Spatial and Temporal Variability of Key Bio-Temperature Indicators and their Effects on Vegetation Dynamics in the Great Lakes Region of Central Asia

Dryland ecosystems are fragile to climate change due to harsh environmental condition. Climate change affects vegetation growth primarily by altering some key bio-temperature thresholds. Key bio-temperatures are closely related to vegetation growth, and slight changes could produce substantial effects on ecosystem structure and function. Therefore, this study selected the number of days with daily mean temperature above 0 °C (DT 0 ), 5 °C (DT 5 ), 10 °C (DT 10 ), 20 °C (DT 20 ), the start of growing season (SGS), the end of growing season (EGS), and the length of growing season (LGS) as bio-temperature indicators to analyze the response of vegetation dynamics to climate change in the Great Lakes Region of Central Asia (GLRCA) for the period 1982–2014. On the regional scale, DT 0 , DT 5 , DT 10 , and DT 20 exhibited an overall increasing trend. Spatially, most of the study area showed that the negative correlation between DT 0 , DT 5 , DT 10 , DT 20 with annual NDVI increased with increasing bio-temperature thresholds. Especially, more than 88.3% of the study area showed a negative correlation between annual NDVI and DT 20 , as increased DT 20 exacerbated ecosystem drought. Moreover, SGS exhibited insignicantly advanced trend, and EGS experienced a signicantly delayed trend. The overall extending trend in LGS was mainly attributed to the delayed EGS. Besides, our study revealed that about 54.7% of the study area showed a negative correlation between annual NDVI and LGS, especially in the north, indicating a negative effect of climate warming on vegetation growth in the drylands. The results of this study will help assess the stability of vegetation to climate variability, and predict the response of vegetation to future climate change in the GLRCA.


Introduction
Global surface temperature was 1.09°C higher in 2011-2020 than 1850-1900, with larger increases over land (1.59°C) than over the ocean (0.88°C) (IPCC 2021). Climate change characterized by global warming has a greater impact on the structure and function of terrestrial ecosystems (Cheng et al. 2021;Holst et al. 2013). Climate warming has intensi ed glacier melting and permafrost degradation (Bolch et al. 2010;Zheng et al. 2020), resulting in a series of ecological and environmental effects, such as the release of large amounts of deep permafrost carbon and declining water table ). Meanwhile, climate warming enhanced vegetation productivity by promoting photosynthesis, extending growing seasons, especially at high latitudes and altitudes (Piao et al. 2007; Wang et al. 2011;Wei et al. 2018). In addition, plenty of evidence suggests that extreme high temperature events are increased with enhancement of frequency and intensity across the globe (Perkins et al. 2012;Tong et al. 2019). Although climate warming reduces the occurrence of extreme low temperature events, such as frost events, but the extension of the plant growing season due to warming may induce more frequent frost days during the growing season (Baumbach et al. 2017; Liu et al. 2018). Accompanied by changes in temperature, the adaptive capacity of vegetation to environmental changes and ecosystem vulnerability increases, signi cantly affecting the provision of global ecosystem services (Fu et al. 2013). Therefore, monitoring vegetation growth and understanding its response to temperature change is important for quantifying global carbon budget, and has become a hot topic in climate change research (Ballantyne et al. 2012;Sitch et al. 2015).
Ecosystems in arid and semi-arid regions are more fragile to climate change due to harsh environmental condition, and even slight changes in the climate can have a substantial in uence on such ecosystems (Li et al. 2021a; Yuan et al. 2021 Zhou et al. (2015) found that the warming trend in the GLRCA initially enhanced the greenness of vegetation before 1991, but then the continued warming trend inhibited further increase in greenness. In addition, several studies have also demonstrated that the increase in temperature prolonged the growing season, and in turn increased ecosystem productivity in some areas of the GLRCA (Bohovic et al. 2016;Wu et al. 2021). However, it remained uncertain how key bio-temperature thresholds changed over the GLRCA, and different bio-temperature may induce various impacts on terrestrial ecosystems under global warming. Therefore, it is essential to explore the spatial and temporal associations of vegetation growth and key bio-temperature thresholds in the GLRCA.
In this study, we selected the number of days with daily mean temperature above 0°C (DT 0 ), 5°C (DT 5 ), 10°C (DT 10 ), 20°C (DT 20 ), the start of growing season (SGS), the end of growing season (EGS), and the length of growing season (LGS) as key indicators of bioclimatology. Based on these indicators, we rstly investigated the temporal and spatial trends of bio-temperature in the GLRCA during the period 1982-2014. Secondly, Pearson's correlation coe cient was used to explore the correlation between NDVI and bio-temperature indicators, thus detecting the response of vegetation growth to temperature changes. This study provides a scienti c basis for quantitative assessment of vegetation growth changes in dryland ecosystems under global warming, and will be helpful for decision-making in implementing ecological restoration and conservation in arid and semi-arid regions.

NDVI data
Remote sensing data, characterized by time continuity and large spatial scales, is an e cient approach to monitor the growth status and cover of vegetation (Goward 1989), and has been widely used in

Climate data
The daily mean temperature data used in this study was obtained from the GLDAS-2. weather stations), and found that the daily mean temperature data of GLDAS had a fairly high accuracy.

Bio-temperature indicators
Bio-temperature thresholds are closely related to the growth and distribution of vegetation, and 0°C, 5°C, 10°C, 20°C have been widely used to assess regional heat resources (Yang et al. 2019;Yin et al. 2017;Zhao and Wu 2016). 5°C is the minimum temperature for photosynthesis of some tropical and subtropical evergreen broadleaf forests (Larcher and Biederman-Thorson 1980). Meanwhile, 5°C was often employed to quantify growing season length and was also considered to be a key indicator in modelling of global vegetation patterns in previous studies (Prentice et al. 1992;Ruml et al. 2017). Most thermophilic crops begin to grow when the daily mean temperature is steadily above 10°C, and DT 10 is closely associated with sprouting and withering of most arboreal leaves (Huang 1958;Qiu and Lu 1980). Furthermore, the number of days below 0°C (frost days) and DT 20  In the Northern Hemisphere, SGS was de ned as the rst day of the rst 6-day period before July 1st with a daily mean temperature greater than 5°C, and EGS was determined as the rst day of the rst 6-day period after July 1st with a daily mean temperature less than 5°C.
LGS was the number of days between SGS and EGS. Ultimately, seven bio-temperature indicators were selected in this study to analyze the effect of temperature change on vegetation growth across the GLRCA (Table 1).

Trend Analysis and Mann-Kendall test
In this study, a linear regression, calculated by the least-squares method, was performed to detect trends in NDVI and climate variables. A positive slope indicates an increasing trend and a negative slope indicates a decreasing trend. In addition, the Mann-Kendall test was adopted to assess the reliability of time series trends (Kendall 1990;Mann 1945). This non-parametric test, which does not require the data to follow a standard distribution pattern and is not affected by sporadic outliers (Sneyers 1990), has been widely used to examine trends in hydrological and environmental data (Luo et al. 2020;Zhao and Wu 2016). In this study, a P value <0.05 was considered signi cant.

Results
3.1. Spatial and temporal variation of temperature and NDVI Figure 1 illustrates the spatial patterns of trends in annual temperature, annual NDVI, and correlation between annual temperature and annual NDVI over the GLRCA for the period 1982-2014. During the entire study period, annual temperature exhibited an obvious increasing trend and had large spatially heterogeneity. Greater warming (> 0.035°C/yr) occurred mainly in the west and southeast. Meanwhile, approximately 63.8% of study area experienced a decreasing trend in annual NDVI. Spatially, the greater decrease (< -0.0008 yr −1 ) in annual NDVI was observed mainly in the west, and the greater increase (> 0.0008 yr −1 ) was primarily in the east. To investigate the response of vegetation growth to temperature changes, we further analyzed the correlation between annual temperature and annual NDVI based on Pearson's correlation coe cient. From 1982 to 2014, approximately 56.4% of the study area was subject to a positive correlation between annual temperature and annual NDVI, and a greater positive correlation (> 0.2) was observed primarily at high altitudes in the southeast.
3.2. Spatial and temporal variation of DT 0 , DT 5 , DT 10 and DT 20 To further explore the spatial and temporal variation in temperature, we analyzed the trends in DT 0 , DT 5 ,

Spatial and temporal variation of SGS, EGS and LGS
Based on daily mean temperature data in the GLRCA during 1982-2014, we further calculated three biotemperature indicators: SGS, EGS, and LGS. Figure 5 illustrates the interannual trends in SGS, EGS, and LGS at the regional scale in the GLRCA. During the entire study period, SGS exhibited a slight decreasing trend with a regional average rate of -0.040 days/yr. Meanwhile, EGS showed a pronounced increasing trend at a rate of 0.460 days/year, and passed the signi cance test at 0.01 level. Changes in LGS are controlled by changes in SGS and EGS. We further analyzed the interannual trend in LGS and found that LGS across the GLRCA increased at a rate of 0.500 days/yr (P < 0.01) at the regional scale. Figure 6 indicates the spatial distribution of trends in SGS, EGS, and LGS over the GLRCA during the period 1982-2014. Over the past 33 years, most of the study area (about 67.8%) exhibited a negative trend in SGS. Spatially, the greater advanced SGS (< -0.4 days/yr) was mainly observed in the south, while the delayed SGS was mostly in the north. For trends in EGS, more than 99.0% of the study area experienced an increasing trend. The larger delays in EGS occurred mainly in the central and southeast. Accompanied by an earlier SGS and a later EGS, LGS showed a pronounced positive trend over 98.1% of the study area, and a higher extended SGS was observed mainly in the central and south.
Temperature is one of the major drivers of vegetation phenology changes. Therefore, we analyzed the correlations between annual NDVI and three indicators of growing season (SGS, EGS, and LGS) derived from surface air temperature to identify the impact of growing season variability on vegetation dynamics (Figure 7). Over the whole study period, the regions showing a negative correlation between annual NDVI and SGS accounted for about 66.7% of the study area, and the greater negative correlation (< -0.2) occurred mainly in the north and southeast. Meanwhile, approximately 64.1% of the study area experienced a negative correlation between annual NDVI and EGS, and a larger negative correlation (< -0.2) was observed mostly in the north. Then, we analyzed the correlation between annual NDVI and LGS, and found that a negative correlation represented approximately 54.7% of the study area. Spatially, the greater negative correlation (< -0.2) between annual NDVI and LGS was mainly found in the central, and the larger positive correlation (> 0.2) was mainly in the southeast.

Discussion
In this study, we investigated the spatial and temporal variation of bio-temperature indicators (including DT 0 , DT 5 , DT 10  Vegetation phenology, including SGS, EGS, and LGS, is a sensitive signal indicating the response of vegetation dynamics to climate change (Richardson et al. 2012;Wu et al. 2021 LGS was observed at high elevations in the southeast. Furthermore, our study found that annual NDVI was negatively correlated with LGS in some regions of the GLRCA, especially in the north. In dryland ecosystems, precipitation is the major driver of vegetation greening, and increased evapotranspiration controlled by climate warming would lead to drought and exacerbate precipitation limitation (Ma et al. 2015). In addition, the increase in temperature might accelerate the growth of vegetation and thus lead to a shortening of the vegetation growth cycle, particularly for herbaceous plants (Sherry et al. 2007; ). Hence, these could explain the negative effect of extended LGS on vegetation growth across the GLRCA.

Conclusions
This study analyzed the temporal and spatial characteristics of bio-temperature indicators (DT 0 , DT 5 , DT 10 , DT 20 , SGS, EGS, and LGS) in the GLRCA based on surface air temperature data during 1982-2014, and examined the response of vegetation dynamics to climate change. The major ndings are as follows.
(1) With climate warming, DT 0 , DT 5 , DT 10 , and DT 20 all showed a pronounced increasing trend at the regional scale. Spatially, there was signi cant heterogeneity in the four indicators, particularly an obvious decrease in DT 20 was observed in the northeast.
(2) Most of the study area showed that the negative correlation between DT 0 , DT 5 , DT 10 , DT 20 with annual NDVI increased with increasing bio-temperature thresholds. Especially, more than 88.3% of the study area experienced a negative correlation between annual NDVI and DT 20 .
(3) During the entire study period, SGS exhibited insigni cantly advanced trend, and EGS experienced a signi cantly delayed trend. Therefore, the overall extending trend in LGS was mainly attributed to the delayed EGS.
(4) About 54.7% of the study area showed a negative correlation between annual NDVI and LGS due to precipitation limitation exacerbated by climate warming, especially in the north, indicating a negative effect of climate warming on dryland vegetation growth.