Spatio-temporal dynamics of snow cover based on multi-source remote sensing data in China

2 Chinese Academy of Sciences, Cold and Arid Regions Environmental and Engineering Research Institute, Lanzhou 730000, China Correspondence to: Xiaodong Huang (huangxd@lzu.edu.cn) 10 Abstract. Through combining optical remote sensing snow cover products with passive microwave remote-sensing snow depth data, we produced a MODIS cloudless binary snow cover product and a 500-m spatial resolution snow depth product for December 2000 to November 2014. We used the synthesized products to analyze the temporal and spatial variation of the snow cover in China. The results indicated that in the past 14 years, the overall annual number of snow-covered days and average 15 snow depth in China increased. The annual average snow-covered area did not change significantly, and the number of snow-covered days in summer in China decreased. The number of snow-covered days in the winter, spring, and fall seasons all increased. The average snow-covered area in the summer and winter seasons decreased, whereas the average snow-covered area in the spring and fall seasons increased. The average snow depth in the winter, summer, and fall seasons decreased. Only the average 20 snow depth in spring increased. The spatial distribution of the increase and decrease in the annual average snow depth was highly consistent with that of the annual number of snow-covered days. The spatial distributions of the variation of the number of snow-covered days and the average snow depth of each season were also highly consistent. The regional differences in the snow cover variation in China were significant. The snow cover increased significantly in South and Northeast China, decreased 25 significantly in Xinjiang, increased in the southwest edge and southeast of the Tibetan Plateau, and mainly decreased in the north and northwest regions of the plateau.

the coarse resolution of passive microwave products greatly limits the accuracy of regional snow cover monitoring. Therefore, cloud removal and downscaling are effective approaches for enhancing the accuracy of snow cover monitoring using optical and passive microwave products, respectively.
This study used the MODIS daily snow cover product and passive microwave snow depth data to produce a daily cloudless snow-covered area product and a downscaled snow depth product with 105 500-m spatial resolution. We studied the temporal and spatial variation characteristics of the snow cover in China in the past 14 years from December 2000 to November 2014. We quantitatively analyzed the temporal and spatial variation characteristics of the number of snow-covered days, average snow depth, and snow-covered area. We revealed the variation of the snow cover in China under the background of climate change, and we provided the basis for further understanding of the 110 mechanism of interaction between climate change and the temporal and spatial variation of the snow cover in China.

Remote sensing snow products
The snow depth data used in this study were from the 'Environmental and Ecological Science Data Center for West China' (http://westdc.westgis.ac.cn), a database with a long time series of snow depth 130 The Cryosphere Discuss., doi:10.5194/tc-2016-124, 2016 Manuscript under review for journal The Cryosphere Published: 16 June 2016 c Author(s) 2016. CC-BY 3.0 License.
in China (1979China ( -2014 developed by Dr. Che (Che, 2008). It is a daily snow depth database inversed using the brightness and temperature data of the passive microwave remote sensing SMMR (1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987), SSM/I (1987SSM/I ( -2007, and SSMI/S (2008SSMI/S ( -2014. This product is saved in text format. The unit of snow depth is cm, and the spatial resolution is 25 km. Currently, the database is widely acknowledged and used (Dai et al., 2010;Wang et al., 2013;Bai et al., 2015). The snow-covered area 135 product includes the MOD10A1 and MYD10A1 binary snow cover products of the MODIS/Terra and MODIS/Aqua daily V005 version covering China. The data were from NSIDC. The spatial resolution is 500 m, and the time period is from December 2000 to November 2014.

Cloud removal and downscaling algorithms
Following the MODIS cloud removal algorithm developed by Dr. Huang, we produced the daily 140 cloudless binary snow cover data for December 2000to November 2014(Huang et al., 2014. The cloud removal algorithm can be summarized in 3 steps: (1) daily snow cover product synthesis: we combined the two products of MODIS/Terra and MODIS/Aqua; (2) adjacent day analysis: we replaced the cloud pixel on a given day with the pixel values on the days before and after under the cloudless condition; and (3) combination with the passive microwave snow depth product: we used the long time 145 series snow depth database of China to determine cloud pixels, completely reclassified the residual cloud pixels to continent or snow-cover pixels, and produced the MODIS daily cloudless binary snow cover images. Based on the downscaling algorithm for the AMSR-E snow water equivalent product by where SD sp is the 500-m spatial resolution snow depth value, SD i is the 25-km spatial resolution snow depth value in year i, SDY i is the 500-m spatial resolution annual number of snow-covered days in year 155 i calculated from the MODIS binary snow cover product, and SDT i is the sum of the annual number of snow-covered days of the 2500 MODIS pixels within a 25-km range in year i.

Analysis of the snow cover variation
The Mann-Kendall (M-K) method is a nonparametric test method widely used in the analysis of long time series of data (Helsel and Hirsch, 1992). This method monitors the variation of monotonic 160 nonlinear data. It has no requirement for data distribution, and it can avoid the interference of a few anomalies (Mcbean and Motiee, 2008). This study used the M-K method to analyze the trend and significance level of the number of snow-covered days and snow depth in the study region at a pixel scale. For a series i = ( 1 , 2 , … , ) with n samples, the test process is as follows: 165 where: where n is the year count (n = 14), m is the number of nodes (repetitive data groups) in the series, and t i 170 is the node width (the number of repetitive data points in the i th repetitive data group).
When n ≤ 10, we directly used the statistic S for the two-sided trend test. S > 0 represents an increase, S = 0 represents no variation, and S < 0 represents a decrease. At a | | ≥ 2 ⁄ , the series trend is significant; otherwise, it is insignificant.
When n > 10, the statistic S approaches the standardized normal distribution. We used the test statistic

175
Z for the two-sided trend test. Z > 0 represents an increase, Z = 0 represents no variation, and Z < 0 represents a decrease. At a given significance level , we looked up the critical 2 ⁄ in the normal We also used Sen's median method to analyze the slope of the variation of the annual number of snow-covered days. This method calculates the slope median of n(n-1)/2 pairs of combinations in a

Analysis of the snow cover variation in China
We used the M-K method to analyze the variation in the number of snow-covered days in the different seasons of winter (December-February next year, spring (March-May), summer (June-August), and fall (September-November) in the grid cells (Fig. 7). The results indicated that in the past 14 years, the regions with significantly decreased winter snow-covered days in China constituted 5.7% of the whole

270
China area, whereas the areas with significant increases constituted 7.2% of the study region ( Fig. 7(a)).
The regions with significantly decreased spring snow-covered days in China constituted 4.0% of the whole China area, whereas the regions with significant increases constituted 6.2% (Fig. 7(b)). The regions with significantly decreased summer snow-covered days in China constituted 3% of the whole China area, whereas the regions with significant increases constituted 2.9% (Fig. 7(c)). The regions 275 with significantly decreased fall snow-covered days in China constituted 1.8% of the whole China area, whereas the regions with significant increases constituted 5.7% (Fig. 7(d)). The results indicated that in

285
Previous studies indicated that low-elevation regions were susceptible to the influence of precipitation, whereas high-elevation regions were more susceptible to the influence of temperature (Xu et al., 2007).
As temperature rises, precipitation increases, leading to acceleration in snow melting rates in high-elevation regions and a decrease in snow-covered area. However, more moisture participates in the atmospheric water cycle process because of this pattern, which increases the precipitation in  In the past 14 years, the regions with significantly decreased winter snow depth constituted 10.6% of the area of China, whereas the regions with significant increases constituted 9.3% ( Fig. 8(a)). The regions with significantly decreased spring snow depth constituted 7.9% of the area of China, whereas the regions with significant increases constituted 9.8% ( Fig. 8(b)). The regions with significantly 305 decreased summer snow depth constituted 1.9% of the area of China, whereas the regions with significant increases constituted 0.9% (Fig. 8(c)). The regions with significantly decreased fall snow depth constituted 7.8% of the area of China, whereas the regions with significant increases only constituted 1.8% (Fig. 8(d)). Overall, the regions with significantly increased and decreased average snow depth in winter and spring in China were essentially the same. The regions with increased snow