Snow, Ice and Environment Over the Tibetan Plateau Zone II Versions EN3 Vol 2 (2) : 0 2017
Download
Daily fractional snow cover dataset over High Asia
 >>
: 2017 - 07 - 11
: 2017 - 07 - 31
: 2017 - 09 - 25
4333 25 0
Abstract & Keywords
Abstract: High Asia, defined as the high altitude area of Asia (including mainly the Tibetan Plateau), is an important area of snow distribution in low- and mid-latitude regions. The dynamic changes of snow in this area have notable effects on water, energy balance and regional climate. As the seasonal snow in the Tibetan Plateau is both typically shallow and exhibits short duration, it is necessary to understand changes in snow cover on a daily time scale. This article describes a novel snow cover dataset named “daily fractional snow cover (FSC) data set over High Asia (2002 – 2018)”. The daily snow cover data of this data set are derived from the MODIS normalized snow index data with a spatial resolution of 500 meters. Data processing involves the application of additional terrain data and a variety of snow cover cloud elimination algorithms. The new dataset provides a cloud-free estimate (cloud cover < 10%) of FSC over High Asia, which meets the input requirements of energy models. Binary snow products under cloud-free conditions are selected as reference data to conduct an inter-comparison of cloud distribution and total cloud area in a time series analysis. Results show that fractional snow cover products are mostly consistent with binary snow products in both space and time during snow accumulation, though some deviation appears in the snow ablation period. In the case of winter 2013, when the fractional snow cover was greater than 50%, the correlation coefficient of the two products was up to 0.8628. This data set is expected to provide high temporal-resolution data for the dynamic monitoring of snow over High Asia, as well as research on climate environment, hydrological and energy balance and disaster assessment.
Keywords: Tibetan Plateau; fractional snow cover; MODIS; cloud cover
Dataset Profile
Chinese title高亚洲逐日积雪覆盖度数据集
English titleDaily fractional snow cover dataset over High Asia
AuthorsQiu Yubao, Wang Xingxing, Han Lulu, Chang Li, Shi Lijuan
Corresponding authorQiu Yubao (qiuyb@radi.ac.cn)
Time spanJuly 2002 - June 2018
Geographical scopeHigh Asia (26° - 46°N, 62° - 105°E) is a high altitude area of Asia mainly comprising of the Tibetan Plateau. It covers the Qinghai-Tibet Plateau, the Himalayas, the Pamirs, the Tianshan Mountains and the Qaidam Basin. The following countries and regions are included: China, Myanmar, Nepal, Bhutan, India, etc.
Spatial resolution500 mData volume27.2 GB
Data formatGeotiff
Data Service System<http://www.sciencedb.cn/dataSet/handle/457>
Souces of fundingSupported by the Special Fund for Meteorological Scientific Research in the Public Interest “Constructing a Remote Sensing Product Dataset for Snow Pack Over Tibetan Plateau” (No.GYHY 201206040), the International Partnership Program of the Chinese Academy of Sciences (No.131CllKYSB20160061) and the National Natural Science Foundation of China (No.41371351)
Dataset compositionThe dataset consists of 17 packages (2002 – 2018, stored by year): HMA_MODIS_FSC_2002.zip, HMA_MODIS_FSC_2003.zip …
HMA_MODIS_FSC_2018.zip, etc., with a total data amount of 27.2 GB.
1.   Introduction
Snow cover is one of the fastest changing natural elements on earth, which has important effects on the circulation of hydropower, the construction of snowmelt runoff models and ecological models, the detection and evaluation of snow disaster, and the research of regional climate.1 High Asia, with the Tibetan Plateau being the main area affected by snow, is notably a snow-concentrated area at low and middle latitudes. Snow cover is one of the major supply sources of surface runoff and groundwater.2 The dynamic monitoring of snow cover in High Asia will contribute to the understanding of mountain environment and global climate change.3 As a result, it is very important to have accurate datasets for monitoring snow cover dynamics in High Asia.14
At present, satellite remote sensing plays a significant role in the dynamic monitoring of snow cover with advantages in timeliness, periodicity, low cost and ability in generating massive data. In this regard, the MODIS (Moderate-Resolution Imaging Spectroradiometer) instrument has been widely applied, which features moderate spatial and high temporal resolution.5The binary and fractional snow cover (FSC) products derived from MODIS have been proved to exhibit high recognition accuracy.6 However, as daily snow monitoring is influenced by cloud cover occurrence, re-estimation of snow cover under cloudy conditions has been addressed in MODIS binary snow products,7 while fractional snow cover products provide a more accurate description for snow. Compared with binary snow products, it can effectively reflect the extent of snow cover, and the features of snow variation.8 Cloud-free FSC data are thus of high interest.
Optical remote sensing is highly sensitive to weather, which restricts the estimation of snow, in particularly for parts of High Asia with frequent water vapor exchange, short snow occurrence and quick snow changes. Therefore, it is necessary to conduct the monitoring by a snow cover dataset with high temporal resolution.
The dataset presented here combines several cloud processing methods presented in literature for MODIS snow products.921 It takes geographic and climatic complexity in High Asia into account, and utilizes information derived from spatio-temporal interpolation and snow distribution. These are applied in the processor for cloud cover mitigation, thereby optimizing the accuracy of snow cover information while reducing the cloud cover to less than 10% for the daily FSC product. The dataset is based on the Normalized Difference Snow Index (NDSI) of MODIS version 6, and introduces traditional algorithms to estimate FSC. MODIS/Terra and MODIS/Aqua were applied for cloud cover mitigation in a five-step scheme, in order to resolve the FSC under cloud cover. Finally, the FSC products are validated against binary snow products for High Asia. The dataset contains daily FSC data in the recent 15 years (2002 – 2016).
2.   Data acquisition and processing
2.1   MODIS data
The MODIS data are provided by the National Snow and Ice Data Center of the US (NSIDC: https://modis.gsfc.nasa.gov/data/) – MODIS Snow Cover Daily L3 Global 500m Grid data, including MODIS/Terra AM data (MOD10A1) and MODIS/Aqua PM data (MYD10A1) of the Normalized-Difference Snow Index (NDSI) product. These data are provided in hdf format and in the sinusoidal map projection mode. Figure 1 shows geographical coverage of the snow cover products for High Asia.
The Digital Elevation Model (DEM) data adopted during data processing are derived from the SRTM dataset provided by the Geospatial Data Cloud (http://www.gscloud.cn/search#1084). The data are presented in the format of Geotiff with a resolution of 90 m.


Figure 1   Range of High Asia and coverage of MODIS tracks
2.2   Data processing
2.2.1   Data preprocessing
Under the WGS84 coordinates of UTM projection, we processed the NDSI data of MODIS version 6 by splicing, projection transform and clipping. Then NDSI algorithm was employed to calculate the MODIS fractional snow cover (in %). The equation and rules are described as follows:
FSC = [–0.01 + 1.45 × (NDSI)] × 100.0 (0.0≤NDSI≤1.0) (1)
(1) If the FSC was equal or less than 0, the pixel value was assigned as FSC = 0;
(2) If the FSC was larger than 0 and equal or less than 100, the pixel value was assigned as the FSC calculated;
(3) If the FSC was larger than 100, pixel value was assigned as FSC = 100;
MODIS normalized difference snow index of version 6 (C6) is an amendment to the index in version 5 (C5).7 C6 applied the quantitative image restoration (QIR) algorithm to restore the lost snow testing data – MODIS/Aqua band 6 data, so that MODIS/Aqua and MODIS Terra had the same quality and precision.7 Finally, the reflectance ratio of the difference between the visible (VIS, band 4) and short wave infrared (SWIR band 6) was used to estimate NDSI, namely NDSI = (band4 – band6)/(band4 + band6).
Since there was no fractional snow cover data in C6, we had to calculate FSC by NDSI using equation (1). FSC was calculated based on the regression relationship between NDSI and the FSC developed from MODIS (Terra and Aqua, 500 m resolution) and Landsat TM data (30 m resolution).22




Figure 2   Comparison of snow cover between MODIS product version 6 and version 5
As shown in Figure 2, by comparing the spatial distribution and scatterplot of snow pixel elements between C5 data (MOD) and C6 data (MYD), it can be seen that the correlation of FSC calculated by NDSI in C6 is 0.98, and it is of higher reliability and consistency when compared with the FSC of C5.
In addition, DEM data was spliced, clipped and processed through projection transform, and was resampled into auxiliary data that matched the MODIS Normalized Difference Snow Index.
2.2.2   Estimation of snow cover under cloud-cover conditions
The following five steps were used to estimate the snow coverage under cloud cover after the daily MODIS snow cover data were obtained. The workflow is shown in Figure 3.


Figure 3   Flow chart showing the elimination of snow cover under cloud cover
Step 1: Synthesis of MODIS/Terra AM data and MODIS/Aqua PM data
For observations influenced by cloud cover, it was assumed that the situation of snow cover remained unchanged for a given, short period of time; MODIS/Terra AM data and MODIS/Aqua PM data for this period were synthesized by the algorithm described as follows:
(1) If either or both of the pixels of the same position in Terra or Aqua is a water body, the synthetic pixel is a water body;
(2) If the snow cover of the same place in Terra and Aqua ranges between 0 and 100, the synthetic pixel is the mean of the snow cover in Terra and Aqua;
(3) As the accuracy of Terra data is higher than that of Aqua data,23 if the same location pixels of Terra and Aqua are non-cloud pixels, the Terra data should prevail;
(4) If the same position in Terra and Aqua is covered by cloud and land or snow, the land or snow coverage will replace the pixel under snow cover;
(5) If there is data loss or invalid value in the synthesis pixel, the pixel is assigned as cloud.
Step 2: Adjacent three-day temporal synthesis
As the probability of snow melting is low during cloudy days, it was assumed that the snow cover would remain unchanged for three days, and set the time window accordingly. Applying the temporally adjacent data of snow cover, the cloudy pixels for a given day can be reclassified. The algorithm is described as follows:
(1) If a pixel of the same place is presented as snow the day before and the day after, the cloud pixel of synthetic snow products on that day is the mean value of the snow cover of two days before and two days after.
(2) If a pixel of the same place is presented as land the day before and the day after, the cloud pixel on that day is classified as land.
(3) If either of the pixels of the same place one day before or after is presented as water body , the cloud pixel on that day is reclassified as water body.
Step 3: “Minimum” snow cover in short time
For purposes of examining snow cover characteristics, a “snow year” was defined as the period from July 1 of a given year to June 30 of the following year. Based on the period of snow cover occurrence, a “snow year” was divided into three parts: summer from July 1 to September 30; winter from October 1 to April 30; summer from May 1 to June 30.
In the high-altitude mountainous areas of High Asia, snow cover is estimated in a short period of time, and the hinterland of the plateau can be judged by setting conditions so that certain cloud pixels are identified as land. “Minimum snow-covered area” and “minimum land area” were analyzed and calculated for the three periods of a “snow year”. Judgment rules were prescribed as follows:
(1) Determination of the “minimum snow cover area” cloud pixel during a given period of time: When the altitude is more than 5,800 m and the pixel is covered with snow, the pixel mean is taken as the snow cover of the cloud pixel in the given time. When the altitude is from 3,000 to 5,800 m, and the number of days covered by cloud and snow altogether is greater than 90% of the total number of the research days, the mean value will be the snow cover under cloud cover.
(2) Determination of the “minimum land area” cloud pixel during a given period of time: The number of days when the pixel is covered by cloud is less than 20% of the total number of days during the period, and the total number of days when the pixel is covered by cloud and days when the pixel is the land is equal to the total number of days, the pixel is a land pixel without cloud during this period of time.
Finally, the cloud pixels of corresponding time periods are replaced by the “minimum” snow and the “minimum” land pixels of the time.
Step 4: Neighboring three-pixel method
Based on the spatial continuity of snow cover, the snow cover can be re-estimated by information on the cloud-free pixel around the cloud pixels. Neighboring three-pixel method is a method which considers three or more of the four pixels adjacent to a cloud pixel are the same. The rules for identifying cloud pixels are listed as follows:
(1) If three or more of the four pixels adjacent to the cloud pixel are snow, the central cloud pixel is assigned the mean value of the eight adjacent cloud pixels.
(2) If three or more of the four pixels adjacent to the cloud pixel are land, the central cloud pixel is assigned as land.
(3) In other cases, the original cloud pixel value is preserved;
Step 5: Eight-day maximum land masking
Through the above steps, the distribution of cloud pixels has been minimized. Finally, the masking of maximum land range in eight days was employed to further judge the pixels under cloud cover. The judgment rules are listed as follows:
(1) If one or more of the pixels of the same position within eight days is water body, the corresponding pixel of the synthetic product is water body;
(2) If one or more of the pixels of the same position within eight days is land, the corresponding cloud cover pixel of synthetic products is land. This step, which re-estimates on the principle of “land first under the condition of cloud cover”, decreased cloud cover to the utmost. That is to say, as long as there are cloudless circumstances (or minimal snow cover) within eight days, it is supposed that the snow is unstable during this period, or the snow cover is small. Statistics show that the cloud fraction decreased most through this step (as shown in Figure 4), which can reach 10.1%. However, as the seasonal snow in High Asia rapidly changes and exhibits short duration, there are some uncertainties in this step for monitoring daily FSC.
3.   Sample description
3.1   Naming format
The daily FSC dataset over High Asia is named in the following format: HMA_MODIS_FSC_YYYYDDD.tif. The file name contains the following information:
(1) HMA_MODIS_FSC: Fractional snow cover dataset over High Asia;
(2) YYYY: year described with four numbers;
(3) DDD: the production time being the DDD day of the YYYY year (January 1 of each year is regarded as the first day). For example, HMA_MODIS_FSC_2012006.tif indicates an MODIS fractional snow cover product over High Asia produced on January 6, 2012.
3.2   Category description
Table 1 shows pixel information of the daily FSC dataset over High Asia.
Table 1   Pixel information
MODIS valueTerrain attribute
0 – 100Snow cover
225Land
237Water body
250Cloud
4.   Quality control and evaluation
4.1   Changes of cloud cover percentages at each step
Taking the year 2014 as an example, Figure 4 reveals the changes of estimated fractional snow cover under cloudy conditions.


Figure 4   Cloud cover changes in the year of 2014
The statistical results of Figure 4 show that the cloud fractions of original MOD10A1 and MYD10A1 were 39.3% and 43.6%, respectively, indicating that these snow products cannot be directly used for real-time dynamic monitoring of daily snow. Instead, re-estimation under cloudy conditions is required, including to use spatio-temporal interpolation to remove cloud cover. Through the synthetic processing of Terra and Aqua data, the average cloud fraction decreased to 32.8%, and the decreased cloud fraction was approximately 6.5% of the Terra data and 10.8% of the Aqua data. After synthesis of adjacent three days, the cloud fraction decreased to 25.7%, by approximately 7.1%. However, this was still unable to meet the demands for dynamic monitoring of snow cover, and further processing of cloud pixels was required. By applying the spatial interpolation (namely, cloud pollution removal by the method of neighboring pixel), a further 0.57% of the cloud fraction was removed. In view that High Asia has areas with relatively steady snow cover and areas with little snow (taken as non-snow covered areas) , such a feature can be used to dispose of fractional snow cover under cloud, by which the cloud fraction decreased approximately by 5.2%. The masking of maximum land range in eight days is to remove cloud pollution by using the relativity of snow cover in a short time, by which the cloud fraction approximately decreased by 10.1%. There were still a few clouds in the final synthesized products, but the cloud cover had already been controlled within 10%.
4.2   Spatio-temporal comparison between fractional and binary snow cover products
In order to verify the credibility of the snow cover products generated, we selected the products randomly from the binary snow product of the same period for analyzing the consistency between the products.9 Figure 5 shows that the two products are consistent in spatial distribution, while the fractional snow cover products show more details at the edge of the snow area. However, it is difficult for the fractional snow cover product to achieve entirely cloudless daily estimates. In the comparative analysis, there still exists a small amount of cloud cover, especially in the hinterland of the plateau where snow changes faster. Despite this, Figure 4 shows that the overall proportion of cloud cover pixels is less than 10%. Therefore, in general, the product has a high degree of credibility.










Figure 5   Spatio-temporal comparison of fractional (right) and binary (left) snow cover products
Data of a winter period of a “snow year” were selected from the binary and fractional snow cover products to compare the total snow area (number of pixels classified as snow) as a time series. As shown in Figure 6, the number of snow pixels for a long time from October 1, 2013 to April 30, 2014 (indirectly reflecting the area) indicates that the snow and binary products are strongly consistent in time, especially when the fractional coverage exceeds 50%.


Figure 6   Comparison of the number of snow pixels in the two snow products
A comparison of space and time series shows that the fractional and binary snow products are highly consistent in monitoring snow of the same period. Moreover, the FSC product effectively conveys information on the proportion of snow cover, which is significantly superior to binary products.
5.   Usage notes
This dataset is a supplement to the snow binary products over High Asia. Data files are compressed in Geotiff format by year, and users can choose to download according to their needs. This dataset can reflect the changing characteristics of snow cover in the High Asia region at the daily timescale. It can be used to analyze local climate change effects through the spatio-temporal changes of snow. It is of great significance to studies of climate, hydrological and energy balance, disaster assessment in High Asia.
Authors and contributions
Qiu Yubao, Contribution: algorithm development.
Wang Xingxing, Contribution: data preprocessing and production.
Han Lulu, Contribution: data preprocessing and production.
Chang Li, Contribution: data preprocessing and production.
Shi Lijuan, Contribution: data collection and preprocessing.
Acknowledgments
We would like to express our heartfelt gratitude to the National Snow and Ice Data Center (http://nsidc.org) for providing the MOD10A1 and MYD10A1 products, and Geospatial Data Cloud (http://www.gscloud.cn) for providing the DEM.
1.
Tang Z, Wang J, Yan L et al. Estimating sub-pixel snow cover from MODIS in Qinghai-Tibet Plateau. Journal of Arid Land Resources and Environment 27 (2013): 33 – 38.
2.
George AR & Dorothy KH. MODIS Snow Products Collection 6 User Guide. Available at: <http://nsidc.org/sites/nsidc.org/files/files/MODIS-snow-user-guide-C6.pdf> [Accessed July 18, 2017].
3.
Tang Z, Wang J, Li H et al. Accuracy validation and cloud obscuration removal of MODIS fractional snow cover products over Tibetan Plateau. Remote Sensing Technology and Application 28 (2013): 423 – 430.
4.
Li P. Distribution of snow cover over the High Asia. Journal of Glaciology and Geocryology 17 (1995): 291 – 298.
5.
Bai Y & Feng X. Introduction to some research work on snow remote sensing. Remote Sensing Technology and Application 12 (1997): 59 – 65.
6.
Hall D, Riggs G & Barton J. Algorithm theoretical basis document (ATBD) for the MODIS snow and sea ice mapping algorithms. NASAGSFC Technical Report, 2001.
7.
Zhang H, Qiu Y, Zheng Z et al. Comparative study of the feasibility of cloud removal methods based on MODIS seasonal snow cover data over the Tibetan Plateau. Journal of Glaciology and Geocryology 38 (2016): 714 – 724.
8.
Nolin A. Recent advances in remote sensing of seasonal snow. Journal of Glaciology 56 (2010): 1141 – 1150.
9.
Qiu Y, Guo H, Chu D et al. MODIS daily cloud-free snow cover products over Tibetan Plateau (2002 – 2015). Science Data Bank. DOI: 10.11922/csdata.170.2016.0003
10.
Hall D, Riggs G, Salomonson V et al. MODIS snow-cover products. Remote Sensing of Environment 83 (2002): 181 – 194.
11.
Furtado J, Cohen J, Butler A et al. Eurasian snow cover variability and links to winter climate in the CMIP5 models. Climate Dynamics 45 (2015): 2591 – 2605.
12.
Yu X, Qiu Y, Guo H et al. Cloud removing method for daily snow mapping over Central Asia and Xinjiang, China. IOP Conf. Series: Earth and Environmental Science 57 (2017). DOI: 10.1088/1755-1315/57/1/012048
13.
Dietz A , Wohner C & Kuenzer C. European snow cover characteristics between 2000 and 2011 derived from improved MODIS daily snow cover products. Remote Sensing 4 (2012): 2432 – 2454.
14.
López-Burgos V, Gupta H & Clark M. Reducing cloud obscuration of MODIS snow cover area products by combining spatio-temporal techniques with a probability of snow approach. Hydrology and Earth System Sciences 7 (2013): 1809 – 1823.
15.
Gao Y, Xie H, Lu N et al. Toward advanced daily cloud-free snow cover and snow water equivalent products from Terra-Aqua MODIS and Aqua AMSR-E measurements. Journal of Hydrology 385 (2010): 23 – 35.
16.
Thirel G, Salamon P, Burek P et al. Assimilation of MODIS snow cover area data in a distributed hydrological model. Hydrology & Earth System Sciences Discussions 8 (2011): 1329 – 1364.
17.
Parajka J & Bloschl G. Spatio-temporal combination of MODIS images-potential for snow cover mapping. Water Resources Research 440 (2008): 79 – 88.
18.
Gafurov A & Bárdossy A. Cloud removal methodology from MODIS snow cover product. Hydrology and Earth System Sciences 13 (2009): 1361 – 1373.
19.
Gao Y, Xie H, Yao T et al. Integrated assessment on multi-temporal and multi-sensor combinations for reducing cloud obscuration of MODIS snow cover products of the Pacific Northwest USA. Remote Sensing of Environment 114 (2010): 1662 – 1675.
20.
Paudel K & Andersen P. Monitoring snow cover variability in an agropastoral area in the Trans Himalayan region of Nepal using MODIS data with improved cloud removal methodology. Remote Sensing of Environment (2001): 1234 – 1246.
21.
Dietz A, Kuenzer C & Conrad C. Snow cover variability in Central Asia between 2000 and 2011 derived from improved MODIS daily snow cover products. International Journal of Remote Sensing 34 (2013): 3879 – 3902.
22.
Salomonson V & Appel I. Estimating fractional snow cover from MODIS using the normalized difference snow index. Remote Sensing of Environment 89 (2004): 351 – 360.
23.
Wang X, Xie H, Liang T et al. Comparison and validation of MODIS standard and new combination of Terra and Aqua snow cover products in northern Xinjiang, China. Hydrological Processes 23 (2010): 419 – 429.
Data citation
1. Qiu Y, Wang X, Han L et al. Daily fractional snow cover dataset over High Asia. Science Data Bank. DOI: 10.11922/sciencedb.457
Article and author information
How to cite this article
Qiu Y, Wang X, Han L et al. Daily fractional snow cover dataset over High Asia. China Scientific Data 2 (2017), DOI: 10.11922/csdata.170.2017.0146
Qiu Yubao
qiuyb@radi.ac.cn
Qiu Yubao, PhD, Associate Professor, engaged in research on environmental remote sensing applications.
Wang Xingxing
Wang Xingxing, MSc, focusing on remote sensing applications.
Han Lulu
Han Lulu, BS, focusing on remote sensing applications.
Chang Li
Chang Li, MSc, focusing on remote sensing applications.
Shi Lijuan
Shi Lijuan, PhD, mainly engaged in snow remote sensing applications.
Special Fund for Meteorological Scientific Research in the Public Interest (No.GYHY 201206040); International Partnership Program of the Chinese Academy of Sciences (No.131CllKYSB20160061); National Natural Science Foundation of China (No.41371351)
Publication records
Published: Sept. 25, 2017 ( VersionsEN3
Released: July 31, 2017 ( VersionsZH1
Published: Sept. 26, 2017 ( VersionsZH2
Updated: Sept. 25, 2017 ( VersionsZH4
References
中国科学数据
csdata