Response of vegetation phenology to climate factors in the source region of the Yangtze and Yellow Rivers

Abstract

Exploring the impact of climate factors on vegetation phenology is crucial to understanding climate–vegetation interactions as well as carbon and water cycles in ecosystems in the context of climate change. In this article, we extracted the vegetation phenology data from 2002 to 2021 based on the dynamic threshold method in the source region of the Yangtze and Yellow Rivers. Trend and correlation analyses were used to investigate the relationship between vegetation phenology and temperature, precipitation and their spatial evolution characteristics. The results showed that: (i) From 2002 to 2021, the multi-year average start of growing season (SOS), end of growing season (EOS) and length of growing season (LOS) for plants were concentrated in May, October and 4–6 months, with a trend of 4.9 days (earlier), 1.5 days (later), 6.3 days/10 a (longer), respectively. (ii) For every 100 m increase in elevation, SOS, EOS and LOS were correspondingly delayed by 1.8 days, advanced by 0.8 days and shortened by 2.6 days, respectively. (iii) The impacts of temperature and precipitation on vegetation phenology varied at different stages of vegetation growth. Influencing factors of spring phenology experienced a shift from temperature to precipitation, while autumn phenology experienced precipitation followed by temperature. (iv) The climate factors in the previous period significantly affected the vegetation phenology in the study area and the spatial variability was obvious. Specifically, the temperature in April significantly affected the spring phenology and precipitation in August widely affected the autumn phenology.

摘要

长江黄河源区植被物候与气候因素的关系

探明水热条件对植被物候的影响，对于了解气候变化背景下，气候与植被之间的相互作用以及生态系统碳水循环至关重要。本文基于动态阈值法提取了长江黄河源区2002–2021年的植被物候数据，采用趋势分析、相关分析的方法探究了植被物候与气温、降水的响应关系及其空间演变特征。结果表明：(i) 2002–2021年，长江黄河源区植被多年平均生长开始时间、生长结束时间、生长季长度分别集中在5月、10月和4–6个月，并且分别以4.9、1.5和6.3 d/10 a的趋势提前、推迟和延长。(ii)垂直方向上，海拔每升高100 m，植被生长开始时间、结束时间和生长季长度分别推迟1.8 d、提前0.8 d和缩短2.6 d。(iii)在植被生长的不同阶段，气温和降水对植被物候的影响存在差异，植被生长开始时间的影响因素经历了从气温到降水的转变，而结束时间的影响因素经历了从降水到气温的转变。(iv)前期气象因子显著影响源区植被物候，空间差异显著。具体地，4月气温显著影响源区东部植被生长开始时间，而8月降水广泛影响植被生长结束时间。

INTRODUCTION

Climate change can cause changes in regional moisture–heat conditions, affecting vegetation microhabitats (Gang et al. 2016), soil environments and then bring about changes in vegetation growth dynamics. Plant phenology is a life cycle phenomenon (Vrieling et al. 2018; Wu et al. 2021), including three indicators of the start of growing season (SOS), end of growing season (EOS) and length of growing season (LOS). To some extent, it can reflect the changing conditions of regional climate (Xu et al. 2020). Meanwhile, phenology also affects terrestrial ecosystems by altering the water cycle and soil carbon balance as well as plant respiration (Wang et al. 2017b, 2021). Thus, exploring the mechanism of phenology changes plays an important role in the correct understanding of ecosystems.

The source region of the Yangtze and Yellow Rivers (SRYY) is an ecologically fragile area in China and the harsh external environment makes it difficult to obtain ground observation data (Li et al. 2021; Meng et al. 2021), so vegetation remote sensing data are used for our study. Remote sensing monitoring has a wide monitoring range, saves labor and has less interference from harsh environments compared with traditional ground observation. This means that remote sensing monitoring can effectively record the growth status and seasonal change information of vegetation in a long time series (Piao et al. 2019a; Yao et al. 2019). However, clouds and aerosols can interfere with optical remote sensing, so acquired vegetation index data need to be denoised, which increases the research difficulty to some extent. Although there are various methods for extracting vegetation phenology based on vegetation indices, the dynamic threshold method has been applied for its property of recognizing differences in vegetation types. And vegetation indices and the range of threshold are different depending on the regional climate conditions and sensor types in various areas or even the same (Chen et al. 2018; Deng 2021; Jia et al. 2016; Wang et al. 2021; Xia 2020). Therefore, it is crucial for conducting regional vegetation phenology studies to determine the thresholds in SRYY.

In many studies, the overall mean value of vegetation phenology has been taken as the object of study, which can ignore the internal spatial variation in the study area. We started with pixels to improve the accuracy of vegetation phenology to some extent because of the higher resolution in our study. Besides, the diversity of climate factors and differences in research methods have made scholars disagree on the trend and extent of vegetation phenology, response relationships and dominant factors in existing research (An et al. 2020; Piao et al. 2019b; Ren et al. 2020; Wang et al. 2017a; Yuan et al. 2024). Meanwhile, it also increases the difficulty of studying response mechanisms between vegetation phenology and moisture–heat conditions (Chen et al. 2023). Scholars generally agree that temperature and precipitation are the main driving factors of vegetation phenology (Du et al. 2019; Piao et al. 2019a; Shen et al. 2022; Sun et al. 2023). Thus, we take them as the main climate indicators to refine the spatial response relationship between vegetation phenology and climate factors.

Based on the vegetation index data, we extracted the vegetation phenology of the SRYY, focusing on two aspects: the spatial and temporal trends of the vegetation phenology and its response to moisture–heat conditions. The aim is to clarify the spatial and temporal patterns of phenology in the research area, to explore the differences in dominant climate factors on a spatial scale and to enrich the study of phenology in alpine ecosystems.

MATERIALS AND METHODS

Study area

The SRYY is located in the hinterland of the Qinghai-Tibet Plateau, roughly between 90°–104° E and 32°–37° N, covering an area of about 2.7 × 10⁵ km² (Hu et al. 2022), spanning the provinces of Qinghai, Gansu and Sichuan, while the main body is located in the southern of Qinghai Province. The average elevation of the SRYY is 4465 m, and the terrain has a large undulation, generally showing a variable trend of high in the west and low in the east (Fig. 1). The multi-year average spring temperature and cumulative precipitation are −3.4°C and 83.8 mm, respectively. There exists a well-developed river system and extensive permafrost (Ahmed et al. 2022; You and Li 2021). And marsh widely spread in the eastern of the SRYY. The SRYY has a typical plateau-continental climate, with distinct wet and dry seasons and alternating hot and cold seasons. The vegetation type is dominated by alpine grassland and meadow, with a few forests and shrubs in the east.

Data acquisition

Normalized Difference Vegetation Index (NDVI) data

The MODIS dataset (MOD13A2) provided by NASA (https://www.nasa.gov/) was used, which has a spatial resolution of 1km and a temporal resolution of 16 days. It is a level 3 product.

Meteorological data

Temperature and precipitation data provided by the National Tibetan Plateau Science Data Center (https://data.tpdc.ac.cn/) were ‘1-km monthly precipitation (mean temperature) dataset for China (1901–2022)’, with a spatial resolution of 0.0083333° (about 1 km), and had a time span from January 1901 to December 2022. The data format is nc file and the unit of data is 0.1 mm (0.1°C). Besides, the datasets based on the Delta downscaling method were evaluated by 496 national weather stations across China, which means the result was credible.

Land use and digital elevation model data

Land use data were provided by the Resource and Environment Science and Data Center (https://www.resdc.cn/), based on remote sensing images of land satellites from the USA. The dataset includes time from 1980, 1990, 1995, 2000, 2005, 2008, 2010, 2013, 2015, 2018, 2020, with a spatial resolution of 30 m. The dataset was categorized into six primary classifications and 25 secondary classifications, with the primary classifications including cropland, forest land, grassland, watershed, construction land and unused land. We selected data on cropland, forest land and grassland for the years 2005, 2010, 2015 and 2020 as the study data for vegetation cover areas. Meanwhile, the digital elevation model with a spatial resolution of 30 m was obtained from the Geospatial Data Cloud (https://www.gscloud.cn/).

Vegetation phenological extraction

First, with the help of the Timesat platform, the NDVI data were smoothed by S-G filtering (the NDVI curves smoothed by S-G filtering were highly consistent with the original NDVI curves). Second, SOS and EOS were extracted by the dynamic threshold method, while the LOS was obtained by calculating the difference between the EOS and the SOS. Third, the three vegetation phenology indicators were determined. Finally, the ArcGIS was used to crop the vegetation phenology dataset in SRYY. It is worth noting that Timesat required the support of Matlab.

The basic principle of the dynamic threshold method is to define the time of SOS and EOS when the NDVI increases or decreases by a certain percentage of the current year’s NDVI variation. Referring to the study of Wang, Jonsson and Eklundh et al. (Jonsson and Eklundh 2002; Wang and Yang 2023), in this article, the thresholds were set to 20% and 50% to determine the start and end of the vegetation growth period. The calculation formula is as follows:

The NDVI is the smoothed NDVI value, and NDVI_max and NDVI_min are the maximum and minimum values of NDVI in a year, respectively. When NDVI_ratio is greater than 0.2 (20%) into the SOS and when NDVI_ratio is less than 0.5 (50%) into the EOS.

While performing the vegetation phenology extraction, we found some missing values and in order to exclude the interference of it, we filtered the 20 years of vegetation phenology data. Assuming that there are 10 or more years of missing data on a unit pixel, we no longer study the pixel but take the mean value as the study value for the multi-year average phenology.

Data analysis

Trends analysis

A simple linear regression analysis was used to analyze the trend of changes in the dynamics of vegetation phenology in SRYY from 2002 to 2021, which revealed the changes in tendencies of value in pixels, with the following equations:

Slope is the interannual trend, n is the length of the study time (n = 20, in our work), i is the year serial number (i = 1, 2, …, 20) and y is the value of the vegetation phenology parameter (SOS or EOS or LOS) in the ith year. When Slope >0, it indicates an increasing change trend, i.e. SOS or EOS is delayed while LOS is prolonged. When Slope <0, it indicates a decrease in vegetation phenology, i.e. SOS or EOS is advancing, and LOS is getting shorter. Besides, the significance of changes in phenological indicators is tested using the F-test, which is expressed as:

n is the number of years (i.e. n = 20), $y^$ is the regression value. $y¯$ indicates the average of SOS (EOS and LOS) for 20 years, and y_i is the SOS (EOS and LOS) of the ith year.

Correlation analysis

The simple correlation coefficient can only reflect the relationship between individual climate factors and vegetation phenology, which is one-sided. To compensate for this deficiency, we used the multiple and partial correlation coefficient to investigate the relationship and degree of influence of multiple climate variables on phenology. Combining the results of correlation analysis, we selected March–May and August–October as the study periods.

In this study, multiple correlation analysis was used to interpret the degree of closeness between a dependent variable (SOS and EOS) and several independent variables (monthly mean temperature and monthly mean precipitation). That can be expressed by the following equation:

R_x,yz is the composite correlation coefficient between dependent variable x and independent variables y, z; R_xz,y is the partial correlation coefficient of variables x and z to fixed variable y, and r_xy is the correlation coefficient between variables x and y. In our work, the F-test was adopted to determine the significance with the following formula:

where n is the number of samples and k is the number of independent variables.

In the research of multifactorial effects, partial correlation analysis can be used to measure the relationship between two elements while keeping other factors constant. In this study, it was used to analyze the relationship between SOS (EOS) and temperature, and SOS (EOS) and precipitation as follows:

r_xy,z is the partial correlation coefficient of variables x and y under the condition of constant z factors. r_xy, r_xz and r_yz denote the correlation coefficients of variables x and y, x and z, y and z, respectively. In this study, a t-test was used to determine the significance of partial correlation with the following formula:

n shows the number of samples, m indicates the number of independent variables.

On the basis of the correlation and significance above, we established the classification criteria of climate factors on vegetation phenology according to the regional characteristics of SRYY and combined with the previous scholars (Chen et al. 2017) whose conclusions are credible, so as to determine the dominant climate factors, as shown in Table 1.

RESULTS

Analysis of spatial and temporal variation in vegetation phenology

Characteristics of spatial variation in vegetation phenology

Fig. 2 shows the average vegetation phenology (SOS, EOS and LOS) changes in the SRYY from 2002 to 2021. The spatial pattern of phenology changes is clear in the SRYY, with obvious differences between east and west. Horizontally, the trend of SOS was gradually advancing, while EOS was gradually delaying, and LOS was gradually prolonging from northwest to southeast. About 80% of SRYY, SOS was concentrated in early May to early June (130–160 DOY) and EOS occurred during mid-to-late October (Fig. 2b), DOY means Julian day, 1 January of each year is counted as 1 DOY and a year is 365 DOY. In the southeastern part of the SRYY, SOS appeared before May (<120 DOY) and its area ratio was 5.9% (Fig. 2a and b). About 87.8% of the study area, LOS was concentrated in 120–180 days and its geographic dispersion was basically consistent with SOS, with the LOS shorter than 4 months in the Tuotuo River and Dangqu River in the northwest of SRYY, and more than 6 months in Hongyuan county in the southeast (Fig. 2c). Vertically, for each 100 m increase in altitude, there was a rate of 1.8 days in SOS, −0.8 days in EOS, −2.6 days in LOS within the altitude range of 3500–5000 m (Fig. 3a–c). EOS first appeared in September and its area accounted for less than 2% in the alpine zone above 5000 m, latest appeared in November with an area ratio of about 1.3% below 4000 m. In terms of trends, roughly 80.7% of the pixels in SRYY showed an earlier SOS, 67.4% displayed a delay EOS and 80.8% showed a longer LOS, while the percentage of the areas with significant changes was 13.6%, 5.9% and 18% (P < 0.05), respectively (Fig. 2d–f). Overall, areas with significant changes in SOS have corresponding changes in LOS in SRYY. And it was mainly concentrated in valleys or low-elevation regions with wetter and warmer environments.

Interannual trends in vegetation phenology

SOS, EOS and LOS of the vegetation in the source area advanced with a trend of 4.9 days/10 a, delayed with a trend of 1.5 days/10 a, and extended with a trend of 6.3 days/10 a (P < 0.05) over the past 20 years, respectively. Vegetation EOS of the source area had been at a relatively stable level, whereas the changes in LOS and SOS were asynchronous (the shape of the fold line was fundamentally symmetrical) and the overall variability was large (Fig. 3d). Fig. 4 depicts the interannual variation of vegetation phenology in SRYY. Vegetation in source region began to grow as early as mid-March (78 DOY) and as late as the end of July (211 DOY) (Fig. 4); the leaves of plants turned yellow first in mid-September (253DOY), and this yellowing could last until mid-November (314 DOY) (Fig. 4). LOS of SRYY was the longest more than 7 months (224 days) and shortest less than 3 months (65 days) (Fig. 4). It was noteworthy that there was an abrupt change in spring phenology as well as the length of vegetation growing season in 2006, and the vegetation SOS (LOS) was delayed (shortened) by about 30 days compared with the previous year. Generally, SOS, EOS and LOS in SRYY showed trends of advancement, postponement and lengthening, respectively, but the variation varied widely in the last 20 years.

Relationship between vegetation phenology and moisture–heat conditions

Partial correlation between phenology and climate factors

Fig. 5 is the partial correlation between monthly average temperature (precipitation) and vegetation phenology in SRYY. Changes in the value of the partial correlation coefficient reflect alterations in the degree of correlation between the variables. On a monthly scale, the response relationship between vegetation phenology and climatic variables (temperature and precipitation) in the SRYY was consistent, then SOS negatively correlated with both, while EOS, on the contrary, showed a positive correlation. In spring, the multi-year average SOS in SRYY had a significant negative correlation (P < 0.01) with April air temperature, while the bias correlation with climate variables at other times was not statistically significant (P > 0.05); the multi-year average EOS in the source area, on the other hand, showed a strong positive correlation (P < 0.01) with August precipitation. The correlation between climate factors and vegetation phenology changed over time. And April temperature and August precipitation were highly correlated with SOS and EOS, respectively.

Spatial variation in the response of phenology to moisture–heat conditions

We focused on the partial correlations of climate factors with SOS and EOS in April and August as a reflection of the spatial variation in response relationships.

In April, SOS was highly influenced by temperature in the eastern SRYY, whereas the western high-elevation zone was significantly affected by precipitation. Most of the SRYY (88.56%) showed a negative correlation between monthly mean temperature and SOS. 21.68% of the area exhibited a significant negative correlation (P < 0.05) and these areas were concentrated in the northeastern of the SRYY (Fig. 6a). Additionally, 11.44% of the areas were positively correlated, and less than 1% showed obvious positive correlation, which was dispersed in the high-elevation zones at the edge of the source region. Meanwhile, the area ratio of mean monthly precipitation negatively correlated with SOS was 79.26%, and about 13.92% of the area showed a significantly negative correlation (P < 0.05), concentrated in the western Qiangtang Plateau. 20.74% was positively correlated and only 0.14% showed a significantly positive correlation (Fig. 6c).

In August, the percentage of area was 86.82% where EOS was positively correlated with monthly average precipitation. And about 27.99% of the region was clearly positive correlation (P < 0.05), which was dispersed throughout the entire source area. A significantly negative correlation was found in the plateau lake and the percentage of area was 13.18% (Fig. 6d). The percentages of regions positively or negatively correlated with monthly mean temperature were basically equivalent (53.11% and 46.89%) and both were obviously correlated by about 1% (Fig. 6b). Comparing the effect of temperature and precipitation on EOS during this period, the spatial difference was more pronounced in EOS under the influence of precipitation, while the effect of temperature was not significant.

Identifying climate factors on phenology changes on a spatial scale

The dominant factors affecting vegetation phenology were spatially and temporally differentiated in SRYY. In April, temperature had a considerable effect on vegetation growth in the east. About 30.14% of the area of SOS changed significantly (P < 0.1) under the interaction of temperature and precipitation. And the temperature-dominated area was more widely distributed, nearly half of the total significant area of change (49.44%) and concentrated in the northeastern part of the SRYY (Fig. 6e). Then 26.27% of the total significant change zone was dominated by precipitation mainly in Qiangtang Plateau and the southeast of the SRYY. Regions concurrently influenced by temperature and precipitation account for 16.77% (P < 0.1). In August, 29.88% of the entire source area experienced substantial alterations owing to the interaction between temperature and precipitation, where the impact of precipitation predominated and constituted more than 80% (84%) of the total significance (Fig. 6f).

DISCUSSION

Vegetation phenology in SRYY

In the past 20 years, the vegetation SOS, EOS and LOS in SRYY had advanced with a trend of 4.9 days/10 a, delayed with a trend of 1.5 days/10 a, and lengthened with a trend of 6.3 days/10 a, respectively, from west to east. Spatial evolution characteristics and interannual trends in phenology were found to agree with previous studies conducted by Wang and Yang (2023) and Shen et al. (2022) but differed in significance of changes. In this study, it was proved that the vegetation phenology in SRYY was dominated by nonsignificant changes, which was verified by Li et al. (2023) but contrary to Shen et al. (2022). This might be due to the differences in data sources or resolution as well as the extent of the study area (Meng et al. 2021). Different satellite products collect varying vegetation information (Shen et al. 2022) and vegetation phenology is extracted based on these values. Additionally, resampling at the same pixel size alters the original vegetation information, resulting in different vegetation phenology (Meng et al. 2021). This illustrates the impact of differences in data sources on vegetation phenology studies.

The study showed that the relationship and degree of influence between vegetation phenology and climate factors may be diverse or even opposite in different growth stages. Influencing factors of SOS had shifted from temperature to precipitation over time. Similarly, the impact factors of EOS had changed from precipitation to temperature. This variation may be related to the alternation of seasons. Temperature is the main limiting factor for vegetation growth in alpine areas. While the temperature begins to rise in the spring, prompting plants to enter the budding period extensively that vegetation has a strong demand for water. Therefore, impact factors become precipitation. While autumn temperature is cooler, vegetation growth is more influenced by temperature.

Vegetation phenology in SRYY showed anomalies in 2006, characterized by a significant delay in SOS and a longer growing season (Fig. 3). The length of the vegetation growing season depended on the changes in SOS and EOS. Over the past two decades, SOS was more variable while EOS was stable in SRYY. Thus, variation in LOS of the SRYY was more obviously influenced by SOS. Vegetation SOS was affected by both temperature and precipitation. The research analyzed the temperature and precipitation data and found that the cumulative precipitation in the spring of 2006 (65.9 mm) had decreased about 17.9 mm less than the multi-year average cumulative precipitation in the spring (83.8 mm), while the temperature had dropped 0.73°C during the same period (Fig. 7). It suggested that moisture and heat conditions were below the average level in the spring of 2006, especially thermal conditions. Relatively poor moisture and heat conditions can delay normal vegetation development and may bring a significant postponement in vegetation SOS. The research conducted by Li et al. (2021) and Jiang et al. (2021) proved that moisture–heat factors were a limiting factor for vegetation growth in spring, indirectly corroborating the above assertion. Besides, it was worth noting that May was a concentrated phase for SOS. The temperature in May had increased 0.27°C and precipitation had decreased about 15.18 mm in 2006, compared with previous years (Fig. 8), which could exacerbate the loss of soil moisture. At the same time, the Qinghai-Tibet Plateau region has a rainy season from May to September and a dry season from October to April. The lack of water in the dry season, coupled with the lack of water in May, caused the spring phenology to be significantly later than the multi-year average SOS in 2006. Therefore, the inequality of thermal and hydrological conditions will cause abnormal changes in vegetation phenology.

Spatial and temporal variability in vegetation phenology

For this study, six indicators (total nitrogen, total phosphorus, total potassium, organic matter, pH value and soil moisture) were examined in 252 specimens from sampling points (Fig. 9) selected from the SRYY. Eventually, we obtained measured soil physical and chemical information data of the study area. By analyzing the data, we found that there was a distinctly spatial variation in the indicators along the meridian direction (Fig. 10). Specifically, the soil condition in the east was better than in the west, indicating a more favorable living environment for vegetation in the east. That is why the spatial distribution of phenology was so different from east to west.

There were visibly spatial transitions of vegetation phenology in SRYY. From the west to the east, vegetation phenology experienced the process of SOS advance, EOS delay and LOS prolongation (Fig. 2a–c). There was the same pattern of vegetation change in the range of 3500–5000 m confirmed by Hu et al. (2022). The variability of vegetation phenology in vertical and horizontal directions reflected the changing patterns of regional moisture–heat conditions, further suggesting that moisture–heat conditions were limiting factors for phenological development. There were poorer soil properties and vegetation growth conditions in the west than in the east, resulting in a shorter LOS (less than 4 months) in the west. For example, the vegetation turned green earlier and the duration of the growing season was longer in Zeku county in the east, while LOS was significantly shorter in the border zone of the source area in the west (Fig. 2d–f). Especially, the change of vegetation phenology within the lake’s distribution zone differed from the overall SRYY. This area was at a high elevation but SOS started earlier or even at the same level as the east (Fig. 2a). The reason for this was that lakes could regulate the regional microclimate and provide a relatively stable environment for vegetation growth and development, while the lake provided moisture replenishment for soil also set the stage for the advancement of SOS (Chen et al. 2017; Li 2022). Nevertheless, in terms of the response relationship between phenology and climate factors, we discovered that 20.74% of the area in April and 13.18% of the area in August had a delayed SOS and an advanced EOS with rising precipitation (Fig. 6c and d). These regions were mainly located in the lakes in the western and northern parts of SRYY. The above explained that excessive soil moisture was possibly harmful to plant growth. Therefore, it was necessary to define the moisture threshold for vegetation growth and development.

The evolution of vegetation phenology in the study area was mostly insignificant, but the effects of climatic variables on phenology were spatially significant. The vegetation showed significant change only with an area ratio of 13.6% (SOS) and 5.9% (EOS) in forest and shrub distribution areas of the SRYY. And the vast meadow area had less change. Differential sensitivity of vegetation types to climate factors may explain the reason why vegetation phenology change is not generally evident in SRYY. Under the influence of the early climate factors, SOS in the eastern and southwestern high-altitude areas of SRYY was significantly affected by temperature and precipitation, respectively, and showed a negative correlation (Fig. 6a–c), which was verified in the results of previous research (Meng et al. 2021). While precipitation had a salient effect on EOS and showed a positive correlation over a large area of the research zone. Further studies revealed that SOS was dominated by earlier temperatures, and preliminary precipitation governed EOS (Fig. 6e and f). The spatial variability of dominant factors might be related to elevations and vegetation types. For example, forest and shrub vegetation had an early onset of the growing season in the eastern part of SRYY. Heat was a major factor in vegetation growth at high altitudes (Chen et al. 2014; Cong et al. 2012; Zhang et al. 2022). And the elevation of SRYY was generally higher (average altitudes over 4000 m). Hence, temperature rebound would have a greater impact on SOS. However, the EOS when heat was relatively abundant and precipitation became the main driver. This conclusion was verified in Jia’s study of the Qilian Mountain region (Jia et al. 2016). Meanwhile, Sun et al. (2023) believed that the optimal temperature and precipitation for the growth of alpine grassland and meadows were lower than those of forests and shrubs, thereby revealing the reason for phenology variances induced by vegetation types (Peng et al. 2021).

Hysteresis effects of temperature and precipitation on vegetation phenology

The time of the strongest climatic response was inconsistent with the intensive period of SOS or EOS, exhibiting a hysteresis phenomenon. Through analyzing the interplay between vegetation phenology and climate factors in SRYY, we found that the temperature in April and precipitation in August had a significantly positive effect on the growth and development of vegetation (Fig. 5). While the SOS of the vegetation extensively appeared in May in SRYY, and the EOS concentrated in October (Fig. 6a and b). There was a delayed phenomenon in time, with 1-month delay for temperature and 2-month delay for precipitation in the study area. That was to say, the accumulation changes in heat and precipitation during the early period would have an impact on the growth and development of vegetation, which reflected the lag effect of water and heat on the evolution of phenology to some extent.

Some scholars have found that changes in preliminary temperature and preliminary precipitation affected vegetation phenology (Du et al. 2019), and climate elements (temperature and precipitation) had a backward effect on SOS of 1–2 months (Li et al. 2023). In addition, the study of Zhang and Liu (2014) confirmed that seed germination required certain water and heat conditions, and the increase in temperature could accelerate the decomposition of organic matter in the soil, the thawing of permafrost, and the improvement of soil enzyme activities (Qiao et al. 2022). All of the above made it possible for high-quality growth of plants. However, these processes were inseparable from the accumulation of heat. To further explore the influence of climatic elements on preliminary phenology (before the concentration period of vegetation phenology), we calculated the monthly mean temperature and cumulative precipitation in March–April and August–September, respectively. Subsequently, we formed a spatial distribution chart of the dominant factors (Fig. 11). As can be seen from the figure, differences in area ratios were small between the effects of single or multiple climate factors on SOS, but the significance of the spatial patterns was obvious. And the impacts of preliminary precipitation on EOS were widespread. This indirectly reflected the complexity of the response mechanism between vegetation and factors such as temperature, humidity and moisture conditions (Li 2020).

Deficiencies

In this article, based on MODIS NDVI data and the dynamic threshold method, we obtain vegetation phenology data of the SRYY, and analyze the response relationship between landform, climate and phenology. More scientific research results have been obtained, but there are still deficiencies.

At first, vegetation types, an interfering factor, are ignored when we determine the thresholds. For various plant types, the range of thresholds is diverse (Deng 2021; Jia et al. 2016; Xia 2020). Although we refer to previous studies (Wang and Yang 2023) and use 0.2 and 0.5 as the thresholds for SOS and EOS, respectively, they all treat plants as a whole (i.e. there was only one vegetation type by default). In addition, NDVI data can reflect the distribution of vegetation but cannot distinguish or identify differences in plant types. Therefore, using NDVI data to determine the phenological period may affect the accuracy of the results.

Secondly, we analyze the relationship between climate factors and phenology only from the perspective of single natural factors (temperature and precipitation). It doesn’t combine multiple factors for analysis, and nor does it consider the influence of human factors. Furthermore, the climatic variables selected are monthly average temperature and precipitation, without observing the influence of extreme temperature or precipitation on the development of vegetation.

Furthermore, the precipitation data we utilized is generated through the Delta spatial downscaling method and the dataset has undergone accuracy validation from 496 independent meteorological observation stations data. However, due to sparse gauges in SRYY, the precipitation has a large uncertainty. In future studies, we will confirm the precipitation in this region using ‘A high-resolution near-surface meteorological forcing dataset for the Third Pole region (1979–2022)’ (https://data.tpdc.ac.cn/zh-hans/data/44a449ce-e660-44c3-bbf2-31ef7d716ec7). The purpose is to reduce uncertainties in precipitation data and improve the reliability of the study with the help of multi-source datasets.

Finally, the focus of previous studies has been on the Qinghai-Tibet Plateau region, with relatively limited research on SRYY. And the literature references lack regional specificity to some extent. However, our study is a better addition to vacancies in the study area and can provide a reference for phenological studies in this river source area. Successive studies can be refined from human activities and extreme weather.

CONCLUSIONS

Based on NDVI data and dynamic thresholds method, this article comprehensively identified the spatial and temporal variation patterns of vegetation phenology in SRYY and its response relationship with moisture–heat changes, and drew the following conclusions: (i) From 2002 to 2021, the vegetation of the SRYY had an earlier SOS, a later EOS and a longer LOS in the east than in the west, with area ratios changing significantly of 13.6%, 5.9% and 18% (P < 0.05), respectively. In terms of time, there was a corresponding trend of −4.9 (advanced), 1.5 (delayed) and 2.6 (lengthened) days/10 a. (ii) Vegetation phenology was influenced by terrain factors. In the range of 3500–5000 m, for every 100 m increase in elevation, SOS was delayed by 1.8 days, EOS was advanced by 0.8 days and LOS was shortened by 2.6 days. (iii) The effects of temperature and precipitation on vegetation phenology were various at different stages of vegetation growth. Influence factors of SOS experienced a shift from temperature to precipitation, while EOS experienced precipitation followed by temperature. (iv) The moisture–heat conditions in the previous period significantly affected the vegetation phenology in SRYY and the spatial variability was obvious. Specifically, the temperature in April significantly affected SOS in the eastern part of the SRYY, and the precipitation in August widely affected EOS.

Funding

This work was supported by the National Key Research and Development Project (2022YFC3201704), the National Natural Science Foundation of China (52079008, 52009006, 52109038), the Research Fund of Key Laboratory of Water Management and Water Security for Yellow River Basin, Ministry of Water Resources (2023-SYSJJ-10), the Natural Science Foundation of Hubei Province (2022CFB554, 2022CFD037) and National Public Research Institutes for Basic R&D Operating Expenses Special Project (CKSF2023311/SZ).

Acknowledgements

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

Conflict of interest statement. The authors declare that they have no conflict of interest.

REFERENCES