Diverse Responses of Vegetation Dynamics to Snow Cover Phenology Over the Boreal Region

Article Diverse Responses of Vegetation Dynamics to Snow Cover Phenology over the Boreal Region

Tao Xiong 1 , Hongyan Zhang 1,*, Jianjun Zhao 1 , Zhengxiang Zhang 1 , Xiaoyi Guo 1 , Zhenhua Zhu 1 and Yu Shan 1,2

1 School of Geographical Sciences, Northeast Normal University, Changchun 130024, China; [email protected] (T.X.); [email protected] (J.Z.); [email protected] (Z.Z.); [email protected] (X.G.); [email protected] (Z.Z.); [email protected] (Y.S.) 2 Inner Mongolia Key Laboratory of Remote Sensing and Geographic Information System, Hohhot 010022, China * Correspondence: [email protected]; Tel.: +86-431-850-99550

Received: 9 March 2019; Accepted: 27 April 2019; Published: 30 April 2019

Abstract: Snow cover phenology plays an important role in vegetation dynamics over the boreal region, but the observed evidence of this interaction is limited. A comprehensive understanding of the changes in vegetation dynamics and snow cover phenology as well as the interactions between them is urgently needed. To investigate this, we calculated two indicators, the start of the growing season (SOS) and the annual maximum enhanced vegetation index (EVImax), as proxies of vegetation dynamics using the Moderate Resolution Imaging Spectroradiometer (MODIS) enhanced vegetation index (EVI). Snow cover duration (SCD) and snow cover end date (SCE) were also extracted from MODIS snow cover datasets. Then, we quantified the spatial-temporal changes in vegetation dynamics and snow cover phenology as well as the relationship between them over the boreal region. Our results showed that the EVImax generally demonstrated an increasing trend, but SOS varied in different regions and vegetation types from 2001 to 2014. The earlier onset of SOS was mainly concentrated in the Siberian boreal region. In the Eurasian boreal region, we observed an advance in the SCE and decrease in the SCD, while in the North American boreal region, the spatial distribution of the trends exhibited substantial heterogeneity. Our results also indicated that the snow cover phenology had significant impacts on the SOS and the EVImax, but the effects varied in different regions, vegetation types, and climate gradients. Our findings provide strong evidence of the interaction between vegetation dynamics and snow cover phenology, and snow cover should be considered when analyzing future vegetation dynamics in the boreal region.

Keywords: vegetation dynamics; snow cover phenology; MODIS; boreal region

1. Introduction Both modelling and observational studies suggest that the mid to high latitudes of the Northern Hemisphere have undergone substantial warming in recent decades [1,2]. There is mounting scientific evidence that the raising temperature has a great impact on different aspects of vegetation, such as phenology, biomass, and productivity [3–6]. For changes in vegetation phenology, one of the most noteworthy findings was a wide range of warm-induced advances in the start of the growing season (SOS) that occurred in the Northern Hemisphere in the 1980s and 1990s, such as in North America [7], Eurasia [8], Europe [9], and the Arctic [10], although the magnitude of the trend varied among different research. Interestingly, multiple reports from both in situ and satellite observations indicated that since 2000, the advancing trends of the SOS have slowed down, with even a reverse trend occurring in some regions [3,11,12]. Meanwhile, for changes in vegetation greenness, a slowed and even reverse trend

Forests 2019, 10, 376; doi:10.3390/f10050376 www.mdpi.com/journal/forests Forests 2019, 10, 376 2 of 20 was also reported in different regions over the Northern Hemisphere after 2000 [13,14]. Considering that changes in vegetation may also lead to multiple feedbacks to the climate system [15], it is necessary to conduct further research on vegetation dynamics after 2000 to understand global climate change. Temperature has been widely reported as a key factor that drives the changes in vegetation dynamics at global and regional scales [16–18]. However, for areas with seasonal snow cover, in addition to temperature, snow cover also affects vegetation dynamics, but the knowledge of the response of vegetation dynamics to snow cover phenology is still limited [19–21]. Over the past several decades, remarkable shifts in snow cover phenology have also been reported in many regions [22,23]. In situ observations indicated that between 1980 and 2006, the snow cover end date (SCE) showed a rapid advance in Eurasia at a rate of 6 5.6 days decade 1, and no trend was found in North America [24]. ± − Satellite observations from 1982 to 2013 showed that the changes in the SCE in the Northern Hemisphere, 1 Eurasia, and North America were 6.12, 7.17, and 3.78 days decade− , respectively; the snow − − − 1 cover duration (SCD) maintained a shortening trend of 1.04, 1.71, and 1.33 days decade− in the Northern Hemisphere, Eurasia, and North America, respectively [22]. The changes in snow cover phenology may have significant impacts on vegetation dynamics for several reasons. First, although there is no direct relationship between snow and vegetation growth [21], changes in snow cover can have cascading effects on the known drivers of vegetation growth, such as the chilling requirement, forcing temperatures, and soil moisture status [13,15,16]. Second, changes in snow cover play a very important role in energy feedback through albedo, surface heating, and water budgeting [23,25]. In seasonal snow-covered regions, snow–vegetation feedback is considered as a key mechanism behind snow-driven growth changes [26]. Therefore, an understanding of the relationship between vegetation dynamics and snow cover phenology is vital for the projection of ecosystem dynamics in the future. The sharp contrast in albedo between snow cover and vegetation, the potential release of large carbon stocks, and strong feedback on ocean-atmosphere circulation make the boreal region play a very important role in terrestrial ecosystems and regulation of the global climate system [27,28]. Therefore, studying the changes in vegetation dynamics in the boreal region and its relationship with snow phenology is very important to understanding the interaction between snow and vegetation under climate change. To understand the influence of snow in the boreal region on the spring phenology of vegetation, observations from passive microwave data and in situ flux data indicated that early snow melting led to the early arrival of spring recovery and enhanced boreal springtime carbon uptake [29]. Evidence from the community land model, which examines the biogeochemical and biogeophysical processes in terrestrial ecosystems, also showed that an increase in winter precipitation could alleviate the strong advancing trend of spring vegetation growth [27]. These conclusions are important, but additional driving mechanisms of different responses between snow cover phenology and vegetation dynamics need to be further explored. In this study, based on the Moderate Resolution Imaging Spectroradiometer (MODIS) 0.05◦ data for the period 2001–2014, we first calculated the SOS and annual maximum enhanced vegetation index (EVImax), which indicates the maximum level of seasonal photosynthetic capacity [30], as the main indicators of vegetation dynamics over the boreal region. SCD and SCE were also derived to represent the snow cover phenology variability. We then assessed the spatial-temporal trends of vegetation dynamics and snow cover phenology over the boreal region and identified the relationships between them. The objectives of this study are therefore to: (1) Quantify the spatial patterns of vegetation dynamics and snow cover phenology across the boreal region; (2) evaluate the temporal trends of vegetation dynamics and snow cover phenology over the whole boreal region and in different land cover types; (3) analyze the relationships between vegetation dynamics and snow cover phenology over the whole boreal region and in different land cover types; (4) explore the climate controls on the response of vegetation dynamics to snow cover phenology. Forests 2019, 10, 376 3 of 20

2. Materials and Methods Forests 2019, 10, x FOR PEER REVIEW 3 of 20 2.1. Study Area The boreal region with an area of approximately 15.09 × 108 ha covers one of the largest forest The boreal region with an area of approximately 15.09 108 ha covers one of the largest forest biomes on Earth and accounts for approximately 14.5% of the× total land area (excluding glaciers) and biomes on Earth and accounts for approximately 14.5% of the total land area (excluding glaciers) and 30% of the forest area [28,31]. This region forms a broad east-west belt stretching over northern 30% of the forest area [28,31]. This region forms a broad east-west belt stretching over northern regions regions of the Eurasian and North American continents, and 60% of this region occurs in Russia, 28% of the Eurasian and North American continents, and 60% of this region occurs in Russia, 28% in Canada, in Canada, and the remaining 12% is distributed in more than 10 countries, such as China, Finland, and the remaining 12% is distributed in more than 10 countries, such as China, Finland, and Norway. and Norway. In terms of the latitudinal division, the boreal region is located at latitudes of 45~75° In terms of the latitudinal division, the boreal region is located at latitudes of 45~75◦ (Figure A1). The (Figure A1). The mean annual temperature ranges from approximately −5 to +5 °C with annual mean annual temperature ranges from approximately 5 to +5 ◦C with annual precipitation ranging precipitation ranging from 20 to almost 200 cm [31]. − from 20 to almost 200 cm [31]. Most of the boreal region is occupied by needleleaf forest, which consists of evergreen needleleaf Most of the boreal region is occupied by needleleaf forest, which consists of evergreen needleleaf forests broadly distributed throughout the study area and deciduous needleleaf forests distributed forests broadly distributed throughout the study area and deciduous needleleaf forests distributed in the East Siberian Taiga. The border to the north is in the transition zone between the boreal forest in the East Siberian Taiga. The border to the north is in the transition zone between the boreal forest biome and the subarctic biome. To the south, the needleleaf forests are gradually replaced with biome and the subarctic biome. To the south, the needleleaf forests are gradually replaced with various various types of vegetation, such as deciduous forest, montane coniferous forest, and grassland, types of vegetation, such as deciduous forest, montane coniferous forest, and grassland, depending on depending on the geographical location (Figure 1). the geographical location (Figure1).

Figure 1.1. Land cover types for the boreal regionregion basedbased onon thethe InternationalInternational Geosphere-BiosphereGeosphere-Biosphere Programme (IGBP) classiﬁcation,classification, derivedderived fromfrom MCD12C1MCD12C1 landland covercover typetype product.product.

2.2. Data Source and Data Processing

2.2.1. MODIS EVI Product

To cover such a large area (approximately 15.09 × 108 ha), we chose the MODIS enhanced vegetation index (EVI, MOD13C1) product (Collection 6) from 2001 to 2014, which has a spatial resolution of 0.05° (approximately 5 km) and a temporal resolution of 16 days (Table A1). This product can be acquired from NASA’s Earth Observing System data gateway (https://search.earthdata.nasa.gov/s-earch). The EVI was developed to be more sensitive in high

Forests 2019, 10, 376 4 of 20

2.2. Data Source and Data Processing

2.2.1. MODIS EVI Product To cover such a large area (approximately 15.09 108 ha), we chose the MODIS enhanced vegetation × index (EVI, MOD13C1) product (Collection 6) from 2001 to 2014, which has a spatial resolution of 0.05◦ (approximately 5 km) and a temporal resolution of 16 days (Table A1). This product can be acquired from NASA’s Earth Observing System data gateway (https://search.earthdata.nasa.gov/s-earch). The EVI was developed to be more sensitive in high biomass regions and minimize the disturbance of the soil background [32,33]. Furthermore, by utilizing a constrained-view angle-maximum value composite (CV-MVC) method, the EVI dataset was processed to reduce the disturbances from clouds, atmosphere, sensors, and surface bidirectional reﬂectance [32,34].

2.2.2. MODIS Snow Cover Product The MODIS snow cover product used in this study was acquired from the National Snow and Ice Data Center (https://nsidc.org). Considering the persistent cloudiness, particularly at high latitudes in the Northern Hemisphere, we selected the 8-day composite data rather than the daily resolution data, which can eliminate cloud obscuration and provide more consistent and cloud-free data. Moreover, to remain consistent with the spatial resolution of the EVI data, we selected MOD10C2 snow cover fraction products to derive the snow cover phenology parameters across the boreal region (Table A1). The MOD10C2 snow cover fraction products are directly created from MOD10A2. In MOD10A2, each 500 m pixel is deﬁned as snow, no snow, cloud, etc. When the data are transformed to 0.05◦ spatial resolution, the original MOD10A2 pixels and the total numbers of the corresponding pixels are considered to derive the fraction or percentage of snow cover [35]. We selected the MOD10C2 fractional snow cover product from January 2000 to December 2014 to derive the snow cover phenology of the boreal region.

2.2.3. Land Cover and Climate Data To explore the relationship between vegetation dynamics and snow cover phenology in different vegetation types, we classified the land surface using MODIS land cover products. To remain consistent with the spatial resolution of the snow cover and vegetation data, we selected MCD12C1 products with the same resolution of 0.05◦ (Table A1). These products were derived using the same algorithm used to produce the Global 500 m Land Cover Type Product (MCD12Q1) and include three classification schemes that describe the land cover characteristics derived from the annual Terra and Aqua MODIS data observations. The product contains the following three classification schemes, which are derived from the supervised decision tree classification: 1. The International Geosphere-Biosphere Programme (IGBP) global vegetation classification scheme, 2. University of Maryland (UMD) Program, and 3. the LAI/FPAR scheme based on MODIS. As the IGBP classification scheme more prominently reflects the importance of natural vegetation in various land cover types [36], we chose the IGBP classification scheme. There are 16 land cover types in the boreal region according to the IGBP classification, excluding evergreen broadleaf forest. After eliminating the nonvegetated area and croplands, we reclassified the land cover types into 7 major land cover types (Table A2) in this study: ENF (evergreen needleleaf forest), DNF (deciduous needleleaf forest), MF (mixed forest), OSH (open shrublands), SA (savannas), GL (grasslands), and WET (permanent wetlands). The distribution of the major vegetation types in the boreal region is depicted in Figure1. The monthly mean temperature and precipitation data with a spatial resolution of 0.5◦ were extracted from the CRU-TS 4.02 climate dataset [37], covering the period from 2001 to 2014 (Table A1). CRU-TS databases have been widely used in recent climate change and vegetation growth dynamic studies [21,38]. In order to match the MODIS products, we used the nearest-neighbor interpolation method to interpolate climate data into the resolution of 0.05◦. Forests 2019, 10, 376 5 of 20

2.3. Vegetation Dynamics Metrics Considering the influence of noise in the original synthesized EVI time series data, many scholars have proposed several effective means to gap-fill and smooth the raw datasets, such as the Savitzky-Golay filtering method [39], the asymmetric Gaussian fitting method [40], the double logistic function [41], and the harmonic analysis of time series method (Hants) [42]. The double logistic method has been proven to be more suitable than other algorithms at high latitudes, such as Savitzky-Golay and asymmetric Gaussian [16,43]. Here, we used the double logistic method to smooth and reconstruct the EVI time series dataset over the boreal region. The basic function of the double logistic algorithm is expressed as: 1 1 g(t; x1, ... , x4) = x t + x t , (1) ( 1− ) ( 3− ) 1 + e x2 1 + e x4 where t is the Julian day, x1 and x3 determine the left and right inflection points, respectively, while x2 and x4 indicate the corresponding rates of change. Declining snow cover may be responsible for the change of EVI associated with the greening trend. To avoid the effects of snow melt, we assigned the negative EVI value to 0 [43,44]. Then, we employed TIMESAT software, which includes a double logistic algorithm and has been widely used to investigate vegetation dynamics [17,44,45]. It allows various vegetation parameters to be derived, including the SOS and EVImax (the maximum value of the annual EVI time series fitted by the double logistic curve). We used the following parameters: Adaptation strength of 2, number of envelope iterations of 2, seasonal parameter of 1, and amplitude season start and end of 20%, following Karkauskaite et al. [46].

2.4. Snow Cover Phenology Metrics Based on the seasonal cycle of snow cover over the boreal region, we used the hydrological year instead of the calendar year to derive the snow cover phenology parameters [22,47]. Here, we defined the hydrological year of the boreal region as the period from September 1st of a given year (t 1) to − August 31st of the next year (t). Then, we defined the onset season of snow cover from September 1st of a given year (t 1) to the last day in February of the following year (t), while the melting − season spanned from March to August of the same year (t). Referring to the definition of snow cover parameters in Chen et al. [22], during the onset season, we searched for the interval (i to i + 7) when the snow cover fraction was equal to 0 and subsequently superior to 0. Then, we defined the date (i + 3.5) as the snow cover onset date (SCO). Likewise, during the melting season, we searched for the interval (i to i + 7) when the snow cover fraction was above 0 and subsequently below 0. Then, we defined the date (i + 3.5) as the SCE. The SCD was defined as the interval between SCO and SCE. This method is one of the easiest ways to calculate snow cover phenology, and can avoid the impact of ephemeral snow [22,24,48].

2.5. Mann–Kendall Trend Test The nonparametric Mann-Kendall (MK) trend test [49,50] was used to analyze the annual trends in both vegetation dynamics and snow cover phenology. Compared with parametric tests, the nonparametric MK trend test is applicable to non-normally distributed data and is not impacted by a few abnormal values, and this method has been widely applied for trend detection [34,51,52]. The Theil-Sen median slope estimator was used to analyze the slope of the annual variations in vegetation dynamics and snow cover phenology. The equation is: ! xi xj β = median − , i > j, (2) i j − where β > 0 indicates an increasing trend in SCD and EVImax and an advancing trend in SOS and SCE, and β < 0 indicates a reverse trend. Forests 2019, 10, 376 6 of 20

2.6. Partial Correlation Analysis The relationship between multivariate related variables is complex, and there are often different degrees of simple correlation between any two variables [53]. Partial correlation analysis is a statistical method that studies the correlation between two related variables by fixing other related variables, which can eliminate the influence of other variables [38]. This method has been successfully applied to previous studies involving climate change and vegetation phenology [24,38,53]. The partial correlation coefficient was calculated as follows:

rxy ryzrxz r = − xy.z q , (3) 2 2 1 ryz (1 rxz ) − − where rxy.z is the partial correlation coefficient between variable x and y after fixing variable z, and rxy, ryz, and rxz are the correlation coefficients between two variables. In addition, a t-test was performed to reflect the significance level (p-values) of the relationship.

3. Results

3.1. Spatial Patterns of Vegetation dynamics and Snow Cover Phenology To analyze the spatial distribution of the vegetation dynamics and snow cover phenology in the boreal region, we calculated the average values of the SOS, EVImax, SCD, and SCE from 2001 to 2014 (Figure2). The average SOS ranged from approximately 100 to 160 days (Figure2a), accounting for 94.6% of the area of the boreal region. The SOS is generally delayed with increasing latitude. In addition, due to the inﬂuence of the altitude, climate, and vegetation types, the distribution of the SOS at the same latitude has great spatial variability. Furthermore, the average spring phenology based on vegetation types indicates that the evergreen needle forest has the earliest SOS, while the grasslands and open shrublands have the latest SOS. Compared with the SOS, the EVImax showed an almost opposite trend in spatial distribution. The average EVImax mainly ranged from 0.1 to 0.7 (Figure2b), accounting for 98.6% of the area of the boreal region. The maximum mean EVImax occurred in the mixed forest and deciduous needle forest, and the minimum mean EVImax occurred in grasslands (Table A3). The average spatial patterns of annual SCD mainly ranged from 130 to 250 days (Figure2c), accounting for 97.4% of the area of the boreal region. Similar to SOS, SCD increased with latitude, but the trend of the change in SCD with latitude was more uniform. The shortest SCD occurred in mixed forests and evergreen needle forests; the longest SCD mainly occurred in open shrublands. Compared with the SCD, the SCE presented nearly the same trend in spatial distribution, except for some diﬀerences in North America. The average SCE mainly ranged from 100 to 160 days (Figure2d), accounting for 94.4% of the boreal region. The latest SCE was mainly distributed in open shrublands; the earliest SCE was mainly distributed in mixed forests and evergreen needle forests (Table A3). Forests 2019, 10, 376 7 of 20 Forests 2019, 10, x FOR PEER REVIEW 7 of 20

Figure 2. AveragesAverages of the start of of the the growing growing season season (SOS, (SOS, aa),), annual annual maximum maximum enhanced enhanced vegetation vegetation b c d index (EVImax, (EVImax, b),), snow snow cover cover duration duration (SCD, (SCD, c), and), and snow snow cover cover end date end (SCE, date (SCE,d) across) acrossthe boreal the boreal region. region. 3.2. Temporal Trends in Vegetation Dynamics and Snow Cover Phenology 3.2. Temporal Trends in Vegetation Dynamics and Snow Cover Phenology Figure3 shows the spatial patterns of temporal trends in the SOS, EVI max, SCD, and SCE across the borealFigure region. 3 shows Significant the spatial trends patterns were of determinedtemporal trends by the in Mann-Kendallthe SOS, EVImax test., SCD, For and the SCE vegetation across the boreal region. Significant trends were determined by the Mann-Kendall test. For the vegetation dynamics, the temporal trends of the SOS and EVImax showed very different spatial distribution dynamics,patterns. Inthe the temporal North American trends of andthe EuropeanSOS and EVI borealmax showed regions, very the SOS different in most spatial parts distribution exhibited a patterns.delayed trend,In the while North most American of Siberia and showed European an advancingboreal regions, trend, the and SOS the in significant most parts changes exhibited were a delayed trend, while most of Siberia showed an advancing trend, and the significant changes were concentrated in specific areas. However, for the EVImax, most of the boreal region showed a significant concentrated in specific areas. However, for the EVImax, most of the boreal region showed a significant increasing trend. Similarly, the SOS trend varied in different vegetation types, but the EVImax showed increasingan increasing trend. trend Similarly, in every the vegetation SOS trend type. varied The in SOSdifferent showed vegetation a nonsignificant types, but advancing the EVImax trendshowed in andeciduous increasing needle trend forests in every and vegetation open shrublands type. Th ande aSOS nonsignificant showed a nonsignificant delayed trend advancing in evergreen trend needle in deciduous needle forests and open shrublands and a nonsignificant delayed trend in evergreen forests and grasslands. However, the EVImax showed a significant increasing trend in most vegetation needletypes except forests deciduous and grasslands. needle forestHowever, (Figure the4 a,b).EVImax showed a significant increasing trend in most vegetation types except deciduous needle forest (Figure 4a, Figure 4b). Unlike the different spatial patterns observed in SOS and EVImax, the trends of SCD and SCE sharedUnlike almost the thedifferent same spatial pattern,patterns butobserved the extents in SOS of and the EVI trendsmax, the varied. trends The of SCDSCD and SCE shared almost the same spatial pattern, but the extents of the trends varied. The SCD and SCE demonstrated clear spatial variability throughout the boreal region. A visible increasing SCD and demonstrated clear spatial variability throughout the boreal region. A visible increasing SCD and advancing SCE was found in most of the Eurasia boreal region except for some regions near the Sea of advancing SCE was found in most of the Eurasia boreal region except for some regions near the Sea of Okhotsk and some regions in the Scandinavian and Russian taiga. However, the SCD and SCE trends were non-uniform in the North American boreal region. A visible increasing SCD and

Forests 2019, 10, 376 8 of 20

Okhotsk and some regions in the Scandinavian and Russian taiga. However, the SCD and SCE trends wereForests non-uniform2019, 10, x FOR PEER in the REVIEW North American boreal region. A visible increasing SCD and advancing8 of SCE 20 was found in Northern Cordillera forests and mid-continental Canadian forests, however, a visible decreasingadvancing SCDSCE andwas delayedfound in SCE Northern was found Cordillera in most offorests the northeast and mid-continental and northwest Canadian North American forests, borealhowever, region. a visible Interestingly, decreasing for diSCDfferent and vegetation delayed types,SCE was the temporalfound in trendsmost inof SCDthe northeast and SCE wereand notnorthwest as consistent North asAmerican the spatial boreal patterns. region. The Interestingl SCD showedy, for different a nonsignificant vegetation decreasing types, the trend temporal in all vegetationtrends in types,SCD and however, SCE awere nonsignificant not as consistent advancing as SCE the trend spatial was foundpatterns. in evergreen The SCD needle showed forests a andnonsignificant grasslands decreasing (Figure4c,d). trend in all vegetation types, however, a nonsignificant advancing SCE trend was found in evergreen needle forests and grasslands (Figure 4c, Figure 4d).

Figure 3. Spatial distributions of the temporal trends in (a) the start of the growing season (SOS), (b) Figure 3. Spatial distributions of the temporal trends in (a) the start of the growing season (SOS), (b) annual maximum enhanced vegetation index (EVImax), (c) snow cover duration (SCD), and (d) snow annual maximum enhanced vegetation index (EVImax), (c) snow cover duration (SCD), and (d) snow cover end date (SCE) over the boreal region. The dotted regions indicate that the trends were signiﬁcant cover end date (SCE) over the boreal region. The dotted regions indicate that the trends were at the 95% level. significant at the 95% level.

Forests 2019, 10, 376 9 of 20 Forests 2019, 10, x FOR PEER REVIEW 9 of 20

Figure 4. Temporal trends in (a) the start of the growing season (SOS), (b) annual maximum enhanced Figure 4. Temporal trends in (a) the start of the growing season (SOS), (b) annual maximum enhanced vegetation index (EVImax), (c) snow cover duration (SCD), and (d) snow cover end date (SCE) in vegetation index (EVImax), (c) snow cover duration (SCD), and (d) snow cover end date (SCE) in different vegetation types over the boreal region. DNF represents deciduous needleleaf forest; ENF different vegetation types over the boreal region. DNF represents deciduous needleleaf forest; ENF represents evergreen needleleaf forest; GL represents grasslands; MF represents mixed forest; OSH represents evergreen needleleaf forest; GL represents grasslands; MF represents mixed forest; OSH represents open shrublands; SA represents savannas; WET represents permanent wetlands. represents open shrublands; SA represents savannas; WET represents permanent wetlands. 3.3. Response of Vegetation dynamics to Snow Cover Phenology 3.3. Response of Vegetation dynamics to Snow Cover Phenology Figure5 shows the spatial distribution of the partial correlation coe fficients between vegetation dynamicsFigure and 5 shows snow the cover spatial phenology. distribution Interestingly, of the partial the partial correlation correlations coefficients between between SOS vegetation and snow coverdynamics phenology and snow showed cover twophenology. very di ffInterestinglerent patterns,y, the but partial the EVIcorrelationsmax and between snow cover SOS phenology and snow exhibitedcover phenology almost the showed opposite two spatial very distributionsdifferent patterns, (Table A4but). SOSthe EVI andmax SCD and were snow positively cover phenology correlated inexhibited 56.32% almost of the borealthe opposite region, spatial but more distributi pronouncedons (Table in Eurasia A4). SOS (61.32%) and thanSCD inwere NA positively (45.26%). Additionally,correlated in the56.32% most of significant the boreal positive region, correlationbut more pronounced was found in in the Eurasia northeast (61.32%) and thethan northern in NA region(45.26%). of EastAdditionally, Siberia, while the significantmost significant negative positive correlations correlation were scatteredwas found throughout in the northeast North America and the (Figurenorthern5a). region The positive of East correlationSiberia, while between significan SOS andt negative SCE accounted correlations for 62.88%were scattered of the boreal throughout region. However,North America unlike (Figure the spatial 5a). The distribution positive correlati of the correlationon between between SOS and SOS SCE and accounted SCD, the for correlation 62.88% of betweenthe boreal SOS region. and However, SCE is higher unlike in the North spatial America distribution (68.41%) of the than correlation in Eurasia between (60.34%). SOS Significant and SCD, positivethe correlation correlations between were SOS mainly and SCE concentrated is higher in theNorth northern America part (68.41%) of east andthan westin Eurasia Siberia (60.34%). and the northwestSignificant partpositive of North correlations America were and regionsmainly nearconcen Jamestrated Bay. in However, the northern negative part correlationsof east and werewest rare,Siberia mainly and inthe northeastern northwest part Siberia of (FigureNorth 5Americab). For all and vegetation regions types,near James the partial Bay. correlationHowever, betweennegative SOScorrelations and snow were cover rare, phenology mainly in showed northeastern no significant Siberia (Figure partial 5b). correlation For all vegetation (Figure6a,b). types, the partial correlation between SOS and snow cover phenology showed no significant partial correlation (Figure 6a, Figure 6b).

Forests 2019, 10, 376 10 of 20

Forests 2019, 10, x FOR PEER REVIEW 10 of 20

The EVImax was negatively correlated with SCD throughout the boreal region (52.95%), EurasianThe EVI borealmax was region negatively (50.72%), correlated and North with American SCD througho borealut region the boreal (57.89%) region (Figure (52.95%),5c). However, Eurasian theboreal relationship region (50.72%), between and the EVINorthmax Americanand SCE was boreal almost region the opposite,(57.89%) with(Figure a positive5c). However, correlation the demonstratedrelationship between throughout the theEVImax boreal and region SCE was (57.71%), almost Eurasian the opposite, boreal regionwith a (56.06%),positive andcorrelation North Americandemonstrated boreal throughout region (61.34%) the boreal (Figure region5d). The (57.71%), spatial distribution Eurasian boreal of the significantregion (56.06%), partial correlationand North betweenAmerican the boreal EVImax regionand snow (61.34%) cover (Figure phenology 5d). was The similar spatial and distribution was discretely of distributedthe significant throughout partial max thecorrelation boreal region. between Similarly, the EVI the and partial snow correlation cover phenology between was the similar EVImax andand was snow discretely cover distributed phenology ofthroughout different vegetationthe boreal typesregion. also Similarly, showed the the partial opposite correlation pattern (Figure between6c,d). the InEVI addition,max and snow among cover the partialphenology correlations of different between vegetation the EVI typesmax alsoand SCDshowed of dithefferent opposite vegetation pattern types, (Figure only 6c, mixedFigure forests6d). In max showedaddition, significant among the negative partial correlations,correlations andbetween among the the EVI partial and correlations SCD of different between vegetation the EVImax types,and SCE,only grasslandsmixed forests and showed mixed forests significant showed negative significant correlations, positive correlations. and among the partial correlations between the EVImax and SCE, grasslands and mixed forests showed significant positive correlations.

Figure 5. Spatial distribution of the partial correlation coeﬃcient between (a) the start of growing Figure 5. Spatial distribution of the partial correlation coefficient between (a) the start of growing season (SOS) and snow cover duration (SCD); (b) start of growing season (SOS) and snow cover end season (SOS) and snow cover duration (SCD); (b) start of growing season (SOS) and snow cover end date (SCE); (c) annual maximum enhanced vegetation index (EVImax) and snow cover duration (SCD); date (SCE); (c) annual maximum enhanced vegetation index (EVImax) and snow cover duration (SCD); (d) annual maximum enhanced vegetation index (EVImax) and snow cover end date (SCE). The dotted (d) annual maximum enhanced vegetation index (EVImax) and snow cover end date (SCE). The dotted regions indicate correlations that were signiﬁcant at the 95% level. regions indicate correlations that were significant at the 95% level.

ForestsForests 20192019,, 1010,, 376x FOR PEER REVIEW 1111 of 2020

Figure 6. Partial correlation coefficient between (a) the start of growing season (SOS) and snow cover Figure 6. Partial correlation coefficient between (a) the start of growing season (SOS) and snow cover duration (SCD); (b) start of growing season (SOS) and snow cover end date (SCE); (c) annual maximum duration (SCD); (b) start of growing season (SOS) and snow cover end date (SCE); (c) annual enhanced vegetation index (EVImax) and snow cover duration (SCD); (d) annual maximum enhanced maximum enhanced vegetation index (EVImax) and snow cover duration (SCD); (d) annual maximum vegetation index (EVImax) and snow cover end date (SCE) in different vegetation types. DNF represents enhanced vegetation index (EVImax) and snow cover end date (SCE) in different vegetation types. DNF deciduous needleleaf forest; ENF represents evergreen needleleaf forest; GL represents grasslands; MF represents deciduous needleleaf forest; ENF represents evergreen needleleaf forest; GL represents represents mixed forest; OSH represents open shrublands; SA represents savannas; WET represents grasslands; MF represents mixed forest; OSH represents open shrublands; SA represents savannas; permanent wetlands. WET represents permanent wetlands. 3.4. Climate Controls on the Response of Vegetation Dynamics to Snow Cover Phenology 3.4. Climate Controls on the Response of Vegetation Dynamics to Snow Cover Phenology To study the control of the climate on the response of vegetation dynamics to snow cover phenology, we firstTo usedstudy the the annual control temperature of the climate and precipitationon the response data of from vegetation 2000 to 2014dynamics to obtain to snow the annual cover averagephenology, temperature we first used and the annual annual cumulative temperature precipitation and precipitation of these 14data years from (Figure 2000 7to). 2014 As shown to obtain in Figurethe annual7, the average temperature temperature gradient and in theannual northern cumulative region precipitation clearly varied of with these latitude, 14 years and (Figure the coldest 7). As regionsshown in were Figure located 7, the in temperature eastern and northeasterngradient in the Siberia. northern The region trend in clearly the precipitation varied with gradient latitude, with and latitudethe coldest was regions not obvious. were located Eurasia in showed eastern high and precipitation northeastern in Siberia. the west The and trend low in precipitation the precipitation in the east,gradient while with North latitude America was showed not obvious. high precipitation Eurasia showed in the high east andprecipitation low in the in west. the west and low precipitation in the east, while North America showed high precipitation in the east and low in the west.

Forests 2019, 10, 376 12 of 20 Forests 2019, 10, x FOR PEER REVIEW 12 of 20

Figure 7. 7. DescriptionDescription of ofthe the (a) (annuala) annual mean mean temperature, temperature, (b) annual (b) annual total precipitation total precipitation in the boreal in the region.borealregion. Using the obtained temperature and precipitation gradients, we calculated the partial correlation Using the obtained temperature and precipitation gradients, we calculated the partial coefficients between the vegetation dynamics and snow cover phenology under each hydrothermal correlation coefficients between the vegetation dynamics and snow cover phenology under each gradient (Figure8). As shown in Figure8, under di fferent hydrothermal gradients, the effects of snow hydrothermal gradient (Figure 8). As shown in Figure 8, under different hydrothermal gradients, the cover phenology on vegetation dynamics showed clear aggregation and heterogeneity. SOS and SCD effects of snow cover phenology on vegetation dynamics showed clear aggregation and mainly showed positive partial correlations in different precipitation gradients below 10 C. Above heterogeneity. SOS and SCD mainly showed positive partial correlations in different− precipitation◦ 10 C, with the increase in the precipitation gradient, the partial correlation between SOS and SCD gradients− ◦ below −10 °C. Above −10 °C, with the increase in the precipitation gradient, the partial appeared to change from negative to positive and then negative. The significant partial correlations correlation between SOS and SCD appeared to change from negative to positive and then negative. were mainly concentrated in areas with temperatures below 10 C and where the temperature ranged The significant partial correlations were mainly concentrated− in◦ areas with temperatures below −10 from 5 to 0 C and the precipitation was approximately 1000 mm (Figure8a). However, SOS and °C and− where◦ the temperature ranged from −5 to 0 °C and the precipitation was approximately 1000 SCE showed positive correlations in most areas; the two mainly showed significant positive partial mm (Figure 8a). However, SOS and SCE showed positive correlations in most areas; the two mainly correlations below 0 C, negative partial correlations above 0 C and below 1000 mm precipitation, showed significant positive◦ partial correlations below 0 °C, negative◦ partial correlations above 0 °C but strong heterogeneity above 1000 mm precipitation (Figure8b). Interestingly, the partial correlation and below 1000 mm precipitation, but strong heterogeneity above 1000 mm precipitation (Figure 8b). between the EVImax and snow cover phenology also showed opposite distribution patterns in different Interestingly, the partial correlation between the EVImax and snow cover phenology also showed hydrothermal gradients. Below 10 ◦C, the EVImax and SCD mainly showed positive partial correlations opposite distribution patterns in− different hydrothermal gradients. Below −10 °C, the EVImax and SCD and significant positive correlations were mainly concentrated in the 10 C region. The EVImax mainly showed positive partial correlations and significant positive− correlations◦ were mainly and SCE mainly showed nonsignificant negative partial correlations below 10 ◦C. Above 10 ◦C, concentrated in the −10 °C region. The EVImax and SCE mainly showed nonsignificant− negative− partial the EVImax and SCD mainly exhibited negative partial correlations, while the EVImax and SCE mainly correlations below −10 °C. Above −10 °C, the EVImax and SCD mainly exhibited negative partial exhibited positive partial correlations, and significant partial correlations were mainly concentrated in correlations, while the EVImax and SCE mainly exhibited positive partial correlations, and significant the region of 5 to 0 C. partial correlations− were◦ mainly concentrated in the region of −5 to 0 °C.

Forests 2019, 10, 376 13 of 20 Forests 2019, 10, x FOR PEER REVIEW 13 of 20

Figure 8. The variation in the partial correlation coefficient between snow cover phenology and spring Figure 8. The variation in the partial correlation coefficient between snow cover phenology and spring phenology and annual maximum enhanced vegetation index along the temperature and precipitation phenology and annual maximum enhanced vegetation index along the temperature and precipitation gradient in the boreal region. (a): Partial correlation coefficient between start of growing season (SOS) gradient in the boreal region. (a): Partial correlation coefficient between start of growing season (SOS) and snow cover duration (SCD); (b): start of growing season (SOS) and snow cover end date (SCE); (c): and snow cover duration (SCD); (b): start of growing season (SOS) and snow cover end date (SCE); annual maximum enhanced vegetation index (EVImax) and snow cover duration (SCD); (d): annual (c): annual maximum enhanced vegetation index (EVImax) and snow cover duration (SCD); (d): annual maximum enhanced vegetation index (EVImax) and snow cover end date (SCE). The circles indicate the maximum enhanced vegetation index (EVImax) and snow cover end date (SCE). The circles indicate pixels with a significant correlation (p < 0.05). the pixels with a significant correlation (p < 0.05). 4. Discussion 4. Discussion 4.1. Temporal Trends in Vegetation Dynamics and Snow Cover Phenology 4.1. TemporalFor the temporal Trends in trend Vegetation in the SOS,Dynamics the major and Snow advancing Cover SOSPhenology trends occurred in the Eurasian boreal region,For while the temporal a reversed trend delayed in the SOS SOS, trend the occurred major ad invancing the Scandinavian SOS trends and occurred Russian in taiga the and Eurasian some partsboreal of region, North while America. a reversed These resultsdelayed were SOS consistent trend occurred with thein the conclusions Scandinavian of Barichivich and Russian [54 ]taiga and Wangand some [55]. parts Moreover, of North SOS America. also showed These di ffresultserent trendswere consistent in different with vegetation the conclusions types. The of causeBarichivich of the two[54] diandfferent Wang SOS [55]. trends Moreover, may bethe SOS combined also showed effects diff of aerent warmer trends November in different and Decembervegetation (late types. winter) The andcause colder of the January two different and March SOS (early trends spring), may which be the delayed combined the springeffects phenology of a warmer of earlier-unfolding November and plantsDecember and accelerated(late winter) the and spring colder phenology January of and later-unfolding March (early plants, spring), respectively which delayed [11,56]. the However, spring therephenology were significant of earlier-unfolding increases in plants the EVI andmax acceleratedin different the regions spring as phenology well as in diofff later-unfoldingerent vegetation plants, types, whichrespectively indicated [11,56]. that However, the increasing there temperature were significant enhanced increases the growth in the ofEVI vegetationmax in different in summer regions over as thewell boreal as in region.different For vegetation snow cover types, phenology, which indi mostcated advancing that the SCE increasing trends were temperature distributed enhanced in Eurasia, the withgrowth a few of distributedvegetation in thesummer high latitudesover the ofboreal boreal region. NA (greater For snow than cover 60◦ N); phenology, simultaneously, most advancing the major delayedSCE trends SCE were trends distributed were distributed in Eurasia, in the with middle a few latitudes distributed of boreal in the NA high (less thanlatitudes 60◦ N). of boreal This result NA is(greater consistent than with 60° N); the simultaneously, conclusion of Chen the major et al. [ 22delayed,35]. The SCE spatial trends distribution were distributed pattern in of the SCD middle was latitudes of boreal NA (less than 60° N). This result is consistent with the conclusion of Chen et al.

Forests 2019, 10, 376 14 of 20 very similar to that of SCE, but the degrees of the changes in SCD and SCE in different vegetation types were very different. This difference was because the change in SCD was more complicated than the change in SCE, while SCD was affected by SCE, the SCO, and the snow depth [24]. Throughout the boreal region, SCD mainly showed a decreasing trend, which was consistent with the previous findings of Choi et al. [57], Chen et al. [22,35], and Peng et al. [24].

4.2. Influence of Snow Cover Phenology on Vegetation Dynamics Previous studies suggested that spring vegetation phenology was principally affected by temperature [4,45,53,58] and many other factors, such as permafrost degradation [59], photoperiod [60], and winter chilling [53,61]. Our results also showed that the spring vegetation phenology was associated with snow cover phenology across the boreal region. This result was supported by Wang et al. [62] and Wang et al. [20], who showed that snow cover phenology had an important impact on SOS for the alpine plants over the Tibetan Plateau. In many parts of the Eurasian boreal region, SCD had a significant positive partial correlation with SOS. These results were consistent with previous conclusions that longer snow cover in winter delayed the arrival of spring phenology [53]. Moreover, in some parts of the North American boreal region, significant negative correlations between SCD and SOS could also be observed, and grasslands and mixed forests also exhibited nonsignificant negative partial correlations among different vegetation types. On the one hand, this result may be due to the strong correlation between SCD and winter chilling and the inconsistent demands of different vegetation types for winter chilling [54]. On the other hand, snow cover is also an effective insulator, which can protect the soil from exposure to harsh climatic conditions, such as low temperature and strong winds [43]. Therefore, the soil temperature in snow-covered areas was usually higher than that in snow-free areas [47], which was conducive to the SOS. Similar to SCD, SCE showed an inconsistent correlation with SOS over the boreal region, but the regions that showed significant positive correlations with SOS were larger than the areas that showed significant correlations between SCD and SOS. Compared to SCD, SCE had a more complex influence mechanism on the spring phenology of vegetation. In addition to affecting winter chilling, the change in SCE had a certain impact on the early spring temperature and soil moisture. On the one hand, the advancing trend of SCE increased the temperature in early spring and provided moisture for the growth of vegetation, which further promoted the arrival of vegetation in spring [21,63]. On the other hand, early snow cover melting may lead to more frost, and vegetation may choose to delay greening to protect against the damage caused by low temperatures [64]. Although snow cover and plant productivity were not directly related, snow cover changes can have a cascade effect on the drivers of known growth season productivity, such as soil moisture conditions [21]. Fortunately, snow manipulation experiments in the boreal region can help us understand how snow cover affects vegetation productivity. On the one hand, the melting time of the snow cover was closely related to the length of the vegetation growth period; on the other hand, the amount of snow was closely related to the soil moisture in the vegetation growing season [65]. Both processes had an important impact on vegetation productivity. Therefore, due to different vegetation types and climatic conditions, the relationship between the EVImax and snow cover phenology showed great heterogeneity.

4.3. Impact of Climate in Vegetation Dynamics Response to Snow Cover Phenology As indicated above, the response of vegetation dynamics to snow cover phenology showed strong heterogeneity in both spatial distribution and different vegetation types. By calculating the partial correlation between vegetation dynamics and snow cover phenology under different climatic gradients, we found that the partial correlation between them showed clear aggregation in different gradients. Interestingly, when the temperature was below 10 C, there were positive partial − ◦ correlations between SCD and SOS and SCD and EVImax, which may be due to the shortening of SCD and the consequent increases in soil freezing and root mortality, resulting in delayed SOS and reduced vegetation growth [43]. When the temperature was greater than 10 C, negative partial correlations − ◦ Forests 2019, 10, 376 15 of 20 began to appear under different precipitation gradients. Because SCD was closely related to winter chilling and less chilling accumulation delayed the SOS and further affected EVImax [66], significant correlations between SCE and SOS as well as between SCE and EVImax were mainly concentrated at 5 C to 0 C, with a significant positive partial correlation. This relationship may be due to the earlier − ◦ ◦ SCE, which caused the spring phenology to advance and led to drier soil moisture, thereby reducing the EVImax [47]. In other climate gradients, both positive and negative correlations were found, which were not significant. This result may be because the lag effects of snow cover phenology on the EVImax in these regions were limited, and the EVImax was more likely to be affected by the climate of the growing season.

5. Conclusions Over the past few decades, the climate has undergone significant changes. Understanding the spatial-temporal changes in vegetation dynamics and snow cover phenology in the boreal region and their links is of great significance for global change research. Using remote sensing data from 2000 to 2014, we found that the annual maximum EVI (EVImax) generally showed an increasing trend, but the start of growing season (SOS) varied in different regions and vegetation types, and the advancing trends were mainly concentrated in the Siberian boreal region. In the Eurasian boreal region, snow cover duration (SCD) mainly showed decreasing trends, snow cover end date (SCE) mainly showed advancing trends, while the spatial distribution of the two trends exhibited great heterogeneity in the North American boreal region. Our results also indicated that the snow cover phenology had a significant impact on the SOS and EVImax, but the effect varied in different regions, vegetation types, and climate gradients. The uneven effect of snow cover phenology on vegetation dynamics emphasized the important role of snow cover in spring and summer vegetation physiology and the challenge of accurately characterizing the growth of vegetation with only temperature and precipitation in most current ecosystem models. Therefore, snow must be regarded as a non-negligible variable. The study of snow cover should not be limited to its feedback on air temperature, and more research should be conducted on other effects on soil temperature in winter and soil moisture in the growing season as well as further impacts on the structure and function of ecosystems. In the northern high latitude, a widespread decline in winter snow accumulation has been observed in recent decades and climate models unanimously project a decrease in the snow-to-rain ratio during the winter season in a warmer future [1,67]. For future research, it is very important that the dynamics of vegetation growth, changes in both snow cover phenology and snow accumulation, and their interactions under different vegetation types, regions, and climatic gradients are evaluated through a variety of models and different kinds of data sources, such as ground data, remote sensing data, and flux towers data.

Author Contributions: All authors contributed significantly to this study. H.Z. designed the study. T.X., X.G., and J.Z. processed the data. T.X., Z.Z. (Zhengxiang Zhang), Z.Z. (Zhenhua Zhu) and Y.S. analyzed the data. T.X. wrote the original manuscript. T.X. and H.Z. are responsible for the revision of the manuscript and polished the language of the paper. Funding: This work was supported by the National Natural Science Foundation of China (Grant numbers 41871330, 41571489, 41771450, 41501449 and 41601438) and the Fundamental Research Funds for the Central Universities (Grant No. 2412019BJ001). Conflicts of Interest: The authors declare no conflict of interest. Forests 2019, 10, 376 16 of 20

Forests 2019, 10, x FOR PEER REVIEW 16 of 20 Appendix A Appendix A

Figure A1. Boreal forest biome based on the Terrestrial Ecoregions of the World dataset. Figure A1. Boreal forest biome based on the Terrestrial Ecoregions of the World dataset. Table A1. Detailed information of the data sources used here. Table A1. Detailed information of the data sources used here. Dataset Resolution Cover Time Download Website Cover MOD13C1Dataset Resolution 0.05◦ 2001–2014 https://Downloadsearch.earthdata.nasa.gov Website /search MOD10C2 0.05◦ 2000–2014Time https://nsidc.org MOD13C1MCD12C1 0.050.05◦ ° 2001–20142001 https://search.earthdata.nasa.gov/searchhttps://search.earthdata.nasa.gov/search CRU-TS 4.02 0.5 2001–2014 https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_4.02/ MOD10C2 0.05◦ ° 2000–2014 https://nsidc.org MCD12C1All MODIS data belong0.05 to°the Level 3 product,2001 which is expressed https://search.earthdata.nasa.gov/search in a uniform time-space grid and usually has certain integrity and consistency. CRU-TS 4.02 0.5° 2001–2014 https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_4.02/ All MODIS data belong to the Level 3 product, which is expressed in a uniform time-space grid and usually has certain integrity and consistency.

Table A2. Harmonized legend used and correspondence with the original product classes.

Classes Value Number of pixels Deciduous needleleaf forest 3 25,171 Evergreen needleleaf forest 1 37,649 Grasslands 10 9538 Mixed forest 5 63,022 Open shrublands 6, 7 117,121 Savannas 8, 9 80,000

Forests 2019, 10, 376 17 of 20

Table A2. Harmonized legend used and correspondence with the original product classes.

Classes Value Number of Pixels Deciduous needleleaf forest 3 25,171 Evergreen needleleaf forest 1 37,649 Grasslands 10 9538 Mixed forest 5 63,022 Open shrublands 6, 7 117,121 Savannas 8, 9 80,000 Permanent wetlands 11 8475 The non-vegetated parts and crops were masked out.

Table A3. Average and standard deviation (in brackets) of the start of the growing season (SOS, in Julian days), annual maximum EVI (EVImax), snow cover duration (SCD, days), and snow cover end date (SCE, in Julian days) for diﬀerent land cover types.

Land Cover Types SOS EVImax SCD SCE Deciduous needleleaf forest 130 (8) 0.43 (0.05) 193 (13) 134 (7) Evergreen needleleaf forest 124 (19) 0.35 (0.08) 174 (19) 126 (14) Grasslands 139 (21) 0.33 (0.14) 190 (37) 134 (28) Mixed forest 132 (16) 0.48 (0.08) 165 (21) 120 (14) Open shrublands 145 (12) 0.36 (0.08) 213 (22) 144 (14) Savannas 136 (15) 0.38 (0.08) 198 (23) 138 (17) Permanent wetlands 127 (13) 0.34 (0.10) 190 (18) 134 (14)

Table A4. Percentages of the partial correlation coeﬃcient between the start of growing season and snow cover duration (SCD_SOS), start of growing season and snow cover end date (SCE_SOS), annual maximum EVI and snow cover duration (SCD_EVImax), and annual maximum EVI and snow cover end date (SCE_EVImax) in the entire boreal region (All), North American boreal region (NA), and Eurasian boreal region (EA).

Partial Correlation All NA EA Positive 47.05% 42.11% 49.28% SCD_EVImax Negative 52.95% 57.89% 50.72% Positive 56.32% 45.26% 61.32% SCD_SOS Negative 43.68% 54.74% 38.68% Positive 57.71% 61.34% 56.06% SCE_EVImax Negative 42.29% 38.66% 43.94% Positive 62.88% 68.41% 60.34% SCE_SOS Negative 37.12% 31.59% 39.66%

References

1. Stocker, T.F.; Qin, D.; Plattner, G.K.; Tignor, M.M.B.; Allen, S.K.; Boschung, J.; Nauels, A.; Xia, Y.; Bex, V.; Midgley, P.M. Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of IPCC the Intergovernmental Panel on Climate Change; Cambridge University Press: New York, NY, USA, 2014. 2. Wallace, J.M.; Fu, Q.; Smoliak, B.V.; Lin, P.; Johanson, C.M. Simulated versus observed patterns of warming over the extratropical Northern Hemisphere continents during the cold season. Proc. Natl. Acad. Sci. USA 2012, 109, 14337–14342. [CrossRef][PubMed] 3. Fu, Y.H.; Zhao, H.; Piao, S.; Peaucelle, M.; Peng, S.; Zhou, G.; Ciais, P.; Huang, M.; Menzel, A.; Peñuelas, J.; et al. Declining global warming eﬀects on the phenology of spring leaf unfolding. Nature 2015, 526, 104–107. [CrossRef] 4. Wu, C.; Hou, X.; Peng, D.; Gonsamo, A.; Xu, S. Land surface phenology of China’s temperate ecosystems over 1999–2013: Spatial-temporal patterns, interaction eﬀects, covariation with climate and implications for productivity. Agric. For. Meteorol. 2016, 216, 177–187. [CrossRef] Forests 2019, 10, 376 18 of 20

5. Yao, Y.; Wang, X.; Li, Y.; Wang, T.; Shen, M.; Du, M.; He, H.; Li, Y.; Luo, W.; Ma, M.; et al. Spatiotemporal pattern of gross primary productivity and its covariation with climate in China over the last thirty years. Glob. Chang. Biol. 2018, 24, 184–196. [CrossRef] 6. Zhang, Y.; Wang, G.; Wang, Y. Response of biomass spatial pattern of alpine vegetation to climate change in permafrost region of the Qinghai-Tibet Plateau, China. J. Mt. Sci. 2010, 7, 301–314. [CrossRef] 7. Piao, S.; Wang, X.; Ciais, P.; Zhu, B.; Wang, T.; Liu, J. Changes in satellite-derived vegetation growth trend in temperate and boreal Eurasia from 1982 to 2006. Glob. Chang. Biol. 2011, 17, 3228–3239. [CrossRef] 8. Delbart, N.; Picard, G.; Le Toan, T.; Kergoat, L.; Quegan, S.; Woodward, I.; Dye, D.; Fedotova, V. Spring phenology in boreal Eurasia over a nearly century time scale. Glob. Chang. Biol. 2008, 14, 603–614. [CrossRef] 9. Vitasse, Y.; François, C.; Delpierre, N.; Dufrêne, E.; Kremer, A.; Chuine, I.; Delzon, S. Assessing the eﬀects of climate change on the phenology of European temperate trees. Agric. Forest Meteorol. 2011, 151, 969–980. [CrossRef] 10. Zeng, H.; Jia, G.; Forbes, B.C. Shifts in Arctic phenology in response to climate and anthropogenic factors as detected from multiple satellite time series. Environ. Res. Lett. 2013, 8, 035036. [CrossRef] 11. Fu, Y.H.; Piao, S.; Op de Beeck, M.; Cong, N.; Zhao, H.; Zhang, Y.; Menzel, A.; Janssens, I.A. Recent spring phenology shifts in western Central Europe based on multiscale observations: Multiscale observation of spring phenology. Glob. Ecol. Biogeogr. 2014, 23, 1255–1263. [CrossRef] 12. Park, T.; Ganguly, S.; Tømmervik, H.; Euskirchen, E.S.; Høgda, K.-A.; Karlsen, S.R.; Brovkin, V.; Nemani, R.R.; Myneni, R.B. Changes in growing season duration and productivity of northern vegetation inferred from long-term remote sensing data. Environ. Res. Lett. 2016, 11, 084001. [CrossRef] 13. Jin, J.; Wang, Y.; Jiang, H.; Cheng, M. Recent NDVI-Based Variation in Growth of Boreal Intact Forest Landscapes and Its Correlation with Climatic Variables. Sustainability 2016, 8, 326. [CrossRef] 14. Wang, X.; Piao, S.; Ciais, P.; Li, J.; Friedlingstein, P.; Koven, C.; Chen, A. Spring temperature change and its implication in the change of vegetation growth in North America from 1982 to 2006. Proc. Natl. Acad. Sci. USA 2011, 108, 1240–1245. [CrossRef] 15. Zhao, L.; Li, Y.; Xu, S.; Zhou, H.; Gu, S.; Yu, G.; Zhao, X. Diurnal, seasonal and annual variation in net

ecosystem CO2 exchange of an alpine shrubland on Qinghai-Tibetan plateau. Glob. Chang. Biol. 2006, 12, 1940–1953. [CrossRef] 16. Zeng, H.; Jia, G.; Epstein, H. Recent changes in phenology over the northern high latitudes detected from multi-satellite data. Environ. Res. Lett. 2011, 6, 45508–45518. [CrossRef] 17. Cong, N.; Wang, T.; Nan, H.; Ma, Y.; Wang, X.; Myneni, R.B.; Piao, S. Changes in satellite-derived spring vegetation green-up date and its linkage to climate in China from 1982 to 2010: A multimethod analysis. Glob. Chang. Biol. 2013, 19, 881–891. [CrossRef] 18. Shen, M.; Cong, N.; Cao, R. Temperature sensitivity as an explanation of the latitudinal pattern of green-up date trend in Northern Hemisphere vegetation during 1982–2008. Int. J. Climatol. 2015, 35, 3707–3712. [CrossRef] 19. Chen, X.; An, S.; Inouye, D.W.; Schwartz, M.D. Temperature and snowfall trigger alpine vegetation green-up on the world’s roof. Glob. Chang. Biol. 2015, 21, 3635–3646. [CrossRef] 20. Wang, K.; Zhang, L.; Qiu, Y.; Ji, L.; Tian, F.; Wang, C.; Wang, Z. Snow eﬀects on alpine vegetation in the Qinghai-Tibetan Plateau. Int. J. Digit. Earth 2015, 8, 58–75. [CrossRef] 21. Wang, X.; Wang, T.; Guo, H.; Liu, D.; Zhao, Y.; Zhang, T.; Liu, Q.; Piao, S. Disentangling the mechanisms behind winter snow impact on vegetation activity in northern ecosystems. Glob. Chang. Biol. 2017, 24, 1651–1662. [CrossRef] 22. Chen, X.; Liang, S.; Cao, Y. Satellite observed changes in the Northern Hemisphere snow cover phenology and the associated radiative forcing and feedback between 1982 and 2013. Environ. Res. Lett. 2016, 11, 084002. [CrossRef] 23. Zhang, Y.; Ma, N. Spatiotemporal variability of snow cover and snow water equivalent in the last three decades over Eurasia. J. Hydrol. 2018, 559, 238–251. [CrossRef] 24. Peng, S.; Piao, S.; Ciais, P.; Friedlingstein, P.; Zhou, L.; Wang, T. Change in snow phenology and its potential feedback to temperature in the Northern Hemisphere over the last three decades. Environ. Res. Lett. 2013, 8, 1880–1885. [CrossRef] 25. Chen, X.; Liang, S.; Cao, Y.; He, T. Distribution, attribution, and radiative forcing of snow cover changes over China from 1982 to 2013. Clim. Chang. 2016, 137, 363–377. [CrossRef] Forests 2019, 10, 376 19 of 20

26. Robinson, D.A.; Kukla, G. Maximum Surface Albedo of Seasonally Snow-Covered Lands in the Northern Hemisphere. J. Clim. Appl. Meteorol. 1985, 24, 402–411. [CrossRef] 27. Yun, J.; Jeong, S.-J.; Ho, C.-H.; Park, C.-E.; Park, H.; Kim, J. Influence of winter precipitation on spring phenology in boreal forests. Glob. Chang. Biol. 2018, 24, 5176–5187. [CrossRef] 28. Bonan, G.B.; Pollard, D.; Thompson, S.L. Effects of boreal forest vegetation on global climate. Nature 1992, 359, 716–718. [CrossRef] 29. Pulliainen, J.; Aurela, M.; Laurila, T.; Aalto, T.; Takala, M.; Salminen, M.; Kulmala, M.; Barr, A.; Heimann, M.; Lindroth, A.; et al. Early snowmelt significantly enhances boreal springtime carbon uptake. Proc. Natl. Acad. Sci. USA 2017, 114, 11081–11086. [CrossRef] 30. Gonsamo, A.; Chen, J.M.; Ooi, Y.W. Peak season plant activity shift towards spring is reflected by increasing carbon uptake by extratropical ecosystems. Glob. Chang. Biol. 2018, 24, 2117–2128. [CrossRef] 31. Taggart, R.E.; Cross, A.T. Global greenhouse to icehouse and back again: The origin and future of the Boreal Forest biome. Glob. Planet. Chang. 2009, 65, 115–121. [CrossRef] 32. Huete, A.; Didan, K.; Miura, T.; Rodriguez, E.P.; Gao, X.; Ferreira, L.G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 2002, 83, 195–213. [CrossRef] 33. Di Mauro, B.; Fava, F.; Busetto, L.; Crosta, G.F.; Colombo, R. Post-fire resilience in the Alpine region estimated from MODIS satellite multispectral data. Int. J. Appl. Earth Obs. Geoinf. 2014, 32, 163–172. [CrossRef] 34. Luo, Z.; Yu, S. Spatiotemporal Variability of Land Surface Phenology in China from 2001–2014. Remote Sens. 2017, 9, 65. [CrossRef] 35. Chen, X.; Liang, S.; Cao, Y.; He, T.; Wang, D. Observed contrast changes in snow cover phenology in northern middle and high latitudes from 2001–2014. Sci. Rep. 2015, 5, 16820. [CrossRef][PubMed] 36. Guo, X.; Zhang, H.; Wu, Z.; Zhao, J.; Zhang, Z. Comparison and Evaluation of Annual NDVI Time Series in China Derived from the NOAA AVHRR LTDR and Terra MODIS MOD13C1 Products. Sensors 2017, 17, 1298. [CrossRef][PubMed] 37. Harris, I.; Jones, P.D.; Osborn, T.J.; Lister, D.H. Updated high-resolution grids of monthly climatic observations-the CRU TS3.10 Dataset. Int. J. Climatol. 2014, 34, 623–642. [CrossRef] 38. Liu, Q.; Fu, Y.H.; Zhu, Z.; Liu, Y.; Liu, Z.; Huang, M.; Janssens, I.A.; Piao, S. Delayed autumn phenology in the Northern Hemisphere is related to change in both climate and spring phenology. Glob. Chang. Biol. 2016, 22, 3702–3711. [CrossRef] 39. Chen, J.; Jönsson, P.; Tamura, M.; Gu, Z.; Matsushita, B.; Eklundh, L. A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky-Golay filter. Remote Sens. Environ. 2004, 91, 332–344. [CrossRef] 40. White, M.A.; de Beurs, K.M.; Didan, K.; Inouye, D.W.; Richardson, A.D.; Jensen, O.P.; O’Keefe, J.; Zhang, G.; Nemani, R.R.; van Leeuwen, W.J.D.; et al. Intercomparison, interpretation, and assessment of spring phenology in North America estimated from remote sensing for 1982–2006. Glob. Chang. Biol. 2009, 15, 2335–2359. [CrossRef] 41. Jönsson, P.; Eklundh, L. TIMESAT—A program for analyzing time-series of satellite sensor data. Comput. Geosci. 2004, 30, 833–845. 42. Zhou, J.; Jia, L.; Menenti, M. Reconstruction of global MODIS NDVI time series: Performance of Harmonic ANalysis of Time Series (HANTS). Remote Sens. Environ. 2015, 163, 217–228. [CrossRef] 43. Yu, Z.; Liu, S.; Wang, J.; Sun, P.; Liu, W.; Hartley, D.S. Effects of seasonal snow on the growing season of temperate vegetation in China. Glob. Chang. Biol. 2013, 19, 2182–2195. [CrossRef] 44. Li, H.; Wang, C.; Zhang, L.; Li, X.; Zang, S. Satellite monitoring of boreal forest phenology and its climatic responses in Eurasia. Int. J. Remote Sens. 2017, 38, 5446–5463. [CrossRef] 45. Zhao, J.; Wang, Y.; Zhang, Z.; Zhang, H.; Guo, X.; Yu, S.; Du, W.; Huang, F. The Variations of Land Surface Phenology in Northeast China and Its Responses to Climate Change from 1982 to 2013. Remote Sens. 2016, 8, 400. [CrossRef] 46. Karkauskaite, P.; Tagesson, T.; Fensholt, R. Evaluation of the Plant Phenology Index (PPI), NDVI and EVI for Start-of-Season Trend Analysis of the Northern Hemisphere Boreal Zone. Remote Sens. 2017, 9, 485. [CrossRef] Forests 2019, 10, 376 20 of 20

47. Wang, X.; Wu, C.; Peng, D.; Gonsamo, A.; Liu, Z. Snow cover phenology affects alpine vegetation growth dynamics on the Tibetan Plateau: Satellite observed evidence, impacts of different biomes, and climate drivers. Agric. For. Meteorol. 2018, 256–257, 61–74. [CrossRef] 48. Wang, X.; Wu, C.; Wang, H.; Gonsamo, A.; Liu, Z. No evidence of widespread decline of snow cover on the Tibetan Plateau over 2000–2015. Sci. Rep. 2017, 7, 14645. [CrossRef] 49. Kendall, M.G. A New Measure of Rank Correlation. Biometrika 1938, 30, 81. [CrossRef] 50. Mann, H.B. Nonparametric Tests against Trend. Econometrica 1945, 13, 245–259. [CrossRef] 51. Sönmez, I.; Tekeli, A.E.; Erdi, E. Snow cover trend analysis using Interactive Multisensor Snow and Ice Mapping System data over Turkey. Int. J. Climatol. 2014, 34, 2349–2361. [CrossRef] 52. Huang, X.; Deng, J.; Ma, X.; Wang, Y.; Feng, Q.; Hao, X.; Liang, T. Spatiotemporal dynamics of snow cover based on multi-source remote sensing data in China. Cryosphere 2016, 10, 2453–2463. [CrossRef] 53. Fu, Y.H.; Piao, S.; Zhao, H.; Jeong, S.-J.; Wang, X.; Vitasse, Y.; Ciais, P.; Janssens, I.A. Unexpected role of winter precipitation in determining heat requirement for spring vegetation green-up at northern middle and high latitudes. Glob. Chang. Biol. 2014, 20, 3743–3755. [CrossRef][PubMed] 54. Barichivich, J.; Briffa, K.R.; Myneni, R.B.; Osborn, T.J.; Melvin, T.M.; Ciais, P.; Piao, S.; Tucker, C. Large-scale variations in the vegetation growing season and annual cycle of atmospheric CO2 at high northern latitudes from 1950 to 2011. Glob. Chang. Biol. 2013, 19, 3167–3183. [CrossRef] 55. Wang, X.; Piao, S.; Xu, X.; Ciais, P.; MacBean, N.; Myneni, R.B.; Li, L. Has the advancing onset of spring vegetation green-up slowed down or changed abruptly over the last three decades? 30-year change of spring vegetation phenology. Glob. Ecol. Biogeogr. 2015, 24, 621–631. [CrossRef] 56. Schuster, C.; Estrella, N.; Menzel, A. Shifting and extension of phenological periods with increasing temperature along elevational transects in southern Bavaria. Plant Biol. 2014, 16, 332–344. [CrossRef] 57. Choi, G.; Robinson, D.A.; Kang, S. Changing Northern Hemisphere Snow Seasons. J. Clim. 2010, 23, 5305–5310. [CrossRef] 58. Menzel, A.; Sparks, T.H.; Estrella, N.; Koch, E.; Aasa, A.; Ahas, R.; Alm-KüBler, K.; Bissolli, P.; Braslavská, O.; Briede, A.; et al. European phenological response to climate change matches the warming pattern. Glob. Chang. Biol. 2006, 12, 1969–1976. [CrossRef] 59. Yang, M.; Nelson, F.E.; Shiklomanov, N.I.; Guo, D.; Wan, G. Permafrost degradation and its environmental effects on the Tibetan Plateau: A review of recent research. Earth-Sci. Rev. 2010, 103, 31–44. [CrossRef] 60. Vitasse, Y.; Basler, D. What role for photoperiod in the bud burst phenology of European beech. Eur. J. For. Res. 2013, 132, 1–8. [CrossRef] 61. Clark, J.S.; Salk, C.; Melillo, J.; Mohan, J. Tree phenology responses to winter chilling, spring warming, at north and south range limits. Funct. Ecol. 2014, 28, 1344–1355. [CrossRef] 62. Wang, S.; Wang, X.; Chen, G.; Yang, Q.; Wang, B.; Ma, Y.; Shen, M. Complex responses of spring alpine vegetation phenology to snow cover dynamics over the Tibetan Plateau, China. Sci. Total Environ. 2017, 593–594, 449–461. [CrossRef][PubMed] 63. Trujillo, E.; Molotch, N.P.; Goulden, M.L.; Kelly, A.E.; Bales, R.C. Elevation-dependent influence of snow accumulation on forest greening. Nat. Geosci. 2012, 5, 705–709. [CrossRef] 64. Liu, Q.; Piao, S.; Janssens, I.A.; Fu, Y.; Peng, S.; Lian, X.; Ciais, P.; Myneni, R.B.; Peñuelas, J.; Wang, T. Extension of the growing season increases vegetation exposure to frost. Nat. Commun. 2018, 9, 462. [CrossRef] 65. Blume-Werry, G.; Kreyling, J.; Laudon, H.; Milbau, A. Short-term climate change manipulation effects do not scale up to long-term legacies: Effects of an absent snow cover on boreal forest plants. J. Ecol. 2016, 104, 1638–1648. [CrossRef] 66. Man, R.; Lu, P.; Dang, Q.-L. Insufficient Chilling Effects Vary among Boreal Tree Species and Chilling Duration. Front. Plant Sci. 2017, 8, 1354. [CrossRef][PubMed] 67. Bintanja, R.; Andry, O. Towards a rain-dominated Arctic. Nat. Clim. Chang. 2017, 7, 263–267. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).