See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344881396

Impacts of snow cover duration on vegetation spring phenology over the

Tibetan Plateau

Article in Journal of Plant Ecology · May 2019

DOI: 10.1093/jpe/rty051

CITATIONS READS

18 168

6 authors, including:

ke Huang Jiaxing Zu
Chinese Academy of Sciences Chinese Academy of Sciences
71 PUBLICATIONS 1,512 CITATIONS 28 PUBLICATIONS 944 CITATIONS

SEE PROFILE SEE PROFILE

Yangjian Zhang Nan Cong

Institute of Geographic Science and Natural Resources Research, Chinese Academ… Peking University
177 PUBLICATIONS 5,886 CITATIONS 54 PUBLICATIONS 3,811 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Yangjian Zhang on 11 September 2022.

The user has requested enhancement of the downloaded file.

Journal of
Plant Ecology Impacts of snow cover duration on
VOLUME 12, NUMBER 3,
PAGES 583–592 vegetation spring phenology over
JUNE 2019

doi: 10.1093/jpe/rty051
the Tibetan Plateau
Advanced Access published
1 December 2018

available online at Ke Huang1,†, , Jiaxing Zu1,2,†, Yangjian Zhang1,3,4,*,

www.jpe.oxfordjournals.org
Nan Cong1, Yaojie Liu1,2 and Ning Chen1,2

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

1
Lhasa Plateau Ecosystem Research Station, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of
Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
2
University of Chinese Academy of Sciences, Beijing 100190, China
3
CAS Center for Excellence in Tibetan Plateau Earth Sciences, Beijing 100101, China
4
College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100190, China
*Correspondence address. Institute of Geographic Sciences and Natural Resources Research, Chinese Academy
of Sciences, 11 Datun Road, Chaoyang District, Beijing 100101, China. Tel: +86-010-64889703; Fax: +86-010-
64889703; E-mail: [email protected]
†
These two authors contributed equally to this work.

Abstract

Aims trends in SOS were observed in the eastern plateau, and the delayed
Snow cover occupies large percentage of land surface in Tibetan trends were mainly founded in western plateau. Snow cover area
Plateau. Snow cover duration (SCD) during non-growing seasons exhibited two separate peaks during autumn and late winter over
plays a critical role in regulating alpine vegetation’s phenology by the plateau. Extended SCD regions mainly distributed in middle-
affecting the energy budgets of land surface and soil moisture con- east of the plateau, while shrunken SCD distributed in other regions
ditions. Different period’s snow cover during non-growing season of the plateau. SCD of different seasons caused distinct effects on
may have distinct effect on the vegetation’s phenology. Start of vegetation SOS. Lengthened autumn SCD advanced SOS over the
season (SOS) has been observed advanced under the ongoing cli- eastern plateau. The slightly lengthened SCD postponed SOS over
mate change in the plateau, but it still remains unclear how the the western plateau. In the wet meadow regions, advanced SOS
SCD alters the SOS. This study attempts to answer the following was positively associated with SCD during the entire non-growing
questions: (i) What is the pattern of spatial and temporal variations season, whereas for the dry steppe, SCD over the preseason played
for SCD and grassland SOS? (ii) Which period’s SCD plays a critical a more dominant role. The SCD of previous autumn and winter also
role in grassland’s SOS? showed lag effect on SOS over meadow regions to a certain extent.
This study confirmed the importance of SCD to phenological pro-
Methods
cesses at the beginning of growing season and further suggested that
The remote sensing datasets from the Moderate Resolution Imaging
role of SCD should be discriminated for different periods and for dif-
Spectroradiometer (MODIS) were utilized to compute the SOS
ferent heat-water conditions. With the lag effects and SCD’s distinct
and SCD on the Tibetan Plateau over 2003–2015. The Asymmetric
effect of different seasons considered, predictions on the Tibetan
Gaussian function was applied to extract SOS. We also explored
Plateau’s spring phenology could be improved.
the spatial pattern and temporal variation of SOS and SCD. Then,
by using linear correlation coefficients, we investigated the driving
effects of different period’s non-growing season SCD on SOS. Keywords: snow cover duration, start of season, remote sensing,
Tibetan Plateau
Important Findings
The non-growing season SCD slightly decreased during 2003– Received: 1 May 2018, Revised: 14 November 2018,
2015, while SOS exhibited an overall advancing trend. Advanced Accepted: 29 November 2018

atmosphere (Beniston et al. 2018; Chapin et al. 2005, 2008).

INTRODUCTION In arctic and alpine regions, snow cover occupies a big pro-
Seasonal snow cover plays a vital role in cryosphere with portion of areas and persists for a long time during winter
large inner-annual variation and has significant influ- and spring. Especially for most of arid or semi-arid alpine eco-
ences on water and energy cycling between ecosystems and systems, seasonal snow cover is the major water resource for

© The Author(s) 2018. Published by Oxford University Press on behalf of the Institute of Botany, Chinese Academy of Sciences and the Botanical Society of China.
All rights reserved. For permissions, please email: [email protected]
584 Journal of Plant Ecology

soil moisture and runoff, which are necessary resources for Snow cover may be the auxiliary driving factor. To gain deep
vegetation growth and human being livings (Xu et al. 2009; insights into snow–phenology interactions, we focused on
Yang et al. 2010). The timing, depth, extent and duration, the Tibetan Plateau as the target. The Tibetan Plateau is dif-
which define the main attributes of snow cover, have been ferent from the arctic and other alpine regions due to its high
widely used for local- and regional-scale studies as indicators altitude, arid and cold climates. Land surfaces are not entirely
of climate change, as well as driving factors on biological and covered by snow during the whole non-growing season. Soil
physiological characteristics and ecosystem functions (Chen freeze/thaw processes may be affected by the snow cover
et al. 2008, 2017; Yuan et al. 2016). change during non-growing season, which can further me-
Before the era of ‘satellite meteorology’, ground-based me- diate vegetation green-up date under the arid climates on the
teorological networks have been monitoring snow for a long Tibetan Plateau. Furthermore, water availability also regu-
time, but their detectability is limited to local scales (Pu et al. lates spring phenology (Dorji et al. 2013). Melting water of

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

2007). Satellite remote sensing has greatly advanced scien- snowpack and active frozen soil layer influences soil mois-
tific understanding of snow cover over harsh environmental ture anomaly. Most previous studies on the Tibetan Plateau
areas at both regional and global scales. Numerous research has focused on snow melt timing effects on SOS, while neg-
work utilizing satellite dataset have presented that climate lected the lag effects of snow on phenology (Wang et al. 2015,
change is altering snowfall regime and then influences snow 2017a). It remains unclear when the snow cover plays a key
cover pattern and dynamics (Scherrer et al. 2004). According role in spring phenology (the critical period of snow cover
to daily radiance data acquired with the Advanced Very High duration [SCD]) and how spatially divergent these processes
Resolution Radiometer and Moderate Resolution Imaging interact in controlling different biomes’ snow–phenology re-
Spectroradiometer (MODIS) series of satellite-borne optical lationships? In brief, the objectives of this study were to (i)
sensors, a drastically decreasing of snow cover in northern quantify the spatial and temporal variations of SCD during
hemisphere has been observed since the 1970s due to in- non-growing season and grassland start of season (SOS) dy-
creased temperature (Hori et al. 2017). Also, decreasing ice namics in the corresponding period; (ii) identify the critical
and snow cover in the Arctic and high mountain region have period of SCD on grassland’s SOS; and (iii) reveal the effects
been suggested (Huang et al. 2017; Peng et al. 2013). Despite of previous winter and current spring SCD on SOS of the
the warming climate, winter precipitation will also increase current year.
the amount of snow accumulation (Saha et al. 2006).
The shifts in timing and extent of snow cover can influ-
ence vegetation seasonality. Some studies indicate that yearly
dynamics of vegetation productivity are regulated by the
MATERIALS AND METHODS
inter-annual variations of snow-free duration (Buus-Hinkler Study area
et al. 2006; Choler 2015; Wang et al. 2015). Alpine vegeta- The Tibetan Plateau is located in the southwestern part of
tion and Arctic tundra are seasonal snow-covered ecosystem China (74–104°E, 25–40°N), including Tibet Autonomous
and are supposed to adapt to limited period of growing season Region, Qinghai Province, and parts of Xinjiang, Gansu and
(Ernakovich et al. 2014; Rammig et al. 2010). Previous stud- Yunnan provinces (Fig. 1). It is home to numerous moun-
ies have highlighted that snow cover dynamics is closely tains over the southwest edge and the southeastern region.
related to surface temperature and soil moisture during dor- The plateau has an inner flat terrain with an average eleva-
mancy break period and early growth period (Matsumura and tion over 4000 m. Though the Tibetan Plateau lies in the
Yamazaki 2012; Westergaard-Nielsen et al. 2017). The insula- semi-arid/arid alpine climate zone, it owns larger snow and
tion effect from snow acts to prevent heat loss from soil during ice coverage when compared with other region on the same
winter and provide seeds and underground tissue like roots latitude due to its high altitude. The annual precipitation is
against winter winds and subzero cold injury (Comerford about 250 mm and ranges along a gradient from 150 mm in
et al. 2013; Repo et al. 2014). Snow regimes can affect under- the southeast to 300–500 mm in the northwest. Vegetation
story soil mineral nitrogen content and nitrate leaching losses, on the plateau mainly consists of meadow and steppe over
then alter plant phenology, growth and plant community the flat plateau and mountain sparse vegetation over the
composition for boreal forests (Kreyling et al. 2012), and arctic high mountain regions, shrublands in the valleys and small
tundra/alpine meadows (Rixen et al. 2008; Rogers et al. 2011). percentage of forest in the southern plateau. The grow-
Snow and wind have non-negligible ecological significances ing season in the plateau mainly covers May–September,
for litter redistribution in arctic landscape (Fahnestock et al. whereas during the non-growing season, the main land
2000). Earlier spring snowmelt is reported to advance spring cover is snow, glacier and bare soil. As the water tower of
phenology and increase frost damage to flower buds of herb- Asia, melted water from snow and glacier of the mountain
aceous species (Inouye 2008). region are the main runoff sources of many important rivers
The spring phenology of grassland was not consistently re- in China, such as the Yellow River, the Yangtze River and
lated to winter warming on the Tibetan Plateau (Shen 2011). Lantsang River.
Huang et al.     |     Impacts of snow cover duration on vegetation spring phenology585

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

Figure 1: the valid phenology extracting grassland types and ecoregions of the Tibetan Plateau, Ecoregion, was defined by different tempera-
ture belt, humidity zone and natural environment zone. HIB1E, eastern part of high mountain and high-cold shrub-meadow zone; HIB1W
western part of high mountain and high-cold shrub-meadow zone; HIC1, southern TP high-cold meadow-steppe zone; HIC2, Qiangtang high-
cold steppe zone; HID1, Kunlun high mountain and high-cold desert zone; HIIA/B1, Western Sichuan and eastern Tibet montane coniferous
zone; HIIC1, Southern Tibet montane shrub-steppe zone; HIIC2, Eastern Qinghai-Qilian montane steppe zone; HIID1, Ngari montane desert-
steppe and desert zone; HIID2, Qaidam montane desert zone; HIID3, Northern slopes of Kunlun montane desert zone.

Data source series of NDVI data (MOD13A2) was produced at sampling inter-
Snow cover dataset vals of 16 days using maximum value composite for 13 years of
2003–2015 by using MODIS instruments onboard Terra. This
The MOD10A1 and MYD10A1 (500-m spatial resolution and
dataset has been improved in both consistency and accuracy in
daily) provide snow cover information of land surface since
time series comparison, the quality and the directional normal-
the early 21st century. Optical remote sensing can detect
ization. It also limits the amount of missing data due to cloud
snow cover to some extent and also is sensitive to cloud con-
and atmosphere contamination to some extent. Zhang (2005)
dition. Qiu et al. (2017) applied temporal and spatial filter on
proved that vegetation phenology can be estimated with a high
MODIS snow cover dataset and produced cloud-free snow
precision from time series with the 16-day temporal resolutions
cover data. The data keep the original spatial resolution and
even if daily data contains some uncertainties. We employed
span from 2002 to 2016 (http://www.dx.doi.org/10.11922/
this dataset (https://ladsweb.nascom.nasa.gov/search/) to re-
sciencedb.170.55). This dataset has been field validated and
trieve vegetation SOS on the Tibetan Plateau.
proved their high performance in extracting snow cover over
the plateau. Phenology extraction and snow character
As for most regions in the plateau, snow-covered periods extraction
span October–December of the last year and January–May Because vegetation follows a definite annual cycle in most
of the current year. We define this period as the non-grow- regions of the plateau, accurate timing of phenological events
ing season of each year. In this study, the time span was set (SOS, growing season length etc.) can be extracted from NDVI
to 2003–2015 in accordance with phenology data at a spatial time series after noise removing. TIMESAT program, which is
resolution of 1 km. We used this data to obtain monthly SCD developed by Jönsson and Eklundh (2004) providing a tem-
at regional and pixel scales during the non-growing seasons of porally smoothed and spatially gap-filled algorithm for recon-
2003–2015. structing remote sensing time series. We used the standard
Asymmetric Gaussian demonstrated in TIMESAT to recon-
Remote sensing vegetation index time series struct the NDVI time series and extract SOS for the plateau
As an indicator of vegetation greenness, normalized difference (Eklundh and Jönsson 2015). The parameters were set as fol-
vegetation index (NDVI) and its time series reflect the inner- lows: the envelope iterations of 3, strength of 2 and threshold
annual dynamics of vegetation growth. A multi-annual time for SOS of 0.2. As for NDVI, the uncertainty is relatively large
586 Journal of Plant Ecology

when a pixel is barely vegetated or densely vegetated. We set Hypothetically, the effect of snow cover on spring phen-
several criterions to increase retrieving accuracies and ob- ology varies with month during non-growing season. To
tain valid SOS: (i) the maximum NDVI value should occur explore the effects of different periods SCD on SOS, the
during July and September; (ii) the multiyear average NDVI Spearman’s rank correlation coefficients were used to deter-
value during July and September is larger than 0.2; (iii) the mine the preseason dominant SCD, which was defined as the
non-growing season mean NDVI is no greater than 0.3 (Ding period when the highest correlation coefficients occurred be-
et al. 2016); and (iv) excluding forest ecosystems. These rules tween the date of SOS and SCD ahead of SOS with a 1-month
can minimize data uncertainties. Figure 1 shows the valid step. The maximum range of this period was defined as start-
phenology extraction areas and main vegetation type over ing from last October and ending to the multiyear average
the plateau. In this study, we excluded ecoregions with low date of SOS (Jeong et al. 2011; Liu et al. 2016a, 2016b). We
grassland coverage, resulting in six ecoregions on the Plateau also calculated the maximum positive and the negative co-

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

(Fig. 1). The HIIC1, HIIA/B1 and HIB1 are relatively humid efficients separately to identify the strongest SCD effects
region. The HIIC1 and HIC1 account for 87.89% of steppe. on spring phenology. The positive coefficients indicate that
The Tanggula Mountains divides the HIBI into two subre- shrunken SCD is parallel to advanced SOS, while the positive
gions: HIB1W and HIB1E. The HIB1W is characterized by ones indicate extended SCD is associated with delayed SOS.
slightly dryer climates than HIB1E. The ensemble mean of the correlation coefficients was calcu-
The SCD is defined as the number of snow covering days lated for each and across all vegetation types.
during a period. To characterize snow duration across the
Tibetan Plateau, daily snow cover data are aggregated to
monthly interval dataset, which were used to find the most
RESULTS
correlated SCD with SOS during non-growing seasons. The spatial pattern of SOS exhibited a great variation over
the whole plateau. Remotely sensed estimated results showed
Statistical analysis that the SOS of grassland gradually delayed from about
The phenological and SCD were used to calculate both yearly 100–120 day of year (DOY) in the east plateau to 200–210
averages and trends. The annual mean was produced on a DOY (later June) in the west plateau (Fig. 2a). Across all
per-pixel basis and also for each region and vegetation type. grassland types, the mountain sparse vegetation’s SOS oc-
Trends in SOS and SCD were examined using the Theil-Sen curred the earliest and the shrublands’ SOS occurred the
non-parametric regression (Hoaglin et al. 1983). This method latest. Meadow’s SOS was earlier than steppe (Fig. 2a). The
is more robust than ordinary least squares linear regression SOS advanced for the Tibetan Plateau with a Sen’s slope of
for short time-series data (Fernandes and Leblanc 2005) and −0.37 day/year (P < 0.1) during 2003–2015. Two distinct
also able to withstand up to 29% of samples being highly trends in SOS were observed between the western and the
variable within a dataset (Hoaglin et al. 1983). Fernandes and eastern plateau. Nearly 69.7% of the study area experienced
Leblanc (2005) reported that the Theil-Sen regression is suit- advanced trends of SOS, with approximately 27.3% of these
able for deriving trends from remotely sensed data, where the advanced areas being statistically significant (Fig. 2b). The sig-
underlying requirements for parametric regressions are often nificantly advanced phenology pixels were distributed over
not satisfied. The trend statistical significance was tested using HIB1, HIIC1 and HIIA/B1. Delayed SOS mainly occurred in
the Mann–Kendall test.

Figure 2: (a) the spatial pattern of multiyear average green-up date, while the line chart in the upper left-hand corner of main panel represents
the percentage of green-up pixel numbers for each vegetation type from early spring (early April) to middle summer (late June). (b) Inter-
annual variation of spring phenology, while the figure in the upper left-hand corner of main panel represents the pixels that exhibit significant
trend at the 0.01 level and marginal significant trend at the 0.1 level.
Huang et al.     |     Impacts of snow cover duration on vegetation spring phenology587

the southwest (HIIC2) and accounted for 31.3%, with 21.3% (steppe) (Fig. 4a and b). For HIIC1, HIBIE and HIIA/B1, SOS
of these delayed areas displaying significant trends. showed negative relationships with SCD. For HIB1W and
The spatial patterns of multiyear average non-growing HIC1, SCD showed positive relationships with SOS. Seasons
season SCD were depicted in Fig. 3a. The mean non-growing of dominant snow cover effects exhibited a heterogeneous
season SCD ranged from less than 30 days of the flat inner pattern. The positive and dominant explanatory effects from
plateau (HIC2 and HIIC2) to longer than 120 days of high SCD occurred mainly during winter (from January to March)
mountain regions and Sanjiangyuan (HIB1). Most of vege- and spring (from April to May), while the negative ones
tated areas exhibited approximately two snow cover peaks mainly related to autumn’s (from October to December) snow
(Fig. 3a). The first peak appeared in later November and the accumulation.
second appeared in middle February. Shrublands exhibited In terms of immediate and lag effects, they also exhibited
one sole peak on December. Meadow has the longest SCD a heterogeneous pattern over the plateau. Two parameters

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

over the whole plateau, while steppe has a shorter SCD. (i.e. the start month and the length of dominant month)
67.5% of the study area’s SCD experienced a slight decrease (Fig. 4c and d) were used to depict the dominant period of
during 2003–2015, with 23.4% of them being statistically sig- SCD effects on SOS. For relatively humid HIBIE, HIIA/B1 and
nificant (Fig. 3b). HIIC1, dominant snow accumulation period mainly occurred
Extended SCD regions were mainly distributed in middle- in autumn and lasted for more than 2 months. Extended
east of the plateau (HIB1E), accounting for 21.5% of the pla- autumn SCD advanced SOS over most areas of HIBIE, HIIA/
teau with a maximum Sen’s slope of −0.4 day/year. The most B1 and HIIC1. In relatively arid regions of HIB1W and HIC1,
pronounced SCD decreasing regions were mainly observed SOS was mainly regulated by instant snow cover just before
in the southwest plateau near Himalaya Mountain (HIIC2). vegetation dormancy breaking. These two regions experi-
Figs 3b and 2b showed that most of the plateau experienced enced shorter spring SCD and advanced SOS. For HIIC2 and
a reduced non-growing season SCD and an advanced grow- HIC2, shorter spring SCD occurred along with a delayed SOS.
ing season, except HIIC2 and HIB1E. For HIIC2, the SCD of To further explore and quantify the temporal pattern
non-growing season showed reduction, while SOS showed of snow cover’s driving effects on SOS, we summarized
delay trend. For HIB1E, extended SCD was in parallel to the average maximum correlation coefficients and corres-
advanced SOS. ponding pixel numbers, as well as starting month and length
By restricting to pixels with non-growing season SCD of SCD dominant period (Fig. 4e). In general, shorter spring,
greater than 5 days, we analyzed the relationships between extended autumn and winter snow durations advanced SOS.
SCD and SOS using correlation analysis. Both seasonal and For different dominant SCD on SOS, the temporal patterns
monthly SCD were used to distinguish the dominant snow of positive and negative coefficients pixels were distinctive.
cover period on SOS. The maximum correlation coefficient From previous October to the current May, the correlation
values between snow duration and SOS varied strongly across coefficient changed from negative to positive. During pre-
the Tibetan Plateau. We defined their positive relationships vious October and December’s dominant SCD, the higher
as vegetation phenology postponed by extended SCD, while correlated areas exerted longer time duration. For areas
the negative relationships as vice versa. For most areas in the where starting months are later than current January, the
East (meadow), SCD exerted negative effects on SOS, while correlation coefficients were positive and the effect lasted for
these relationships became positive in most parts of the west less than two months. As illustrated in Fig. 4f and g, snow

Figure 3: (a) the spatial pattern of average inter-annual snow cover duration (SCD), while the line chart in the upper left-hand corner of main
panel represents the percentage of snow cover pixel numbers for each vegetation type. (b) Inter-annual variation of SCD during entire non-
growing season, while the figure in the upper left-hand corner of main panel represents the pixels that exhibit significant trend at the 0.01 level
and marginal significant trend at the 0.1 level.
588 Journal of Plant Ecology

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

Figure 4: the effect of the dominant snow cover duration(SCD) on the start of season(SOS): (a) maximum correlation coefficient during the
dominant period, with top inset showing the significant correlation region; (b) the dominant SCD season; (c) start month of the dominant SCD
period, while the figure in the upper left-hand corner of main panel represents the pixels that exhibit significant correlation at the 0.01 level and
marginal significant correlation at the 0.1 level; (d) Length of the dominant SCD period; the temporal pattern and pixel number of (e) average
maximum correlation coefficient (unit: 102 pixels); (f) average maximum positive correlation coefficient and (g) average maximum negative
correlation coefficient and pixel numbers for dominant period over the plateau (unit: 103 pixels). Each block or circle’s color is the value of aver-
age maximum correlation coefficient, the number on the block and the size of circle indicate the number of pixels during respective dominant
period. Red indicates the negative coefficient, while the green indicates the positive.

cover during the period close to SOS obviously postponed mostly those areas where dominant SCD lasted for more
SOS. Closer to the previous end of season, advancing ef- than 3 months.
fects of snow cover on current SOS became more signifi- We further analzyed the spatial pattern of negative and
cant (Fig. 4g). Furthermore, the highly correlated areas were positive relationship pixels for meadow and steppe separately.
Huang et al.     |     Impacts of snow cover duration on vegetation spring phenology589

Compared with steppe, higher percentage of negative corre- MODIS vegetation index time series are effective in cap-
lated pixels was found for meadow (Table 1). Longer autumn turing spring phenology over most grasslands of the Tibetan
SCD generated greater advancing effects on SOS of the next Plateau. In this study, MODIS data showed a significant ad-
year. For steppe, positive correlation pixels in spring accounts vanced trend of spring phenology, which is in accordance
for a higher percentage than negative correlation pixels in with the findings extracted from other remote sensing data
autumn with instantaneous effect. (Shen et al. 2014, 2015; Zhang et al. 2013). Temperature is
In further exploring SCD’s effects on the SOS, we mapped suggested to have been increasing over the whole plateau in
inter-annual variations of dominant SCD (online supplemen- the past decades (Tao et al. 2014; Wang 2018), and its effects
tary Fig. S1a). For regions where SCD displayed negative on vegetation spring phenology are obvious (Cao et al. 2018;
effects on the SOS, elongated SCD during autumn advanced Cong et al. 2017; Liu et al. 2016c). SCD also can regulate re-
the spring phenology. For meadow and steppe together, the- sponses of spring phenology to warming by changing mois-

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

dominant SCD exhibited a marginal shortened trend, while ture conditions.
the SOS delayed for most areas of steppe (Fig. 1). For most The relationships between dominant SCD and spring
areas of steppe (HIIC2 and HIC2), the dominant SCD was phenology vary considerably among different seasons on
short (online supplementary Fig. S1b). the Tibetan Plateau. The finding that shrunk spring SCD is
correlated with advanced spring phenology indicates that
SCD influences the start of growing season to some extent.
DISCUSSION Previous similar studies have also suggested that less snow
This study revealed that SCD on the Tibetan Plateau has been cover considerably advances start of carbon uptake period
shortening due to a warming climate despite increased pre- in subalpine and start of growing season in alpine range-
cipitation. Non-growing season temperature was reported to land (Galvagno et al. 2013; Paudel and Andersen 2013). The
increase since the late 20th century for the Tibetan Plateau reason lies in that earlier snow melting is related to warmer
(Yang et al. 2006). The winter temperature was below zero temperature and earlier nutrients utilization by plants. The
on the plateau, and increasing temperature neutralized changing environment caused by spring snow provides
increased precipitation by thawing more snow (Gao et al. immediate beneficial conditions for vegetation growth and
2012; Peng et al. 2010). Snow cover extent on the plateau their effects on start of growing season hardly have any
normally has two peaks during non-growing seasons, and time lag.
average spring SCD is longer than that of autumn. The find- For SCD of previous autumn and winter, the relationships
ings reported by the present study are in accordance with between SCD and start of growing season are mainly nega-
other studies both based on MODIS datasets (Li et al. 2018) tive and also display certain lag effect. The longer previous
and weather stations records (Duo et al. 2017; Xu et al. 2017). autumn SCD is paralleled with an earlier start of growing
The SCD during autumn, winter and spring exhibited dif- season. The autumn and winter temperatures are below zero
ferent trends. For heavy snow area over the high mountain, over most regions of the plateau. A longer snow cover period
snow cover dramatically reduced, and in Sanjiangyuan, it can provide longer insulating of wind and prevent exposing
slightly increased. The distinct pattern indicates that their soils to extremely low temperature, thereby decreasing plant
air-water mass were originated from different water vapor root mortality during autumn and winter (Wang et al. 2018;
channels (Krishnamurti and Kishtawal 2000), as snow fall Zhang 2005). This should be further validated by soil tem-
is strongly linked to available moisture (Pu and Xu 2009). perature dataset and SCD. Some researchers suggested that
The plateau is situated in a semi-arid climate zone with mean when soils are frozen, both liquid precipitation and substan-
annual precipitation less than 500 mm, as well as relatively tial snowpack melt can run off via overland flow from upland
low snowfall. As a kind of solid rainfall, snow cover condition to aquatic ecosystems (Qiao et al. 2017). This phenomenon
was regulated by climates. In turn, snow also changed the is extremely common over the ‘carpet’ grassland of alpine
water budget for growing season vegetation (Choler 2018), meadows region with waterproof dense shoot and root litter
and further influenced the energy balance of the plateau and layers over the top soil layer. However, autumn snow cover
East Asia (Wang et al. 2017b). for previous year can keep melting water through freezing

Table 1: statistic of positive and negative pixel numbers of steppe and meadow

Negative Positive

Less lag effect Long time lag Less lag effect Long time lag

October–January February–May October–January February–May October–January February–May October–January February–May

Steppe 19.95% 18.39% 10.88% 3.61% 13.15% 14.03% 7.26% 2.44%

Meadow 29.15% 12.94% 16.59% 2.20% 9.06% 18.11% 3.71% 2.88%
590 Journal of Plant Ecology

soil, which would be a part of soil water reserve for moisture FUNDING
supply during early growing seasons.
This study was financially supported by the ‘Strategic Priority Research
Different vegetation types and species take distinct strate-
Program (A)’ of the Chinese Academy of Sciences (XDA20050102),
gies in utilizing snowfall and accumulating snow resources
a science and technology project of State Grid Corporation of
(Kreyling 2010; Legault and Cusa 2015), thereby exhibit- China (SGxzzzlwzhbGCJS1700095), which hosted by the State
ing varied responses to snow cover. The spring phenology of Grid Corporation of China, National Natural Science Foundation
meadow and steppe on the Tibetan Plateau and different sub- of China (Young Scientists Fund 41801083, China National Funds
regions’ vegetation showed varied responses to SCD. Over the for Distinguished Young Scientists 41725003 and Young Scientists
northern plateau, the frozen soil temperature is less sensitive Fund 41501103) and the Postdoctoral Research Foundation of China
(2018M631560).
to snow cover, and steppe showed less response to SCD vari-
Conflict of interest statement. None declared.
ation during autumn (Luo et al. 2016). Alpine Steppe on the

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

Tibetan Plateau is mainly composed of short longevity spe-
cies, and their active growing seasons are short. They nor-
mally develop strategy of starting growth once snow melts. REFERENCES
Meadows distributed in the head of three-river region, which
Beniston M, Farinotti D, Stoffel M, et al. (2018) The European moun-
is the wetter region of the plateau, showed expanded SCD tain cryosphere: a review of its current state, trends, and future
during the whole non-growing season, whereas the spring challenges. Cryosphere 12:759–94.
phenology still advanced. The possible reason lies in that Buus-Hinkler J, Hansen BU, Tamstorf MP, et al. (2006) Snow–vege-
snow is relatively deep in these areas. Under climate warm- tation relations in a high arctic ecosystem: inter-annual variability
ing, deep snow insulates vegetation buds from frost and inferred from new monitoring and modeling concepts. Remote Sens
melting water under warmer temperature facilitates vegeta- Environ 105:237–47.
tion growth during early growing season (Huang et al. 2017). Cao R, Shen M, Zhou J, et al. (2018) Modeling vegetation green-
For relatively arid regions, snow falls in fairly low frequency up dates across the Tibetan Plateau by including both seasonal
and with rather small water equivalent, snow fall exerts less and daily temperature and precipitation. Agr Forest Meteorol
impacts on vegetation spring phenology when compared with 249:176–86.
other environmental factor. Chapin FS, Randerson JT, McGuire AD, et al. (2008) Changing feed-
backs in the climate-biosphere system. Front Ecol Environ 6:313–20.

CONCLUSION Chapin FS III, Sturm M, Serreze MC, et al. (2005) Role of land-surface
changes in arctic summer warming. Science 310:657–60.
This study explored how the spatiotemporal variability of non- Chen XN, Long D, Hong Y, et al. (2017) Observed radiative cooling
growing season SCD correlates with vegetation spring phenol- over the Tibetan Plateau for the past three decades driven by
ogy on the Tibetan Plateau. According to the outcome of the snow cover-induced surface albedo anomaly. J Geophys Res Atmos
correlation analysis of SCD and SOS, this study confirms the 122:6170–85.
hypothesis that snow cover influences on alpine vegetation Chen WN, Wu Y, Wu N, et al. (2008) Effect of snow-cover duration on
phenology differs with periods and vegetation types. In gen- plant species diversity of alpine meadows on the eastern Qinghai–
eral, the SCD in previous autumn and early winter plays a Tibetan Plateau. J Mt Sci 5:327–39.
positive role in advancing the start of growing season in rela- Choler P (2015) Growth response of temperate mountain grasslands
tively wet and meadow regions. In arid regions, steppe showed to inter-annual variations in snow cover duration. Biogeosciences
delayed spring phenology in response to shortened non-grow- 12:3885–97.
ing season SCD. The shortened SCD in spring tend to induce Choler P (2018) Winter soil temperature dependence of alpine plant
direct effects in advancing start of vegetation growing season. distribution: implications for anticipating vegetation changes
under a warming climate. Perspect Plant Ecol Evol Syst 30:6–15.
This study overviewed effects of various time lag and dominant
period’s SCD on vegetation phenology. Future work should Comerford DP, Schaberg PG, Templer PH, et al. (2013) Influence of ex-
perimental snow removal on root and canopy physiology of sugar
address the relationships between vegetation phenology and
maple trees in a northern hardwood forest. Oecologia 171:261–9.
snow accumulation features (e.g. snow depth or snow water
Cong N, Shen M, Piao S (2017) Spatial variations in responses of
equivalent) and includes other climate variables (e.g. tempera-
vegetation autumn phenology to climate change on the Tibetan
ture and rainfall) in examining the mechanism behind snow–
Plateau. J Plant Ecol 10:744–52.
vegetation interactions on the Tibetan Plateau. Furthermore,
Ding MJ, Li LH, Nie Y, et al. (2016) Spatio-temporal variation of spring
effective combination of ground and satellite observations for
phenology in Tibetan Plateau and its linkage to climate change
detailed long-term investigations should be envisaged. from 1982 to 2012. J Mt Sci-Engl 13:83–94.
Dorji T, Totland O, Moe SR, et al. (2013) Plant functional traits me-
SUPPLEMENTARY MATERIAL diate reproductive phenology and success in response to experi-
mental warming and snow addition in Tibet. Global Change Biol
Supplementary material is available at Journal of Plant Ecology 19:459–72.
online.
Huang et al.     |     Impacts of snow cover duration on vegetation spring phenology591

Duo C, Quzhen L, Zhigang Y, et al. (2017) Spatiotemporal variations Li CH, Su FG, Yang DQ, et al. (2018) Spatiotemporal variation of snow
of snowfall days over the Tibetan Plateau from 1981 to 2010. J cover over the Tibetan Plateau based on MODIS snow product,
Appl Meteorol Sci 28:292–305. 2001–2014. Int J Climatol 38:708–28.
Eklundh L, Jönsson P (2015) TIMESAT: a software package for Liu Q, Fu YH, Zeng Z, et al. (2016a) Temperature, precipitation, and
time-series processing and assessment of vegetation dynamics. In insolation effects on autumn vegetation phenology in temperate
Kuenzer C, Dech S, Wagner W (eds). Remote Sensing Time Series. China. Global Change Biol 22:644–55.
Remote Sensing and Digital Image Processing, vol 22. Springer, Cham, Liu Q, Fu YH, Zhu Z, et al. (2016b) Delayed autumn phenology in
141–58. the Northern Hemisphere is related to change in both climate and
Ernakovich JG, Hopping KA, Berdanier AB, et al. (2014) Predicted spring phenology. Global Change Biol 22:3702–11.
responses of arctic and alpine ecosystems to altered seasonality Liu LL, Zhang XY, Donnelly A, et al. (2016c) Interannual vari-
under climate change. Global Change Biol 20:3256–69. ations in spring phenology and their response to climate change
Fahnestock J, Povirk K, Welker J (2000) Ecological significance across the Tibetan Plateau from 1982 to 2013. Int J Biometeorol

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

of litter redistribution by wind and snow in Arctic landscapes. 60:1563–75.
Ecography 23:623–31. Luo S, Fang X, Lyu S, et al. (2016) Frozen ground temperature trends
Fernandes R, Leblanc SG (2005) Parametric (modified least squares) associated with climate change in the Tibetan Plateau three river
and non-parametric (Theil-Sen) linear regressions for predicting source region from 1980 to 2014. Clim Res 67:241–55.
biophysical parameters in the presence of measurement errors. Matsumura S, Yamazaki K (2012) A longer climate memory carried
Remote Sens Environ 95:303–16. by soil freeze–thaw processes in Siberia. Environ Res Lett 7:045402.
Galvagno M, Wohlfahrt G, Cremonese E, et al. (2013) Phenology and Paudel KP, Andersen P (2013) Response of rangeland vegetation
carbon dioxide source/sink strength of a subalpine grassland in to snow cover dynamics in Nepal Trans Himalaya. Clim Change
response to an exceptionally short snow season. Environ Res Lett 117:149–62.
8:025008.
Peng S, Piao S, Ciais P, et al. (2010) Change in winter snow depth and
Gao J, Williams MW, Fu X, et al. (2012) Spatiotemporal distribution of its impacts on vegetation in China. Global Change Biol 16:3004–13.
snow in eastern Tibet and the response to climate change. Remote
Peng S, Piao S, Ciais P, et al. (2013) Change in snow phenology and
Sens Environ 121:1–9.
its potential feedback to temperature in the northern hemisphere
Hoaglin DC, Mosteller F, Tukey JW (1983) Understanding robust and over the last three decades. Environ Res Lett 8:014008.
exploratory data analysis. In Hoaglin DC, Mosteller F, Tukey JW
Pu ZX, Xu L (2009) MODIS/Terra observed snow cover over the
(eds). Wiley Series in Probability and Mathematical Statistics. New York,
Tibet Plateau: distribution, variation and possible connection with
NY: Wiley, 129–63.
the East Asian Summer Monsoon (EASM). Theor Appl Climatol
Hori M, Sugiura K, Kobayashi K, et al. (2017) A 38-year (1978–2015) 97:265–78.
northern hemisphere daily snow cover extent product derived
Pu ZX, Xu L, Salomonson VV (2007) MODIS/Terra observed seasonal
using consistent objective criteria from satellite-borne optical sen-
variations of snow cover over the Tibetan Plateau. Geophys Res Lett
sors. Remote Sens Environ 191:402–18.
34:6.
Huang XD, Deng J, Wang W, et al. (2017) Impact of climate and eleva-
Qiao H, Tian YQ, Yu Q, et al. (2017) Snowpack enhanced dissolved
tion on snow cover using integrated remote sensing snow products
organic carbon export during a variety of hydrologic of events
in Tibetan Plateau. Remote Sens Environ 190:274–88.
in an agricultural landscape, Midwestern USA. Agric For Meteorol
Inouye DW (2008) Effects of climate change on phenology, frost 246:31–41.
damage, and floral abundance of montane wildflowers. Ecology
Qiu Y, Zhang H, Chu D, et al. (2017) Cloud removing algorithm for
89:353–62.
the daily cloud-free MODIS-based snow cover product over the
Jeong SJ, Ho CH, Gim HJ, et al. (2011) Phenology shifts at start vs. Tibetan Plateau. J Glaciol Geocryol 39:515–26.
end of growing season in temperate vegetation over the nor-
Rammig A, Jonas T, Zimmermann NE, et al. (2010) Changes in al-
thern hemisphere for the period 1982–2008. Global Change Biol
pine plant growth under future climate conditions. Biogeosciences
17:2385–99.
7:2013–24.
Jonsson P, Eklundh L (2004) Timesat—a program for analyzing time-
Repo T, Sirkia S, Heinonen J, et al. (2014) Effects of frozen soil on
series of satellite sensor data. Comput Geosci 30:833–45.
growth and longevity of fine roots of Norway spruce. Forest Ecol
Kreyling J (2010) Winter climate change: a critical factor for tem- Manag 313:112–22.
perate vegetation performance. Ecology 91:1939–48.
Rixen C, Freppaz M, Stoeckli V, et al. (2008) Altered snow density and
Kreyling J, Haei M, Laudon H (2012) Absence of snow cover reduces chemistry change soil nitrogen mineralization and plant growth.
understory plant cover and alters plant community composition in Arct Antarct Alp Res 40:568–75.
boreal forests. Oecologia 168:577–87.
Rogers MC, Sullivan PF, Welker JM (2011) Evidence of nonlinearity
Krishnamurti TN, Kishtawal CM (2000) A pronounced continental- in the response of net ecosystem CO2 exchange to increasing levels
scale diurnal mode of the Asian summer monsoon. Mon Weather of winter snow depth in the high arctic of northwest Greenland.
Rev 128:462–73. Arct Antarct Alp Res 43:95–106.
Legault G, Cusa M (2015) Temperature and delayed snowmelt jointly Saha SK, Rinke A, Dethloff K, et al. (2006) Influence of a complex
affect the vegetative and reproductive phenologies of four sub- land surface scheme on Arctic climate simulations. J Geophys Res
Arctic plants. Polar Biol 38:1701–11. Atmos 111:14.
 592 Journal of Plant Ecology

Scherrer SC, Appenzeller C, Laternser M (2004) Trends in Swiss Wang K, Zhang L, Qiu YB, et al. (2015) Snow effects on alpine
alpine snow days: the role of local- and large-scale climate vari- vegetation in the Qinghai–Tibetan Plateau. Int J Digital Earth
ability. Geophys Res Lett 31:4. 8:56–73.
Shen M (2011) Spring phenology was not consistently related to Westergaard-Nielsen A, Lund M, Pedersen SH, et al. (2017) Transitions
winter warming on the Tibetan Plateau. Proc Natl Acad Sci USA in high-Arctic vegetation growth patterns and ecosystem product-
108:E91–2. ivity tracked with automated cameras from 2000 to 2013. Ambio
Shen M, Piao S, Cong N, et al. (2015) Precipitation impacts on vege- 46:39–52.
tation spring phenology on the Tibetan Plateau. Global Chang Biol Xu J, Grumbine RE, Shrestha A, et al. (2009) The melting Himalayas:
21:3647–56. cascading effects of climate change on water, biodiversity, and live-
Shen MG, Zhang GX, Cong N, et al. (2014) Increasing altitudinal gra- lihoods. Conserv Biol 23:520–30.
dient of spring vegetation phenology during the last decade on the Xu W, Ma L, Ma M, et al. (2017) Spatial–temporal variability of
Qinghai–Tibetan Plateau. Agric For Meteorol 189:71–80. snow cover and depth in the Qinghai–Tibetan Plateau. J Clim

Downloaded from https://academic.oup.com/jpe/article/12/3/583/5224557 by Nanjing University user on 02 April 2021

Tao J, Zhang Y, Zhu J, et al. (2014) Elevation-dependent temperature 30:1521–33.
change in the Qinghai–Xizang Plateau grassland during the past Yang MX, Nelson FE, Shiklomanov NI, et al. (2010) Permafrost deg-
decade. Theor Appl Climatol 117:61–71. radation and its environmental effects on the Tibetan Plateau: a
Wang SJ (2018) Spatiotemporal variability of temperature trends on review of recent research. Earth Sci Rev 103:31–44.
the southeast Tibetan Plateau, China. Int J Climatol 38:1953–63. Yang X, Zhang Y, Zhang W, et al. (2006) Climate change in Mt.
Wang S, Wang X, Chen G, et al. (2017a) Complex responses of spring Qomolangma region since 1971. J Geog Sci 16:326–36.
alpine vegetation phenology to snow cover dynamics over the Yuan WP, Xu WF, Ma MN, et al. (2016) Improved snow cover model
Tibetan Plateau, China. Sci Total Environ 593–594:449–61. in terrestrial ecosystem models over the Qinghai–Tibetan Plateau.
Wang X, Wang T, Guo H, et al. (2018) Disentangling the mechanisms Agric For Meteorol 218:161–70.
behind winter snow impact on vegetation activity in northern eco- Zhang TJ (2005) Influence of the seasonal snow cover on the ground
systems. Global Chang Biol 24:1651–62. thermal regime: an overview. Rev Geophys 43:26.
Wang CH, Yang K, Li YL, et al. (2017b) Impacts of spatiotemporal Zhang G, Zhang Y, Dong J, et al. (2013) Green-up dates in the Tibetan
anomalies of Tibetan Plateau snow cover on summer precipitation Plateau have continuously advanced from 1982 to 2011. Proc Natl
in eastern China. J Clim 30:885–903. Acad Sci USA 110:4309–14.

View publication stats