Three large cities in Northeast China (Shenyang, Changchun and Harbin) and their surrounding rural areas are taken as the study units. The wavelet transform method is used to reduce the noise data of long time series Moderate-resolution Imaging Spectroradiometer (MODIS) enhanced vegetation index (EVI) data and reconstructed smooth EVI curves in pixel scale. Based on EVI curves, the key vegetation phenological parameter indexes from 2009 to 2016, namely the start of growing season (SOS) and the end of growing season (EOS) are extracted by using the dynamic threshold method. The spatio-temporal change of vegetation phenology and their responses to surface temperature in each study unit are analyzed. The results show that the differences in the spatial distribution of the SOS and the EOS in each study unit are significant between urban area and rural area. The area where a smaller value of SOS is taken is consistent with the area where a larger value of EOS is taken, and vice versa. The spatial distribution of vegetation phenology indexes and their geographical locations (either urban or rural) are highly consistent. The closer the actual location of each pixel is to its urban center, the smaller the SOS value is and the larger the EOS value is, which indicates an early start and a late end of the growing season, that is, a longer growing season. From 2009 to 2016, the average annual SOS values in these 3 cities form an early center in the urban areas and show a tendency of gradual delay from the urban area to the rural area, whereas the average annual EOS values form a late center in urban areas and show a tendency of gradual advance from the urban area to the rural area. The SOS is between the 99th and the 112nd day in urban areas, while the SOS is between the 125th day and the 135th day in rural areas. The time difference of urban-rural SOS is about 3 weeks. The EOS is between the 293rd and the 300th day in urban areas, while the EOS is between the 284th and the 286th day in rural areas. The urban EOS is approximately 2 weeks later than that in rural areas. The inter-annual change trend of vegetation phenological parameter indexes in each study unit has some similarities, that is, SOS shows an early trend with time, and the trend of SOS inter-annual change in urban and rural areas is consistent but with different rate. The maximum of the SOS value during the study period occurred in 2012 in the study areas, which was the coldest year in terms of vegetation phonological indexes and this coincided with the climatic phenomena of extreme weather such as blizzard and low temperature, which were affected by the cold wave in the year. The average annual land surface temperature (LST) of each study unit is significantly correlated with the corresponding vegetation phenological indexes. There is a significant negative correlation between SOS and LST, and a high positive correlation between EOS and LST. The higher the average temperature of the same vegetation phenological period, the earlier the start time of vegetation growth season, and the later the end time of vegetation growth season.