of anthropogenic factors in the study area were obtained by the statistics of population per unit area and GDP per capita of each sub-region each year (Figure 3; Supplementary Tables S1, and S2). The population increased from 28.45 million per unit area to 43.21 million per unit area, and the corresponding per capita GDP increased from 0.018 to 0.145.

Beijing is the most populous and developed, with unit area population and unit area GDP of 631.5 × 10⁴ km⁻² and 244.2 × 10⁸ RMB km⁻² in 1988, respectively. The northernmost province of Heilongjiang is the most sparsely populated, with a unit area population size of approximately 76.2 × 10⁴ km⁻² in 1988. Its unit area GDP is slightly higher than that of Guizhou, only 12.1 × 10⁸ RMB km⁻². In terms of quantitative sorting, the order of population and GDP in the eastern provinces of China was basically the same in 1988. In other words, it shows that densely populated cities are more developed.

The population growth and development of different provinces were quite different from 1963 to 1988. The population growth rate of Beijing was the greatest in all provinces, reaching 9.33 × 10⁴ km⁻² year⁻¹. Heilongjiang also has the lowest population growth rate of 0.95 × 10⁴ km⁻² year⁻¹. Although the population of the unit area in Henan is not particularly large, its growth rate is only second to that of Beijing. The fastest and slowest growth in GDP is Beijing and Heilongjiang, approximately 13.38 × 10⁸ and 0.63 × 10⁸ RMB km⁻² year⁻¹, respectively. We found that the rate of urban development was almost synchronous with the rate of population growth.

The temporal dynamics of plant phenology in 1963–1988 were calculated and analyzed based on the germination stage, wilting stage, and the length of the growing season of the woody plants (Figure 4; Supplementary Figures S1–S3). During the period from 1963 to 1988, the length of the growth season of the whole region ranged from 183.0 to 213.2 days, with an average of 200.1 days. The germination stage was concentrated on days 92.5–116.8 of each year, with an average of approximately 102.1, and the years with the largest variation were concentrated in 1976, 1980, and 1987. The wilting stage occurred on days 277.0–296.5 of each year, with an average of approximately 286.5, and the years with the largest variation were concentrated from 1974 to 1978.

The spatial dynamics of plant phenology were calculated and analyzed in sub-regions (Figure 5). Spatially, the germination stage showed an obvious pattern of gradual delays from north to south. Although there was a similar pattern for the wilting stage, the delay was less pronounced in the interior southwest. This reduced the increases in the length of the growing season in these southern subregions. Germination stages were delayed in all other sub regions except Guizhou. The variations in germination stages in northern subregions, such as Heilongjiang, Shanxi, and Liaoning, were not significant, while those in southern sub-regions, such as Guizhou, Guangxi, and Zhejiang, were significant, indicating regional differences between the north and the south. Coastal areas were more likely to wither earlier. Inland Chongqing and Hunan, however, experienced a significant delay in wilting. The growing season lengthened only in two northeast and two southwest subregions, especially in Chongqing and Guizhou.

Climate change has widely affected urban plant phenology (Ren et al., 2018; Wohlfahrt et al., 2019). We found a significant correlation between plant phenology and warming, consistent with other studies, which indicated that higher temperatures lengthen growing seasons (Menzel, 2003; Badeck et al., 2004; Cleland et al., 2007). However, germination has been suggested to be more sensitive to daytime warming than nighttime warming, both at the species and ecosystem levels (Piao et al., 2015; Rossi and Isabel, 2017). Due to the acceleration of nighttime warming over the past few decades (Davy et al., 2017), any method of reflecting this asymmetric warming may lead to an underestimation of the temperature sensitivity at the start of the growing season.

Compared with non-populated western China (Chen et al., 2011; Shen et al., 2011), our study found that precipitation has no significant influence on plant phenology, unlike temperature, in eastern China (Figure 6B). In addition, our results suggested that the influence of precipitation in the north was more significant than in the south, with regional differences. The differences are probably affected by water status (Jaworski and Hilszczański, 2013). In the regions with abundant groundwater, the dependence on precipitation was weak, leading to a weak significant correlation (Ballestrini et al., 2010).

Studies have shown that some species are resilient to climate change, but different species respond at different rates (Burns et al., 2003; Visser et al., 2004; Visser and Both, 2005), reflected in our results by the fact that plant phenology is affected by the different climates and vegetation in the sub-regions. Temperature has a great influence on the length of the plant-growing season in Guizhou, especially the preseason temperature, with a significant correlation (germination and wilting). This may be caused by the special geographical location and climate conditions of Guizhou. Temperature there was lower at high altitudes (with an average altitude of 1,100 m); thus, its plant phenology was more sensitive to temperature fluctuations (Ke et al., 2016). In Zhejiang, Guangxi, and Sichuan, the negative correlation between temperature and growing season length of local plants suggested that the earlier start and end of the growing season caused by climate warming cannot compensate for each other. The germination stage of plants may not be temperature-restricted in these regions (Meng, 1988). The variation in growing season length in Beijing is significantly correlated with temperature but not significantly correlated with precipitation, and the reason for this may be that the large population of Beijing and economic development caused more heat (Zhou and Zhang, 2003; Li et al., 2005). Now that plant phenology in subregions such as Beijing is mainly affected by local climate change, including the urban heat island effect, and light effect (Jochner and Menzel, 2015; Li F. et al., 2017), more data support is needed to separate these human activities from climate change.

Urbanization affects plant phenology in different ecosystems (Roetzer et al., 2000; Zhou and Zhang, 2003; Li et al., 2005). Similar to other regions of the world, this phenomenon has occurred in China; for example, satellite data showed that urbanization in the Yangtze River Delta region also lengthens the growing season (Han et al., 2008). On the one hand, phenology is affected by the urban heat island effect (Zhang et al., 2004; Jochner and Menzel, 2015). On the other hand, our results showed that the changed population had a greater impact on urban plant phenology than GDP, which indirectly reflected the effect of heat increase caused by population change, consistent with previous studies (Li X. et al., 2017).

We found that plant phenology is more influenced by climate than human activity in some urbanized areas (Figure 9). The effects of urbanization on phenology vary with vegetation types (Li X. et al., 2017; Ren et al., 2018), but these regional differences are not just due to differences in the types of vegetation. Combined with our results and some results from other researchers (Yan-Ming et al., 2004; Jochner and Menzel, 2015; Gunawardena et al., 2017), the human management of urban plants has a critical impact on urban plant phenology. The industrial development of the emerging urban agglomeration was late, with a scientific economic structure, and the change in the proportion of the tertiary industry had a great impact on the length of growing season. This suggests that these cities may have invested more in urban plant management, which counteracted the effects of climate change on urban plant phenology (Supplementary Figure S4).

The rapid growth of the region despite a small population base gave rise to the lengthening of the growing season being most affected by population changes in Guangxi (Figure 9). With the approval of the central government, Guangxi urgently built and/or rebuilt 33 frontier highways in 1979, which greatly influenced local plant phenology (Wei, 1998). The subregion with the most significant impact of GDP change is Chongqing. Due to its unique location and strategic position, although this region is located in western inland China, the GDP increased significantly, which had a key important impact on the lengthened growing season of urban plants (Chen et al., 2007). In the context of global warming, management and policies (e.g., urban greening) should pay more attention to maintaining the stability of urban plant phenology.

Although we tried our best to select plant data from the same species for analyses, our data cannot fully represent the whole region due to the differences in vegetation types in large-scale cross-regional studies. Second, this study obtained data through field observations, and the sample size of selected sites and samples was smaller than that of other methods, such as remote sensing.

In addition, since the statistical data of the urban population and GDP cannot be obtained, the inter-annual changes in population and GDP per unit area of the sample point on the provincial scale were used to measure the changes in population and GDP. Such data are used to analyze their impact on plant phenology, which probably underestimates their impact on plant phenology due to the higher population and GDP in cities than in the countryside. In addition, the establishment of measurement standards for human factors is also of great research value, such as the establishment of urbanization index and land-use type change index, all of which need further research and exploration.