4.1 Trends in grassland phenology
We analyzed the dynamics of different grasslands phenology on the QTP from 2000 to 2019. The results showed that the SOS had a significant advance trend in the eastern and central areas with lower altitude, but an extremely insignificant change in the southwestern areas with altitude above 5000 m. This study suggested that climate variables (such as precipitation, extreme climate, and evapotranspiration) might be related to the changes in the advanced SOS and global warming, and the spring phenology changes could not be used directly as an indicator of climate warming. Therefore, it is impossible to answer climate warming questions using only phenological metrics, especially in high altitude area of southwest QTP. EOS was significantly delayed, especially in the Southwestern regions, but not obvious in the central and northern regions (main grassland types are desert, temperate steppe, alpine steppe and alpine meadow). Previous studies found insignificant change in EOS, but temperate steppe showed a significant change (Li et al.2018; Yang et al.2015).In this study, it was found that the desert EOS was delayed, and the trend was more obvious, but the changes in other regions were not obvious. At the same time, a recent study by Professor He of the State Key Laboratory of Grassland Agroecosystems of Lanzhou University found that the QTP had experienced long-term and rapid climate change over the past 50 years, and climate warming had promoted the early SOS of alpine vegetation, which resulted in an increase biomass in the spring but accelerated growth in the middle of the growing season, leading to a water deficit and accelerating the progression of the EOS. This conclusion was consistent with the findings of many researchers such as Huang et al. (Huang et al.2019) and Ma et al. (Ma et al.2016), but it contradicts with the results of Meng et al. (Meng et al.2017) and Zhou et al. (Zhou et al.2018). Scholars have different results on the grassland phenology, whose reason might be that the study period was different, which leaded to inconsistent results. Secondly, it might be due to diverse temporal and spatial distribution pattern of environmental factors in different regions. Although warming temperatures had greatly changed spring phenological events, different regions and multi-grassland respond differently to this change. For example, climate warming promotes spring phenology of cold grasslands, especially in high latitude regions, while spring phenology in seasonally dry grasslands advanced slowly or even delayed with the climate warming. This was mainly because the spring phenology of cold and dry grasslands was more sensitive to temperature and moisture, respectively (Kong et al.2017; Fu et al.2021; Shen et al.2014).
4.2 Relationship between grassland phenology metrics and altitude
Grassland phenology change was periodic and continuous dynamic process. There are many factors impacting phenological changes (Ren et al.2018; Shen et al.2014). Studies have found that the grassland phenology changes significantly and exist differently in grassland growth rates on different altitudinal gradients on the QTP (Liu et al.2020; Shen et al.2014). We also found that SOS was gradually advanced at a rate of 1.9 d per 100 m with the increase of altitude in areas below 3100 m. In areas with altitude higher than 3300 m, SOS was gradually delayed and the change rate of SOS showed significant differences with changing altitude. Grassland EOS delayed with increasing altitude, but there were obvious differences in the rate of delay. Our research results were more consistent with the conclusions of Li et al. (Li et al.2017) and Park et al. (Piao et al.2003) that altitude had an important influence on grassland phenology on the QTP. Not only did temperature decrease with increasing altitude, but related studies have also found that the rate of warming was greater at high altitude than low altitude (Li et al.2017). In this study, it was also found that the response of grassland SOS to temperature was more significant than that of precipitation and showed a significant positive correlation with temperature. In the lower altitudes of Qinghai, Gansu and Xinjiang, SOS was positively correlated with temperature, but SOS showed a significant negative correlation with temperature at higher altitudes of Tibet. In Gansu around Qaidam Basin, temperature and precipitation have a lagging effect on EOS. There were some spatial differences in the grassland phenology at different altitudinal gradients on the QTP due to the spatial differences of temperature and precipitation in latitudinal zone. In the eastern regions with lower altitude, grassland SOS showed a significant advance trend, while in these areas of the south and west higher altitude, grassland SOS showed delaying trend, which may be the result of the combined effect of increasing temperature.
4.3 Relationships between grassland phenology metrics and environmental factors
Temperature and precipitation were considered important climatic factors affecting grassland phenology (Gao et al.2015; Wang et al.2019), but compared to average climate change, the occurrence of extreme weather events and droughts was sudden, predictable and destructive, and might have a greater impact on the grassland phenology. Therefore, it might affect the structure, composition and function of terrestrial ecosystems severely due to sudden environmental factors change, thereby affecting the grassland growth cycle, grassland productivity and yield (Yuan et al.2020; Shen et al.2020; Nagy et al.2013). But there were few studies on the effects of drought, evapotranspiration, solar radiation and extreme climate on grassland phenology on the QTP. This study analyzed the sensitivity of grassland SOS and EOS to various environmental factors based on the random forest algorithm. The results showed that grassland SOS was mainly influenced by TXN, TNX, RX5day, TNN, TVDI and SR, while EOS was mainly influenced by TVDI, TNX, MAT_MEAN, MAT_MAX, MAT_MIN and TXX (Fig.6).
The relationship between the sensitive environmental factors and phenology was analyzed: ①SOS of multi-grassland was positively correlated with TXN, TNX, RX5day, TNN and TVDI, and negatively correlated with SR. According to the changes in environmental indicators (Fig.7), the decrease in TNN, TNX, TVDI and RX5day and the increase in TXN and SR could advance the grassland SOS. This conclusion also further confirmed that higher temperatures were required for grassland dormancy and spring regeneration, and the increase in temperature increases the warmth of grassland germination and leaf expansion (Luedeling et al.2009; Zhang et al.1995). And some studies have also found that the increase in temperature increases the decomposition rate of soil organic matter and the nutrients in the soil are more easily mineralized and made available to the grassland, which was conducive to the entry of grassland to phenological phase (He et al.2018). Some studies have found that spring temperature increased affects grassland growth in arid areas due to water stress in the soil, resulting in restricted grassland growth and delayed phenology [54], which depends on temperature, solar radiation and precipitation matching grassland phenology with factors such as soil type. These above conclusions could provide a theoretical basis for the research conclusions of this paper. ②EOS was mainly negatively correlated with TVDI, TNX, MAT_MEAN, MAT_MAX, MAT_MIN and TXX. According to the changes in environmental indicators (Fig.7), the main reason for the grassland EOS change might be the decrease in average annual TVDI and the increase in annual average temperature, annual maximum temperature, annual minimum temperature and extreme daily maximum temperature on the QTP. Some studies have shown that grassland in arid and semi-arid areas consists mainly of xerophytes and strong xerophytes. Therefore, the influence of precipitation on EOS was much greater than the influence of temperature on EOS in the growth phase, mainly because precipitation could alleviate water stress in the soil so that grass cannot enter the yellowing phase earlier, and the increase in temperature decreases the moisture content of the soil, causing grassland to enter the yellowing phase earlier (Wang et al.2016). Grasslands were mainly located in dry areas in eastern and northern regions of the QTP. The growth of grassland in summer was highly stressed by drought. After autumn, the temperature dropped rapidly and the growth period of grassland was shorter. The results of this study further showed that the main influence to EOS was temperature on the QTP, and Liu et al (Liu et al.2016) also confirmed that the grassland demand of water decreases in autumn and the temperature rose increases the activity of photosynthetic enzymes and slowed down the decomposition of chlorophyll. This conclusion also showed that increasing in different temperature were also important factors that lead to grassland EOS changes.
4.4 Differences of grassland phenology between ground-monitoring, previous studies and remote-sensing inversion
On the QTP, the limiting precipitation, strong solar radiation and high evaporation result in sparse grasslands growth in some regions. Grassland coverage in these regions is low, less than 20% and even less than 5% in some regions. Therefore, the background information may be contaminated by the target signal (mainly soil). In desert coverage regions, the sensitivity of the sensor to the detection of grassland spectral information is reduced, with weak absorption valleys and reflection peaks, which leads to the grassland spectral information collected from remote sensing images to be extremely weak and difficult to detect. MOD13Q1 product covers a large area, and there may be some errors in the process of inversion. Thus, in Fig.8 and Table 2, the correlation coefficient (r) between remotely sensed grassland SOS and ground-measured SOS was 0.637 (P<0.05), the Bias was 0.321 and the RMSE was 5.95 d. The r between remotely sensed EOS and ground-measured EOS was 0.776 (P<0.05), the Bias was 0.241 and RMSE was 5.07 d. Validation results showed that remotely sensed SOS and EOS provided earlier data than ground-based observations, but the results were well correlated. At the same time, the results of the whole QTP were compared with the results of Deng et al. (Deng et al.2020), Ma et al. (Ma et al.2016), Li et al. (Li et al.2014) and Kong et al. (Kong et al.2017). Li et al. (Li et al.2014) found that the EOS was concentrated at 240-300 d in the Qinghai Basin, which was quite different from our results, but the results of this study were similar to those of Deng et al. (Deng et al.2020) and Kong et al. (Kong et al.2017). The distribution was more consistent. The reason could be that Li et al. (Li et al.2014) used EVI data and the main area, but this article and other researchers used NDVI data. The used of difference VI led to large differences in the phenology extraction. In addition, the phenological data from ground stations were used to verify the calculated phenological results. These results indicated that phenology data can be used to obtain the grassland SOS and EOS on the QTP.
Table 2 Comparison of phenological results between this paper and other studies
Study area
|
study phase
|
SOS(d)
|
EOS(d)
|
Data Resources
|
Literature Reference
|
QTP
|
1999~2009
|
120~170
|
250~300
|
SPOT
|
Deng et al.2020
|
QTP
|
1982~2015
|
130~180
|
280~330
|
GIMMS NDVI
|
Ma et al.2016
|
Qinghai lake watershed
|
2001~2012
|
110~150
|
240~300
|
MODIS EVI
|
Li et al.2014
|
QTP
|
1982~2013
|
115~125
|
260~300
|
GIMMS NDVI3g
|
Kong et al.2017
|
QTP
|
2000~2019
|
140~170
|
250~280
|
MODIS NDVI
|
This paper
|
4.5 Limitations and future directions
This study attempted to explore how multi-grassland were influenced by multiple environmental factors on the QTP. However, the grassland type are abundant, with significant vertical zonality, and different grassland types and flowering functional groups have diverse sensitivity to different environmental factors change on the QTP, such as sherry et al. (Sherry et al.2007) found that grasslands of early flowering functional group were more sensitive to cooling. Wang et al. (Wang et al.2014) found that grassland mid-flowering functional group were more sensitive to temperature increasing than decreasing, and the phenological period of species had the different response to environmental factors. Even under the same hydrothermal conditions, especially in the alpine zone, the germination and phenological characteristics of multi-grassland were also different. Although we conducted some useful research on the response of multi-grassland to environmental factors change based on relatively remote sensing data on the QTP, we neither considered the differences in individual characteristics of grasslands affected by environmental factors, nor the influence of factors such as rainfall duration and soil water-holding capacity on grassland phenology such as the findings of Cong et al. (Cong et al.2013) and Jin et al. (Jin et al.2009). This study presented different environmental factors on an annual scale without considering seasonal characteristics. In the future research, we should also consider whether phenological periods of multi- grassland have different responses to seasonal environmental factors from the seasonal scale.