Climate and topographic controls on snow phenology dynamics in the Tienshan Mountains, Central Asia

Climate and topographic controls on snow phenology dynamics in the Tienshan Mountains, Central Asia

Atmospheric Research 236 (2020) 104813 Contents lists available at ScienceDirect Atmospheric Research journal homepage: www.elsevier.com/locate/atmo...

5MB Sizes 0 Downloads 23 Views

Atmospheric Research 236 (2020) 104813

Contents lists available at ScienceDirect

Atmospheric Research journal homepage: www.elsevier.com/locate/atmosres

Climate and topographic controls on snow phenology dynamics in the Tienshan Mountains, Central Asia Yupeng Lia,b,c, Yaning Chena, Zhi Lia,

T



a

State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, China University of Chinese Academy of Sciences, Beijing 100049, China c Department of geography, Dartmouth college, NH 03220, USA b

A R T I C LE I N FO

A B S T R A C T

Keywords: Snow phenology Tienshan Mountains Climate change Snow cover

Although previous studies have characterized changes in seasonal snow cover or made predictions about snow cover in a changing climate, no comprehensive spatiotemporal analysis of snow phenology has been presented for the Tienshan Mountains, Central Asia. Relying on daily cloud-free snow cover fraction products originating from Moderate Resolution Imaging Spectroradiometer (MODIS) for 2002/03–2017/18, the snow phenology parameters (i.e., snow cover duration [SCD], snow onset date [SOD] and snow end date [SED]) are derived for each hydrological year within the study period and the characteristics analyzed for the Tienshan Mountains. The spatiotemporal changes of snow phenology have strong altitude dependence. The mean gradients of SCD, SOD, and SED with elevation are 6.0, −2.55, and 3.44 d/100 m, respectively. Because of differences in solar radiation and water vapor sources, the north-facing areas generally have a higher SCD, earlier SOD, and later SED than south-facing areas. Also, the trends of the snow phenology parameters at high and low altitudes show opposite changes. Consistent with the increase in snow cover area in recent years, SCD for the entire Tienshan region showed a clear uptick, especially in Northern Tienshan. This prolonged SCD was more related to advanced SOD than to SED, as decreased temperature and increased precipitation in autumn are conducive to snow accumulation.

1. Introduction

the number of snow cover duration (SCD) days, with a delayed snow onset date (SOD) and an advanced snow end date (SED). These phenomena have been well-documented in the literature and evaluated by satellite data and in-situ observations (Choi et al., 2010; Peng et al., 2013). For example, Choi et al. (2010) investigated the spatial and temporal patterns in the onset, offset, and length of the snow season across NH continents using a weekly snow cover dataset derived from NOAA satellite data. The results showed that full snow cover duration decreased by 5.3 d/decade from 1972 to 2007, which was mostly attributable to a progressively earlier snowmelt date (−5.5 d/decade). On a continental scale, the snow end date showed a rapid advance over Eurasia (−2.6 ± 5.6 d/decade) but virtually no trend over North America (0.1 ± 5.8 d/decade). Meanwhile, the SOD exhibited a trend towards later dates in Eurasia (1.3 ± 4.9 d/decade) and North America (1.1 ± 4.9 d/decade) (Peng et al., 2013). However, the response of snow cover to global warming exhibits significant differences across different regions. Recent research conducted by Chen et al. (2015) found contrasting changes in snow cover phenology observed in the northern middle latitudes, where snow cover duration was

Snow cover is a major component of the cryosphere and has relatively large intra-annual and inter-annual variations. The variability in both the extent and the amount of seasonal snow cover exerts a considerable influence on the world's climatological, hydrological, ecological and anthropological processes, resulting in equally variable social, economic and ecological consequences (Allchin and Déry, 2017; Brown et al., 2007; Wulder et al., 2007). Satellite measurements and surface observations provide evidence that the spatial extent of annual snow cover in the Northern Hemisphere (NH) has undergone significant reductions in recent decades (Brown and Robinson, 2011; Déry and Brown, 2007; Dye, 2002; Estilow et al., 2015). Climate projections suggest that this trend will continue and that snow cover extent in the NH will further decrease by 7 to 25% by the end of the 21st century, depending on climate scenarios (Alexander et al., 2013; Brown and Robinson, 2011). In response to the variability of snow cover extent, snow cover phenology has also experienced notable changes, such as a reduction in



Corresponding author at: Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, No.818 South Beijing Road, Urumqi 830011, China. E-mail addresses: [email protected] (Y. Li), [email protected] (Y. Chen), [email protected] (Z. Li).

https://doi.org/10.1016/j.atmosres.2019.104813 Received 26 February 2019; Received in revised form 9 December 2019; Accepted 19 December 2019 Available online 20 December 2019 0169-8095/ © 2019 Elsevier B.V. All rights reserved.

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

~2325 m a.s.l., with some elevations exceeding 3000 m a.s.l., while land cover types include grassland (58.7%), barren and sparsely vegetated regions (32.8%), cropland (3.7%), forest (0.4%) urban (0.25%), and scrublands (0.03%) (Li et al., 2019). The Tienshan Mountains are dominated by a typical temperate continental arid and semi-arid climate, which is characterized by temperature extremes in summer and winter. The characteristic aridity of the region is manifest in the surrounding deserts and dry regions. Based on mountain ranges, drainage and climate features, the Tienshan region is divided into four sub-regions: Northern Tienshan, Western Tienshan, Central Tienshan, and Eastern Tienshan. Most of the precipitation falls on the windward western and northwestern slopes, which are open to cold northerly and northwesterly air inflows and are also influenced by westerly moist influxes from the North Atlantic (Liu et al., 2017). Precipitation amounts vary from 710 to 790 mm at one extreme to 1500 to 2000 mm at the other. In the east and interior regions of the Tienshans, total precipitation decreases to between 200 and 400 mm due to the rain-shadow effect of the mountains.

increased by 9.74 days from 2001 to 2014, and in high latitudes, where snow cover duration was decreased by 5.57 days during the same time period. Changes in snow cover are not only closely related to climatic conditions (temperature and precipitation) but also vary with topography (elevation, aspect, and slope). These differences are related to variations in radiation and energy balances, and potentially to different accumulation regimes caused by windward/leeward effects (Tong et al., 2009). The snow cover trends can also depend on elevation. In Central Asia, the time of snow cover melting occurs later below 1800 m but earlier above 2500 m (Dietz et al., 2014). The role of temperature and precipitation in snow cover variations can also differ according to elevation bands. In the upper Heihe River Basin, for instance, a threshold altitude of 3650 ± 150 m was found by Bi et al. (2015). In the same study, temperature was discovered to be a negative factor controlling snow cover (except in the winter season below the threshold altitude), while precipitation acted as a positive factor (except in summer above the threshold altitude) (Bi et al., 2015). As the water tower of Central Asia and also an important ecological barrier, the Tienshan Mountains provide the source for the glacier and snow cover meltwater that supplies the major rivers in the area (e.g., Syr Darya River, Ili River, Tarim River and others) (Chen et al., 2016b; Tang et al., 2017). The average annual warming rate over the past halfcentury has been 0.34 °C/decade (Jiang et al., 2013), which is significantly higher than both the average global warming rate and that of the NH for the same period. However, the precipitation in the Tienshan Mountains remained stable or slightly increased during the same time frame (Chen et al., 2017). We anticipate that rapid warming could alter snowmelt water processes, accelerate the melting of snow, shift precipitation from snow to rain, and cause the mutual feedback of decreased surface albedo of snow and snowmelt water, thereby affecting the recharge of runoff and water resources (Chen et al., 2016b). Based on snow cover fraction data from the cloud-cleared Moderate Resolution Imaging Spectroradiometer (MODIS), the seasonal trend for snow cover extent during 2001–2015 was analyzed for the Tienshan Mountains by Tang et al. (2017). The results indicate that the mean snow cover area in summer and winter shows a decreasing rate of −0.016%/a and 0.114%/a, respectively (Tang et al., 2017). Chen et al. (2016a, 2016b) further pointed out that the maximum snow cover area decreased at a rate of −0.17%/a, with the largest decreases occurring in Central (−0.32%/a) and Eastern (−0.28%/a) Tienshan Mountains. Nevertheless, the study by Yang et al., 2008), based on observational data, showed that maximum snow depth deepened at a rate of 1.15 cm/ decade from 1959 to 2003, making it approximately 16% higher than the average during 1991–2003 (Yang et al. (2008). From the above analyses, we can see that previous studies mainly focused on changes in the extent and depth of snow cover, while spatial and temporal patterns of snow cover start and melt dates for the Tienshan Mountains have received less attention. Moreover, the combined effects of climate and topography on snow phenology have not yet been described in detail. In the present study, MODIS daily snow cover fraction products were used to produce daily cloud-free snow cover products over the Tienshan Mountains for each hydrological year (September 1st to the following August 31st) during 2002/03–2017/18. The spatial and temporal patterns of SCD, SOD, and SED in different elevation zones were analyzed and attributions of the driving factors of snow phenology anomalies were identified using climate data.

3. Data and methodology 3.1. Data sources Daily MODIS snow fraction cover data from September 1, 2002, to August 31, 2018, were downloaded from the Science Data Bank produced by Qiu et al. (2017). The snow cover fraction (SCF) data were generated from Normalized-Difference Snow Index data from MODIS Version 6, using traditional algorithms (Riggs et al., 2006). The snow cover mapping algorithm applied to the MODIS data exploits the strong reflectance in the visible (MODIS band 4 at 0.545–0.565 μm) and the strong absorption capacity in the short-wave infrared (MODIS band 6 at 1.628–1.652 μm) part of the spectrum by the Normalized-Difference Snow Index (Hall et al., 2002). Considering the geographic and climatic complexity, and meeting the production requirements of daily nearcloud-free (< 10%) data products in the high Asia region, temporal and spatial information on snow is used in cloud removal algorithms to realize the maximization of snow cover and minimization of cloud pollution. The SCF values are interpolated to the remaining cloudy pixels using the cubic spline interpolation algorithm based on observations (SCF values) of cloud-free days. Validated by in-situ observed SCD, MODIS-derived SCD shows a high consistency of 85.79% and a mean absolute error of 4.17 days (Tang et al., 2017), which ensures that this snow product can be confidently used to monitor snow dynamics, climatic environment, hydrological and energy balance, and disaster assessment. Snow cover variation is largely determined by snow season temperature and precipitation. Due to the scarcity of meteorological sites data in the Tienshan Mountains, a monthly temperature and precipitation dataset from 1960 to 2018 derived from the Climatic Research Unit time series (CRU TS v. 4.00) (http://www.cru.uea.ac.uk/ cru/data/hrg/) was used to detect the effect of changes in precipitation and temperature on snow dynamics. The dataset, which incorporated many process changes and contained all land areas (excluding Antarctica) at a 0.5° resolution, is based on records from the climate stations, which have been subjected to extensive manual and semi-automatic quality control measures. Newly acquired data records have been added, while poor quality or suspect records or values have been deleted (Harris et al., 2014). To explore the relationship between snow phenology and elevation variations, the Digital Elevation Model (DEM) data are derived from the Shuttle Radar Topography Mission website (http://srtm.csi.cgiar.org/) and feature an absolute vertical accuracy of < 16 m and an absolute level precision of < 20 m (Huang et al., 2014). The resolution of DEM (90 m) data was aggregated to 500 m to match the spatial resolution of the snow cover data (Huang et al., 2014).

2. Study area The Tienshan Mountains are located in the heart of the Eurasian continent within an area bounded by 66°E, 96°E, 37°N, and 46°N (Fig. 1). They stretch over 2500 km from west-southwest to eastnortheast and occupy a total area of around 591,350 km2, covering portions of Uzbekistan, Kyrgyzstan, south-eastern Kazakhstan, and the Xinjiang Uyghur Autonomous Region in China. The mean altitude is 2

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 1. Location of study area. (I) Eastern Tienshan (II) Northern Tienshan (III) Western Tienshan and (IV) Central Tienshan.

the time series of area-mean snow phenology indices and climatology indices. A t-test was performed to evaluate the significance level (pvalues). Additionally, the Mann-Kendall test for a trend with Sen's (1968) slope estimates was conducted on each series of snow statistics (SCF, SCD, SOD, and SED) to estimate trend magnitudes and their significance. This method does not assume a specific distribution of the data and is not sensitive to outliers.

3.2. Methodology 3.2.1. Snow cover phenology retrieval Based on the daily cloud-free snow product for each hydrological year during 2002/03–2017/18, the snow cover phenology parameters (SCD, SOD and SED) for each pixel were retrieved. SCD represents the total number of days with snow cover in a hydrological year, while SOD is defined as the date of the first snow cover day and SED as the date of the last snow cover day in a hydrological year. Among the methods proposed for determining the dates of both SOD and SED, few account for the fact that transient snowfall events always occur in the early fall and late spring. These transient snowfalls not only cause multiple beginning-dates and ending-dates of snow cover but skew the results of trend analyses, if failing to account for this phenomenon (Thompson, 2016). In the present study, the snow cover phenology parameters are determined using the algorithm described by Wang and Xie (2009), with slightly modified parameters. This method accounts for early- and late-season ephemeral snow-cover events. The results indicate that MODIS-derived SCD agree well (90%) with in situ SCD, and the SOD and SED maps also had good agreement with in-situ measurements with mean values of a 1-week forward shift and a 1-week backward shift, respectively (Wang and Xie, 2009). SCD is calculated as:

4. Results 4.1. Spatiotemporal variations in SCF Fig. 2 shows spatiotemporal variations in SCF across the entire Tienshan Mountains region. The 16-year annual mean SCF values were 31.1 ± 2.3% and the snow cover distribution was spatially uneven due to the high topography and complex terrain in combination with atmospheric circulation. The spatial distribution of SCF showed a general reduction from the high mountains to the low mountains and from northwest to southeast. The overall trend of SCF across the entire region showed a non-significant increase (0.07%/a, p > .05). Furthermore, areas with decreasing and increasing trends accounted for 52.7% and 47.3%, respectively, while areas showing significant decreases occupied about 8.2% and were mainly distributed to the east of the Eastern Tienshan sub-region and south of the Central Tienshan sub-region. The areas of significant increase were scattered, accounting for about 3.6%. Seasonally, the snow cover reached the maximum extent in January (66.9 ± 6.2%) and then progressively decreased as the snow melted, reaching the lowest extent in August (2.4 ± 0.3%). Both the maximum and minimum snow cover indicated a declining trend at a rate of 0.62%/a (p < .05) and 0.04%/a (p > .05), respectively. Detailed analysis of the inter-annual variability of SCF determined by the Mann-Kendall test revealed an increase in SCF from October to the following May (except for January) and a decrease in SCF from June to September.

n

SCD =

∑ H (Di − 50) 0

where n is the total number of days (images) within a year, Di is the snow cover fraction (%) in a pixel (0 ≤ Di ≤ 100), and the Heaviside function equals 0 (1) for negative (positive) arguments. SOD is calculated as:

SOD = D1 − SCD1′ where SCD1′ represents snow-covered days within the period from September 1 to January 20 of the following year. In this study, D1 is the Julian day of January 20, when snow cover reaches the maximum extent in the Tienshan Mountains. SED is calculated as:

4.2. Spatial changes in snow cover phenology Fig. 3 illustrates the spatiotemporal variations of snow cover phenology over the Tienshan Mountains and its four sub-regions. The mean annual SCD, SOD and SED values from MODIS for the extensive snowcovered regions during the 16-year period were 113.9 ( ± 8.6), 331.2 ( ± 4.5), and 81.1 ( ± 6.0) in day of year (DOY), respectively. As can be seen in Fig. 3, SCD is clearly associated with altitude. Mountainous regions such as the Bogda Mountains, the Tormti Mountains, and the Alatao Mountains stand out due to long SCD, early SOD, and late SED. In addition, valleys like the Yourdusi Basin are often characterized by much lower SCD than the surrounding ranges. That the regional snow cover in the Northern and Western Tienshan persists longer, starts earlier, and ends later than in the Eastern and Central Tienshan is likely due to the northern slope of the Tienshan Mountains being the

SED = D2 + SCD2′ where SCD2′ indicates snow-covered days within the period from January 21 to August 31 of the same year. In this study, D2 is the Julian day of January 21. 3.2.2. Statistical analyses To analyze how climate controls snow cover phenology, we calculated the partial correlation coefficient between SOD and autumn (September to November) temperature (Ta) and precipitation (Pa), as well as SED and spring (March to May) temperature (Tm) and precipitation (Pm). The partial correlation coefficient R was computed from 3

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 2. Depictions of (a) annual-mean, (b) trend and (c) corresponding significance levels of SCF on a pixel scale for the Tienshan Mountains from 2002/03–2017/18 are shown. The red and blue colors in (c) indicate changes significant at the 0.05 level. Depictions of (d) the annual cycle of SCF, (e) the time series of maximum and minimum SCF, and (f) monthly trend of SCF are also shown (Unit: %/a). * indicates a significant change at the 95% statistical confidence level. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

delayed SOD and advanced SED were observed. For example, in Eastern Tienshan, SOD was postponed 0.14 d/a and SED started earlier −0.12 d/a, resulting in a shorter snow period. Meanwhile, a large area (53.9%) showed increasing SCD (slope > 0), with the increase significant in 3.6% of that zone. Of particular note were the increasing rates (at above 4 days per year) in the Yourdusi Basin and the increased SCD in Northern and Western Tienshan, the latter of which appeared to be more related to earlier SOD than delayed SED. In Northern Tienshan, SCD increased by 1.14 d/a (p < .05), which was related to earlier SOD

windward slope, uplifting the westerly airflow and causing abundant precipitation. In contrast, the southern slope is dominated by a descending air current, resulting in much less precipitation. For the entire Tienshan region, SCD showed a non-significant increasing trend at a rate of 0.31 day per year (p > .05), associated with both earlier SOD (−0.25 d/a, p > .05) and SED (−0.001 d/a, p > .05) (Table 1). However, spatial patterns for SCD, SOD and SED changes were heterogeneous. SCD decreased (slope < 0) across 38.6% of the region, primarily south of Central and Eastern Tienshan, where

Fig. 3. Climatological annual-mean and changes of duration (a, b), start (d, e), and melt (g, h) of snow cover over the Tienshan Mountains in 2002/03–2017/18 are also shown, as are the significance levels denoted by (c), (f) and (i). 4

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

the change of SCD. In addition to elevation, aspect also can affect snow phenology by altering local solar radiation and moisture conditions, as well as potentially the accumulation regime, due to windward/leeward effects. The effect of aspect is shown in Fig. 5. The north-facing areas generally have a higher SCD, earlier SOD, and later SED than south-facing areas. Meanwhile, the increase of SCD in the north-facing areas is greater than that in the south-facing areas, which can be explained by the stronger decrease of SOD and increase of SED in the north-facing areas than that in the south-facing areas. As we know, south-facing aspects receive more solar radiation, which often enhances snowmelt between storms, resulting in less accumulation on these aspects. However, snow cover on the north-facing areas receives less insolation and thus melts slower than over south-facing areas. Besides, most of the precipitation falls on the windward western and northwestern slopes also conducive to snow accumulation. In order to study the response of snow cover phenology to climate change, we used partial correlation to analyze the relationship between SOD and autumn temperature and precipitation at the pixel level. We also analyzed the relationship between SED and spring temperature and precipitation using the same method. As shown in Fig. 6, 13.7% of the pixels have a significant positive correlation between SOD and Ta across the entire study area, with an area-mean correlation of R = 0.57. The four sub-regions of the Tienshan Mountains all exhibit a positive correlation between SOD and Ta, with Northern Tienshan showing the largest correlation (Table 2). This means that the higher the temperature, the later the starting time for snow accumulation. The correlation also revealed that autumn precipitation likely controls SOD in several regions (such as high mountain areas in Eastern Tienshan and some lowlands in Western Tienshan), with about 8.5% of the pixels having a significant negative correlation between SOD and Pa. SED was also more controlled by temperature than precipitation in spring. The area-mean correlation coefficient between SED and Tm is −0.46, while the area-mean correlation between SED and P m is only −0.05. SED was significantly negatively correlated with Tm in about 16.4% of the entire region, and areas with significant positive correlation only account for 0.2%. From the above analysis, we can see that temperature plays a more important role than precipitation in SOD and SED changes.

Table 1 Changes in SCD, SOD and SED for the Tienshan Mountains and its four subregions (Unit: d/a). Regions

SCD

SOD

SED

Whole Region Eastern Tienshan Northern Tienshan Western Tienshan Central Tienshan

0.31 −0.16 1.14⁎ 0.40 −0.12

−0.25 0.14 −0.66 −0.55 0.11

−0.001 −0.12 0.37 0.06 0.06

Note:



indicate statistical significance at the 0.05 level.

(−0.66 d/a, p > .05) and postponed SED (0.37 d/a, p > .05).

4.3. The effect of climate and topography on snow phenology The snow phenology over the Tienshan Mountains is closely associated with elevation. To further quantitatively describe this relationship, the 16-year-mean values and trends of SCD, SOD and SED during 2002/03–2017/18 were calculated in Fig. 4 as a function of elevation, for elevation bins of 100 m. Due to the fact that areas above 4000 m may be affected by all-year snow cover (glaciers), these areas are excluded from our consideration to avoid spurious results related to mixing of seasonal snow cover and glaciers. Below 4000 m, SCD becomes longer, SED later, and SOD earlier as the altitude increases. The mean gradients of SCD, SOD and SED with elevation are 6.0, −2.55, and 3.44 d/100 m, respectively. Trend analysis shows that SCD decreases below 1500 m a.s.l., but increases between 1500 and 4000 m a.s.l., indicating SCD at high and low altitudes show a contrasting response to climate change. Besides, below 1500 m a.s.l., with the decrease of altitude, the decreasing trend of SCD increases. Over 1500 m a.s.l., the increase of SCD reaches its maximum at 2500–3000 m a.s.l.. This characteristic is consistent with the change of SOD at different altitudes, that is, SOD postpones below 1500 m a.s.l., but SOD advances over 1500 m a.s.l.. The response of SED to climate change at different altitudes is generally similar to that of SCD except at 1500–3000 m a.s.l., in which the change of SED is subtle and the effect of the SED on SCD is also much smaller than that of SOD. The above analysis further confirms that the change of SOD dominates

Fig. 4. Depiction of (a) mean and trend of duration (b), start (c), and melt (d) of snow cover as a function of elevation in the Tienshan Mountains during 2002/ 03–2017/18. 5

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 5. The mean of SCD, SOD and SED at different aspects for four subregions of the Tienshan Mountains (left panel) and the trend of SCD, SOD and SED at different aspects for the whole Tienshan Mountains (right panel).

3000 m a.s.l.. The relationship between SED and Tm was negative at all altitudes with significant relationships occurring at 2000–2500 m a.s.l.. Furthermore, SED positively correlated with Pm in almost all elevation zones, with the exception of elevations below 2000 m a.s.l.. At the same time, the trend of temperature and precipitation in spring and autumn at different altitudes was also analyzed. Thus, in autumn, temperature decreases and precipitation increases at all altitudes, while in spring, temperature increases and precipitation decreases.

To further understand the mechanism of climate impacts on snow phenology, temperature and precipitation trends during autumn and spring were calculated based on the CRU series dataset from 2002/ 03–2017/18 (Fig. 7). Against the backdrop of global warming, the Ta and Tm show a contrasting change during 2002/03–2017/18, with a cooler Ta (−0.06 °C/a) and a warmer Tm (0.04 °C/a). The precipitation trends during autumn and spring also show contrasting changes, albeit weak ones. Pa increased at 0.01 mm/a, while Pm decreased at 0.17 mm/ a. This change in climate explains both the advance in snow start time and ablation time for the Tienshan Mountains. Decreased temperatures and increased precipitation are conducive to snow accumulation and thus resulted in advanced SOD throughout the region. Meanwhile, advanced SED is only found in East Tianshan. The significant negative correlation between SED and spring temperature, indicating warming during the melt season is the main reason leading to earlier snow melting in this area. To further investigate the spatial pattern of climatic factors affecting SOD and SED, Fig. 8 shows an analysis of these two parameters in relation to elevation variations. The relationship between SOD and temperature was positive at all altitudes, with significant relationships occurring at 2000–3500 m a.s.l.. Conversely, SOD was generally negatively correlated with precipitation, except for elevations below 1500 m a.s.l. It is worth mentioning that the relationships between SOD and precipitation and temperature were strongest at about

5. Discussion and conclusion Relying on daily cloud-free snow cover fraction products originating from MODIS, the snow phenology parameters (SCD, SOD, and SED) were derived for each hydrological year from 2002/03–2017/18. The characteristics of snow phenology were then analyzed and attributions of the driving factors of snow phenology anomalies were identified. 5.1. Comparisons with previous studies In contrast to reductions in snow cover experienced by most of the NH, snow cover in the Tienshan Mountains showed a subtle non-significant increase during 2002/03–2017/18. This unique change is nonetheless consistent with results found by Chen et al. (2016a), whose study revealed that mainland China displayed an overall significant 6

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 6. Partial correlation between (a) SOD and autumn temperature (mean value for September–November) and (b) areas of significant correlation (p < .05); (c) SOD and autumn precipitation (sum for September–November) and (d) areas of significant correlation; (e) SED and spring temperature (mean value for March–May) and (f) areas of significant correlation; and (g) SED and spring precipitation (sum for March–May) and (h) areas of significant correlation.

Snow cover duration has also been correspondingly prolonged over the Tienshans, indicating that increased SCD is more related to an advanced SOD than an almost unchanged SED. This is consistent with changes observed in latitudes lower than 40°N, where SOD advanced 2.70 ( ± 1.97) days (2001–2014) (Chen et al., 2015). Similar results of earlier SOD were reported in the Tibetan Plateau (Chen et al., 2018a, 2018b) and Central Asia (Dietz et al., 2014) as well, with Chen et al. (2018a, 2018b) positing that earlier SOD accounted for 67.7% of increase in SCD in the Tibetan Plateau between 2001 and 2014, while Dietz et al. (2014) revealed a shift towards earlier snow cover start in Central Asia, especially in high elevation regions. A recent study conducted by Yang et al. (2019) used passive microwave satellite data to derive their snow depths and snow phenology over Tienshan Mountains from 1979 to 2016. This research indicated that SCD and SED across the Tienshan experienced a significant decrease, and that SED dominated the variability of the snow season. Considering the different research periods, our conclusions do not conflict with Yang et al.'s (2019) research. On the contrary, the snow phenology for the study region occurs precisely in response to climate change. Our results showed that decreasing temperatures and increasing precipitation in autumn are conducive to snow cover

Table 2 Partial Correlation Between SOD (SED) and autumn (spring)temperature and precipitation in the Tienshan Mountains and four subregions. Regions Whole Region Eastern Tienshan Northern Tienshan Western Tienshan Central Tienshan

Note:



RSOD-Ta ⁎

0.57 0.35 0.62⁎ 0.45 0.40

RSOD-Pa

RSED-Tm

RSED-Pm

−0.13 −0.51 −0.26 −0.42 0.05

−0.46 −0.68* −0.55 −0.52 −0.34

−0.05 −0.42 −0.05 −0.21 0.08

indicate statistical significance at the 0.05 level.

increasing trend of 0.29% decade−1 from 1982 to 2013. Recent research conducted by Tang et al. (2017) indicated that snow cover in the Tienshan Mountains during 2001–2015 showed an increasing trend in spring and autumn but a decreasing one in summer and winter. However, both the maximum and minimum snow cover showed a declining trend. The decline in minimum snow cover, represented by glaciers and permanent snow cover, was in line with recent reports regarding substantial glacier shrinkage in the Tienshan Mountains (Chen et al., 2016b). 7

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 7. Changes in autumn temperature (Ta) and spring temperature (Tm) during 1960/61–2017/18 (a, d) and 2002/03–2017/18 (b, e), and corresponding time series of Ta (c) and Tm (f). Changes in autumn precipitation (Pa) and spring precipitation (Pm) during 1960/61–2017/18 (g, h) and 2002/03–2017/18 (j, k), and corresponding time series of Pa (i) and Pm (l).

8

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

Fig. 8. Partial correlation coefficient between (a) snow onset date and autumn temperature and precipitation, and partial correlation coefficient between (b) snow end date and spring temperature and precipitation in different elevations. Trend of (c) autumn temperature and spring temperature, and (d) autumn spring precipitation and spring precipitation for 2002/03–2017/18 in different elevations. The solid symbols indicate that the correlations are significant at 95% confidence level.

change in SCD is highly dependent on altitude. At the same time, trend analysis shows that SCD at high and low altitudes show a contrast respond to climate change with SCD decreases below 1500 m a.s.l. but increases above 1500 m a.s.l.. This difference may come from the temperature in high altitude rise faster than that at low altitudes (Fig. 8). Besides, the average temperature at high altitudes is relatively low, and the recent warming does not make the average temperature higher than 0 °C. The change of SOD and SED also shows different at high and low altitudes, which not unique in Tienshan Mountain but also reported in Central Asia. Andreas et al. (2014) found that later snow cover melts below 2100 m a.s.l. and above 4600 m a.s.l. and earlier snow cover melts between 2100 and 4600 m a.s.l.. Moreover, the effect of aspect on snow phenology also is analyzed. In general, the northfacing areas show a higher SCD, earlier SOD, and later SED than those in south-facing areas. It's easy to understand because the south slope is

maintenance and thus prolong the duration of the snow. It should be pointed out that the advance of snow melting in the East Tienshan and the postponement of snow melting in the North Tienshan cause little change in the time of snow melting in the whole Tienshan Mountains.

5.2. Possible mechanisms for recent snow phenology changes Topography has a major effect on weather and climate in the Tienshan Mountains, and elevation and aspect, therefore, play important roles in the snow phenology distribution. For SCDs bellowing 4000 m a.s.l., they exhibit a strong linear relationship with elevation, 6 days for each 100 m increase. This change is stronger than the findings in an alpine watershed of western Canada (4.3 days/100 m for the snowmelt season) and Tizinafu watershed located in the northern slope of the western Kunlun Mountains (She et al., 2015), indicating that the 9

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

(Barnett et al., 2005). Such a shift would inevitably influence the accumulation and melting processes of snow and glaciers and thus further affect the streamflow and terrestrial total water storage in the Tienshan Mountains. Given these potentialities for rapid and significant changes in hydrological processes, it is important to investigate the impacts and mechanisms of this precipitation shift on runoff, which is the focus of our ongoing work. In addition, snow phenology variations have critical consequences for vegetative processes in the biosphere, as vegetation and snow are tightly linked in cold biomes, especially the beginning of growth and snowmelt (Klein et al., 2016). This also represents a crucial area for further research.

sunny and thus it receives more solar radiation energy. Also, since western and northwestern slopes are open to cold northerly and northwesterly air inflows and are also influenced by westerly moist influxes from the North Atlantic, most of the precipitation falls on north-facing areas. Our results also show that the effects of temperature and precipitation on snow cover phenology vary at different altitudes. The positive relationship between SOD and autumn temperatures and the negative relationship between SED and spring temperatures at different altitudes is easy to understand, given that decreasing temperatures enable snowfall in autumn while increasing temperatures cause snowmelt in spring. It is noteworthy that precipitation is a negative factor below 2000 m in spring. This may be because temperatures at low altitudes are typically higher than 0 °C, so the warmer rains would wash away any remaining snow cover and accelerate snowmelt (Bi et al., 2015). The increase of SCD and earlier SOD in the Tienshan Mountains are related to a cooler Ta in recent years, which is beneficial for snow accumulation and early snowfall. Decreased Ta was also observed in the Tibetan Plateau (Chen et al., 2018) and mainland China (Chen et al., 2016a), which is essentially consistent with the long-term tendency of large-scale cooling trends of land surface temperature in winter over mid-latitudes, observed since the 1990s (Cohen et al., 2017). The Ta decline in mid-latitude was attributed to changes in the Arctic system through changes in storm tracks, the jet stream, and planetary waves and their associated energy propagation (Cohen et al., 2017). Moreover, the recent Pacific Ocean cooling effect in winter has also been proven to be a major driving factor of a decline in Ta over northern midlatitudes (Kosaka and Xie, 2013). Accompanied by the decline in Ta, Pa likewise increased across most of the Tienshan region, which coincides with the earlier findings of large-scale cold snaps and heavy snowfall in high to middle latitudes of the NH during boreal winter (Cohen et al., 2012). Previous studies concluded that with the strengthened North America Subtropical High and West Pacific Subtropical High, more water vapor was brought to the Tienshan Mountains (Li et al., 2016; Yao et al., 2012).

The authors declared that they have no conflicts of interest to this work. We declare that we do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.

5.3. Limitations and future work

Acknowledgments

Compared with snow phenology distribution obtained from groundbased meteorological records, satellite-based observations have the advantage of large spatial coverage and high temporal resolution. Although this study provides a fine-resolution analysis of the snow phenology for the Tienshan Mountains, sixteen years is still too short a time frame to fully understand variations in snow cover and amount under climate change. Limited by incomplete spatial coverage, low spatial resolution, and short time span of the current snow cover products, the long time series of snow products across the Tienshan Mountains are urgently needed to investigate the snow response to climate change. In order to achieve this goal, recent research provides ideas for the realization of long time series snow products in Tianshan region by integrating snow cover extent from the Advanced Very HighResolution Radiometer (AVHRR) surface reflectance climate data record and or existing snow cover data sets. For example, Using AVHRR surface reflectance climate data record and several existing snow cover products, Chen et al. (2018) generates a composite daily SCE record (5 km) of the TP from 1981 to 2016. This will be one of our next steps. Moreover, in snow-dominated regions, changes in snow cover can have serious implications for hydrological regimes. Increased snowmelt due to higher temperatures can increase runoff, but if glaciers and snow continue to decrease, water storage will also decrease. Within a few decades or less, this may lead to water shortages during dry seasons (Chen et al., 2016b). Meanwhile, earlier melting times can change the seasonal distribution of runoff, leading to intensified conflicts between water supply and demand and increasing the frequency of the occurrence of drought and floods. Another anticipated change in a warming future is a temperature-induced shift of precipitation from snow to rain

The research is supported by, the National Natural Science Foundation of China (41701024; 41630859), the Youth Innovation Promotion Association of the Chinese Academy of Sciences. Youth Innovation Promotion Association of the Chinese Academy of Sciences (No. 2018480).

Author statement I have drafted the work or revised it critically for important intellectual content; AND I have approved the final version to be published. I agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Author contributions Dr. Yupeng Li and Zhi Li conceived and wrote the main manuscript text, Prof. Yaning Chen put forward valuable suggestions for this article. Declaration of Competing Interest

References Alexander, L., Allen, S., Bindoff, N.L., 2013. Climate Change 2013: The Physical Science Basis-Summary for Policymakers. Intergovernmental Panel on Climate Change. Allchin, M., Déry, S., 2017. A spatio-temporal analysis of trends in Northern Hemisphere snow-dominated area and duration, 1971–2014. Ann. Glaciol. 58, 21–35. https://doi. org/10.1017/aog.2017.47. Andreas, D., Christopher, C., Claudia, K., Gerhard, G., Stefan, D., 2014. Identifying changing snow cover characteristics in Central Asia between 1986 and 2014 from remote sensing data. Remote Sens. 6, 12752–12775. https://doi.org/10.3390/ rs61212752. Barnett, T.P., Adam, J.C., Lettenmaier, D.P., 2005. Potential impacts of a warming climate on water availability in snow-dominated regions. Nature 438, 303. https://doi.org/ 10.1038/nature04141. Bi, Y., Xie, H., Huang, C., Ke, C., 2015. Snow cover variations and controlling factors at Upper Heihe River Basin, Northwestern China. Remote Sens. 7, 6741–6762. https:// doi.org/10.3390/rs70606741. Brown, R.D., Robinson, D.A., 2011. Northern hemisphere spring snow cover variability and change over 1922-2010 including an assessment of uncertainty. Cryosphere 5, 219–229. https://doi.org/10.5194/tc-5-219-2011. Brown, R., Derksen, C., Wang, L., 2007. Assessment of spring snow cover duration variability over northern Canada from satellite datasets. Remote Sens. Environ. 111, 367–381. https://doi.org/10.1016/j.rse.2006.09.035. Chen, X., Liang, S., Cao, Y., He, T., Wang, D., 2015. Observed contrast changes in snow cover phenology in northern middle and high latitudes from 2001-2014. Sci. Rep. 5, 16820. https://doi.org/10.1038/srep16820. Chen, X., Liang, S., Cao, Y., He, T., 2016a. Distribution, attribution, and radiative forcing

10

Atmospheric Research 236 (2020) 104813

Y. Li, et al.

cooling. Nature 501, 403. https://doi.org/10.1038/nature12534. Li, B., Chen, Y., Chen, Z., Xiong, H., Lian, L., 2016. Why does precipitation in northwest China show a significant increasing trend from 1960 to 2010? Atmos. Res. 167, 275–284. https://doi.org/10.1016/j.atmosres.2015.08.017. Li, Y., Chen, Y., Li, Z., 2019. Developing daily cloud-free snow composite products from MODIS and IMS for the Tienshan Mountains. Earth Space Sci. 6, 266–275. https:// doi.org/10.1029/2018EA000460. Liu, J., Zhang, W., Liu, T., 2017. Monitoring recent changes in snow cover in Central Asia using improved MODIS snow-cover products. J. Arid Land 9, 763–777. https://doi. org/10.1007/s40333-017-0103-6. Peng, S., Piao, S., Philippe, C., Pierre, F., Zhou, L., Wang, T., 2013. Change in snow phenology and its potential feedback to temperature in the Northern Hemisphere over the last three decades. Environ. Res. Lett. 8, 014008. https://doi.org/10.1088/ 1748-9326/8/1/014008. Qiu, Y., Wang, X., Han, L., Chang, L.., Shi, L., 2017. Daily Snow Cover Dataset in High Asia. Science Data Bankhttps://doi.org/10.11922/sciencedb.457. Riggs, G.A., Hall, D.K., Salomonson, V.V., 2006. MODIS Snow Products User Guide to Collection. pp. 5. http://modis-snow-ice.gsfc.nasa.gov/?c=userguides. Sen, P.K., 1968. Estimates of the regression coefficient based on Kendall's tau. J. Am. Stat. Assoc. 63, 1379–1389. She, J., Zhang, Y., Li, X., Feng, X., 2015. Spatial and temporal characteristics of snow cover in the Tizinafu Watershed of the Western Kunlun Mountains. Remote Sens. 7, 3426–3445. https://doi.org/10.3390/rs70403426. Tang, Z., Wang, X., Wang, J., Wang, X., Li, H., Jiang, Z., 2017. Spatiotemporal variation of snow cover in Tianshan Mountains, Central Asia, Based on Cloud-Free MODIS Fractional Snow Cover Product, 2001-2015. Remote Sens. 9, 1045. https://doi.org/ 10.3390/rs9101045. Thompson, J.A., 2016. A MODIS-derived snow climatology (2000-2014) for the Australian Alps. Clim. Res. 68, 25–38. https://doi.org/10.3354/cr01379. Tong, J., Déry, S.J., Jackson, P.L., 2009. Topographic control of snow distribution in an alpine watershed of western Canada inferred from spatially-filtered MODIS snow products. Hydrol. Earth Syst. Sc. 13, 319–326. https://doi.org/10.5194/hessd-52347-2008. Wang, X., Xie, H., 2009. New methods for studying the spatiotemporal variation of snow cover based on combination products of MODIS Terra and Aqua. J. Hydrol. 371, 192–200. https://doi.org/10.1016/j.jhydrol.2009.03.028. Wulder, M.A., Nelson, T.A., Derksen, C., Seemann, D., 2007. Snow cover variability across Central Canada (1978–2002) derived from satellite passive microwave data. Clim. Chang. 82, 113–130. https://doi.org/10.1007/s10584-006-9148-9. Yang, Q., Cui, C., Sun, C., Ren, Y., 2008. Snow cover variation in the past 45 years (19592003) in the Tianshan Mountains, China. Advance in Climate Change Research 4, 13–17. Yang, T., Li, Q., Ahmad, S., Zhou, H., Li, L., 2019. Changes in Snow Phenology from 1979 to 2016 over the Tianshan Mountains, Central Asia. Remote Sens. 11, 1–16. https:// doi.org/10.3390/rs11050499. Yao, T., Thompson, L., Yang, W., Yu, W., Gao, Y., Guo, X., Yang, X., Duan, K., Zhao, H., Xu, B., 2012. Different glacier status with atmospheric circulations in Tibetan Plateau and surroundings. Nat. Clim. Chang. 2, 663–667. https://doi.org/10.1038/ nclimate1580.

of snow cover changes over China from 1982 to 2013. Clim. Chang. 137, 363–377. https://doi.org/10.1007/s10584-016-1688-z. Chen, Y., Li, W., Deng, H., Fang, G., Li, Z., 2016b. Changes in Central Asia’s water tower: past, present and future. Sci. Rep. 6, 35458. https://doi.org/10.1038/srep35458. Chen, Y., Li, Z., Fang, G., Deng, H., 2017. Impact of climate change on water resources in the Tianshan Mountians, Central Asia. Acta Geol. Sin. 72, 18–26. https://doi.org/10. 11821/dlxb201701002. Chen, X., Long, D., Hong, Y., Hao, X., Hou, A., 2018a. Climatology of snow phenology over the Tibetan plateau for the period 2001-2014 using multisource data. Int. J. Climatol. 38, 2718–2729. https://doi.org/10.1002/joc.5455. Chen, X., Long, D., Liang, S., He, L., Zeng, C., Hao, X., Hong, Y., 2018b. Developing a composite daily snow cover extent record over the Tibetan Plateau from 1981 to 2016 using multisource data. Remote Sens. Environ. 215, 284–299. https://doi.org/ 10.1016/j.rse.2018.06.021. Choi, G.Y., Robinson, D.A., Kang, S.K., 2010. Changing Northern Hemisphere snow seasons. J. Clim. 23, 5305–5310. https://doi.org/10.1175/2010JCLI3644.1. Cohen, J.L., Furtado, J.C., Barlow, M.A., Alexeev, V.A., Cherry, J.E., 2012. Arctic warming, increasing snow cover and widespread boreal winter cooling. Environ. Res. Lett. 7, 014007. https://doi.org/10.1088/1748-9326/7/1/0140070. Cohen, J., Screen, J.A., Furtado, J.C., Barlow, M., Whittleston, D., Coumou, D., Francis, J., Dethloff, K., Entekhabi, D., Overland, J., 2017. Recent Arctic amplification and extreme mid-latitude weather. Nat. Geosci. 7, 627–637. https://doi.org/10.1038/ NGEO2234. Déry, S.J., Brown, R.D., 2007. Recent Northern Hemisphere snow cover extent trends and implications for the snow-albedo feedback. Geophys. Res. Lett. 34, 60–64. https:// doi.org/10.1029/2007GL031474. Dietz, A., Conrad, C., Kuenzer, C., Gesell, G., Dech, S., 2014. Identifying changing snow cover characteristics in Central Asia between 1986 and 2014 from remote sensing data. Remote Sens. 6, 12752–12775. https://doi.org/10.3390/rs61212752. Dye, D.G., 2002. Variability and trends in the annual snow-cover cycle in Northern Hemisphere land areas, 1972–2000. Hydrol. Process. 16, 3065–3077. https://doi. org/10.1002/hyp.1089. Estilow, T.W., Young, A.H., Robinson, D.A., 2015. A long-term Northern Hemisphere snow cover extent data record for climate studies and monitoring. Earth Syst. Sci. Data 7, 137–142. https://doi.org/10.5194/essd-7-137-2015. Hall, D.K., Riggs, G.A., Salomonson, V.V., DiGirolamo, N.E., Bayr, K.J., 2002. MODIS snow-cover products. Remote Sens. Environ. 83, 181–194. Harris, I., Jones, P.D., Osborn, T.J., Lister, D.H., 2014. Updated high-resolution grids of monthly climatic observations-the CRU TS3.10 Dataset. Int. J. Climatol. 34, 623–642. https://doi.org/10.1002/joc.3711. Huang, X., Hao, X., Feng, Q., Wang, W., Liang, T., 2014. A new MODIS daily cloud free snow cover mapping algorithm on the Tibetan Plateau. Sciences in Cold and Arid Regions 6, 0116–0123. https://doi.org/10.3724/SP.J.1226.2014.00116. Jiang, Y., Chen, Y., Zhao, Y., Chen, P., Yu, X., Fan, J., Bai, S., 2013. Analysis on changes of basic climatic elements and extreme events in Xinjiang, China during 1961–2010. Adv. Clim. Chang. Res. 4, 20–29. https://doi.org/10.3724/SP.J.1248.2013.020. Klein, G., Vitasse, Y., Rixen, C., Marty, C., Rebetez, M., 2016. Shorter snow cover duration since 1970 in the Swiss Alps due to earlier snowmelt more than to later snow onset. Clim. Chang. 139, 1–13. https://doi.org/10.1007/s10584-016-1806-y. Kosaka, Y., Xie, S., 2013. Recent global-warming hiatus tied to equatorial Pacific surface

11