**4. Discussion**

Plant phenology is one of the key regulators of ecosystem processes, which is sensitive to environmental change. The acceleration of urbanization in recent years has produced substantial impacts on vegetation phenology over urban areas, such as the local warming induced by the urban heat island effect. This study explored impacts of urbanization on SOS and distinguished contributions of ΔLST and other factors to ΔSOS based on a temperature contribution separation model. We found that the SOS was negatively correlated with the daytime LST in spring, and the ΔSOS dominated by the ΔLST and other factors contributed 72% and 28% to the ΔSOS, respectively. Previous studies showed that there were lots of aspects besides temperature were different between urban and rural under urbanization, which had a certain impact on plant phenology. (1) The land cover changes under urbanization changed the soil properties extremely in the urban, affecting the relationship between plants, water and nutrients [49]. (2) There were high concentration of greenhouse gases (such as CO2) and major pollutants (such as NO, NO2, CO, SO2 and particulates with a diameter of 10μm or less) in the urban, resulting from the emissions produced by factories and automobiles [50]. In this regard, many studies showed that pollutants in the urban environment could cause the advance or delay of plant phenology [51–53]. (3) Due to the increasing artificial light caused by the human activities at night in the urban, the growth of plants was seriously influenced [54,55]. Therefore, there were various differences between the urban areas and their rural surroundings, not only in terms of temperature, but also in other aspects that affect plant phenology.

The LST in spring from February to April was selected to explore the relationship between LST and SOS in this study. Previous studies showed that meteorological parameters such as temperature and precipitation in a period of time before the phenological event were important determinants affecting the occurrence of phenology. The period of time is significant to the study of the relationship between phenology and climate [56,57], which called preseason duration. Polgar et al. found that the temperature in late winter and spring or preseason temperature played an important role in the occurrence of SOS [58]. Zhou et al. and Jia et al. found that the LST showed a statistically significant correlation with SOS (*p* < 0.05, R<sup>2</sup> > 0.8) [7,15]. In addition, different temperature indicators were used to explore its effect on the SOS, such as daily maximum temperature and diurnal temperature difference. (1) Piao et al. showed that the SOS was more sensitive to the preseason daily maximum temperature in the northern hemisphere, and 68% of the European Union and 83% of the United States had a preseason duration of 0–3 months [59]. (2) The results of Huang et al. showed that 77.2% of the northern hemisphere, the SOS had the strongest correlation with the average diurnal temperature difference in the preseason period of 1–3 months [60]. In general, the results above indicated that the temperature in spring was relevant to SOS with the preseason duration of 0–3 months.

At the same time, there were some limitations in this study. (1) As shown in Figure 10, the LST during the daytime showed a statistically significantly negative correlation with SOS both in the urban (*p* < 0.05) and the rural (*p* < 0.01), while the R<sup>2</sup> was low. The low R<sup>2</sup> might be related to the impurity of the data where existed many mixed pixels in the urban sample. To explore the changes in plant phenology caused by the urban heat island effect, we needed to focus on forest where is severely affected by urbanization. However, the MODIS EVI data with a resolution of 250 m was the data with long time series, the highest resolution and we could currently obtain. Therefore, the problem of mixed pixels inevitably existed in the urban, which affected the correlation between SOS and LST. Although the R<sup>2</sup> was relatively low, they were both statistically significant (*p* < 0.05), which was meaningful in a certain degree. In the future, data with higher resolutions should be used to reduce the uncertainty caused by the data and make results more reliable. (2) In this study, all types of forest in Hangzhou were used to explore the impact of urbanization on plant phenology. Previous studies found that the phenology and the response to urbanization varied in the different types of vegetation. However, due to the limitation of the accuracy of data, we could only exclude shrub, farmland and grassland. Relatively uniform and stable forests were extracted as the study object to weaken the impact of different vegetation types to a certain extent. Reliably, previous studies utilized all vegetation types for the study and go<sup>t</sup> reliable results, having a certain significance [7,33]. In future research, we hope to more finely distinguish the vegetation types and improve the accuracy of the study, in order to obtain more reliable results. (3) As mentioned above, there were lots of aspects besides temperature that were different between urban and rural under urbanization, which had a certain impact on plant phenology. Although other factors were taken into consideration in this study, no specific data analysis was carried out on other factors such as greenhouse gases. In future studies, a more quantitative and detailed discussion on the effects of other factors on vegetation phenology should be advanced. With the development of high-quality data and online data processing platforms (e.g., Google Earth Engine), it is feasible to apply the methodology of our present work to other study areas even to a global scale.
