Comparison and analysis of snow cover data based on different definitions of snow cover days

Di An , DongLiang Li , Yun Yuan

1. Nanjing University of Information Science & Technology, Key Laboratory of Meteorological Disaster of Ministry of Education,Nanjing, Jiangsu 210044, China

2. Meteorological Bureau of Xuzhou, Xuzhou, Jiangsu 221000, China

*Correspondence to: Prof. DongLiang Li, Nanjing University of Information Science & Technology, Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing, Jiangsu 210044, China. Email: lidl@nuist.edu.cn

Comparison and analysis of snow cover data based on different definitions of snow cover days

Di An1,2, DongLiang Li1*, Yun Yuan1

1. Nanjing University of Information Science & Technology, Key Laboratory of Meteorological Disaster of Ministry of Education,Nanjing, Jiangsu 210044, China

2. Meteorological Bureau of Xuzhou, Xuzhou, Jiangsu 221000, China

*Correspondence to: Prof. DongLiang Li, Nanjing University of Information Science & Technology, Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing, Jiangsu 210044, China. Email: lidl@nuist.edu.cn

In order to analyze the differences between the two snow cover data, the snow cover data of 884 meteorological stations in China from 1951 to 2005 are counted. The data include days of visual snow observation, snow depth, and snow cover durations, which vary according to different definitions of snow cover days. Two series of data, as defined by "snow depth" and by "weather observation," are investigated here. Our results show that there is no apparent difference between them in east China and the Xinjiang region, but in northeast China and the Tibetan Plateau the "weather observation" data vary by more than 10 days and the "snow depth" data vary by 0.4 cm. Especially in the Tibetan Plateau, there are at least 15 more days of "weather observation" snow in most areas (sometimes more than 30 days). There is an obvious difference in the snow cover data due to bimodal snowfall data in the Tibetan Plateau, which has peak snowfalls from September to October and from April to May. At those times the temperature is too high for snow cover formation and only a few days have trace snow cover. Also, the characteristics and changing trends of snow cover are analyzed here based on the snow cover data of nine weather stations in the northeast region of the Tibetan Plateau,by the Mann-Kendall test. The results show significantly fewer days of snow cover and shorter snow durations as defined by"snow depth" compared to that as defined by "weather observation." Mann-Kendall tests of both series of snow cover durations show an abrupt change in 1987.

weather observation; days of snow cover; depth of snow; durations

1. Introduction

In the current studies on snow cover in China, the definitions of snow cover days have different standards. As defined by Cuiet al. (2005), Donget al. (2004), and Weiet al.(2002), only observed snow depths greater than 0.5 cm are considered as snow cover. Xiet al. (2009) recorded a snow cover day when more than half of the surface within the visual field around the weather stations was covered by snow (including snow grains, graupel, and ice particles) to comply with the observable condition of snow depth. Ma(2008) recorded the snow depth of all the trace snow cover days as 0.5 cm for statistical purposes. The above different standards were found in our process of analyzing the data statistics. Some experts have pointed out that these different definitions and statistical methods have influenced past research results on snow cover. For example, research results of snow cover in the Tibetan Plateau based on different types of data were compared and analyzed by Chenet al.(1999),and significant differences were found to exist in the definition results of the anomalous years of snow cover. Therefore,in view of the fact that past data used different definitions of snow cover days and different statistical standards in different regions, two kinds of data were compared and analyzed in this study to determine their effects on the results.

According to the China Meteorological Administration’s specifications for surface meteorological observation (China Meteorological Administration, 2003), the observation of snow depth is required when more than half of the surface within the visual field around a weather station is covered by snow (including snow grains, graupel, and ice particles). On the days of observable snow depth, a snow stake is inserted vertically into the snow in the observation location at 8:00 a.m. every day (without inserting it into the soil). Based on which gradation line the snow surface reaches, the snow depth is recorded in integers of centimeters. For example,when the daily snow depth is less than 0.5 cm, the snow depth is recorded as 0 cm; when the snow depth is equal to or greater than 0.5 cm, the values are rounded off to a minimum of 1.0 cm. An average snow depth of less than 0.5 cm is recorded as 0 cm.

However, two standards of definition of snow cover days exist in the current studies on snow cover in China.One definition of snow cover day is based on the weather observation,i.e., a snow cover day is recorded when more than half of the surface within the visual field of the observation field is covered by snow (hereinafter referred to as the definition by "weather observation"). The other definition of snow cover day is based on the snow depth,i.e., a snow cover day is recorded when the snow cover area reaches the observational requirement depth of 1.0 cm(hereinafter referred to as the definition by "snow depth").The snow depth on snow cover days satisfying the observational requirements but without an average snow depth reaching 0.5 cm is recorded as 0; this is considered a "trace snow cover day" (China Meteorological Administration,2003; Yeet al., 2007).

Given these two definitions, differences exist in the statistics of the annual snow cover days, the annual average snow depth, and the interval days between the first and last snow. When defined by weather observation, the annual snow cover days and the interval days between the first and last snow will be greater than those defined by snow depth,and the annual average snow depth defined by weather observation will be less than that defined by snow depth.

2. Data and methods

This study used data on snow cover days from 963 stations in China from 1951 to 2005. Based on the standards of definition by weather observation and by snow depth, statistical analyses were completed on two sets of data pertaining to annual average snow cover days, annual average snow depths, and the number of interval days between the first and last snow. Only those stations were selected that did not have zero values for multi-annual average snow cover days,snow depth, and interval days (Figure 1), resulting in usable data from 884 stations. The stations with values of 0 were basically concentrated in the southern region of Nanling, so stations in that region were not selected.

The annual average values of the above three elements were counted by "snow cover year," which is from July 1 of a year to June 30 of the next year. For "weather observation"snow cover days, all snow cover days that conformed to the definition of weather observation were tallied in the snow cover year. For "snow depth" snow cover days, all snow depths of snow cover days were tallied and then divided by the number of snow cover days (the snow depth of "trace snow cover" days were recorded as 0 cm). The interval days between the first and last snow are the days between the first day and the last day of snow cover in the snow cover year.According to these three elements within the definition by snow depth, the trace snow cover days are not considered in the statistics here.

Figure 1 Distribution of the meteorological stations (the representative stations are emphasized)

3. Comparisons of snow cover days by two statistical methods

The multi-annual average snow cover days with the two definitions are shown in figure 2. The multi-annual average snow cover days as defined by snow depth are shown in figure 2(a). The three regions with the largest number of snow cover days are the northeast region, the Xinjiang region, and the Tibetan Plateau. There were more than 30 snow cover days in the northeast region, and the number of days increased with the rise in latitude. However, the contour in the Northeast Plain shows a northward bend because there were fewer snow cover days there than on the lateral Daxinganling Mountain and Changbai Mountain. In the northern part of the northeast region there were more than 100 snow cover days, and in the most northern region there were more than 160 snow cover days. Tianshan Mountain in the Xinjiang region had more than 30 snow cover days(multi-annual average) (Li, 1993; Wei and Lu, 1995; Guoet al., 2004; Qinet al., 2006) and the number of snow cover days gradually increased northward. In the region adjacent to the Altai Mountains there were more than 120 snow cover days.

The stations with observable snow cover in the Tibetan Plateau were mostly concentrated in the eastern part, with few such reports in the western area. In figure 2, a representative high-value center was Qingshuihe Station in the Tibetan Plateau (33.8°N, 97.13°E) with 137 snow cover days as defined by weather observation and 105 snow cover days as defined by snow depth. Representative low-value stations were Andehe Station (37.07°N, 82.72°E) with 3.9 snow cover days as defined by weather observation and 3.8 snow cover days as defined by snow depth, and Mangya (38.25°N,90.8°E) with 6.4 snow cover days as defined by weather observation and 2.9 snow cover days as defined by snow depth. In most regions of the eastern part there were more than 20 snow cover days, and in some regions of the Bayan Har Mountains there were more than 40 snow cover days.Except for these three regions, the number of snow cover days of the central and southern areas in China declined with the decrease of latitude. The winter temperature in the Sichuan Basin was higher than in other regions at the same latitude with little snow. Thus, there were less than four snow cover days there.

Figure 2(b) shows the multi-annual average snow cover days in China as defined by weather observation. It can be seen from figure 2(a) and (b) that, except for the significant differences in snow cover days, based on the two definitions,in the Tibetan Plateau the distribution status was basically consistent in China.

Figure 2 Average days of snow cover days as defined by snow depth (a) and by weather observation (b)

Based on the two definitions, the snow cover days varied statistically in degrees in different regions. The more trace snow cover days there were, the greater the statistical difference was. Figure 3(a) shows the multi-annual average trace snow cover days that resulted from deducting snow covers days as defined by snow depth from those as defined by weather observation. It can be seen from the figure that the high values (the most snow cover days) appeared in the northeastern region of Inner Mongolia, Changbai Mountain,and the central region of the Tibetan Plateau. The annual snow cover days as defined by weather observation ranged between 40 and 90, with the trace snow cover days ranging mostly between 10 and 20.

The ratio of the trace snow cover days and the snow cover days as defined by snow depth is shown in figure 3(b). The region with the largest ratio was located in the Tibetan Plateau;the ratio was more than 30 percent in most of that area, and occasionally more than 40 percent. Comparing figure 3(a) and figure 3(b), the trace snow cover days of the Tibetan Plateau were high in both value and proportion. For the Xinjiang region and the northeast area with their relatively large numbers of snow cover days, the trace snow cover days and propor-tions were less than those of the Tibetan Plateau. In these regions there were 10 trace snow cover days with a ratio of about 20 percent; in specific parts of northeastern Inner Mon-

golia, the proportion was 30 percent. In north China, east China, and south China, there were less than four trace snow cover days with a proportion of less than 20 percent.

Figure 3 Average days of trace snow cover days (a) and ratio of days of trace snow cover days to days of snow cover as defined by weather observation (b)

The largest difference in the results with the two definition methods appeared at Maduo Station of the Tibetan Plateau (98°13′E, 34°55′N, 4,273.3 m elevation; the station is marked with a diamond in figure 1). Here there were 32 multi-annual average trace snow cover days and the ratio to snow cover days as defined by snow depth was 65 percent.The comparison graph of multi-annual snow cover days at Maduo Station is shown in figure 4. It can be seen from figure 4(a) that, statistically, snow cover days based on the two definitions differed significantly. The smallest difference

appeared in 1974 with 174 snow cover days as defined by snow depth and 164 days as defined by weather observation,as well as 10 trace snow cover days. The greatest difference appeared in 1995 with 126 and 65 snow cover days, respectively, as defined above, as well as 61 trace snow cover days.Figure 4(b) demonstrates that annual changes of the trace snow cover days fluctuated greatly, and its ratio to the snow cover days as defined by snow depth was not stable. The more trace snow cover days there were, the higher the ratio was, and vice versa.

Figure 4 Variations of snow cover days at Maduo station (a) for statistical contrast of the two definitions: the solid line is the snow cover days as defined by weather observation and the dashed line is the snow cover days as defined by snow depth; (b) annual changes of trace snow cover days: the solid line is the trace snow cover days and the dashed line is the ratio of trace snow cover days to snow cover days as defined by snow depth (%).

4. Comparisons of statistical snow depths with the two definitions

Our statistical method of assessing snow cover days as defined by snow depth was to tally all the snow depths in a snow cover year (the trace snow cover was not included in this tally), and then divide this sum by the snow cover days that had snow depth recorded as at least 1 cm. To assess snow cover days as defined by weather observation, the accumulated value of snow cover days in a year was not variable; thus, this is simply the sum of snow cover days and trace snow cover days. Therefore, the difference between two statistical methods is that the statistical value of snow cover days as defined by weather observation is smaller than that of snow cover days as defined by snow depth.

The multi-annual average snow depths based on these two definition standards are shown in figure 5. It can be seen from figure 5(a) that the snow depths in China had three zones with large values: the northeast region, the Xinjiang region, and the southwestern Tibetan Plateau (Bamzai and Shukla, 1999; Guoet al., 2004). The snow depths in north China and east China were mostly around 3 cm, although a small high-value center existed in the Dabie Mountain area.As shown in figure 5(b), the snow depths in various regions reduced by different degrees, and there were fewer areas with snow depth of 3 cm in north China and east China.

Comparisons of the snow depths nationwide are shown in figure 6. Figure 6(a) was drawn with the data obtained by the subtracted values between the tallied results of the two definition standards. It can be seen that the difference of the two types of snow depths of most areas in China was less than 0.4 cm; stations that had snow depths greater than 0.4 cm were mainly in northeast China and western Tibet. In some areas near Changbai Mountain there were snow depths greater than 1 cm. Therefore, in terms of differences, the snow depths determined by the two statistical methods were not very different and were usually less than 1 cm. The ratio between the snow cover days as defined by snow depth and those as defined by weather observation is shown in figure 6 as follows:

whereHsnowis the depth of snow on snow cover days as defined by snow depth, andHweatheris the is the depth of snow on snow cover days as defined by weather observation.The numerator is the absolute difference (cm), and the fraction is the relative difference(%).

Figure 5 Average snow depths on snow cover days as defined by snow depth (a) and on snow cover days as defined by weather observation (b)

It can be seen from figure 6(b) that the ratios in the southern region of the Yangtze River were mostly less than 10 percent. Due to the small amount of snow in this area, the multi-annual average snow depths were mostly less than 2 cm. The ratios in north China and east China were mostly 15 percent. High values appeared in the central Northeast Plain and the Tibetan Plateau. Comparing figure 6(a) and figure 6(b), the difference of the northeast area was greater than in other areas in China, so a high value in this area was normal.The snow depth observed by the stations in the Tibetan Plateau was very small, with an annual average snow depth generally less than 6 cm. This was related to the weather stations in the Tibetan Plateau having been built in an area with a relatively low elevation and a relatively flat terrain,and thus this cannot represent the snow cover status in the mountain area with its relatively high elevation.

Previous analysis indicates that there were a very large number of trace snow cover days in the Tibetan Plateau.Therefore, the accumulated snow depth was not variable in the snow depth data of snow cover days as defined by weather observation, even though the number of snow cover days increased greatly. Accordingly, these results were quite different from the analytic results of snow cover days as defined by snow depth.

Figure 6 The absolute differences (a, in cm) and relative differences (b, in %) of average snow depths between the two definitions

Comparing the two kinds of data, there was a great difference between two statistical results at Pagri Station(89°05′E, 27°44′N, 4,301.2 m elevation). Here the snow depths based on the two definitions were 4.1 cm and 5.7 cm,respectively, and there was a difference in the multi-annual average depth of 1.6 cm. The average snow depth on snow cover days as defined by weather observation was 28 percent less than that on snow cover days as defined by snow depth.

It can be seen from figure 7 that the two definitions of snow cover days generally showed identical statistical trends.

When the annual snow depth was relatively small, such as in 1958, the two statistical results were 1 cm and 4 cm, respectively, and the number of snow cover days as defined by weather observation was 75 percent less than that as defined by snow depth. The maximum value appeared in 1995,when several large-scale snowfalls occurred in that snow cover year; the maximum value of the daily snow depth reached 87 cm. In that year the statistical results based on the two definitions were 37 cm and 40 cm, respectively, and the bias of the two results was only 7.5 percent, which accounted for only 1/10 of that in 1958.

Figure 7 Variations of snow depths at Pagri Station, as defined by snow depth (solid line ) and by weather observation (dotted line)

5. Comparisons of the interval days between the first and last snow, by the two definition standards

For snow cover days as defined by weather observation,the interval days are those days between the first date and the last date with the record of visually observed snow cover in a snow cover year. For snow cover days as defined by snow depth, the interval days are those between the first date and the last date with a snow depth up to or more than 1 cm recorded in a weather station in a snow cover year. Therefore, there are more interval days of weather observation snow than interval days of snow depth.

Figure 8(a) shows that there was a center with a low value of interval says,i.e., Tazhong Station (39°N, 83.67°E)in the southern Xinjiang region, with 13.3 interval days of multi-annual average snow cover. The zones with large values were located in northeast China, the rest of the Xinjiang region, and the Tibetan Plateau. For northeast China, the results obtained based on the two definitions shown in the figure were not very different. There were 120 interval days in most parts of the region, and 180 days in the northernmost part. The interval days in the Xinjiang region counted by the two definitions showed similar distributions and gradually increased northward, from the beginning value of 60 days to the northernmost part having 150 days. The greatest difference existed in the Tibetan Plateau, and the specific differences are presented in figure 9.

The difference between the statistical results based on the two definitions is shown in figure 9(a). The differences in most regions south of the Yangtze River were very small,ranging between 1 day and 5 days. Because the winter temperature of the Sichuan Basin was higher than that of other areas at the same latitude, a small number of snowfalls occurred in this region and the snow melted rapidly, which resulted in trace snow cover days that were easy to tabulate.Therefore, an area with a small value of interval days appeared there. The differences in east China and north China were about 10 days; in the northeast region the range was between 10 days and 15 days; that of the Xinjiang region was around 10 days; that of the Tibetan Plateau was generally 15 days; and the maximum value reached more than 35 days.

It can be seen from the ratios in figure 9(b) that the areas with the greatest differences were the northwestern Tibetan Plateau and the Sichuan Basin. The multi-annual average interval days between the first and last snow was less than 30 days in the Sichuan Basin, with a large number of trace snow cover days. Thus, a high-value area formed there. The ratio of the eastern Tibetan Plateau was about 20 percent, but the ratio in most regions in the northwestern Tibetan Plateau was up to 40 percent; this was also caused by a large number of trace snow cover days in this region.

Figure 8 Average snow cover durations of snow depths as defined by depth (cm) of snow (a) and by weather observation (b)

Figure 9 The absolute differences (a, in days ) and relative differences (b, in % ) of snow cover durations between the two definitions of snow cover days

6. Change trend analyses of the statistical snow cover data with the two definitions

Nine stations on the northeastern slope of the Tibetan Plateau were selected to analyze the similarities and differences in the change trends of different snow cover data(these stations are marked with triangles in figure 1, and the specific locations are shown in table 1). The basis for selecting these stations was the relatively significant difference between the two data sets and the relatively complete data sequences.

The snow depths, the snow cover days, and the interval days of the nine representative stations were averaged for comparison. It can be seen from figure 10 that the two statistical results showed declining trends with slightly different degrees. The statistical change trends of the snow cover days based on the two definitions in figure 10(a) were basically consistent. The slope of the trend line of the snow cover days as defined by weather observation was -0.84 day/(10 years), whereas that defined by snow depth was -0.85 day/(10 years). The slopes of the trend lines of the snow depth with the two definitions were -0.08 cm/(10 years) and-0.1 cm/(10 years), respectively, as shown in figure 10(b).This indicates that the declining trend of the snow depth as defined by weather observation was more significant than that as defined by snow depth. As shown in figure 10(c), the slope of the trend line of the interval days between the first and last snow as defined by snow depth was -2.54 day/(10 years) and was more significant than that defined by weather observation.

Table 1 Station Information

Figure 10 Variations of snow cover days (a), snow depth (b), and snow cover durations (c). The solid line is snow cover days as defined by snow depth and the dotted line is snow cover days as defined by weather observation.

In order to further analyze the changes in various elements of snow cover, Mann-Kendall (M-K) tests were conducted on the snow cover data based on the two definitions (Fu and Wang,1992). No significant catastrophe points appeared in the results of the M-K test on the snow cover days and snow depths. The graphs of M-K test results of the interval days of snow cover are shown in figure 11, where significant catastrophe points oc-

curred in the interval days of snow cover as defined by snow depth in 1987 (Li, 1995). Figure 11(a) and (b) show the relatively consistent change in the curves; significant catastrophe points appeared in 1987, and the number of interval days declined from 1987 onward. This is consistent with the hypothesis about the abrupt change of the northwest China climate in 1987 proposed by Liet al.(2003) and Shiet al. (2003).

Figure 11 The M-K tests of snow cover durations as defined by snow depth (a) and as defined by weather observation (b).Solid lines are the UF (Forward_u) curves and dashed lines are the UB (Backward_u) curves.The intersection of the curves represents the time when a trend or change starts.

7. Analyses of the trace snow cover days in the Tibetan Plateau

The statistical results of the snow cover days, the snow depths, and the interval days between the first and last snow in the Tibetan Plateau showed great differences based on the two definitions, which were related to the characteristics of snow cover in the Tibetan Plateau. In terms of snow depth,the proportion of the trace snow cover days observed in the stations of the Tibetan Plateau was very high, with most regions evidencing more than 40 percent, especially at both ends of the snow season. Numerous studies have proven that the snowfalls in the Tibetan Plateau are different from other regions, showing a bimodal shape, and the snowfall peaks often occur in the transition periods between summer and winter,i.e., April-May and September-October (Li and Mi,1983; Zou and Cao, 1989; Zou and Cao, 1991; Ke and Li,1998). During these periods, the temperature is relatively high and does not facilitate the formation of snow cover,although many snowfalls occur in the plateau. Therefore, a large number of trace snow cover days have appeared. The plateau is influenced by the cold air mass in winter with low temperature and little water vapor, thus making it hard to form snowfalls. However, once a snowfall is formed, the snow does not easily melt, resulting in a sustained large number of snow cover days. For the above reasons, the data show that there were more trace snow cover days in the Tibetan Plateau than in other regions. Additionally, the trace snow cover days were concentrated at both ends of snow season, which led to great differences in the statistical snow depths, the snow cover days, and the interval days between first and last snow based on the two definitions.

8. Conclusions and discussions

Statistical analyses of the data pertaining to the snow cover days, the snow depths, and the interval days between the first and last snows in China, and the differences analysis,indicate that the there were smaller differences in the statistical results in south China, east China, and north China based on the two definitions of snow cover days. The difference in the Xinjiang region was small,i.e., the number of snow cover days as defined by weather observation was only 2 to 5 days more than that defined by snow depth. The difference of the snow depth based on the two definitions and the ratio to that defined by weather observation was around 10 percent. The interval days between the first and last snow increased within 10 days. The greatest differences existed in northeast China and the Tibetan Plateau. There the difference of snow cover days was more than 6 days with a relative difference of more than 20 percent (up to 30 percent in most areas and even reaching 50 percent in the Tibetan Plateau). In terms of the interval days between the first and last snow, the statistical difference in northeast China ranged between 10 and 15 days, and those in the Tibetan Plateau and the Sichuan Basin were rather significant. Due to its warm climate, snowfalls do not form easily in the Sichuan Basin without very thick snow cover. Thus, trace snow cover days in that region accounted for a large proportion,resulting in the greater difference of the statistical results. As determined by the predominant climate characteristics of the snow season in the Tibetan Plateau, the statistical differences in most parts of that region were greater than 20 days, and the maximum differences among the stations were more than 35 days. A large number of trace snow cover days ap-peared at both ends of the snow season, and the amplitude of the advanced first snow date and the delayed last snow date increased accordingly.

Nine representative stations on the northeastern slope of the Tibetan Plateau were selected to compare their snow cover trends and abrupt trend changes. The results proved that the statistical snow cover results based on the two definitions showed declining trends with slightly different degrees. Statistical catastrophe points in the interval snow cover days based on both definitions occurred in 1987.

We conclude that for these two kinds of data (based on the two definitions), if the study results of the abrupt change of snow cover are consistent, either data can be used. For studies on the change trends of snow cover days, we recommend using the statistical snow cover data as defined by weather observation. Especially in studies on the snow cover days of the Tibetan Plateau, there were relatively significant differences in the snow cover days and the interval days between first and last snow. The statistical snow cover days and interval days between the first and last snow as defined by weather observation were more factual. However, for the calculation of snow depth, the values of the two statistical results were rather small, and that defined by weather observation was smaller. Therefore, which data are suitable for studies on snow depth of the Tibetan Plateau merits further discussion.

This work was supported by the National Basic Research Program of China (2007CB411506) and the State Key Laboratory of Cryospheric Science (SKLCS08-06). The authors are very grateful to all the experts and editors.

Bamzai AS, Shukla J, 1999. Relation between Eurasian snow influence of Tibetan snow on monsoons and precipitation-cover, snow depth, and the Indian summer monsoon: an observation. Journal of Climate, 12:3117-3132.

Chen QJ, Gao B, Sun AJ, 1999. Study on relation between abnormal snow cover over Qinghai-Xizang Plateau and ENSO cycle. Journal of Plateau Meteorology, 18(2): 147-161.

China Meteorological Administration, 2003. Specifications for Surface Meteorological Observation. Meteorology Press, Beijing. 61-63.

Cui CX, Yang Q, Wang SL, 2005. Comparative analysis on the long-term variation of snow cover in the mountainous regions and plains in Xinjiang during the period of 1960-2003. Journal of Glaciology and Geocryology, 27(4): 486-490.

Dong AX, Guo H, Wang LP, Liang TG, 2004. CEOF analyses on the variation of seasonal snow cover duration in north Xinjiang since recent 40 years. Journal of Plateau Meteorology, 23(6): 937-940.

Fu CB, Wang Q, 1992. The definition and detection of the abrupt climatic change. Chinese Journal of Atmospheric Sciences, 16(4): 482-490.

Guo YJ, Zhai PM, Li W, 2004. Snow cover in China, derived from NOAA satellite remote sensing and conventional observation. Journal of Glaciology and Geocryology, 26 (6): 755-760.

Ke C, Li PJ, 1998. Spatial and temporal characteristic of snow cover over the Qinghai-Xizang Plateau. Acta Geographica Sinica, 53(3): 209-215.

Li DL, Wei L, Cai Y, Zhang CJ, Feng JY, Yang Q, Yuan YJ, Dong AX, 2003.The present facts and the future tendency of the climate change in northwest China. Journal of Glaciology and Geocryology, 25(2):135-142.

Li PJ, 1995. Distribution of snow covers over the High Asia. Journal of Glaciology and Geocryology, 17(4): 292-295.

Li PJ, 1993. Dynamic characteristic of snow cover in western China. Acta Geographica Sinica, 48(6): 503-515.

Li PJ, Mi DS, 1983. Distribution of snow cover in China. Journal of Glaciology and Geocryology, 5(4): 12-18.

Ma LJ, 2008. Spatial-temporal variation of snow cover over the Tibetan Plateau and the relationship with the atmospheric circulation factors in the last 50 years. Ph.D. dissertation, Chinese Academy of Sciences, Beijing.

Qin DH, Liu SY, Li PJ, 2006. Snow cover distribution, variability, and response to climate change in western China. Journal of Climate, 19:1820-1833.

Shi YF, Shen YP, Ding YJ, Zhang GW, 2003. An Assessment of the Issues of Climatic Shift from Warm-Dry to Warm-Wet in Northwest China. Meteorology Press, Beijing.

Wei ZG, Huang RH, Chen W, Dong WJ, 2002. Spatial distributions and interdecadal variations of the snow at the Tibetan Plateau weather stations. Chinese Journal of Atmospheric Sciences, 26(4): 496-506.

Wei ZG, Lu SH, 1995. Distribution of snow covers on the Qinghai-Xizang plateau and its influence on surface albedo. Journal of Plateau Meteorology, 14(2): 67-73.

Xi Y, Li DL, Wang W, 2009. Study of the temporal-spatial characteristics of snow cover days in Hetao and its vicinity. Journal of Glaciology and Geocryology, 31(3): 446-456.

Ye BS, Yang DQ, Ding YJ, Han TD, 2007. A bias-corrected precipitation climatology for China. Acta Geographica Sinica, 29(1): 3-13.

Zou JS, Cao CZ, 1991. Some factors of impacts on snowfall over Qinghai-Tibet plateau. Advances in Water Science, 2(1): 42-49.

Zou JS, Cao CZ, 1989. Climatology analysis of snowfall over Qinghai-Xizang Plateau. Chinese Journal of Atmospheric Sciences, 13(4):400-409.

10.3724/SP.J.1226.2011.00051

22 July 2010 Accepted: 25 October 2010