Snow effects on alpine vegetation in the Qinghai-Tibetan Plateau

Abstract

Understanding the relationships between snow and vegetation is important for interpretation of the responses of alpine ecosystems to climate changes. The Qinghai-Tibetan Plateau is regarded as an ideal area due to its undisturbed features with low population and relatively high snow cover. We used 500 m Moderate Resolution Imaging Spectroradiometer (MODIS) datasets during 2001–2010 to examine the snow–vegetation relationships, specifically, (1) the influence of snow melting date on vegetation green-up date and (2) the effects of snow cover duration on vegetation greenness. The results showed that the alpine vegetation responded strongly to snow phenology (i.e., snow melting date and snow cover duration) over large areas of the Qinghai-Tibetan Plateau. Snow melting date and vegetation green-up date were significantly correlated (p < 0.1) in 39.9% of meadow areas (accounting for 26.2% of vegetated areas) and 36.7% of steppe areas (28.1% of vegetated areas). Vegetation growth was influenced by different seasonal snow cover durations (SCDs) in different regions. Generally, the December–February and March–May SCDs played a significantly role in vegetation growth, both positively and negatively, depending on different water source regions. Snow's positive impact on vegetation was larger than the negative impact.

Keywords:

• phenology
• snow cover duration
• snow melting date
• NDVI
• vegetation green-up date

1. Introduction

Over the past two decades, terrestrial vegetation has undergone dramatic changes from regional to global scale monitored with long-term satellite records (Li et al. 2010; Stockli et al. 2011; Beck and Goetz 2011, 2012; Fensholt et al. 2012). Recent reviews confirm that climate changes are likely to affect several aspects of vegetation, such as plant composition and diversity (Ozanne et al. 2003; Hallinger, Manthey, and Wilmking 2010), phenology (Nemani et al. 2002), greenness and productivity (Thomey et al. 2011; Zhang et al. 2013), and biomass and vegetation fraction (Zhang and Walsh 2007; Yu, Wang, and Wang 2010). Several studies have explicitly examined factors that control vegetation change and have confirmed that climate is the dominant factor affecting changes in plant communities (Voronin et al. 2003; Millett, Johnson, and Guntenspergen 2009). In northern mid-high latitudinal zones, increasing temperature and precipitation are considered as the primary climatic factors controlling vegetation growth (Tucker et al. 2001; Ichii et al. 2002).

In alpine vegetation areas, snow is a major source of fresh water for vegetation activities (Beniston et al. 1997). In addition, snow insulates soil from cold temperatures and wind, which is beneficial to vegetation growth (Fahnestock et al. 1998; Nobrega and Grogan 2007). Previous studies have suggested that increasing snow depth promotes alpine vegetation growth (Groffman et al. 2001; Dorrepaal et al. 2004). However, most such studies focused only on the response of vegetation activities to snow depth, which results in an unclear view of the relationship between vegetation activities and snow phenology. Furthermore, snow measurements in previous studies have primarily been limited to in situ records and microwave data. In situ snow records are scarce in the region and coarse resolution microwave data have low measurement accuracy. The Moderate Resolution Imaging Spectroradiometer (MODIS) snow products (MOD10A2) provide relatively high temporal frequency (8-day) for monitoring snow cover at 500-m resolution and provide an effective method for monitoring snow phenology parameters, thereby promoting awareness of the effect of snow on vegetation activities.

As a typical alpine ecosystem, the Qinghai-Tibetan Plateau is an ideal area for the study of the effect of snow phenology on vegetation activities. The plateau has a relatively large area covered by snow (Pu and Xu 2009) and is more sensitive to global climatic changes than most other parts of the world (Ye and Wu 1998). Snow cover plays an important role in ecological environment on the Qinghai-Tibetan Plateau (Li et al. 2008; Li et al. 2010; Xiao et al. 2002; Yan and Onipchenko 2006). A warming trend over the plateau, with an average climate increase of 0.37°C/decade, was recently reported by Li et al. (2008). Increasing temperatures may result in a decline in snow cover and enhancement of vegetation activities. A warming trend may also promote interactions between changes in snow cover and vegetation activities (Dye and Tucker 2003). Understanding the interactions between vegetation and snow in alpine ecosystems can support ecological modeling and help deepen understanding of vegetation dynamics.

The objective of this study is to explore the effect of snow phenology on vegetation in alpine ecosystems. In this study, we used MODIS-derived vegetation parameters and snow phenology parameters to explore the interactions between snow cover and vegetation in the Qinghai-Tibetan Plateau during 2001 and 2010. We analyzed (1) the spatial patterns of snow phenology parameters (snow cover duration and snow melting date) and vegetation parameters (green-up date and greenness), (2) the relationships between snow melting date (SMD) and vegetation green-up date, and (3) the relationships between snow cover duration (SCD) and vegetation greenness.

2. Study area and data

2.1. Study area

The Qinghai-Tibetan Plateau is located in southwestern China (74°E–104°E, 25°N–40°N), including Qinghai Province, Tibet Autonomous Region, and parts of Gansu, Xinjiang, Sichuan, and Yunnan provinces (Figure 1). As the ‘roof of the world,’ the Qinghai-Tibetan Plateau has an average elevation over 4000 m and is the highest and largest plateau in the world. It has a well-developed network of watercourses, including the headwaters of the Yangtze River, the Yellow River, and the Lantsang River.

Figure 1. Vegetation types and water source regions of the Qinghai-Tibetan Plateau.Data source: the Institute of Geography Science and Natural Sources Research, Chinese Academy of Sciences (http://www.geodata.cn/Portal/). The plateau is divided into eight water source regions: Western Qilian Mountains, Eastern Qilian Mountains, Central and Western Kunlun Mountains, Northern Qiangtang, Southern Qiangtang, Central Himalaya Mountains, Yigong Tsangpo, and Source of Three Rivers.

The Qinghai-Tibetan Plateau forms a unique climatic feature characterized by strong solar radiation, warm and humid summers, and cold and dry winters (Peng et al. 2012). In most parts of the plateau, the annual mean temperature is below 0°C. Summer precipitation follows a gradient from about 450 mm in the southeast to 800 mm in the northwest (Chen et al. 2012).

The vegetation on the Qinghai-Tibetan Plateau generally follows the moisture and temperature gradient. Based on the Vegetation Atlas of China (1:1,000,000) (Editorial Committee of Vegetation Altas of China, the Chinese Academy of Sciences, 2001), the vegetation includes meadows (26.2%), steppes (28.1%), grasslands (9.7%), forests (7.4%), and others (Figure 1). Overall, 84.2% of the study area is covered by vegetation. The cold and dry northern plateau is covered by desert grasslands, the central area by meadows and steppes, and the southeastern plateau by shrubs and forests. The growing season over the plateau is May to September (Xu et al. 2005).

Snow phenology over the plateau is an indicator of the climate. Beginning in September, snow cover expands rapidly and reaches its peak in December to February, and then rapidly reduces and finally disappears in late May (Qin, Liu, and Li 2006). From May to September, the plateau is snow-free in most areas, except for a few locations in the mountain ranges. The duration of snow cover on the plateau varies with elevations (Pu, Xu, and Salomonson 2007; Pu and Xu 2009). Areas persistently covered by snow are primarily located in the Himalayas, the Kunlun Range, the Karakoram Range, and western part of the Yarlung Zangbo Valley, which include the most heavily glaciated regions in the world. The melt waters of the snow and glaciers are the main runoff sources of the upper reaches of many rivers (Zhu, Shi, and Wang 2012), such as the Yangtze River, the Yellow River, and the Lantsang River, and provide vital water source for most China.

2.2. Data

2.2.1. NDVI data

The normalized difference vegetation index (NDVI) has been used as a surrogate for photosynthetic capacity, because it is highly correlated to the absorbed fraction of photosynthetically active radiation and gross photosynthesis (Asrar et al. 1984; Myneni et al. 1995; Tucker et al. 2001; Zhou et al. 2001; Xiao and Moody 2005). The index is defined as the ratio of the difference between near-infrared reflectance (ρ _nir) and red reflectance (ρ _red) to their summation (Equation 1) (Rouse et al. 1974).

(1)

where ρ_red and ρ_nir are the spectral reflectance measurements acquired in the visible (red) and near-infrared (nir) regions, respectively. NDVI has been widely used for monitoring vegetation changes from regional to global scales (Myneni et al. 1997; Xiao and Moody 2005; Ho et al. 2006; Parent and Verbyla 2010; De Jong et al. 2011; Fensholt et al. 2012). In the study, we used MOD09A1 (500-m spatial resolution and 8-day interval) to monitor vegetation growth during the growing season (May–September) from 2001 to 2010 in the Qinghai-Tibetan Plateau. Each MOD09A1 pixel contains the best possible observation during an 8-day period as selected based on high observation coverage, low view angle, the absence of clouds or cloud shadow, and aerosol loading. The 500-m and 8-day NDVI time-series images from 2001 to 2010 were derived from the red and near-infrared reflectance (bands 1 and 2) in the MOD09A1 products based on Equation (1). The Quality Assurance layer was used to mask out cloud, shadow, and snow/ice pixels that were excluded in the data analysis.

2.2.2. Snow cover data

In comparison to other snow cover products (e.g. NOHRSC and SSM/I), the MODIS snow cover data MOD10A2 (500-m spatial resolution and 8-day interval) has higher spatial and temporal resolutions and higher measurement accuracy (Hall et al. 2002). In addition, MOD10A2 can effectively minimize the effect of cloud contamination in most cases (Hall and Riggs 2007) and has higher classification accuracy for both snow and land than another MODIS snow product (MOD10A1) that has 500-m resolution and daily interval (Zhou, Xie, and Hendrickx 2005). Previous studies have shown that the overall accuracy of MOD10A2 ranges between 84% and 91% (Wang, Xie, and Liang 2008; Pu, Xu, and Salomonson 2007; Pu and Xu 2009). Considering the relatively high accuracy, we adopted MOD10A2 from September 2000 to May 2010 to derive the snow phenology parameters in our study.

3. Methods

3.1. Derivation of snow phenology parameters

Two snow phenology parameters, SCD and SMD, were used to describe snow conditions in the Qinghai-Tibetan Plateau. The SCD is positively associated with soil moisture and improves vegetation growth (Chen et al. 2008). Variation in the snow melting date affects the timing of fresh water availability for vegetation activities (Wang and Xie 2009). In our study, a hydrological year is referred to as the period between 1 September of concurrent year to 31 August of the following years. SCD is defined as the total snow cover days in a hydrological year. The following Equation (2) is used to calculate SCD:

(2)

where n is the number of intervals within a hydrological year and d_i is occurrence of snow cover, with value of 1 indicating snow and value of 0 indicating no snow. Because the snow cover data have an interval of 8 days, d_i is multiplied by 8 to obtain the actual days of SCD within a hydrological year. Then SCD was calculated for each year between 2001 and 2010. Seasonal SCD was calculated based on the images of a particular season using the same method described above. For example, the calculation of spring SCD used images from March to May.

Snow melting date (SMD) was calculated using Equation (3) based on the method described by Wang and Xie (2009):

(3)

where Fd is a fixed date and SCD_aFd is the number of snow-covering days after Fd, calculated using Equation (2). Dietz et al. (2013) modified Fd to indicate the date when the snow cover area reaches the largest. In the Qinghai-Tibetan Plateau, the snow cover area reaches its maximum twice each year, in December and February (Pu, Xu, and Salomonson 2007; Pu and Xu 2009). Similar to the study by Dietz et al. (2013), we set Fd to the date of the second peak. Specifically, Fd was set to February 16 that was the average of the 10 years from 2001 to 2010.

3.2. Derivation of vegetation green-up date

Green-up date is a key phenology parameter, which indicates the time when vegetation begins to grow. Vegetation green-up date is considered as an important indicator of climate change. Changes in vegetation green-up date can alter the length of growing season and affect many properties of terrestrial ecosystems, such as carbon cycles, nitrogen cycles, and vegetation primary production (Walther et al. 2002; Carrara et al. 2003; Barr et al. 2007; Piao et al. 2007). TIMESAT is a software developed for analyzing time-series data to derive phenology parameters, such as green-up date, green-off date, and length of growing season (Jönsson and Eklundh 2004), and has been widely used in various studies (Boschetti et al. 2009; Wang et al. 2011). In this study, we employed the TIMESAT program to derive green-up date. The derivation of green-up date includes three steps: (1) The time-series NDVI data were smoothed using the Savitzky–Golay filtering method. Small moving windows in the Savitzky–Golay filter yielded results that are able to detect rapid changes, whereas increasing moving window leads to smoothed NDVI time-series data. To achieve a balance between responsiveness and smoothness, we used a window size of 2. (2) The filtered time series was fitted using an asymmetric Gaussian method. Those pixels that failed to return a valid result were set to ‘no data’. (3) The fitted NDVI time series was analyzed using the TIMESAT program. The time series at each pixel was analyzed to extract green-up date that represented the date when NDVI increased to 30% of the amplitude of the fitted Gaussian function.

3.3. Correlation analysis between snow cover and vegetation

To explore the sensitivity of vegetation activities to snow cover, two correlation analyses were performed at pixel level to examine the relationships between SMD and green-up date and between SCD and vegetation greenness (indicated by NDVI). All the variables were examined from 2001 to 2010, so the sample size of them was 10. Pearson correlation coefficients (r) were calculated, and the significance of the correlation coefficients (p) was estimated with two-tailed t-test at a significance level of 90%.

There are lag effects in the vegetation response to snow in alpine and arctic ecosystems (Buus-Hinkler et al. 2006; Peng et al. 2010). However, little is known about the lag effects of snow duration on vegetation growth in the Qinghai-Tibetan Plateau. In this study, we examined the lag effects of seasonal SCD on vegetation greenness using the cross-correlation method illustrated in Figure 2. Because the period from September to May was usually dominated by snow cover (Qin, Liu, and Li 2006), we divided SCD into five periods:

1. September–November (fall) accumulation,

2. December–February (winter) accumulation,

3. March–May (spring) accumulation,

4. December–May accumulation, and

5. September–May accumulation.

We firstly investigated the vegetation greenness in each month from May to September to determine the months when vegetation activities were most influenced by SCD, and then investigated which SCD periods had the most influence on vegetation activities. We used the cross-correlation analysis for five snow cover periods and the NDVI of five months. The lag effect analysis between seasonal SCDs and NDVI consisted of three steps as follows.

Figure 2. Illustration of lag effect analyse between snow cover duration and NDVI.

Step 1: Identification of most influenced months of NDVI. The correlations of all five SCDs (September–November, December–February, March–May, December–May, and September–May) with May NDVI were calculated and the maximum (positive and negative) values of the five correlation coefficients were obtained for each pixel. The same procedure was then performed for the NDVI of other four months (June, July, August, and September). The image that had the maximum (positive and negative) correlation coefficients was obtained for each month and the month that had largest number of pixels with significant correlation (p < 0.1) was regarded as the ‘most influenced month of NDVI’ by SCDs (Figure 2, Step 1).

Step 2: Identification of most influencing periods of snow cover. The five SCD periods that was the most correlated with the NDVI in the most influenced months was regarded as the ‘most influencing period’ of snow cover for vegetation greenness (Figure 2, Step 2).

Step 3: The interactions between the most influencing periods of SCDs and the most influenced months of NDVI were analyzed among different water source regions using correlation analysis to evaluate the effects of snow phenology on vegetation activities in the alpine ecosystem (Figure 2, Step 3).

4. Results and discussion

4.1. Spatial distribution of snow cover and vegetation parameters

Snow cover in the Qinghai-Tibetan Plateau is very uneven in space (Figure 3). The maps of SCD from 2001 to 2010 (Figure 3a) show that the longest SCD (> 200 days) occurred in the Himalayas, Karakoram, Kunlun, and Nyainqentanglha Mountains, accounting for 4.3% of the plateau area. Areas with relatively shorter SCD (20 < days < 200) were located mostly in the central plateau, accounting for 71.3% of the plateau. This area is dominated by steppes and meadows where snow cover prevents the vegetation from photosynthesis and thus the conditions for vegetation growth. Due to large-scale shielding from the Himalaya and Karakoram Mountains, areas with short SCDs were located mainly in the southwestern and southeastern plateau, accounting for about 24.4% of the plateau.

Figure 3. Snow cover parameters (2001–2010 average): (a) snow cover duration and (b) snow melting date.

The pattern of SMD in the plateau is similar to SCD (Figure 3b). Relatively early SMD (March) was observed in the areas with shorter SCD scattering over the Qinghai-Tibetan Plateau. These areas cover 70% of the plateau, which are primarily distributed in the lower lands of the central and northeastern plateau. Areas with relatively late SMD (April and May or later) is mainly located in the areas with long SCD in the Himalaya, Karakoram, Kunlun, and Nyainqentanglha Mountains.

Overall, vegetation greenness on the Qinghai-Tibetan Plateau decreased from the southeast to the northwest (Figure 4a). The mean annual growing season (May–September) NDVI during 2001–2010 was highest in the southeastern Plateau (Figure 4a), which is dominated by forests. The mean annual growing season NDVI values decreased to 0.3–0.5 in the central plateau where the dominant vegetation is meadows, and reached the lowest in the northwest part where desert grasslands and steppes are the main vegetation types (Figure 4a). Green-up date shows a similar spatial pattern (Figure 4b) as the growing season NDVI. The green-up date occurred between late April and early May in the eastern plateau and June in the western plateau.

Figure 4. Vegetation parameters (2001–2010 average): (a) growing season (May–September) NDVI and (b) green-up date.

Although alpine ecosystems are influenced by many environmental factors, snow phenology plays a key and complicated role in controlling the spatial pattern of vegetation parameters (Billings and Bliss 1959). The variations in SCD result in a snow melting gradient, which further affects the alternation between freezing and thawing in the eastern Qinghai-Tibetan Plateau (Chen et al. 2008). In the areas with early SMD, without the protection of snow cover, vegetation exposed to cold temperatures cannot grow well because of the harsh environment (Chen et al. 2008). In contrast, in the areas with late SMD, snow insulation protects vegetation from damage caused by freezing and thawing, but the short growing season constrains the vegetation growth (Benedict 1990; Galen and Stanton 1993, 1995; Totland and Alatalo 2002; Chen et al. 2008). Therefore, vegetation may grow better in those areas with intermediate SCD. The above findings explain the patterns for vegetation phenology that may have close relationship with SCDs and SMD.

4.2. Effects of SMD on vegetation green-up date

Figure 5 shows the spatial distribution of correlation coefficients between snow melting date and green-up date in the Qinghai-Tibetan Plateau. Large areas in the southeastern and northwestern plateau have no data due to cloud contamination. We observed that 39.9% of the meadow areas (accounting for 26.2% of vegetated areas) and 36.7% of the steppe areas (28.1% of vegetated areas) had significant correlations (p < 0.1) between SMD and vegetation green-up date, mainly located in the central plateau. Snow cover blocks incoming photosynthetic active radiation (PAR), and thus, the late snowmelt date delays the green-up date (Belzile et al. 2001; Delbart et al. 2006; Buus-Hinkler et al. 2006). In addition, changes in SMD may alter the vegetation green-up date by influencing the temperature for vegetation growth (Grippa et al. 2005). During snow melting period, cold and wet air always takes away large amounts of heat from soil and sometimes leads to low temperature, which, in turn, is harmful for vegetation growth. The reactions of vegetation green-up date to SMD found in our study are similar to other regions of the northern hemisphere. For example, Knight et al. (1979) found that vegetation green-up was delayed by about 7–9 days due to delayed SMD in subalpine meadows in Wyoming, U.S. Levis and Bonan (2004) found a strong correlation between day of leaf emergence and SMD in northern Europe, central Canada, and eastern China.

Figure 5. Correlation coefficients between snow melting date and green-up date in the Qinghai-Tibetan Plateau. Areas with insignificant correlations (p > 0.1) are shown in white. In the areas identified as ‘snow cover before Fd’, no snow was detected after the peak date of February 16.

4.3. Effects of SCDs on growing season NDVI

Snow cover could have both positive and negative impacts on vegetation. Snow promotes vegetation growth by providing water for vegetation growth and protecting soil from exposure to wind and low temperature. On the other hand, it may also negatively affect vegetation growth as it associates with low temperature and delays the snow melting date. We examined both positive and negative effects of snow on vegetation growth in the Qinghai-Tibetan Plateau. Lag effects existed in the correlation between SCDs and vegetation greenness. To determine the lag effect, we employed the cross-correlation method described in Section 3.3 and Figure 2.

The highest positive correlation between SCDs and NDVI between May and September (Figure 2, Step 1) occurred in June and August (Figure 6). There was a significant correlation (p < 0.1) between SCD and NDVI for 57.6% of the vegetated areas in June (Figure 6b). In July, SCD and NDVI were significantly correlated for 52.1% of the vegetated area. The correlation between SCD and NDVI was significant (p < 0.1) for most area (62.2%) in August (Figure 6d). Fewer areas (54.6% of the vegetated areas) were significantly correlated with SCDs (p < 0.1) in May than in June and August. In May, the vegetation begins to grow in the Qinghai-Tibetan Plateau and vegetation conditions are poorer than those in other months. The poor vegetation conditions limit the identification of the relationships between SCD and NDVI (Figure 6a). The lowest percentage of significantly correlated pixels (accounting for 50.2% of the vegetated areas) was found in September. In the Qinghai-Tibetan Plateau, the vegetation begins to wither in September (Xu et al. 2005), and water availability is no longer the dominant factor affecting vegetation development. This may be the reason for the relatively weak correlation between SCD and NDVI in September (Figure 6e).

Figure 6. Maximum positive correlation coefficient between monthly NDVI from (a) May to (e) September and five snow cover durations (September–November, December–February, March–May, December–May, and September–May). The histogram in (f) shows the percentage of vegetated areas with significant positive correlations between snow cover duration and NDVI (p < 0.1) for each month from May to September.

Similar to the distribution of positive correlation, the higher negative correlation between SCDs and May-September NDVI also occurred in June, July, and August (Figure 7). There was a significant correlation (p < 0.1) between SCD and NDVI for 23.3% of the vegetated areas in June (Figure 7b). The correlation between SCD and NDVI was significant (p < 0.1) for 26.0% of the vegetated area in July (Figure 7c). In August, SCD and NDVI were significantly correlated for 24.2% of the vegetated area (Figure 7d). Fewer areas were found to be significantly correlated with SCDs (p < 0.1) in May and September than that in June, July, and August, with 19.4% and 6.9% of the vegetated areas, respectively. The reasons seem to be the same as we pointed out for the positive correlation. Overall, the positive impacts of snow cover on the Qinghai-Tibetan Plateau overweigh its negative impacts. Therefore, June, July, and August were regarded as the most influenced months of NDVI by snow cover durations in the Qinghai-Tibetan Plateau.

Figure 7. Maximum negative correlation coefficient between monthly NDVI from (a) May to (e) September and five snow cover durations (September–November, December–February, March–May, December–May, and September–May). The histogram in (f) shows the percentage of vegetated areas with significant negative correlations between snow cover duration and NDVI (p < 0.1) for each month from May to September.

To further investigate the seasonal SCDs that had the dominant influence on vegetation growth, we examined the snow cover durations had the highest positive and negative correlations with NDVI in June, July, and August (Figure 8). Table 1 lists the percentages of vegetated areas with the significant correlations (p < 0.1) between different stages of snow cover durations and NDVI in June, July, and August. The most influencing period of SCD on June–August NDVI was from December to February for positive relationships and from March to May for negative relationships in the Qinghai-Tibetan Plateau.

Figure 8. Seasonal snow cover durations having the highest positive (a1–a3) and negative (b1–b3) correlations with NDVI in (1) June, (2) July, and (3) August.

Table 1. Percentages of vegetated areas with significant correlation (p < 0.1) between different stages of snow cover duration and NDVI in June, July, and August.

	Snow cover durations
	September–November (%)	December–February (%)	March–May (%)
June NDVI
Positive	15.26	21.93	14.93
Negative	0.03	8.68	14.36
July NDVI
Positive	13.34	19.42	13.08
Negative	0.01	13.35	12.64
August NDVI
Positive	16.00	26.23	14.33
Negative	0.01	11.17	12.98

Note: Bold values indicate p < 0.1.

The correlations between SCDs and NDVI vary among water source regions. In the southwest part of the study area (especially in the Southern Qiangtang and the Central Himalaya Mountains), snow cover in March–May had the most positive influence on NDVI in June and July (Figure 8a1–a2), while snow cover in December–February had more positive impact on NDVI in August in these regions (Figure 8a3). Similar phenomenon was found in the northern part of the Central and Western Kunlun Mountains, and the adjoining area of the Source of Three Rivers and Yigong Tsangpo. In contrast, the snow cover in March–May had most positive influence on NDVI in August and the snow cover in September–November and December–February mostly positively influenced NDVI in June and July in the Northern Qiangtang. In the northern part of the Source of Three Rivers, snow cover in September–November had the greatest positive impact on vegetation growth in June, July, and August. For those areas with longer SCDs, snow cover promotes the vegetation growth through increased water availability and the insulating effects of relatively deep snowpack on the soil (Grippa et al. 2005; Yan and Onipchenko 2006; Peng et al. 2010). In addition, longer SCDs usually indicate deeper snow depth that prevents both wind and low temperature from affecting soil temperature. Therefore, soil temperature below deep snow cover is generally higher than that in snow-free areas (Jones 1999; Walker et al. 1999; Groffman et al. 2001; Freppaz et al. 2008; Kaste et al. 2008), and thus, higher soil temperature could enhance the microbial activities that would benefit vegetation growth through increased soil respiration and nutrient availability (Monson et al. 2006) during the growing season. Moreover, deep snow cover protects vegetation against frost damage, dehydration, and physical damage from wind (Walker et al. 1999). Snow cover also moderates intensive and deep soil freezing and suppresses soil instability caused by frost and weathering.

Negative impacts of snow cover durations on vegetation in June, July, and August existed throughout the Qinghai-Tibetan Plateau. Snow cover in December–February and March–May mainly prevents vegetation growth (Figure 8b1–b3). For vegetation in June and August, snow cover in March–May had more negative influence than that in December–February. Long duration of snow cover in March–May may lead to late snow melting date, which delays vegetation green-up and leads to low accumulated temperature in summer and decreased vegetation growth. For July, snow cover in December–February was the main seasonal SCD that negatively affected vegetation growth.

Overall, the seasonal SCDs with significant impacts on vegetation growth vary across different regions in the study area. When SCDs in different seasons are considered, the impacts could be positive or negative. Based on our analysis, the area with maximum positive correlation coefficient (Figure 6) is much larger than that of maximum negative correlation coefficient (Figure 7). Therefore, the positive influence of SCDs on vegetation growth overweighs the negative influence in the plateau. However, when SCD is considered for the periods from December to May and from September to May, its impact on vegetation growth is offset due to both positive and negative impacts. The reason may be the opposite signs of the correlations for different seasonal SCDs. Another reason may be that only one or two specific seasonal SCDs have significant impact on vegetation growth for a certain pixel, but the relationship turns out to be insignificant when adding other seasonal SCDs.

5. Conclusion

Understanding the relationships between snow cover and vegetation activities is essential for estimating the potential effects of climate change in alpine regions. In this study, we examined the effects of snow phenology on vegetation phenology and greenness in the Qinghai-Tibetan Plateau. Our study shows that snow melting date is identified as an important factor affecting green-up date through its effect on PAR and temperature. Vegetation growth usually has strong responses to seasonal snow cover durations. Generally, the December–February and March–May SCDs play significantly roles in vegetation growth, both positively and negatively, depending on different water source regions. However, the positive influence overweighs the negative influence.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant 41271372) and the National Basic Research Program of China (973) (Grant No. 2009CB723906). The work by Lei Ji was performed under USGS contract G13PC00028. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government. We would like to thank the anonymous reviewers for valuable comments and suggestions for revising and improving the article.