Trend of snow cover fraction over East Asia in the 21st century under different scenarios

Fang Wang , YiHui Ding

National Climate Center, China Meteorological Administration, Beijing 100081, China

Trend of snow cover fraction over East Asia in the 21st century under different scenarios

Fang Wang*, YiHui Ding

National Climate Center, China Meteorological Administration, Beijing 100081, China

Using the snow cover fraction (SNC) output from eight WCRP CMIP3 climate models under SRES A2, A1B, and B1 scenarios, the future trend of SNC over East Asia is analyzed. Results show that SNC is likely to decrease in East Asia, with the fastest decrease in spring, then winter and autumn, and the slowest in summer. In spring and winter the SNC decreases faster in the Qinghai-Xizang Plateau than in northern East Asia, while in autumn there is little difference between them. Among the various scenarios, SRES A2 has the largest decrease trend, then A1B, and B1 has the smallest trend. The decrease in SNC is mainly caused by the changes in surface air temperature and snowfall, which contribute differently to the SNC trends in different regions and seasons.

snow cover; future trend; Qinghai-Xizang Plateau; East Asia

1. Introduction

As the most extensive component and the most rapidly and seasonally changing variable in the cryosphere, snow cover plays an important role in modulating the Earth’s climate. On the one hand, snow cover greatly reflects solar radiation reaching the Earth’s surface, decreasing local air temperature. On the other hand, snow cover affects local water balance and hydrological cycles. Consequently, both effects lead to changes in atmospheric circulation, temperature, and precipitation (Vavrus, 2007).

The snow over Eurasia, especially over the Qinghai-Xizang Plateau (hereafter QXP), which is the most important snow-covered area in the world, has a great effect on Asian climate. Previous studies have focused on the spatial distribution and interdecadal changing of snow cover over Asia and the QXP (Wei and Lu, 1995; Li, 1996; Ke and Li,1998; Weiet al., 2002). A close relationship has been found between Eurasian snow cover and the Asian monsoon. An increase in Eurasian snow cover may cause weakened or delayed Asian summer monsoon; however, this relationship is very complex (Yang and Wu, 2009). A relationship has also been found between Tibetan snow cover and the Asian summer monsoon (Fanet al., 1997; Yang and Yao, 1998;Zhang and Tao, 2001; Zhu and Ding, 2007). Abnormal winter snow cover in Tibet may also cause anomalies in the East Asian winter monsoon and further affects the distribution of the subsequent summer monsoon rainfall in China (Chenet al., 2000; Penget al., 2005; Wuet al., 2009). For example,snow cover during winter and spring over Eurasia and Tibet is linked to the summer flood and drought in China and is considered a key factor for precipitation prediction during the flooding season (Chen and Song, 2000).

On interdecadal timescale, the variation of precipitation pattern ("southern flood and northern drought") in eastern China has a significant correlation to atmospheric heat source and winter and spring snow cover in Tibet (Zhuet al., 2007).Model simulations also show that above-normal snow cover over Tibet is an important reason for the occurrence of the"southern flood and northern drought" precipitation pattern in eastern China (Zhuet al., 2009). Furthermore, abnormal snow cover also strongly influences atmospheric circulation (Chenet al., 2003; Chen and Sun, 2003). Given the importance of snow cover for the climate over Asia, it can be inferred that the future change in snow cover is likely to have a direct impact on the monsoon climate in East Asia.

The Intergovernmental Panel on Climate Change (IPCC)Fourth Assessment Report (AR4) (IPCC, 2007) indicates that snow cover is likely to decrease greatly in the 21st century due to global warming. At the end of the 21st century,the projected reduction in the annual mean snow cover in the Northern Hemisphere is 13% under the SRES B2 scenario.In North America, a robust negative trend has been projected for snow extent by climate models under the SRES A1B scenario (Frei and Gong, 2005). However, how snow cover over East Asia is likely to change in the future has been given scant attention in recent studies. In this study, we first analyze the ability of multiple climate models to simulate snow cover over East Asia, and then depict the possible trend of snow cover over East Asia in the 21st century against the background of global warming, mainly on the seasonal timescale.

2. Data and methods

2.1. Data

There are at least two ways to represent snow extent.One is the actual area, with a unit of km2, and the other is snow cover fraction (hereafter SNC), which is the ratio of snow cover area to grid area in a single grid, with a unit of%. The latter is often used to represent snow cover area in climate models and it will be used in our study. The model data are from the WCRP CMIP3 Multi-Model Database,and only eight models are selected from the total of 23 models for our study because they have complete output for both 20th-century simulation and 21st-century projection under the SRES A2, A1B, and B1 scenarios. The eight models are CGCM3.1 (T47) (Canada), CSIRO-Mk3.0(Australia), CSIRO-Mk3.5 (Australia), GISS-ER (USA),INM-CM3.0 (Russia), MIROC3.2 (medres) (Japan),MRI-CGCM2.3.2 (Japan), and CCSM3 (USA). The variables used include monthly mean SNC, surface air temperature,and snowfall. The model output and detailed model description are available at http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php. In addition, the NOAA monthly mean snow cover data from January 1971 to September 1995 are used for validation purposes. The data are derived from the Northern Hemisphere EASE-Grid Weekly Snow Cover, and are available at http://www.esrl.noaa.gov/psd/data/gridded/data.snowcover.html.

2.2. Analysis method

Due to the large difference in resolution between different models, and for the convenience of ensemble and comparison with validation data, both the model and the validation data are interpolated onto a standard 2.5°×2.5° latitude-longitude grid by using a bilinear interpolation algorithm. The multi-model ensemble is expressed as:

whereNis the number of models (eight in our study).Si,j,t,mis the SNC for themth model at theith longitude, thejth latitude, and at timet.Si,j,t,eis the same asSi,j,t,mbut for the ensemble, with the subscripteindicating the multi-model ensemble.

To obtain the long-term trend of SNC, the simple linear regression method is used. Empirical orthogonal function(EOF) analysis and the multiple linear regression method are also applied. In this study, all curves have been smoothed with the nine-point quadratic smoothing formula.

Considering the features of SNC distribution and seasonality, we selected two key areas, namely, Tibet (27.5°–40°N,70°–105°E; hereafter TB) and northern East Asia land(40°–60°N, 70°–140°E; hereafter NH), to explore the future changes in SNC over East Asia. The seasons are defined as winter (DJF), spring (MAM), summer (JJA), and fall (SON).

3. Ability of climate models to simulate SNC over East Asia

Figure 1 gives a comparison of SNC climatology between NOAA observations and the multi-model ensemble(only figures for summer and winter are shown). It can be seen that the basic SNC distribution is simulated reasonably well. For example, SNC is mainly located in Tibet in summer, and it extends to the northern land area of East Asia in the other seasons. However, large deficiencies can be seen as well, especially the inability of the models to simulate the Tibet SNC pattern in summer when there is more SNC around Tibet from the Pamir Mountains to the Himalayas and the eastern part of the Nyainqentanglha Mountains, but less SNC in central Tibet. This model deficit may be caused by the coarse resolution of the climate models used in our study, which are unable to simulate the sub-grid snow processes in these regions that have complex terrain and underlying surfaces. Furthermore, similar simulation bias can also be found in the non-summer seasons. Also, SNC is underestimated for NH in winter and spring. It should be noted that an obvious positive bias occurs in central China, which may result in spurious trends when climate models are used to project SNC trends in these areas.

As for regional averages, great discrepancies can be seen for all seasons between the different models (Figure 2; only winter features are shown). In spring, the SNC of the multi-model ensemble (NOAA) is 44% (43%) and 40% (25%)in NH and TB, respectively. In fall, it is 31% (25%) and 26% (16%), and in winter it is 76% (80%) and 57% (32%).In summer, SNC is mainly located in TB, which is 8% in the multi-model ensemble, less than that for NOAA (13%).Overall, in NH, SNC simulation is closer to the NOAA observations. However, in TB, it is obviously larger than the NOAA observations in fall, winter, and spring, and is smaller in summer.

Figure 1 Distribution of multi-year mean of snow cover fraction in winter (a, c) and summer (b, d) in East Asia from NOAA observation data (a, b) and the multi-model ensemble (c, d)

Figure 2 Historical variations of snow cover fraction in northern East Asia (a) and the Qinghai-Xizang Plateau (b)in winter from 1961 to 2000

Changes in SNC are usually affected by many factors,such as air temperature, snowmelt rate, and snowfall, which lead to great uncertainties in simulating the interannual variation and long-term trends of SNC by climate models. From Figure 2 we can see large inconsistencies between different models in simulating the interannual variation of SNC in winter. Only a few models show significant correlation to the NOAA observations. As to linear trends (Table 1, calculated from the 24-year data from 1971 to 1994), the multi-model ensemble successfully simulates the signs of the trends except in fall of NH. However, the model values are much smaller than those of NOAA observations.

Table 1 Linear trends (unit: %/decade) of snow cover fraction from 1971 to 1994

It must be noted that the NOAA SNC data have deficiencies, such as low resolution and being susceptible to cloudiness, especially the change in statistical method in 1981 that led to an SNC error from tens to 3×104km2(Weiet al., 2002). Thus, great uncertainty may exist if only NOAA data are used to validate models. In addition, previous studies based on station observation data indicate that the Tibet snow experienced an obvious increase from the 1960s to the 1980s, and then decreased from the 1990s (Weiet al., 2002; Lei and Fang, 2008). This aspect cannot be captured in the climate models used in our study. The station-observed snow depths are not equal to the SNC output by climate models, either. Therefore, these results from station-observed snow depths can only serve as a reference to our results.

It must be noted that climate models have only limited ability to simulate the interannual changes of SNC, however,they show some skills in simulating the longterm trend of SNC. Furthermore, we believe the projected consistent global warming in the 21st century is the strongest influential factor, which may weaken the effects of other uncertainties. So we nevertheless used our model output to project the SNC trends in the 21st century.

4. Possible changes of SNC in East Asia

4.1. SNC trends

Figure 3 provides the SNC curves of the multi-model ensemble for the two key areas, TB and NH, during 2010–2099 under different scenarios. In all seasons, SNC is projected to decrease significantly for both areas in the future. For TB, under the SRES A2 scenario, the linear percentage trends are -1.42, -0.42, -0.90, and -1.16 per decade for MAM, JJA, SON, and DJF, respectively. Under the A1B scenario, the trends are -1.06, -0.39, -0.69, and -0.82 per decade. Under the B1 scenario, they are -0.62, -0.26, -0.44,and -0.50 per decade. As for NH, under SRES A2, the trends are -1.17, -0.93, and -0.51 per decade for MAM,SON, and DJF, respectively (JJA is not considered here due to its negligible SNC). Under A1B, the trends are -0.86,-0.70, and -0.41 per decade, and under B1 they are -0.40,-0.39, and -0.20 per decade (all of these trends significantly exceed the 99% confidence level). It can be seen that the SRES A2 has the largest decrease trend, then A1B, and the B1 has the least decrease, which is applicable to all seasons and areas. Under the same scenario, SNC decreases the fastest in MAM, then DJF and SON, and is slowest in JJA.In MAM and DJF, SNC decreases faster in TB than in NH,while in SON there is little difference between MAM and DJF.

Considering the above-discussed uncertainty of SNC projections, we examined the consistency between the ensemble and the different models. Under A2 (Figure 4),the 8 (8), 8 (8), 7 (5), 8 (8), 8 (8), 8 (8), and 8 (6) models give negative trends for NH_MAM, NH_SON, NH_DJF,TB_MAM, TB_SON, TB_DJF, and TB_JJA, respectively(the numbers in parentheses passed the significance test at the 99% confidence level). Under A1B the 8 (8), 8 (8), 6 (5), 8 (8),8 (8), 7 (7), and 8 (6) models give negative trends, and under B1 the 8 (6), 8 (7), 6 (4), 8 (6), 8 (7), 8 (5), and 8 (7) models give negative trends. Overall, most models show an agreement in projecting the future SNC trend over East Asia, and only small differences exist between the different models.

To further understand temporal and spatial variations, we applied EOF analysis to the annual mean SNC data(2010–2099) over East Asia (Figure 5). The results show that the first EOF, which explains 74.4% of the total variance, is characterized by a uniform positive distribution over East Asia, with the largest value in the Tibet. The corresponding time coefficients decrease significantly in the 21st century. The second EOF, explaining 3.6% of the total variance, displays a distribution of negative values in Tibet and its eastern area, and positive values over the northern land. The third EOF, accounting for 2.4% of the total variance, is negative in northern China and positive in Tibet and the northern part of NH. It is noted that the second and third EOFs may be associated with the interdecadal change of the East Asian monsoon in spite of their smaller contributions to the total variance.

Figure 3 Projected variations of snow cover fraction of the Qinghai-Xizang Plateau and northern East Asia in 2010–2099 for the multi-model ensemble under different scenarios compared to the average of 1961–2000. (a) TB_MAM: spring of Tibet;(b) NH_MAM: spring of northern East Asia; (c) TB_SON: fall of Tibet; (d) NH_SON: fall of northern East Asia;(e) TB_DJF: winter of Tibet; (f) NH_DJF: winter of northern East Asia; (g) TB_JJA: summer of Tibet.

Figure 4 Linear trends of snow cover fraction in 2010–2099 simulated by different models under the SRES A2 scenario.NH_MAM: spring of northern East Asia; NH_SON: fall of northern East Asia; NH_DJF: winter of northern East Asia;TB_MAM: spring of Tibet; TB_SON: fall of Tibet; TB_DJF: winter of Tibet; TB_JJA: summer of Tibet.

Figure 5 EOF analysis of projected snow cover fraction in East Asia from 2010 to 2099. Left for spatial distribution and right for time coefficient. (a), (b) for the first mode; (c), (d) for the second mode; (e), (f) for the third mode.

4.2. Causes of SNC changes

Among the various reasons for SNC changes, surface air temperature (SAT) is perhaps the most important factor.Continuing global warming is likely the direct reason for the SNC decrease. To illustrate this aspect, we selected SRES A2 as an example. The linear trends of SAT and snowfall are given in Table 2 for TB and NH during 2010–2099. It can be seen that SAT is likely to increase at a rate of 0.44–0.58 °C per decade in the future. The fastest warming generally occurs in winter, then spring and autumn, and the least in summer. SNC shows a very significant negative correlation with SAT, indicating that the acceleration of snowmelt due to global warming is the main cause of SNC decrease.

Snowfall also affects the SNC to some extent. According to the trend of snowfall, changes in snowfall depend greatly on seasons and areas. Except for the increasing trend in winter of NH, snowfall decreases consistently, with the fastest decrease in spring and autumn of TB, and the slowest decrease in winter of TB. This difference in snowfall may be reflected in SNC.

Table 2 Linear trends of surface air temperature (T) and snowfall (P) during 2010–2099 under the SRES A2 scenario

Considering the strong linear relationships of SNC with SAT and snowfall, a set of multiple linear regression equations are established for different areas and seasons as follows (to better understand the contribution of SAT and snowfall to SNC, the constant is set to zero):

whereTis air temperature andPis snowfall. The regression coefficients are listed in Table 3. All seven equations pass the significance test at the 99% confidence level and are able to fit the future SNC changes. From these equations it can be seen that temperature rise has a positive contribution to the reduction of SNC. However, an increase in snowfall can play a supplementary role in SNC and slow its decrease.Without considering snowfall, the largest SNC decrease with SAT occurs in spring, then fall and winter, and the least in summer. In contrast, if SAT is not considered, SNC is most sensitive to the snowfall in spring of TB and fall of NH but least sensitive in spring of NH. Overall, the response of SNC to SAT and snowfall shows significant differences between different areas and seasons.

To better understand the role of SAT and snowfall in SNC changes, the contributions to SNC trend are given in Table 4 for different areas and seasons. In spring, the SNC exhibits very large response to the changes of SAT in both NH and TB. However, the contribution of snowfall to SNC in TB is much greater than that in NH, leading to the largest SNC decreasing trend in spring of TB, followed by that in spring of NH. In fall, the SNC trends in NH and TB are closer, mainly due to the small differences in the contributions from SAT and snowfall. In winter, the contribution of SAT in TB is larger than that in NH, combined with the snowfall increase in NH that slows down the decrease in SNC, resulting in much larger change of SNC in TB as compared to that in NH. In summer,the response of SNC to SAT is small, so the SNC change is mainly caused by the decrease in snowfall.

Table 3 Coefficients in Equation (2)

Table 4 Contribution linear trend (unit: %/decade) of surface air temperature (T) and snowfall (P) to snow cover fraction under the SRES A2 scenario

5. Conclusions and discussion

The output of eight WCRP CMIP3 models is used to assess the ability of model simulation to project future SNC trends under three SRES scenarios. Our main conclusions are as follows:

(1) Climate models are somewhat able to simulate the distribution of SNC climatology but there are still obvious deficiencies as compared to observations, especially in Tibet with its complex terrain. Models generally perform poorly in simulating the interannual variation of SNC;however, the ensemble mean can give the correct sign of an SNC trend as observed in most cases, in spite of a large gap in magnitude.

(2) We believe that SNC is likely to decrease consistently in East Asia land in the future. Under the same scenario, SNC decreases the fastest in spring, then winter and autumn, and the slowest in summer. SNC decreases faster in the TB than in NH in spring and winter, while there is little difference in fall. Also, SRES A2 has the largest decrease trend, then A1B, and B1 has the least trend.

(3) SAT and snowfall are the two main causes of SNC decrease; however their contributions are significantly different depending largely on areas and seasons.

As shown by previous studies, decrease in SNC may play a critical role in the climate over East Asia. Firstly, a decrease in SNC reduces surface albedo, which in turn increases the surface radiation balance and local temperature.As a result, rising temperature further reduces SNC, forming a positive feedback loop. However, to what extent the feedback intensity can reach is still unknown. Secondly, as that was pointed out by Sunet al. (2010), both the monsoon circulation and the related precipitation are likely to intensify or increase in the future over East Asia, based on an analysis of the output from IPCC AR4 climate models. Combined with our study, these results may to some degree be related to the projected decrease in SNC over East Asia, which should be confirmed by further investigations.

It is noted that there are still great deficiencies in simulating snow processes by the current climate models, due to both the models themselves and the uncertainty in scenarios.Therefore, results of this study can only represent the ability of the current climate models in simulating the future change in SNC. However, it is believed that our understanding of snow and its impact on climate change will be enhanced with the improved understanding of the physical mechanisms of snow processes.

We acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI),and the WCRP’s Working Group on Coupled Modelling(WGCM) for their roles in making available the WCRP CMIP3 multi-model dataset. Support for this dataset is provided by the Office of Science, U.S. Department of Energy.This study was supported by the National Key Science and Technology Program of Ministry of Science and Technology of China (Grant No. 2007BAC03A01).

Chen HS, Sun ZB, 2003. The effects of Eurasian snow cover anomaly on winter atmospheric general circulation––Part I: observational studies.Chinese Journal of Atmospheric Sciences, 27(3): 304–316.

Chen HS, Sun ZB, Zhu WJ, 2003. The effects of Eurasian snow cover anomaly on winter atmospheric general circulation––Part II: Model simulation. Chinese Journal of Atmospheric Sciences, 27(5): 847–860.

Chen QJ, Gao B, Zhang Q, 2000. Studies on relation of snow cover over the Tibetan Plateau in winter to the winter-summer monsoon change. Chinese Journal of Atmospheric Sciences, 24(4): 477–492.

Chen XF, Song WL, 2000. Analysis of relationship between snow cover on Eurasia and Qinghai-Xizang Plateau in winter and summer rainfall in China and application to prediction. Plateau Meteorology, 19(2):214–223.

Fan GZ, Luo SW, Lu SH, 1997. The preliminary numerical experiments of effect of anomalous snow cover over Plateau in winter on east and south Asian summer monsoon. Plateau Meteorology, 16(2): 140–152.

Frei A, Gong G, 2005. Decadal to century scale trends in North American snow extent in coupled atmosphere-ocean general circulation models.Geophysical Research Letters, 32(18): L18502.

Intergovernmental Panel on Climate Change (IPCC), 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK.

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

Lei J, Fang ZF, 2008. A comparison study on snow cover of the conventional meteorological observation and its variational tendency in Qinghai region. Plateau Meteorology, 27(1): 58–67.

Li PJ, 1996. Monitoring snow cover on the High-Asia. Journal of Glaciology and Geocryology, 18(special issue): 105–114.

Peng JB, Chen LT, Zhang QY, 2005. Multi-scale variations of snow cover over QXP and tropical pacific SST and their influences on summer rainfall in China. Plateau Meteorology, 24(3): 366–377.

Sun Y, Ding YH, 2010. A projection of future changes in summer precipitation and monsoon in East Asia. Science in China, Series D: Earth, (2):284–300.

Vavrus S, 2007. The role of terrestrial snow cover in the climate system.Climate Dynamics, 29(1): 67–73.

Wei ZG, Lu SH, 1995. Distribution of snow cover on the Qinghai-Xizang Plateau and its influence on surface albedo. Plateau Meteorology, 14(1):67–73.

Wei ZG, Huang RH, Chen W, 2002. Spatial distributions and interdecadal variations of the snow at the Tibetan Plateau weather stations. Chinese Journal of Atmospheric Sciences, 26(4): 496–508.

Wu BY, Yang K, Zhang RH, 2009. Eurasian snow cover variability and its association with summer rainfall in China. Advances in Atmospheric Sciences, 26(1): 31–44.

Yang K, Wu BY, 2009. Advances in researches on relationship between Eurasia snow cover and Asian monsoon. Meteorological Science and Technology, 37(3): 263–270.

Yang MX, Yao TD, 1998. A review of the study on the impact of snow cover in the Tibetan Plateau on Asian monsoon. Journal of Glaciology and Geocryology, 20(2): 186–191.

Zhang SL, Tao SY, 2001. Influences of snow cover over the Tibetan Plateau on Asian summer monsoon. Chinese Journal of Atmospheric Sciences,25(3): 372–390.

Zhu YX, Ding YH, 2007. Influences of snow cover over Tibetan Plateau on weather and climate: Advances and problems. Meteorological Science and Technology, 35(1): 1–8.

Zhu YX, Ding YH, Liu HW, 2009. Simulation of the influence of winter snow depth over the Tibetan Plateau on summer rainfall in China. Chinese Journal of Atmospheric Sciences, 33(5): 903–915.

Zhu YX, Ding YH, Xu HG, 2007. The decadal relationship between atmospheric heat source of winter and spring snow cover over Tibetan Plateau and rainfall in East China. Acta Meteorologica Sinica, 65(6):946–958.

10.3724/SP.J.1226.2012.00107

*Correspondence to: Fang Wang, senior engineer of National Climate Center, China Meteorological Administration. No. 46,Zhongguancun South Avenue, Haidian District, Beijing 100081, China. Tel: +86-10-58995335; Email: fangwang@cma.gov.cn

May 22, 2011 Accepted: July 30, 2011