EN中文

Journal of Geographical Sciences >

2023 , Vol. 33 >Issue 3: 464 - 482

DOI: https://doi.org/10.1007/s11442-023-2092-z

Research Articles

Spatiotemporal variations in remote sensing phenology of vegetation and its responses to temperature change of boreal forest in tundra- taiga transitional zone in the Eastern Siberia

  • LI Cheng , 1, 2 ,
  • ZHUANG Dafang 1 ,
  • HE Jianfeng , 1, * ,
  • WEN Kege 1, 2
Expand
  • 1. State Key Laboratory of Resources and Environment Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • 2. University of Chinese Academy of Sciences, Beijing 100049, China
*He Jianfeng (1967-), PhD and Associate Professor, specialized in remote sensing applications. E-mail:

Li Cheng (1991-), PhD, specialized in applications of remote sensing of resources and environment and geographic information system. E-mail:

Received date: 2022-08-20

  Accepted date: 2022-10-25

  Online published: 2023-03-21

Supported by

Major Special Project-The China High-Resolution Earth Observation System(30-Y20A07-9003-17/18)

Fold

Abstract

Phenology is an important indicator of climate change. Studying spatiotemporal variations in remote sensing phenology of vegetation can provide a basis for further analysis of global climate change. Based on time series data of MODIS-NDVI from 2000 to 2017, we extracted and analyzed four remote sensing phenological parameters of vegetation, including the Start of Season (SOS), the End of Season (EOS), the Middle of Season (MOS) and the Length of Season (LOS), in tundra-taiga transitional zone in the East Siberia, using asymmetric Gaussian function and dynamic threshold methods. Meanwhile, we analyzed the responses of the four phenological parameters to the temperature change based on the temperature change data from Climate Research Unit (CRU). The results show that: in regions south of 64°N, with the rise of temperature in April and May, the SOS in the corresponding area was 5-15 days ahead of schedule; in the area between 64°N and 72°N, with the rise of temperature in May and June, the SOS in the corresponding area was 10-25 days ahead of schedule; in the northernmost of the study area on the coast of the Arctic Ocean, with the drop of temperature in May and June, the SOS in the corresponding area was 15-25 days behind schedule; in the northwest of the study area in August and the southwest in September, with the drop of temperature, the EOS in the corresponding areas was 15-30 days ahead of schedule; in regions south of 67°N, with the rise of temperature in September and October, the EOS in the corresponding area was 5-30 days behind schedule; the change of the EOS in autumn was more sensitive to the change of the SOS in spring, because the smaller temperature fluctuation can cause the larger change of the EOS; the growth season of vegetation in the study area was generally moving forward, and the LOS in the northwest was shortened, while the LOS in the middle and south of the study area was prolonged.

Key words： phenology; climate change; Siberia; asymmetric Gaussian function; GeoDetector

Cite this article

LI Cheng , ZHUANG Dafang , HE Jianfeng , WEN Kege . Spatiotemporal variations in remote sensing phenology of vegetation and its responses to temperature change of boreal forest in tundra- taiga transitional zone in the Eastern Siberia[J]. Journal of Geographical Sciences, 2023 , 33(3) : 464 -482 . DOI: 10.1007/s11442-023-2092-z

1 Introduction

A phenological event is a recurring occurrence, which shows how the environment, such as weather, water, and soil, affects plants and animals. The most important factor is the weather. It is a key indication of climate change since it not only represents the current weather but also the accumulation of weather over time. By studying phenology, we may gain a deep understanding of how the local climate changes and how it affects animals and plants (Lieth, 1974; Zhu and Wan, 1975; Zheng, 2002; Xu, 2005; Dai, 2013; Liu, 2018; Zhang, 2018; Li, 2019). Zu and Yang (2016) found that the spring temperature was negatively correlated with the spring phenology of broad-leaved forests, coniferous forests, and meadows while researching the relationship between the phenological changes of various vegetation and climatic factors in Northeast China from 1982 to 2006. Additionally, there was a link between the spring phenology and the winter temperature from the preceding year. Rupiya and Yang (2018) found that the SOS of cotton was substantially correlated with the average temperature at the start of the growing season in their study of the spatio-temporal variations in cotton phenology and temperature in Xinjiang from 2001 to 2016. According to Ma et al. (2016) the length of the growing season showed an increasing trend, and the inter-annual fluctuations of the greening and withering phases of the vegetation on the Qinghai-Tibet Plateau showed an overall trend in advance.
Traditionally, the birth of animals, and the growth and decline of nearby plants are traditionally observed by researchers to record phenology at the phenological observation site (Wan et al., 1979). This method can record local phenological phenomena truthfully and accurately. However, limited by human and material resources, it is difficult to go deep into areas with harsher natural conditions, such as the Eastern Siberian tundra-taiga in transition zone where human activities are less, and phenological records are fewer. Therefore, the phenology in this area, which is almost a pure natural condition, can reveal the response of the growth and development of plants and animals to climate change more naturally (Ebata and Tateishi, 2001; Zeng, 2015; Xue et al., 2016; Li et al., 2017).
Because of its broad coverage, prolonged observation period, robust periodicity, and geographic continuity, remote sensing observation technology offers a new approach to phenological observation that can accurately depict seasonal and inter-annual changes in vegetation development (Huete et al., 2002; Zhang et al., 2003; Chen and Wang, 2009; Wu et al., 2009; Varlamova and Solovyev, 2016; Guo et al., 2017). According to Justice et al. (1985), the NOAA-AVHRR time series data might be utilized to monitor global vegetation phenology and extract phenological markers such as vegetation growth season parameters by computing NDVI. For determining the beginning of the vegetation growth season, Lloyd et al. (1990) presented a threshold method and used a NDVI threshold of more than or equal to 0.099. Fischer (1994) specified a defined threshold to designate the beginning and end of the vegetation growth season. The threshold for the beginning of the vegetation growth season is vary in each study region due to variations in lighting, soil background, and land cover types. Relative Greenness Index (RGI) were used by White et al. (1997) to develop the Seasonal Medial Nodes (SMN), and RGI=0.5 was chosen by him as the dynamic threshold for the beginning and end of the vegetation growth season. In order to more precisely extract vegetation growth season data, the dynamic threshold approach was used to remove variations in environmental background values across research locations. According to studies by Jonsson and Eklundh (2004), the threshold for the beginning of the growth season should be better to set at 10%, which is more general and is particularly appropriate for high-latitude parts of the northern hemisphere where the vegetation growing season is brief and rapid.
Since 2000, MODIS data have increasingly supplanted NOAA-AVHRR as the primary data source for remote sensing monitoring phenology because of their improved geographic resolution, increased number of spectral bands, and shortened revisit durations (Yu and Zhuang, 2006). For related study, it was more practical to produce a variety of themed data products using MODIS raw data. The northern hemisphere’s high latitudes, however, resulted in a colder temperature and more snowfall. According to Picard et al. (2005), the vegetation phenology in Siberia derived from high-resolution satellite remote sensing photos with interference of snow cover may result in significant mistakes. The Resource and Environmental Science and Data Center of the Chinese Academy of Sciences produced a new NDVI data set by further removing cloud, snow, and shadow using MODIS MOD09 data. These NDVI products could more closely present the actual development process of vegetation owe to benchmark growth synthesis and improved interpolation techniques (Liu and Liu, 2013; Liu, 2017; Liu et al., 2017).
The MODIS-NDVI data, on the other hand, are dispersed data with 8-day time intervals that are inevitably going to be disturbed by atmospheric disturbances, changes in land use, and other causes, leading to aberrant points that do not follow the growth law of vegetation. To be consistent with the NDVI change curve during the vegetation growth season, those data must be fitted using a fitting tool. Reed et al. (1994) employed the median filtering method to remove cloud interference and retrieve information about the plant growth season, however this method would also remove information about the vegetation greenness. In order to smoothly fit the SPOT-NDVI data in northern China, Chen et al. (2000) employed the Savitzky-Golay smooth filter fitting approach. They then retrieved the vegetation characteristics of the growth season. The S-G fitting method is a window fitting. Because there are fewer fitting sites and a shorter growth season in the northern hemisphere’s high latitude region, it is challenging to produce more precise fitting results (Chen et al., 2000; Wen, 2016). The asymmetric Gaussian filter fitting approach, which is based on Gaussian filtering, was introduced by Jonsson and Eklundh (2002) taking into account the asymmetry of the vegetation growth season curve. To determine the growth season that was more in line with vegetation’s natural growth rule, the left and right sides of the growth season curve were fitted.
First, this study compared time series fitting functions including S-G smoothing filter fitting, asymmetric Gaussian fitting, and Double-Logistic fitting using MODIS-NDVI data from 2000 to 2017. Second, the dynamic threshold method was combined with the asymmetric Gaussian fitting method to suggest an NDVI curve that was better suited for the vegetation in high latitudes of the northern hemisphere. Lastly, four remote sensing phenological parameters were extracted: the Start of Season (SOS), the End of Season (EOS), the Middle of Season (MOS), and the Length of Season (LOS). The long-term series data of each parameter were linearly regressed in order to look at the spatiotemporal pattern of the parameters of each growing season. Along with examining the temporal and spatial distribution of temperature changes in the research location, changes in vegetation growth season parameters were also correlated with temperature variations.

2 Data and method

2.1 Study area

The research area, which spans from 115°E to 125°E and borders the Olenok Gulf in the Laptev Sea in the north and 60°N in the south, is situated in Central Siberia’s eastern region. The whole area is located on Russian territory owned by the Republic of Sakha (Yakutia). The mountainous region in the south has the maximum elevation, exceeding 500 m, whereas the northern portion is the easternmost portion of the North Siberian Lowland. The transect is located at a height of roughly 300 m on average. In general, the southeast has a higher elevation than the northwest. The Oleniok River, which flows through the northwest, the Lena River, which travels east to west, and its tributaries, the Vilyui River, which flows west to east, are the primary rivers in the transect. The climate in this region is typical of the continent, with wide variations in temperature between winter and summer, large variations in daytime and nighttime temperatures, and little precipitation. Winters are long and summers are short; the northernmost winters can last up to nine or ten months. It takes very little time for temperatures to rise in spring and fall. A Google Earth view of the study area is shown in Figure 1.
Figure 1 Scope and land cover types of the study area in the Eastern Siberia
Tundra, grassland, low shrubs, and deciduous coniferous woods make up the different vegetation cover types in this region, which run from north to south. There are not any major cities or other areas suitable for construction and human activity. Therefore, effects of human activity can be disregarded, and the distinctive phenological fluctuations of the vegetation in this area are approximatively considered the result of natural variability.

2.2 Data measure and process

2.2.1 NDVI data set

The Resource and Environmental Science and Data Center of the Chinese Academy of Sciences provided the MODIS-NDVI data for this study, which had a spatial resolution of 0.005° and a time resolution of 8 days. There are 828 photos altogether across the 18-year period from 2000 to 2017. The original MODIS-NDVI data product, which had a resolution of 0.005°, was resampled to 0.01° and transformed to tif format using the data aggregation method.

2.2.2 Land cover data set

The MODIS land cover product MCD12Q1, which was downloaded from the USGS website and used in this study, is shown in Figure 1. The type of land cover in the research area as determined by Terra and Aqua observation data is represented by this data. There are 17 primary land cover types in the land cover data set. The International Geosphere Biosphere Program (IGBP) divides it into 3 non-vegetable land kinds, 3 land development and mosaic land types, and 11 forms of natural vegetation.
Since the types of plant cover in the study area varied relatively little between 2000 and 2017, it is on the assumption that the vegetation types did not change in this research. As a result, using the MCD12Q1 land cover products from the study’s first year, the grid range consists of the following 12 data products: h19v01, h20v01, h21v01, h22v01, h21v02, h22v02, h23v02, h24v02, h23v03, h24v03, h25v03, h26v03. To produce land cover type data appropriate for the study area, the original data was sliced and spliced using MRT software.

2.2.3 Temperature data set

The CRU meteorological data set was the source of the meteorological information used in this study. This dataset, which integrates numerous well-known weather databases from across the world and covers the entire planet, is a grid climate dataset with a geographic resolution of 0.5° * 0.5° that was created by the Climate Research Unit of the University of East Anglia, UK. The full data set covers the years 1901 through the present. This study trimmed monthly average temperature data compatible with the spatiotemporal range of the NDVI data and used the local NDVI data to mask the global CRU data. There were 216 total issues of the temperature data, with 12 issues published annually from 2000 to 2017. To match the resolution of the NDVI time series data, the data was resampled to 0.01° using the closest neighbor approach.
The difference in temperature change between 2000 and 2017 was determined in accordance with the regression coefficients to characterize the linear change trend of the monthly average temperature after 18 years of monthly average temperatures were extracted using linear regression analysis. The formula is as follows.
$\left\{ \begin{matrix} k=\frac{\mathop{\sum }_{i=1}^{n}({{x}_{i}}-\bar{x})({{y}_{i}}-\bar{y})}{\mathop{\sum }_{i=1}^{n}{{({{x}_{i}}-\bar{x})}^{2}}}=\frac{\mathop{\sum }_{i=1}^{n}{{x}_{i}}{{y}_{i}}-n\bar{x}\bar{y}}{\mathop{\sum }_{i=1}^{n}x_{i}^{2}-n{{{\bar{x}}}^{2}}} \\ b=\bar{y}-k\bar{x} \\\end{matrix} \right.$
$\bar{y}=kx+b$
where k represents the temperature change slope, b represents the intercept of the temperature change regression equation, n represents the time span, i represents the i year(s) after 2000, x represents the $\bar{x}$ year, and y represents the temperature. represents the average value of the year, and $\bar{y}$ represents the average temperature.

2.3 Key remote sensing phenological factors extraction

The definitions of the four vegetation growth season variables examined in this paper are shown in Table 1. Four vegetation growth season characteristics were taken from the fitted and reconstructed NDVI time series data in order to mimic the vegetation growth season curve.
Table 1 Definition of Vegetation growth season parameters extracting from NDVI data
Vegetation growth season parameter Definition
The start of the season (SOS) Time for which the left edge has increased to 10% measured from the left minimum level.
The end of the season (EOS) Time for which the right edge has decreased to 10% measured from the right minimum level.
The mid of the season (MOS) Computed as the mean value of the times for which, respectively, the left edge has increased to the 80 % level and the right edge has decreased to the 80 % level.
Length of the season (LOS) Time from the start to the end of the season.
The NDVI curve of each cycle of the vegetation’s growth season in the high latitudes of the northern hemisphere is asymmetrical; the increase is more pronounced in the early stages of the growth season and less pronounced in the latter stages. In order to fit the two halves before and after each vegetation cycle, this study chose the fitting method of an asymmetric Gaussian function suited for piecewise fitting, using the maximum NDVI value as the midline.
Jonsson and Eklundh (2002) was the first to suggest the asymmetric Gaussian function fitting approach in 2002. The method’s key is to replicate the changes in vegetation growth season by using a combination of piecewise Gaussian functions. The season of vegetation growth is represented by each combination. Finally, a kind of smoothing method was used to connect each Gaussian fitting curve and recreate the time series.
The formula is:
$f(t)=f(t;\ {{c}_{1}},\ {{c}_{2}},\ {{a}_{1}},\ \ldots,\ {{a}_{5}})={{c}_{1}}+{{c}_{2}}g(t;\ {{a}_{1}},\ \ldots,\ {{a}_{5}})$
where c1 and c2 determine the reference plane and amplitude; a1, … a5 are non-linear parameters, which determine the shape of the function g(t; a1, … a5).
Among them:
$g\left( t;\ {{a}_{1}},\ \ldots,\ {{a}_{5}} \right)=\left\{ \begin{matrix} ~exp\left[ -{{\left( \frac{t-{{a}_{1}}}{{{a}_{2}}} \right)}^{{{a}_{3}}}} \right],\ t>{{a}_{1}} \\ ~exp\left[ -{{\left( \frac{{{a}_{1}}-t}{{{a}_{4}}} \right)}^{{{a}_{5}}}} \right],\ t<{{a}_{1}} \\\end{matrix} \right.$
The TIMESAT was used to derive four vegetation growth season characteristics, such as SOS, EOS, MOS, and LOS, based on the data after fitting and smoothing. The threshold setting should adhere to the following guidelines in light of the characteristics of shorter vegetation growing seasons and quick growth starting and ending periods in the high latitudes of the northern hemisphere: The first requirement is that this number should be as close as feasible to the winter NDVI background value; the second is that it cannot be too low in order to introduce noise and affect the NDVI signal. The dynamic criteria for the beginning and end of the growth season were set to 10% of the difference between the maximum and minimum values of the curve based on the findings of earlier studies (Jonsson and Eklundh, 2004). The linear change difference of each parameter over the course of 18 years was then determined by performing a linear regression analysis on each vegetation growing season parameter. Trend analysis was conducted on all pixels in regions where the parameters of each vegetation growth season change significantly, along with temperature changes, and the significance of the change trend was assessed using the P value.

3 Results and analysis

3.1 Distribution pattern of temperature change trend

3.1.1 Temporal and spatial distribution characteristics of monthly average temperature between 2000 and 2017

Based on the computed monthly average temperature for 18 years, create a 3D spatial distribution map with a precision of 0.01° * 0.01°, as shown in Figure 2.
Figure 2 Spatial distribution of monthly average temperature in the study area between 2000 and 2017
Figure 2 illustrates how the spatial characteristics of the average monthly temperature in the research area were largely similar, with clear differences in the latitude gradient and a cooling trend as latitude is increased. There was a monthly temporal variation in the temperature difference between the north and the south. The greatest temperature variations, which reached 14°C, occurred in the months of November, February, and March. In December and January, there was a temperature differential around 10°C between the north and the south. The temperature difference between the north and the south has been gradually decreasing since April, reaching its lowest point in September, at roughly 5°C, before gradually increasing after that.
However, it is important to note that in the Arctic Ocean coastal zone, the northernmost part of the research area, the temperature increases as latitude increases. This trend was particularly noticeable from December to January next year.
There were clear monthly temporal fluctuations in the parameters of temperature change with longitude. The temperature in the west at the same latitude was greater than that in the east from October through February next year. The temperature had the greatest variation from December to January next year. Since March, the temperature has been gradually rising in the east and falling in the west at the same latitude. The slope was the greatest from July to August, i.e., the temperature difference was the greatest; that is, the temperature in the west was lower than the temperature in the east.
At the same time, it is clear that the temperature distribution was more uneven throughout the May to October vegetation growth season than it had been at other times. This unevenness was most pronounced at the southernmost part of the research region.

3.1.2 The temporal and spatial distribution characteristics of the linear change trend of monthly average temperature from 2000 to 2017

The findings of a linear regression analysis on the research area’s monthly average temperature from 2000 to 2017 are displayed in Figure 3.
Figure 3 Linear trend of monthly average temperature in the study area between 2000 and 2017
Figure 3 shows that a temperature decrease of no more than 1°C occurred in the study region generally in July, the northwestern area in August, the southwestern area in September, and the area south of 67°N in November. Other months show a large rise in the research area’s average temperature. However, the range of temperature variations varied depending on the season and the location.
The following chart illustrates how the temperature varies by season: The largest temperature increases occurred in the spring (March to May) and winter (December to February of the next year), but the temperature changes in the summer (June to August) and autumn (September to November) were very different. With an average temperature increase between 3°C and 4°C and a maximum temperature increase of 5°C, March and April among them show the most noticeable temperature increase trend; January, October, and December show average temperature increases between 2°C and 3°C and maximum temperature increases of 4°C; March and May The overall temperature in July showed a slight cooling trend, with an average cooling of 0.5-1°C. In June, the average temperature increased by 1-2°C. There were heating and cooling phenomena in August, September, and November, but the change was not great, and the average temperature was within 1°C.
The following describes the spatial distribution of temperature changes: The temperature increase in the north was noticeably greater than that in the south in January, March, April, and October; the increase in temperature in the south was greater than that in the north in February and December; the research area’s temperature increased only slightly in May and June. The study area’s central region experienced 1°C temperature drop in July, while the southern and northern regions experienced a dip of only 0.5°C; there was a cooling tendency in the northwest and a warming trend in the southeast in August; in September, the southwest experienced a decreace in temperature while the north experienced an uptick; in November, the south had a decreace in temperature while the north experienced an uptick.
An extensive analysis of the temperature changes from April to October during the period of vegetation growth revealed that in April, the temperature increase in the study area south of 65°N was higher than that in other areas, reaching up to 2-3°C. In May, the temperature rose in the area with the largest range, near 65°-70°N, and the highest temperature rise was 2°C. In June, the area with the highest temperature rise was near 62°N and reached up to 3°C, and in October, the research area’s most northern region experienced a maximum temperature rise of 3.5°C; between 62°N and 70°N and between 60°N and 62°N, the temperature increase ranged between 2-3°C and 1-2°C, respectively. The smallest temperature increase range, of roughly 0.3°C, was in the southwest.

3.2 Analysis of the temporal and spatial change pattern of phenological factor change trend and the response of temperature change

3.2.1 The start of season

Figure 4a demonstrates that the SOS in the study region was mostly concentrated around the 120th to 150th day, with the SOS value being most concentrated around 140 days, or the middle and end of May, which marked the start of the plant growth season in the study area. Overall, SOS had a clear latitude distribution in the research area and was moving later and later from south to north. SOS first appeared in the Vilyui River Plain region to the southwest of the research area, and it most recently appeared along Olignok Bay.
Figure 4 Spatial distribution (a) and the linear trend (b) of average SOS in the study area from 2000 to 2017
Figure 4b demonstrates that SOS in the study area generally displayed an advance trend (P=0.0403 < 0.05), with a linear trend in 18 years ranging from -20 to 5 days. The northernmost portion of the area had been delayed along the Arctic Ocean (see Figure 4b in the blue rectangle). A particular law with the shift in latitude was seen in the spatial distribution of the SOS change trend. SOS reached its greatest value in advance around 64°-67°N and 71°N (in the black oval box in Figure 4b), and it can reach its maximum for about 25 days. SOS’s spatial transformation distribution varied in the longitude direction as well. The western portion of the research area experienced a substantial trend difference, whilst the eastern portion was rather flat.
Combining the SOS change with the temperature change trend reveals that the SOS change in the region south of 64°N was mostly influenced by the temperature change in
April and May, which occurred 5-15 days in advance. In the comparable region, the average correlation coefficient with the temperature in April is r = -0.4498, the highest correlation value is r = -0.8762, and the average correlation coefficient with the temperature in May is r = -0.5631, the highest correlation coefficient is r = -0.9275. The average correlation coefficient with the corresponding regional temperature in May is r = -0.5993, the maximum correlation coefficient is r = -0.9357, and the average correlation coefficient with the corresponding regional temperature in June is r = -0.4003, the maximum correlation the coefficient is r = -0.8580. Between 64°N and 72°N, SOS changes were primarily impacted by temperature changes in May and June. SOS in some regions of the research area’s northernmost Arctic Ocean coast (Figure 4b blue box) was delayed by 15 to 25 days, which was connected to the area’s cooling trend in July. Maximum and average correlation coefficients are r = -0.8137 and r = -0.4975, respectively.
SOS altered in advance the most (P = 0.0205 < 0.05) between 20 and 25 days in the 65°-68°N central region (the black box in Figure 4b), where the temperature increased significantly in April, May, and June.

3.2.2 The end of season

Figure 5a shows that the end of the vegetation growth season in the research area was from mid-September to mid- to late-October. EOS was mostly concentrated between 265 and 280 days, and EOS was mostly concentrated around 270 days. Latitude had an impact on EOS. The earlier the EOS is south of 70°N, and the later it is north of 70°N, the greater the latitude. The initial appearance of EOS was in the region of 65°-68°N (Figure 5a black box range).
Figure 5 Spatial distribution of average EOS (a) and the linear trend of EOS (b) in the study area from 2000 to 2017
Figure 5b illustrates how the spatial distribution of EOS had clear signs of latitude change, with a change range of -25 to -25 days and a trend of advancement in the north and southwest and a trend of postponement in the south. As seen in Figure 5b blue box range, the most pronounced EOS advancement was mostly concentrated in the region north of 67°N (P=0.0325), with an advancement range of 15-25 days, or as long as 30 days. The region where EOS was delayed was primarily located south of 67°N (P=0.0403), and the delay was between 5 and 25 days. The black box in Figure 5b represents the highest delay, which was 63°N for 30 days.
It can be seen that when combined with the temperature change trend graph, the region where EOS advanced essentially matched with the region where the temperatures decreased in August and September. The maximum correlation coefficient with the corresponding regional temperature in August is r = 0.8056, and the average correlation coefficient with that temperature is r = 0.4582; similarly, the maximum correlation coefficient with the corresponding regional temperature in September is r = 0.4478. The region where the temperature rises in August, September, and October was the one where EOS was delayed. The average correlation coefficient with the corresponding area temperature in August is r = 0.5007, and the maximum correlation coefficient is r = 0.7709; in September, the average correlation coefficient with the corresponding area temperature is r = 0.3162, and the maximum correlation coefficient is r = 0.8108; and in October, the average correlation coefficient with the corresponding regional temperature is r = 0.5183, and the maximum correlation coefficient is r = 0.8679.

3.2.3 The middle of season

Figure 6a demonstrates that the MOS occurred primarily between the 200th and 220th day, which corresponds to the midst of the study area’s vegetation growth season from late June to early July. To the north of 66°N, the latitude distribution was more pronounced. Its spatial distribution features demonstrated that the time difference of MOS to the south of 66°N was not great, and the maximum difference was within 5 days. The MOS is later the higher the latitude.
Figure 6 Spatial distribution of average MOS (a) and the linear trend of MOS (b) in the study area from 2000 to 2017
Figure 6b demonstrates that the difference in MOS changes was minimal and that the research area’s total MOS exhibits an advance trend (P=0.0986), with an advance range of 3-10 days. The research area’s 65°-68°N region had the biggest advance range (Figure 6b black box range) (P=0.0384), which extended up to 13 days in advance. Combining the temporal and regional distribution of temperature changes, it can be observed that the significant temperature increases in spring, the modest temperature increases in autumn, and the localized temperature decreases were mostly responsible for the overall forward movement of MOS.

3.2.4 The length of season

Figure 7a demonstrates that the majority of the LOS was concentrated in the 120-160 day period, meaning that the growth season was only around 4-5 months long. The region with the shortest LOS may be seen in the blue box in Figure 7a near 71°-72°N. In addition to the research area’s southernmost point, the longest area also encompassed the region about 64°N.
Figure 7 Spatial distribution of average LOS (a) and the linear trend of LOS in the study area from 2000 to 2017
According to the land cover map, the lowland plains and the intersection of deciduous coniferous woods and sparse forests on each side of the Vilyui River are where the longest LOS areas may be located.
The LOS shortening in the central and western parts was the longest between 68°-71°N (Figure 7b black box range), up to 25 days or more, as can be seen from Figure 7b. The LOS in the northern half of the research area revealed a shortening trend (P=0.0301), shortening by 20-25 days. While the prolonged days ranged from 10 to 30 days, the west near 66°N and the southwestern part of the research region had the longest extended days, up to 30 days or more. The central-eastern portion of 65°-68°N and the area south of 65°N exhibited a trend of extension of LOS (P=0.0379).

4 Discussion

There have been noticeable climate changes in recent decades in the eastern Siberian tundra-taiga transition zone, which is situated in the high latitudes of the northern hemisphere and where there is little human activity. In order to investigate how plants respond to climate change in their natural environment, free from man-made disturbances, first large-scale, long-term, high-resolution surface NDVI products were obtained by using the MODIS remote sensing images and the growth of the surface vegetation were characterized. Then, four vegetation remote sensing phenological parameters reflecting the characteristics of vegetation growth season changes were extracted by employing the asymmetric Gaussian fitting method and dynamic threshold method. This study revealed that SOS in the research area generally exhibited an early trend, while EOS varied in their timing due to regional temperature changes, with the majority of LOS exhibiting an extended trend. Comparing the temporal and spatial distribution of SOS with EOS from 2000 to 2017, it can be found that SOS had more pronounced latitude distribution features while EOS was more concentrated and shorter in duration. When the linear change trends of SOS and EOS are compared each other over time and space from 2000 to 2017, it can be found that EOS’s spatial distribution of the change trend clearly exhibited latitude change characteristics. In contrast to SOS’s advance under the same temperature change, EOS responded more to temperature changes. Because EOS is sensitive, even a slight temperature shift can result in noticeable changes. The reduction in LOS in the northern portion of the research area mostly caused by the tiny SOS shift in the area and the early EOS, as can be observed by combining the trend of SOS and EOS. The SOS advance in this area was bigger than the EOS advance, which contributed to the extension of LOS in the central and eastern portion of 65°-68°N (area inside the blue ellipse in Figure 7b). The SOS advance and the EOS delay in this region were the main causes of the extension of LOS in the southwest of the research area (area inside the green box in Figure 7b). Findings in this study are essentially in line with Xue et al. (2016) who investigated the phenology and temperature variations in Siberian coniferous forests and pointed out that there was the greatest length of the growth season and the highest temperature in 2005, and that temperature fluctuations had a significant impact on the length of the season. Additionally, the monthly average temperature spatial distribution map showed that the temperature rises with the rise of latitude in the northernmost Arctic Ocean coastline zone of the research area, particularly from December to January next year. The warm North Atlantic current might be what caused those phenomena.
When we compared the distribution of the various vegetation types in the study area over 18 years, we discovered that while there were minor variations in the vegetation types along the borders of the different vegetation types, overall changes in the distribution of vegetation in the study area were not significant. This resulted in stable experimental conditions for this study. In order to study the overall response of all vegetation in the high latitudes of the northern hemisphere to climate change, this study used the NDVI of all ground features in the study area as the research object, failing to differentiate between different vegetation types. This allowed the study area to reveal that the characteristics of vegetation remote sensing phenology changed corresponding to different climate change characteristic areas simply, intuitively, and clearly. Unquestionably, examining how various vegetation types respond to climate change will aid in a better understanding of the phenological responses of vegetation to climate change. The existing data in this study, however, are constrained according the sampling precision and the response of different vegetation types to climate change has not been studied. Therefore, the focus of our investigation was only on the various climate change characteristic areas and the vegetation remote sensing phenological change characteristic areas. Based on the findings of this study, field-sampling will be carried out in regions that have clearly visible phenological and climate changes, along with higher resolution meteorological data and more precise vegetation coverage data, in order to concentrate on the study of the response of various vegetation types to climate change.
This study was unable to collect enough site-measured meteorological data as experimental data due to the lack of meteorological observation sites in this region. Only a crop with a coarser resolution (0.5° × 0.5°) of the CRU global meteorological data set will yield it. Given that, the resolution of the NDVI data is 0.01°× 0.01°, this study may have made some errors that affected the experimental results. However, given the size of the research area as a whole, climate change was a large-scale, long-term, overall change event rather than a small-scale, quick shift. The research area’s 18-year temperature change trend map (Figure 3) shows that the overall temperature change was comparatively flat and that there was little significant variation in temperature changes across the study area. Therefore, CRU meteorological data sets are too imprecise to reflect the trend of temperature variations in the study region. This article, which was constrained by the resolution of the temperature data, did not discuss the small-scale special temperature-variation regions of vegetation phenology factors, instead focusing on regions with large-scale variation characteristics to describe a general rule of how vegetation phenology characteristics observed through remote sensing respond to temperature changes. Higher-resolution local temperature-vegetation phenological factors will be investigated subsequently with the dense deployment of meteorological sites, the collection of sampling data from field surveys, and the combination of satellite remote sensing photos with higher resolution.
Different NDVI time series curve fitting approaches were discovered to have a significant impact on the outcomes of the derived phenological factors during the process of extracting phenological factors. The most popular S-G fitting approach can keep the local variation of the curve to the greatest extent among them, but the peak shape it produces is narrower. The peak value of the NDVI curve is quite narrow in the northern half of the research region, particularly in the tundra area, where the growing period is short. Overfitting may occur in woods with a long and stable period. There are fewer fitting points and higher noise when using the S-G fitting approach. Similar to the Gaussian function utilized in this study, double logistic fitting is also a piecewise fitting. Its distinctive feature is the change in growth rate from small to large in the left half of the curve, whereas the change in growth rate from large to small in the right half. Numerous studies have utilized this technique to match the NDVI time series of high latitudes in the northern hemisphere since this feature is more consistent with the growth curve of high latitudes in that hemisphere. The Asymmetric-Gaussian function, however, has an excellent fitting effect in higher latitude grasslands and tundra regions, making it appropriate for NDVI time series curve fitting throughout the northern hemisphere. There must be a thorough discussion of the unique differences in the extraction of various vegetation phenological parameters using various fitting techniques.
At high latitudes in the northern hemisphere, temperature plays a significant role in biological processes and the development of vegetation. Most scientists agree that global warming will aid in overcoming productivity barriers. The capacity of vegetation for photosynthesis may be impacted by phenological changes in temperature cycles, however this is unknown. The timing and magnitude of the advance, for instance, will have different effects on the biochemical process of the frozen soil, which will lead to different trends in the physiological state and photosynthesis capacity of the vegetation in the following growing season, despite the fact that warming advances the start of the vegetation-growing season. It is still unknown how different vegetation kinds react to temperature fluctuations and how sensitive they are to them. While changes in vegetation phenology are a result of temperature changes, the same changes in vegetation phenology have a negative impact on temperature, which in turn influences temperature changes. As a result, further research that need done will be the feedback mechanism between temperature and phenological parameters in the vegetation.

5 Conclusions

In the high-latitude transect regions of the northern hemisphere from 2000 to 2017, this study collected change trend information of the vegetation growth season parameters SOS, EOS, MOS, and LOS based on remote sensing data. In parallel, a thorough examination of the impacts of the spatiotemporal characteristics of temperature changes in 18 years on the changes of various phenological parameters of vegetation was conducted in conjunction with the spatial distribution of temperature changes. Research reveals:
(1) Springtime temperatures often rise, which causes SOS in the study location to occur 10-25 days early. Among them, the temperature increased in April and May, advancing SOS in the south of the study area; the temperature increased in May and June, advancing SOS in the northern part; however, the temperature increase in June was modest and decreased in July, delaying SOS along the study area’s northernmost Arctic Ocean.
(2) Various locations show different trends in temperature rise and decline over different months of the autumn season, and the spatial distribution of EOS in the research area varied earlier and later. In August and September, temperatures declined and caused EOS to advance in the northwest and southwest of this area, respectively. The timing of EOS in the southern part of this area was delayed because of an increase in temperature in September and October. Compared to SOS at the start of spring, changes at the conclusion of EOS were more responsive to temperature changes. Smaller temperature fluctuations would cause greater variations in EOS.
(3) The research area’s general vegetation growth season was advancing, LOS in the northwest was getting shorter, and LOS in the middle and southern regions was getting longer.
This study demonstrates how closely Eastern Siberia’s vegetation phenology variations relate to climate change. The relationship between vegetation and climate change can be better understood by researching its coupling mechanism.

Acknowledgements

Thanks to Professor Liu Ronggao, Professor Xu Xinliang, and Dr. Liu Yang in the Institute of Geographic Sciences and Natural Resources Research of the Chinese Academy of Sciences, the Chinese Academy of Sciences Resource and Environmental Science Data Center (http://www.resdc.cn/), for providing data support for this research.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Chen X, Tan Z, Schwartz M D et al., 2000. Determining the growing season of land vegetation on the basis of plant phenology and satellite data in Northern China. International Journal of Biometeorology, 44(2): 97-101. (in Chinese)

PMID

[2]
Chen X Q, Wang L H, 2009. Progress in remote sensing phenological research. Progress in Geography, 28(1): 35-42. (in Chinese)

[3]
Dai J H, Wang H J, Ge Q S, 2013. Changes of spring frost risks during the flowering period of woody plants in temperate monsoon area of China over the past 50 years. Acta Geographica Sinica, 68(5): 593-601. (in Chinese)

DOI

[4]
Ebata M, Tateishi R, 2001. Phenological stage monitoring in Siberia by using NOAA/ AV HRR data. Singapore:The 22nd Asian Conference on Remote Sensing.

[5]
Fischer A, 1994. A model for the seasonal variations of vegetation indices in coarse resolution data and its inversion to extract crop parameters. Remote Sensing of Environment, 48(2): 220-230.

DOI

[6]
Guo J, Chen S, Xu B et al., 2017. Remote sensing monitoring of grassland vegetation greenup based on SPOT-VGT in Xilingol League. Geographical Research, 36(1): 37-48. (in Chinese)

[7]
Huete A, Didan K, Miura T et al., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment, 83(1/2): 195-213.

DOI

[8]
Jonsson P, Eklundh L, 2002. Seasonality extraction by function fitting to time-series of satellite sensor data. IEEE Transactions on Geoscience and Remote Sensing, 40(8): 1824-1832.

DOI

[9]
Jonsson P, Eklundh L, 2004. TIMESAT: A program for analyzing time-series of satellite sensor data. Computers & Geosciences, 30(8): 833-845.

DOI

[10]
Justice C O, Townshend J R G, Holben B N et al., 1985. Analysis of the phenology of global vegetation using meteorological satellite data. International Journal of Remote Sensing, 6(8): 1271-1318.

DOI

[11]
Li L H, Liu L S, Zhang Y L et al., 2017. Elevation-dependent alpine grassland and phenology on the Tibetan Plateau. Geographical Research, 36(1): 26-36. (in Chinese)

[12]
Li X T, Guo W, Ni X N et al., 2019. Plant phenological responses to temperature variation in an alpine meadow. Acta Ecologica Sinica, 39(18): 1-11. (in Chinese)

DOI

[13]
Lieth H, 1974. Phenology and seasonality modeling. Ecological Studies, 120(6): 461.

[14]
Liu R, Liu Y, 2013. Generation of new cloud masks from MODIS land surface reflectance products. Remote Sensing of Environment, 133: 21-37.

DOI

[15]
Liu R, Shang R, Liu Y et al., 2017. Global evaluation of gap-filling approaches for seasonal NDVI with considering vegetation growth trajectory, protection of key point, noise resistance and curve stability. Remote Sensing of Environment, 189: 164-179.

DOI

[16]
Liu R G, 2017. Compositing the minimum NDVI for MODIS Data. IEEE Transactions on Geoscience and Remote Sensing, 55(3): 1396-1406.

DOI

[17]
Liu Y J, Chen Q M, Ge Q S et al., 2018. Spatiotemporal differentiation of changes in wheat phenology in China under climate change from 1981 to 2010. Science China Earth Sciences, 48(7): 888-898. (in Chinese)

[18]
Liu Y J, Ge Q S, Dai J H, 2020. Research progress in crop phenology under global climate change. Acta Geographica Sinica, 75(1): 14-24. (in Chinese)

DOI

[19]
Lloyd D, 1990. A phenological classification of terrestrial vegetation cover using shortwave vegetation index imagery. International Journal of Remote Sensing, 11(12): 2269-2279.

DOI

[20]
Ma X F, Chen S Y, Deng J et al., 2016. Vegetation phenology dynamics and its response to climate change on the Tibetan Plateau. Acta Prataculturae Sinica, 25(1): 17-25. (in Chinese)

[21]
Picard G, Quegan S, Delbart N et al., 2005. Phenology modelling in Siberia and its impact on the carbon budget. Global Change Biology, 11(12): 2164-2176.

DOI

[22]
Reed B C, Brown J F, Vanderzee D et al., 1994. Measuring phenological variability from satellite imagery. Journal of Vegetation Science, 5(5): 703-714.

DOI

[23]
Rupiya X, Yang L, 2018. Monitoring spatial-temporal change of cotton phenology in Xinjiang and its response to climate change. Remote Sensing Technology and Application, 33(5): 923-931. (in Chinese)

[24]
Varlamova E, Solovyev V, 2016. Study of NDVI vegetation index in East Siberia under global warming. XXII International Symposium Atmospheric and Ocean Optics. Atmospheric Physics. International Society for Optics and Photonics.

[25]
Wan M W, Liu X Z, 1979. Methods of Phenological Observation in China. Beijing: Science Press: 42-66. (in Chinese)

[26]
Wen K G, 2016. Vegetation responses to climate change in northern high latitudes (2000-2013)[D]. Beijing: University of Chinese Academy of Sciences. (in Chinese)

[27]
White M A, Thornton P E, Running S W, 1997. A continental phenology model for monitoring vegetation responses to interannual climatic variability. Global Biogeochemical Cycles, 11(2): 217-234.

DOI

[28]
Wu W B, Yang P, Tang H J et al., 2009. Monitoring spatial patterns of cropland phenology in North China based on NOAA NDVI data. Scientia Agricultura Sinica, 42(2): 183-191. (in Chinese)

[29]
Xia C F, Li J, Liu Q H, 2013. Review of advances in vegetation phenology monitoring by remote sensing. Journal of Remote Sensing, 17(1): 1-16. (in Chinese)

[30]
Xu Y Q, Lu P L, Yu Q, 2005. Response of tree phenology to climate change for recent 50 years in Beijing. Geographical Research, 24(3): 412-420. (in Chinese)

DOI

[31]
Xue B L, Guo Q H, Gong Y W et al., 2016. The influence of meteorology and phenology on net ecosystem exchange in an eastern Siberian boreal larch forest. Journal of Plant Ecology, 9(5): 520-530.

DOI

[32]
Yu X F, Zhuang D F, 2006. Monitoring forest phenophases of Northeast China based on MODIS NDVI data. Resources Science, 28(4): 111-117. (in Chinese)

[33]
Zeng L L, 2015. Remote sensing for crop phenology detection: Taking corn and soybeans as examples[D]. Wuhan: Wuhan University. (in Chinese)

[34]
Zhang X D, Zhu W B, Zhang J J et al., 2018. Phenology of forest vegetation and its response to climate change in the Funiu Mountains. Acta Geographica Sinica, 73(1): 41-53. (in Chinese)

DOI

[35]
Zhang X X, Ge Q S, Zheng J Y, 2003. Overview on the vegetation phenology using the remote sensing. Advances in Earth Science, 18(4): 47-57. (in Chinese)

[36]
Zheng J Y, Ge Q S, Hao Z X, 2002. Impacts of climate warming on plants phenophases in China for the last 40 years. Chinese Science Bulletin, 47(21): 1826-1831. (in Chinese)

DOI

[37]
Zhu K Z, Wan M W, 1975. Phenology. Beijing: Science Press, 1-4. (in Chinese)

[38]
Zu J X, Yang J, 2016. Temporal variation of vegetation phenology in northeastern China. Acta Ecologica Sinica, 36(7): 2015-2023. (in Chinese)

Options
Outlines

/

〈 〉