**3. Results**

#### *3.1. Temporal and Spatial Variation in Vegetation Phenology*

The vegetation phenology derived from the satellite data was consistent with ground observations at the Xiying River, Liancheng, Suganhu, and Sidalong stations in 2020 and the Haibei station from 2006 to 2015. The correlation coefficient (R2) between the SOS and field data was 0.536 (*p* < 0.01), the mean absolute error (MAE) was 9 d, the root mean square error (RMSE) was 11 d, the R<sup>2</sup> was 0.533 (*p* < 0.01), the MAE was 5 d, and the RMSE was 6 d between the EOS and field data (Figure S1). Based on the validation results described above, the remote sensing monitoring method adopted in this paper can accurately reflect vegetation phenological characteristics in the QLMs.

The interannual changes in vegetation phenology in the QLMs from 2001 to 2020 showed different fluctuation ranges (Figure 2). There was an advanced SOS trend of 0.510 d/year and an extended LOS trend of 0.580 d/year. There was a delayed EOS trend at a rate of 0.066 d/year, which is only a slight change. However, no significant changes were found in these vegetation phenology parameters (*p* > 0.05).

The vegetation phenology parameters varied with altitude (Figure 3). With an increase in altitude, the SOS showed a gentle upward trend, but the correlation between the SOS and altitude was weak (*p* > 0.05). Conversely, with an increase in altitude, the EOS gradually advanced and the LOS gradually shortened. There was a significant negative correlation between the altitude and both EOS and LOS (*p* < 0.05), and the correlation coefficients were high (R<sup>2</sup> ≥ 0.899). The SOS tended to be delayed by 0.20 d/100 m, while the EOS

tended to advance by 0.60 d/100 m, and the LOS tended to extend by 0.80 d/100 m with increasing altitude.

**Figure 2.** Interannual changes in vegetation phenology in the QLMs.

**Figure 3.** Characteristics of the changes in the SOS (**a**), EOS (**b**), and LOS (**c**) with altitude.

The multiyear average spatial distribution of phenology in QLMs from 2001 to 2020 is shown in Figure 4a,c,e. From east to west, the vegetation phenology showed evident changes. Overall, the SOS in the study area mainly occurred from 115 days to 150 days, which accounted for more than 80% of the vegetation region. The earlier SOS was mainly seen in the eastern and western QLMs, and the later SOS was mainly distributed in the central section. In addition, the multiyear mean EOS of vegetation phenology varied between 255 and 275 d (more than 80% of the overall pixels) from the middle of September to early October. The EOS showed the opposite pattern in terms of spatial distribution compared with the SOS; it was earlier in the central section of the QLMs and later in the western and eastern sections of the QLMs. Due to the combined effects of SOS and EOS, the LOS was mainly between 110 and 160 d. The spatial pattern of LOS was similar to that of EOS, whereby the LOS was shorter in the central section of the QLMs and longer the eastern and western sections of the QLMs.

**Figure 4.** Spatial distribution and the change trend of vegetation phenology parameters. The left side represents the spatial pattern of the average value of the SOS (**a**), EOS (**c**), and LOS (**e**) from 2001 to 2020. The right side represents the spatial distribution of the change trend of the SOS (**b**), EOS (**d**), and LOS (**f**) from 2001 to 2020. The trends are considered significant for pixels according to the MK test (*p* < 0.05).

Figure 4b,d,f and Table 1 show the spatial distribution of the vegetation phenology trend in the QLMs from 2001 to 2020. A total of 72.37% of the vegetation pixels showed an advancing trend of SOS from 2001 to 2020. A total of 13.85% of pixels, which were mainly concentrated in the central and eastern sections of the QLMs, showed a significant advancing trend of SOS. A few areas in the northwest of the QLMs showed a delayed trend of SOS, whereas only 1.44% of the total land area was significantly delayed. Regions with delayed EOS accounted for 47.59% of the vegetation pixels in the study area from 2001 to 2020 and were mainly located in the eastern and central sections of the QLMs. In addition, the areas with advanced EOS were mainly located on the northern margins and at the northwest of Qinghai Lake. Approximately 6.8% of vegetation pixels had a significant delayed trend in terms of the EOS, and 3.9% of vegetation pixels had a significantly advanced trend. There was an overall extended LOS trend for most parts of the vegetation area (71.66% of the vegetation pixels) from 2001 to 2020, with 12.65% being significantly extended and only 1.87% being significantly shortened. The areas with extended LOS trends were mainly distributed in the central and eastern sections of QLMs.

**Table 1.** The percentage of different trends of vegetation parameters based on MK analysis across the QLMs.


#### *3.2. Response of Vegetation Phenology to Seasonal Driving Factors*

The spatial distribution of the partial correlation coefficients between seasonal driving factors and vegetation phenology metrics are displayed in Figure 5. For the QLMs, the SOS was negatively correlated with spring temperature in 73.81% of vegetation pixels, while 21.21% of pixels showed a significant correlation (*p* < 0.05) and were mainly located in the eastern and central parts of the study area (Figure 5a). The percentage of negative and positive correlations between the SOS and spring precipitation was similar (Figure 5c), with a significant negative correlation occurring in the northeast of Hala Lake. More than half of the vegetation pixels (61.34%) of the SOS had a negative correlation with spring soil moisture, of which 9.18% of pixels showed a significant negative correlation (Figure 5e), mainly at the west of Qinghai Lake. The results above indicate that the increases in spring temperature and soil moisture likely cause the SOS to advance in most part of the QLMs.

The partial correlation coefficients between the EOS and autumn temperature showed that the EOS was positively correlated with temperature in 65.20% of vegetation pixels, and 9.89% of the areas passed the significance test (Figure 5b). Approximately 56.19% of the vegetation pixels showed a negative correlation between the EOS and autumn precipitation, and only 4.75% of the areas passed the significance test (Figure 5d). For autumn soil moisture, positive correlations between the EOS and soil moisture occurred in 60.44% of the total vegetation pixels, and approximately 9.50% of the pixels showed a significant positive correlation (*p* < 0.05, Figure 5f), most of which were distributed west of Qinghai Lake and southeast of Hala Lake. In total, the autumn temperature and soil moisture influenced the EOS in most areas, and the increase in autumn temperature and soil moisture likely caused the EOS to be delayed.

**Figure 5.** The partial correlation coefficients between vegetation phenology parameters and seasonal driving factors. (**<sup>a</sup>**,**c**,**<sup>e</sup>**) represent the partial correlation coefficients between the SOS and spring temperature, precipitation, and soil moisture, respectively. (**b**,**d**,**f**) represent the partial correlation coefficients between the EOS and autumn temperature, precipitation, and soil moisture, respec-tively. The inset panels on the bottom left of each subpicture present pixels with a significantly (*p* < 0.05) negative (blue) and positive (red) correlation. The percentages of positive (P) and negative (N) correlations (the values in brackets indicate the percentage of significant correlations) are shown at the top of each subpicture.

#### *3.3. Vegetation Phenology Parameters Response to Seasonal Driving Factors Based on Different Elevation Zones*

The driving factors had different effects on vegetation phenology depending on elevation. The results of the partial correlation analysis between the vegetation phenology parameters (SOS, EOS) and seasonal driving factors (temperature, precipitation and soil moisture) for different elevation zones are shown in Figure 6. A mainly negative correlation occurred between the SOS and spring temperature (the percentage of significant negative correlation ranged from 14.59% to 24.87%) at different elevation zones. At middle elevations (3000–4000 m a.s.l.), more than 76% of the vegetation pixels showed a negative correlation between the SOS and spring temperature (Figure 6a), and more than 24% of the areas passed the significance test. The SOS was mainly negatively correlated with spring soil moisture (more than 61%) at the <4000 m elevation zone, and the percentage of areas that passed the significance test at the 95% level ranged from 7.24% to 12.5% (Figure 6c). However, the SOS had the opposite correlation with spring soil moisture in the highest elevation zone (mainly positive), 7.63% of the areas showed a significant positive correlation (Figure 6c). Compared with spring temperature and soil moisture, spring precipitation had a weaker influence on the SOS at the four elevation zones, and the areas of positive and negative correlation between spring precipitation and the SOS were similar, with few pixels showing a significant correlation (Figure 6b).

**Figure 6.** Percentages of correlation between SOS and three driving factors at different elevation zones. The three driving factors were (**a**) spring temperature (Tspring), (**b**) spring precipitation (Pspring), and (**c**) spring soil moisture (SMspring).

At the lowest elevation zone (<3000 m a.s.l.), the EOS was positively correlated with summer temperatures and precipitation in 62.57% and 72.16% of areas, and 8.03% and 18.1% of the areas showed a significant correlation (*p* < 0.05), respectively (Figure 7b,e). Notably, a negative correlation between EOS and summer soil moisture occurred in 80.96% of vegetation pixels at the lowest elevation zone (<3000 m a.s.l.), which was more than four times larger than the positive correlation (19.04%). Approximately 30.06% of the pixels showed a significantly negative correlation between summer soil moisture and EOS (*p* < 0.05) at the lowest elevation zone (<3000 m a.s.l.), while areas with significant positive correlations represented only 0.59% of the total (Figure 7h), indicating that the EOS was advanced in most low elevation regions with an increase in summer soil moisture. A positive correlation between the EOS and autumn temperature covered more than 60% of the area in the different elevation zones, and more than 7.00% of the pixels exhibited a significant correlation (Figure 7c). For the region with elevations less than 4000 m, more than 8.81% of the areas demonstrated a significantly positive correlation between the EOS and autumn soil moisture (Figure 7i). For the regions with elevations of less than 3500 m, the area with a significant positive correlation between the EOS and autumn soil moisture was greater than the EOS and autumn temperature (Figure 7c,i), so at lower elevations (less than 3500 m), soil moisture played a more important role in vegetation growth in autumn. At higher elevations (higher than 3500 m), temperature played a more important role in vegetation growth in autumn (Figure 7c). Overall, compared with autumn temperature and soil moisture, autumn precipitation had a weaker influence on the EOS at the four elevation zones (Figure 7f).

**Figure 7.** Percentages of correlation between EOS and different driving factors at different elevation zones. The driving factors were (**<sup>a</sup>**,**b**,**<sup>c</sup>**) seasonal temperature (Tspring, Tsummer, and Tautumn are spring, summer, and autumn temperature, respectively), (**d**,**e**,**f**) seasonal precipitation (Pspring, Psummer, and Pautumn are spring, summer, and autumn precipitation, respectively), and (**g**,**h**,**i**) seasonal soil moisture (SMspring, SMsummer, and SMautumn are spring, summer, and autumn soil moisture, respectively).

#### *3.4. Vegetation Phenology Response to Seasonal Driving Factors across Vegetation Types*

Generally, different vegetation types had different responses to driving factors (Figures 8 and 9). For different vegetation types, the partial correlation coefficients between the SOS and spring temperature were mostly negative (Figure 8a), especially for broadleaf forests, needleleaf forests, shrubland, and meadows (more than 23% of regions had significantly negative correlations). The partial correlation coefficients between the SOS and spring soil moisture were also mostly negative (Figure 8c) for different vegetation types, except alpine vegetation, and more than 7.80% of areas had significantly negative correlations. For alpine vegetation, the SOS was negatively correlated with spring temperature in approximately 65.18% of areas, of which 12.14% of areas showed a significant negative correlation (Figure 8a). However, it was positively correlated with the spring soil moisture in 52.04% of areas, with 7.49% of regions showing a significantly positive correlation (Figure 8c). Compared with spring temperature and soil moisture, spring precipitation had a weaker influence on the SOS in most vegetation types. The areas of positive and negative correlation between spring precipitation and the SOS were similar, with few pixels showing a significant correlation (Figure 8b).

Compared with the correlation between the EOS and autumn temperature and soil moisture, there were limited positive and negative correlations between the EOS and autumn precipitation, and there were relatively few significant pixels for most vegetation types (Figure 9f). Autumn temperature and soil moisture had mainly positive correlations with the EOS for most vegetation types (Figure 9c,i), and the EOS had a more significant relation with soil moisture than temperature for grasslands and deserts. There was a negative correlation between EOS and spring temperature in more than 65% of areas of broadleaf forests, and 13.49% of the areas were significantly correlated (Figure 9a). For needleleaf forests, there was a negative correlation between EOS and spring precipitation in 66.11% of areas, and 10.64% of the areas were significantly correlated (Figure 9d). A significant positive correlation between the EOS and summer precipitation was found in 11.51% of pixels for broadleaf forest and 11.36% of pixels for needleleaf forest (Figure 9e). Summer and autumn soil moisture had opposite correlations with the EOS in QLMs, except meadows. The EOS was mainly negatively correlated with summer soil moisture, especially for broadleaf forests, needleleaf forests, and grasslands (approximately 15.58%, 14.71%, and 18.78% of the pixels had significant negative correlations, respectively). The EOS of most vegetation types were positively correlated with autumn soil moisture, especially those of grasslands and deserts (for which approximately 11.35% and 12.97% of the pixels had significant positive correlations, respectively).

**Figure 8.** PerPercentages of correlation between the SOS and three driving factors in different vegetation types. The three driving factors were (**a**) spring temperature (Tspring), (**b**) spring precipitation (Pspring), and (**c**) spring soil moisture (SMspring).

**Figure 9.** Percentages of correlation between EOS and different driving factors in different vegetation types. The driving factors were (**<sup>a</sup>**,**b**,**<sup>c</sup>**) seasonal temperature (Tspring, Tsummer, and Tautumn are spring, summer, and autumn temperature, respectively), (**d**,**e**,**f**) seasonal precipitation (Pspring, Psummer, and Pautumn are spring, summer, and autumn precipitation, respectively), and (**g**,**h**,**i**) seasonal soil moisture (SMspring, SMsummer, and SMautumn are spring, summer, and autumn soil moisture, respectively).
