Phenology Response to Climatic Dynamic across China ’ s Grasslands from 1985 to 2010

Because the dynamics of phenology in response to climate change may be diverse in different grasslands, quantifying how climate change influences plant growth in different grasslands across northern China should be particularly informative. In this study, we explored the spatiotemporal variation of the phenology (start of the growing season [SOS], peak of the growing season [POS], end of the growing season [EOS], and length of the growing season [LOS]) across China’s grasslands using a dataset of the GIMMS3g normalized difference vegetation index (NDVI, 1985–2010), and determined the effects of the annual mean temperature (AMT) and annual mean precipitation (AMP) on the significantly changed phenology. We found that the SOS, POS, and EOS advanced at the rates of 0.54 days/year, 0.64 days/year, and 0.65 days/year, respectively; the LOS was shortened at a rate of 0.62 days/year across China’s grasslands. Additionally, the AMT combined with the AMP explained the different rates (ER) for the significantly dynamic SOS in the meadow steppe (R2 = 0.26, p = 0.007, ER = 12.65%) and typical steppe (R2 = 0.28, p = 0.005, ER = 32.52%); the EOS in the alpine steppe (R2 = 0.16, p < 0.05, ER = 6.22%); and the LOS in the alpine (R2 = 0.20, p < 0.05, ER = 6.06%), meadow (R2 = 0.18, p < 0.05, ER = 16.69%) and typical (R2 = 0.18, p < 0.05, ER = 19.58%) steppes. Our findings demonstrated that the plant phenology in different grasslands presented discrepant dynamic patterns, highlighting the fact that climate change has played an important role in the variation of the plant phenology across China’s grasslands, and suggested that the variation and relationships between the climatic factors and phenology in different grasslands should be explored further in the future.


Introduction
Plant phenology plays an important role in the biogeochemical cycles of ecosystems, referring to periodic events (e.g., bud burst, flowering, and senescence) in the life cycles of living species [1], and is regarded as a 'fingerprint' for climate change [2].Recently, phenology change has been a heated topic in ecology [3], as it is a critical factor contributing to carbon cycle dynamics [4], and the changes in plant phenology have been associated with climate change in several studies (e.g., [5][6][7]).Plant phenology is the study of the timing of seasonal events that are considered to be the result of adaptive responses to climate variations on short-and long-time scales [8]; thus, phenology and climate are intimately linked [2].
For the plant phenology of northwest China, studies have primarily focused on two dimensions.On one hand, the changes in the phenology and their relation to temperature or precipitation have been studied repeatedly (e.g., [10,36,37]).On the other hand, the temporal-spatial patterns of the phenological changes (the variations of dates of the SOS, EOS, POS, and LOS) have been examined (e.g., [26,31,38]).In particular, the starting date of the plant growing season (SOS) has provoked fierce debate (e.g., [34,[39][40][41][42][43][44]).In fact, the changes in phenology across the Tibetan Plateau (TP) have been analyzed in many reports, but these reports might be incomplete, because (1) the distribution patterns of phenological indicators (SOS, POS, EOS, and LOS) might be different in the different grasslands of China, particularly for the different environments on the TP and Inner Mongolia Plateau; and (2) further information is required on the quantitative influence of climate on plant phenology.Therefore, the dynamic mechanisms of phenological change in different grasslands require further exploration.Specifically, the two objectives of this study were as follows: (1) to explore the temporal-spatial patterns of plant phenology in China's grasslands from 1985 to 2010; and (2) to disentangle the response mechanisms of plant phenology in different grasslands to environmental factors.

Study Area
The study area (74.93-120.06• E, 29.25-49.51• N) is primarily located in the northwestern part of China and covers approximately 3.10 × 10 6 m 2 (Figure 1), which accounts for 79.7% of the grasslands in China.The area covers the regions of Inner Mongolia, Gansu, Ningxia, Xinjiang, Qinghai, and Tibet.The primary grassland types in the study area are alpine, temperate, and mountain grasslands.In detail, the alpine steppes and alpine meadows are on the Tibetan Plateau (TP); desert, typical, and meadow steppes are on the Inner Mongolian Plateau; and the mountain meadows are in the Xinjiang area [45].The climate is arid and semiarid with an annual mean temperature (AMT) and annual mean precipitation (AMP), which range from −6.2 to 8.8 • C and from 92.3 to 694.0 mm, respectively, in the study area [45].
primary grassland types in the study area are alpine, temperate, and mountain grasslands.In detail, the alpine steppes and alpine meadows are on the Tibetan Plateau (TP); desert, typical, and meadow steppes are on the Inner Mongolian Plateau; and the mountain meadows are in the Xinjiang area [45].The climate is arid and semiarid with an annual mean temperature (AMT) and annual mean precipitation (AMP), which range from −6.2 to 8.8 °C and from 92.3 to 694.0 mm, respectively, in the study area [45].

NDVI Database
We analyzed the plant phenology trends on the Tibetan Plateau from 1985 to 2010 using GIMMS3g NDVI remote sensing data sets with a spatial resolution of 0.083 • , which were compiled by merging segments (data strips) during a half-month period, using the maximum value composites (MVC) method [46].These data were corrected for calibration, view geometry, and volcanic aerosols, and have been verified using a stable desert control point.

Climate Database
Climatic data were collected from the Meteorology Information Center of the Chinese National Bureau of Meteorology (CNBM) (http://data.cma.cn/).Figure 1 shows the locations of the national standards of meteorology (Figure 1).The primary climatic factors (AMT and AMP) of the National Standard of Meteorology Observatories (1985-2010) were included in the document.Moreover, the spatial interpolation of AMT and AMP was generated in ArcGIS 10.2 (ESRI, Inc., Redlands, CA, USA) with the kriging interpolation method, in order to create 0.05 • resolution climatic factors.

SOS, EOS, POS, and LOS Were Obtained with the Retrieval Method
Four indicators, that is, the start of the growing season (SOS), peak of the growing season (POS), end of the growing season (EOS), and length of the growing season (LOS), were employed in describing the phenology (Figure 2).We derived the phenology indicators using the GIMMS 3g NDVI data set covering the period 1985-2010, because the plants in China's grassland ecosystems only have one growing season in a year.To smooth the NDVI data set, the Fourier Series Model was applied as Equation (1).
In Equation (1), y indicates the fitted, smoothed NDVI value; a 0 models a constant (intercept) term in the data; w is the fundamental frequency of the data series; and n is the number of terms in the series.For forest, shrub, and steppe, acceptable model results are obtained when n is equal to 1.Then, the SOS, POS, and EOS were determined using the derivatives of the fitted Fourier equation.As illustrated in Figure 2, SOS was defined as the Julian day when the derivative reached the highest value; POS was the Julian day when the maximum NDVI estimated by the fitted Fourier function appeared; EOS was the corresponding Julian day when the derivative of the fitted Fourier function obtained its minimum value; and LOS equaled the difference between EOS and SOS.
In Equation (1), y indicates the fitted, smoothed NDVI value; a0 models a constant (intercept) term in the data; w is the fundamental frequency of the data series; and n is the number of terms in the series.For forest, shrub, and steppe, acceptable model results are obtained when n is equal to 1.Then, the SOS, POS, and EOS were determined using the derivatives of the fitted Fourier equation.As illustrated in Figure 2, SOS was defined as the Julian day when the derivative reached the highest value; POS was the Julian day when the maximum NDVI estimated by the fitted Fourier function appeared; EOS was the corresponding Julian day when the derivative of the fitted Fourier function obtained its minimum value; and LOS equaled the difference between EOS and SOS.

Data Analyses
To analyze the temporal and spatial dynamics of the phenology, and to reveal the relationships between the phenology and climatic factors, the least squares method was employed in this analysis.The change tendency of phenology and the spatial correlation relationships between phenology and climatic factors were analyzed by the least square method.
A redundancy analysis (RDA) in the software of R (R Core Development Team, R Foundation for Statistical Computing, Vienna, Austria) was used to determine the climatic factors for the variation of the phenology indicators, with the determinants (p < 0.05).RDA is a linear canonical ordination method and is widely used to identify the patterns of the dynamics in the response variables [47].RDA is an effective way to evaluate the relationships among multiple interacting variables, which are also closely associated with the potential explanatory variables [48,49].Finally, linear regression analysis was applied to develop the correlations between the phenology indicators and climatic factors in the software SigmaPlot 10.0 (Systat Software, Inc., Chicago, IL, USA).

Spatial Distribution of Plant Phenology, AMT, and AMP in China's Grasslands
The spatial distribution patterns and plant phenology in different grasslands are shown in Figure 3. From northeast to southwest, the SOS and EOS were gradually delayed.Moreover, from northwest to southeast, the POS was gradually delayed.From the margin to the center of the study area, the LOS gradually shortened.The SOS was primarily observed from the 115th to 130th day from 1 January, which was distributed in Inner Mongolia and the TP, with the lowest (121) and highest (129) values in the desert steppe and alpine steppe, respectively (Figure 3A).The POS primarily occurred between days 190 and 210 from 1 January, and was also mostly distributed along Inner Mongolia to the TP, with the lowest (197) and highest (204) values from 1 January occurring in the temperate meadow and alpine steppe, respectively (Figure 3B).The EOS was primarily found between days 240 and 290 from January 1, and was presented in most regions in the study area and the lowest (265) and highest (273) values of EOS from 1 January were found in temperate meadow and alpine meadow, respectively (Figure 3C).The longer LOS was centrally distributed on the edge of the study area, with values between days 152 and 201, and the values from days 137 to 152 were primarily observed in the center of the TP and the east and the southwest of Inner Mongolia (Figure 3D).Additionally, the highest (153) and lowest (139) LOS values were found in desert steppe and temperate meadow, respectively.
Furthermore, the results indicated that the value of the AMT (1985-2010) ranged from −4.63 to 15.89 • C. A relatively low AMT was observed in the center of the TP and north of Inner Mongolia, and a relatively high AMT was found in the corridor of Hexi-Xinjiang and south of the TP (Figure 3E).Additionally, the AMP decreased toward the south-north on the TP, toward the north-south in Xinjiang, and toward the east-west in the study area, with the AMP ranging from 4.02 to 1089.01 mm (Figure 3F).

Dynamics of Plant Phenology, AMT, and AMP in China's Grasslands
The SOS showed an advanced trend in most regions across China's grasslands, with primarily advanced values of 0-2 days (Figure 4A).Similarly, the POS and EOS showed an advanced tendency across China's grasslands; advanced values of 1-2 days were primarily found in the center of the TP and Inner Mongolia, and those of 0-1 day were primarily observed in the east and south of Inner Mongolia, in the east and south of the TP, and in the north of Xinjiang (Figure 4B,C).Additionally, the LOS presented a shortened trend over China's grasslands; advanced values of 1-3 days were centrally distributed in the TP and mid Inner Mongolia, and those of 0-1 day were primarily found in the north and south of Inner Mongolia, in the east of the TP, and in most regions in Xinjiang (Figure 4D).In general, the SOS, POS, and EOS came early, with the variation amplitude of 0.54 days/year, 0.64 days/year, and 0.65 days/year, respectively.The LOS shortened at a rate of 0.62 days/year.
the LOS presented a shortened trend over China's grasslands; advanced values of 1-3 days were centrally distributed in the TP and mid Inner Mongolia, and those of 0-1 day were primarily found in the north and south of Inner Mongolia, in the east of the TP, and in most regions in Xinjiang (Figure 4D).In general, the SOS, POS, and EOS came early, with the variation amplitude of 0.54 days/year, 0.64 days/year, and 0.65 days/year, respectively.The LOS shortened at a rate of 0.62 days/year.Moreover, the changed values of AMT showed variation from −0.14 to 0.41 °C, and the AMT presented a tendency of increase across China's grasslands, except for the regions in the north of Inner Mongolia, the south of the TP, and the center of Xinjiang (Figure 4E).The dynamics of AMP Moreover, the changed values of AMT showed variation from −0.14 to 0.41 • C, and the AMT presented a tendency of increase across China's grasslands, except for the regions in the north of Inner Mongolia, the south of the TP, and the center of Xinjiang (Figure 4E).The dynamics of AMP were observed between −15.09 and 7.20 mm during the study period, and the decreases in AMP were primarily distributed in the northeast of Inner Mongolia and the south of the TP (Figure 4F).
Simultaneously, the increases in AMP were primarily in the center of the TP and the north of Xinjiang (Figure 4F).were observed between −15.09 and 7.20 mm during the study period, and the decreases in AMP were primarily distributed in the northeast of Inner Mongolia and the south of the TP (Figure 4F).Simultaneously, the increases in AMP were primarily in the center of the TP and the north of Xinjiang (Figure 4F).

Responses of Plant Phenology to AMT and AMP across China's Grasslands
A significant negative correlation between the SOS and AMT was found in the east-north of Inner Mongolia, the north of Xinjiang, and a part of the area in the TP (Figure 5A).Moreover, a significant positive relationship between the SOS and AMT was revealed in the east-center of Inner Mongolia and the south and center of the TP (Figure 5A).Additionally, a significant negative

Responses of Plant Phenology to AMT and AMP across China's Grasslands
A significant negative correlation between the SOS and AMT was found in the east-north of Inner Mongolia, the north of Xinjiang, and a part of the area in the TP (Figure 5A).Moreover, a significant positive relationship between the SOS and AMT was revealed in the east-center of Inner Mongolia and the south and center of the TP (Figure 5A).Additionally, a significant negative correlation between the POS and AMT was found in the north of Xinjiang, the east-north of Inner Mongolia, and the center-west of the TP (Figure 5B).Similarly, between EOS/LOS and AMT, significant positive and negative relationships were revealed in the north of Xinjiang and the north of the TP (Figure 5C,D), respectively.
ISPRS Int.J. Geo-Inf.2018, 7, x FOR PEER REVIEW 7 of 15 correlation between the POS and AMT was found in the north of Xinjiang, the east-north of Inner Mongolia, and the center-west of the TP (Figure 5B).Similarly, between EOS/LOS and AMT, significant positive and negative relationships were revealed in the north of Xinjiang and the north of the TP (Figure 5C,D), respectively.An obvious significant correlation between SOS and AMP was observed in most regions including the west-center of Inner Mongolia, the north of Xinjiang, and part of the regions on the TP (Figure 6A).Furthermore, a significant relationship between POS and AMP was primarily distributed in the center of Inner Mongolia, the north and east of the TP, and the north of Xinjiang (Figure 6B).A significant negative correlation between POS/EOS and AMP was found in the center of the TP and the east-north of Inner Mongolia (Figure 6B,C), with a significant relationship between EOS and AMP distributed in the center of the TP and the east-north of Inner Mongolia (Figure 6C).A significant negative correlation between LOS and AMP was primarily found in the center of the TP and the north of Inner Mongolia (Figure 6D).Simultaneously, a significant relationship between LOS and AMP was revealed in the east of the TP and part of the region in Inner Mongolia (Figure 6D).An obvious significant correlation between SOS and AMP was observed in most regions including the west-center of Inner Mongolia, the north of Xinjiang, and part of the regions on the TP (Figure 6A).Furthermore, a significant relationship between POS and AMP was primarily distributed in the center of Inner Mongolia, the north and east of the TP, and the north of Xinjiang (Figure 6B).A significant negative correlation between POS/EOS and AMP was found in the center of the TP and the east-north of Inner Mongolia (Figure 6B,C), with a significant relationship between EOS and AMP distributed in the center of the TP and the east-north of Inner Mongolia (Figure 6C).A significant negative correlation between LOS and AMP was primarily found in the center of the TP and the north of Inner Mongolia (Figure 6D).Simultaneously, a significant relationship between LOS and AMP was revealed in the east of the TP and part of the region in Inner Mongolia (Figure 6D).

Figure 6. Significant correlations between SOS and AMP (A); between POS and AMP (B); between EOS and AMP (C); and between LOS and AMP (D)
. The significance level is 0.05.

Dynamics of Plant Phenology in Different Grassland Types
To further reveal the dynamics of plant phenology, we analyzed the change rate of the phenology indicators in different grasslands across northern China.As shown in Table 1, the SOS in the meadow steppe (R 2 = 0.26, p = 0.007) and typical steppe (R 2 = 0.28, p = 0.005); the EOS in alpine steppe (R 2 = 0.16, p <0.05); and the LOS in alpine (R 2 = 0.20, p <0.05), meadow (R 2 = 0.18, p <0.05), and typical (R 2 = 0.18, p <0.05) steppes changed significantly.

Table 1. The variation of vegetation phenology in different grasslands across northern China (1985-2010). y is the phenology indicator (start of the growing season [SOS], peak of the growing season [POS], end of the growing season [EOS], and length of the growing season [LOS]
) ; x is the years 1985-2010; R2 is the fitting optimization index; p is the probability to reject the null hypothesis; y0 is the intercept; and a is the slope.

Dynamics and Relationships in Plant
Phenology, AMT, and AMP in Different Grassland Types

Dynamics of Plant Phenology in Different Grassland Types
To further reveal the dynamics of plant phenology, we analyzed the change rate of the phenology indicators in different grasslands across northern China.As shown in Table 1, the SOS in the meadow steppe (R 2 = 0.26, p = 0.007) and typical steppe (R 2 = 0.28, p = 0.005); the EOS in alpine steppe (R 2 = 0.16, p < 0.05); and the LOS in alpine (R 2 = 0.20, p < 0.05), meadow (R 2 = 0.18, p < 0.05), and typical (R 2 = 0.18, p < 0.05) steppes changed significantly.

Dynamics of AMT and AMP in Different Grassland Types
Based on the results in Table 1, we further revealed the dynamics of the climatic factors (Table 2) in grasslands with significantly changed phenology.The results indicated that the AMT in the alpine, meadow, and typical steppes showed a trend of increase, in which the average change rates of AMT were 0.05 • C/year, 0.06 • C/year, and 0.06 • C/year (Table 2), respectively.Conversely, a trend of decrease in AMP was observed in the grasslands of alpine, meadow, and typical steppes, with the average decrease in values of AMP of 1.75 mm/year, 0.60 mm/year, and 0.72 mm/year, respectively (Table 2).We further explored the response mechanisms between the phenology (significantly changed phenology in Table 1) and climatic factors by RDA.The results from the RDA showed that AMT and AMP together explained 12.65% (Figure 7A) and 32.52% (Figure 7B) of the variation of the SOS in the meadow steppe and typical steppe, respectively.Moreover, AMT and AMP together explained 6.22% of the dynamic of the EOS in the alpine steppe (Figure 7C).Additionally, AMT and AMP together explained 6.06% (Figure 7D), 16.69% (Figure 7E), and 19.58% (Figure 7F) of the variation of the LOS in the alpine, meadow, and typical steppes, respectively.Notably, AMP had a relatively larger influence on the variation of the SOS in the meadow steppe and typical steppe than that of AMT, whereas by contrast, AMT had a greater influence on the dynamics of the EOS in the alpine steppe.Similarly, AMT also had a relatively larger influence than that of AMP on the variation of the LOS in the alpine, meadow, and typical steppes.
Similarly, advanced EOS results are also found in many studies, in which the advanced rates of EOS are 0.37 days/year, 0.61 days/year, 0.23 days/year, and 0.50 days/year for China [58], the Eurasian continent [55], the Northern Hemisphere [56], and the world [57], respectively.Additionally, the inter-annual LOS presented a shortened trend in our results.Consistent with our results, the LOS shortened with the values of 1.16 days/year, 1.33 days/year, 0.56 days/year, and 0.38 days/year for China [54], the Eurasian continent [55], the Northern Hemisphere [27], and the world [55], respectively.In several studies, both observational and experimental, the POS shows an advanced trend in other regions [50][51][52][53]59,60].
In general, plant phenology showed significant changes in alpine, meadow, and typical steppes, which were primarily distributed in the center of the TP and Inner Mongolian Plateau.The maximum increased rates of AMT and AMP in the center of the TP (Figure 4E) and the maximum decrease in the AMP in the center of the Inner Mongolian Plateau (Figure 4F) might explain the corresponding significant changes in the phenology indicators (Table 1).Additionally, the alpine steppe in the north of the TP with a relatively low AMT and AMP was more sensitive to the dynamics of climatic factors than the alpine meadow in the southeast of the TP with a relatively high AMT and AMP, which might
Similarly, advanced EOS results are also found in many studies, in which the advanced rates of EOS are 0.37 days/year, 0.61 days/year, 0.23 days/year, and 0.50 days/year for China [58], the Eurasian continent [55], the Northern Hemisphere [56], and the world [57], respectively.Additionally, the inter-annual LOS presented a shortened trend in our results.Consistent with our results, the LOS shortened with the values of 1.16 days/year, 1.33 days/year, 0.56 days/year, and 0.38 days/year for China [54], the Eurasian continent [55], the Northern Hemisphere [27], and the world [55], respectively.In several studies, both observational and experimental, the POS shows an advanced trend in other regions [50][51][52][53]59,60].
In general, plant phenology showed significant changes in alpine, meadow, and typical steppes, which were primarily distributed in the center of the TP and Inner Mongolian Plateau.The maximum increased rates of AMT and AMP in the center of the TP (Figure 4E) and the maximum decrease in the AMP in the center of the Inner Mongolian Plateau (Figure 4F) might explain the corresponding significant changes in the phenology indicators (Table 1).Additionally, the alpine steppe in the north of the TP with a relatively low AMT and AMP was more sensitive to the dynamics of climatic factors than the alpine meadow in the southeast of the TP with a relatively high AMT and AMP, which might be another reason [35,36,46].The non-significant changes to the phenology indicators in temperate meadows were primarily found in the north of Xinjiang Province, which might be explained because plant phenology at high altitudes is more sensitive and complicated than that at low altitude [31].

Control Mechanisms of Plant Phenology and Climate in Different Grasslands
Our results demonstrated that significant changes in plant phenology were revealed in the meadow steppe (R 2 = 0.26, p = 0.007) and typical steppe (R 2 = 0.28, p = 0.005) for the SOS, primarily distributed in Inner Mongolia (Table 1).Accompanying the differences in temperature and precipitation conditions, the SOS in the meadow steppe and typical steppe advanced gradually (Table 1).For the significantly changed SOS, an increased temperature in the meadow steppe (0.06 • C/year, Table 2) and typical steppe (0.06 • C/year, Table 2), which was primarily found in the center of the Inner Mongolian Plateau, accelerated the accumulation of heat and advanced the resumption of growth activity of plants, which affected the timing of the phenological phases [31].Green-up onset tended to be advanced by the increasing temperature, and precipitation also exerted a stronger effect (Figure 7) on the SOS in the drier regions than that in others [32].Thus, the SOS in the north of Xinjiang Province was significantly correlated with AMT and AMP (Figures 5A and 6A).Similarly, advances in spring phenology are also reported in other studies [50,[60][61][62].Moreover, warming in springs and winters on the TP also continuously advanced the SOS at a rate of 1.04 days/year (1982-2011) [33], and an obvious positive correlation between the SOS and spring temperature was revealed in North America [63].In summary, the starting growth of plant phenology was particularly sensitive to the change of temperature and precipitation [39][40][41][42][43][44].
Significantly advanced EOS was observed in the alpine steppe in the center of the TP (R 2 = 0.16, p < 0.05; Table 1), and the AMT relatively explained most of the variation of the EOS (Figure 7).The warming temperature in the alpine steppe (0.05 • C/year, Table 2) therefore rapidly accelerated plant growth, resulting in an earlier completion of the growth in the steppe species [31].Similarly, most regions in the center of the TP present a significant increase in temperature and a clear advance in the EOS by several weeks [64].Thus, warming tended to be advanced for the EOS on the TP [64].Because the growth of the alpine steppe on the TP is the most temperature-limited [65], plant growth is primarily sensitive to the dynamics in temperature [66 -69].Additionally, the dynamics of AMP in the alpine steppe might lead to the termination of plant growth to some extent, because water is the most important factor that affects plant growth [70][71][72].
The significantly advanced SOS in the meadow steppe and typical steppe, together with no significant change in the EOS, led to the gradual extension of the LOS in the meadow steppe (R 2 = 0.18, p < 0.05) and typical steppe (R 2 = 0.18, p < 0.05; Table 1).By contrast, the significantly advanced EOS, together with no significant change in the alpine steppe, accounted for the shortening of the LOS in the alpine steppe (R 2 = 0.20, p < 0.05; Table 1).The variation of the LOS in the north of Xinjiang Province was positively correlated with AMT (Figure 5A), because of the significant relationship between AMT and SOS (Figure 5D), and warming temperature is beneficial to the extension of plant growth [34,54,61,64].For example, an increase in AMT by 1 • C resulted in an extension of five days in Europe [61]; and across north China, a warming during March to early May by 1 • C led to an extension of the SOS by 7.5 days [54], whereas an increase of AMT by 1 • C from mid-August to early October generated a delay of 3.8 days in the EOS [54].

Limitations of the Current Study
Although a part of the variation in the phenology indicators was explained by climatic factors, in our study, other determinant factors might have had an effect on the dynamics of plant phenology, such as evapotranspiration, solar radiation, and soil properties (soil water content, soil total carbon, soil total nitrogen, soil total phosphorus, soil available nitrogen, soil available phosphorus, and soil microorganisms, among others).Because the effects of climatic factors on soil nutrients and the activity of soil microbes are closely related to the growth of plants, future research should devote more attention to the linkages between plant phenology and soil factors.For example, experiments on warming and simulating extreme rainfall are important to reveal how the phenology indicators (SOS, POS, EOS, and LOS) respond to changes in global climate and soil properties.

Conclusions
This study systematically revealed the spatiotemporal variation of plant phenology indicators (SOS, POS, EOS, and LOS) across China's grasslands during 1985 to 2010.For the average phenology indicators over northern China, from the northeast to southwest, the SOS and EOS were gradually delayed.From the northwest to southeast, the POS was gradually delayed.From the margin to the center of the study area, the LOS gradually shortened.For the dynamics of the phenology indicators, the SOS, POS, and EOS presented an advanced tendency across China's grasslands.Additionally, a shortening trend of the LOS over China's grasslands was observed.
The SOS in the meadow steppe and typical steppe, and the EOS in the alpine steppe exhibited significantly advanced trends from 1985 to 2010.Moreover, the LOS showed significantly shortened trends in the meadow steppe and typical steppe, primarily distributed in Inner Mongolia, and the LOS underwent a lengthening trend in the alpine steppe as primarily revealed on the TP.The inter-annual climate changes of AMT and AMP in China's grasslands accounted for the significantly dynamic SOS in the meadow steppe (12.65%) and typical steppe (32.52%); the EOS in the alpine steppe (6.22%); and the LOS in the alpine (6.06%), meadow (16.69%), and typical (19.58%) steppes.Thus, the changes in the plant phenology under climate change are critical requirements for the carbon cycle dynamics in these grassland ecosystems.

Figure 1 .
Figure 1.The study area is primarily located in the northwestern part of China.

Figure 1 .
Figure 1.The study area is primarily located in the northwestern part of China.

Figure 2 .
Figure 2. Illustration of the normalized difference plant index (NDVI) ratio method used to model the growing season (SOS), peak of the growing season (POS), end of the growing season (EOS), and length of the growing season (LOS).

Figure 5 .
Figure 5. Significant correlations between SOS and AMT (A); between POS and AMT (B); between EOS and AMT (C); and between LOS and AMT (D).The significance level is 0.05.

Figure 5 .
Figure 5. Significant correlations between SOS and AMT (A); between POS and AMT (B); between EOS and AMT (C); and between LOS and AMT (D).The significance level is 0.05.

Figure 6 .
Figure 6.Significant correlations between SOS and AMP (A); between POS and AMP (B); between EOS and AMP (C); and between LOS and AMP (D).The significance level is 0.05.

15 Figure 7 .
Figure 7. Variance in phenology that might be explained (%) by climatic factors (AMT and AMP) using redundancy analysis.Graphs (A-F) refer to the SOS in the meadow and typical steppes; the EOS in the alpine steppe; and the LOS in the alpine, meadow, and typical steppes that might be explained by climatic factors, respectively.

Figure 7 .
Figure 7. Variance in phenology that might be explained (%) by climatic factors (AMT and AMP) using redundancy analysis.Graphs (A-F) refer to the SOS in the meadow and typical steppes; the EOS in the alpine steppe; and the LOS in the alpine, meadow, and typical steppes that might be explained by climatic factors, respectively.

Table 1 .
The variation of vegetation phenology in different grasslands across northern China.y is the phenology indicator (start of the growing season [SOS], peak of the growing season [POS], end of the growing season [EOS], and length of the growing season [LOS]) ; x is the years 1985-2010; R2 is the fitting optimization index; p is the probability to reject the null hypothesis; y0 is the intercept; and a is the slope.
** Represents significance at the level of 0.05.

Table 2 .
The variation of climatic factors (annual mean temperature [AMT] and annual mean precipitation [AMP]) in three grasslands (1985-2010).The Mean, Min, Max. and Std are the average, minimum, maximum, and standard deviation values, respectively, of the AMT or AMP in corresponding grasslands from 1985 to 2010.Responses of Plant Phenology to AMT and AMP in Different Grassland Types