Abstract

Potential benefits or disadvantages of increasing precipitation in high-latitude regions under a warming climate are dependent on how and in what form the precipitation occurs. Precipitation frequency and type are equally as important as quantity and intensity to understanding the seasonality of hydrological cycles and the health of the ecosystem in high-latitude regions. This study uses daily historical synoptic observation records during 1936–90 over the former USSR to reveal associations between the frequency of precipitation types (rainfall, snowfall, mixed solid and liquid, and wet days of all types) and surface air temperatures to determine potential changes in precipitation characteristics under a warming climate. Results from this particular study show that the frequency of precipitation of all types generally increases with air temperature during winter. However, both solid and liquid precipitation days predominantly decrease with air temperature during spring with a reduction in snowfall days being most significant. During autumn, snowfall days decrease while rainfall days increase resulting in overall decreases in wet days as air temperature increases. The data also reveal that, as snowfall days increase in relationship to increasing air temperatures, this increase may level out or even decrease as mean surface air temperature exceeds −8°C in winter. In spring and autumn, increasing rainfall days switch to decreasing when the mean surface air temperature goes above 6°C. The conclusion of this study is that changes in the frequency of precipitation types are highly dependent on the location’s air temperature and that threshold temperatures exist beyond which changes in an opposite direction occur.

1. Introduction

Common sense seems to tell us that, when air temperature increases, more snow will become rain and thus snowfall days will decrease and rainfall days will increase. Owing to the complexity of the climate system and its feedback processes, the relationship between precipitation types and frequency may not be as simple previously thought, especially in high-latitude regions where air temperatures frequently stay below the freezing point during the cold season. Studies of winter snowfall amount in high-latitude regions suggest that snowfall increases with air temperature as long as air temperature stays below a certain point depending on the geographical location (Davies et al. 1999; Ye and Mather 1997). Studies have also found that snow has provided a declining proportion of total annual precipitation in southern Canada, western Siberia, and England, but no changes have been found in regions north of 55°N under a warming climate (Huntington et al. 2004; Trenberth et al. 2007). Furthermore, changes in precipitation amount do not necessarily go in the same direction as changes in frequency of wet days since the frequency and intensity of heavy precipitation account for a larger share of change as air temperature increases.

Studies of precipitation change under a warming climate over the Arctic region have mostly consisted of trend analyses of total precipitation (Groisman et al. 1999; Serreze et al. 2000; Ye 2001a), snow accumulation (Brown and Braaten 1998; Ye 2001b; Ye et al. 1998), and high-intensity rainfall days (Groisman et al. 2005; Khon et al. 2007; Semenov and Bengtsson 2002). Most of these trend studies assumed that air temperature has increased significantly and thus trends are likely to be associated with a warming climate during the study period of their research. A recent study by Groisman et al. (2006) indicated that the outcome of their trend analysis was really dependent on the time period they studied over northern Eurasia during which warming did not materialize until recent decades. For example, if we study trends over northern Eurasia for the time period from 1951 to 1995, increases in air temperature are significant; however, warming is not evident if we examine them for the time period of 1936–95. With the dramatic reduction in weather stations starting in the early 1990s when the former USSR disintegrated (Groisman et al. 2006), a threat was posed to trend analyses due to problems in spatial coverage and thus the continuity of the time series.

Confidence in the validity of the warming trend over the Arctic in recent decades and the future has become very high in the scientific community based on overwhelming observational evidence and the projections of climate models (Alley et al. 2007). An alternative way to examine potential changes in certain environmental parameters is to examine their relationships with air temperature, especially in regions where there are problems with data continuity during recent decades when significant warming has occurred. Of course, this method is based on an assumption that the relationship between these parameters and air temperatures will stay the same in the future and that relationships derived from studies in one location will be applicable in another. Although this is not a perfect scientific assumption, past relationships do provide valuable information about the future and can be used to provide insights into potential changes in the future. This is especially true if large geographical regions are examined and the relationships between different localities with different climate conditions are compared.

Some studies have used the outputs of climate model simulations to describe potential changes under a warming climate. For example, the global climate model output under anthropogenic forcing scenarios suggests increases in precipitation frequency, amount, and intensity during winter over four major river basins in northern Eurasia (Khon et al. 2007). Increases in precipitation and precipitation intensity during winter in Europe are predicted during the years of 2071–2100 based on the Intergovernmental Panel on Climate Change (IPCC) A2 and B2 emissions scenarios (Giorgi et al. 2004) and disproportional increases in heavy precipitation events for 2000–99 using the ECHAM4/Ocean Isopycnal Model (OPYC3) (Semenov and Bengtsson 2002). But models are still not ideal for predicting realistic snow conditions and thus are not satisfactory for separating solid from liquid precipitation types accurately (Frei et al. 2003; Roesch and Roechner 2006). Given the significance of snow to regulating the energy budget and hydrological cycles over Arctic land surfaces, changes in the frequency of solid precipitation under a warming climate require special attention.

This study uses historical synoptic observation records over the former Soviet Union to examine changes in snowfall days, rainfall days, mixed-phase precipitation days, and wet days (including all forms of precipitation) associated with surface air temperatures during 1936–90.

2. Data and methodology

The synoptic weather data are from the 6- and 3-hourly meteorological observations from 223 USSR stations available at the Carbon Dioxide Information Analysis Center (CDIAC), Oak Ridge National Laboratory, Oak Ridge, Tennessee (ORNL/CDIAC-180, NDP-048/R1; http://cdiac.esd.ornl.gov). Each station record consists of 6- (1936–65) and 3-hourly (1966–90) observations of 24 meteorological variables including air temperature, past and present weather type, precipitation amount, cloud amount and type, sea level pressure, relative humidity, and wind speed and direction. The data have undergone extensive quality assurance by the All-Russian Research Institute of Hydrometeorological Information–World Data Centre (RIHMI-WDC), the National Climatic Data Center (NCDC), and the CDIAC (Razuvaev et al. 1995). The changes in observation times through the two different time periods (before and after 1966) have been adjusted based on station time zone.

Snowfall, rainfall, mixed-phase precipitation, and wet days are extracted based on weather codes recorded for the observation time and during the preceding hour. Using weather codes based on visual observations avoids precipitation data problems associated with wetting losses, changes in the definition of trace precipitation, and instrumentation changes over the study region. Snowfall days include any day on which any form of solid precipitation including snow, ice pellets (or sleet), snow showers, and diamond dust, but not hail, has fallen. Rainfall days include any day on which any form of liquid precipitation falls, including drizzle, rain, showers, freezing rain, and thunderstorms associated with liquid products. Any types of mixed solid and liquid precipitation are not included in either the rain or snow category; instead, they are grouped into a mixed-phase precipitation category that also includes days on which both solid and liquid precipitation were observed during different observation hours in the study. For wet days, all forms of precipitation are included.

To stay consistent with the number of observations per day, only four observations per day are used throughout the study time period. Thus, for the later period starting in 1966, only the four observations that occurred closest to the observation times in previous years are used. If there is one observation showing snow or rain, the day is considered as a snow or rain day. There are a very small number of days during which both snow and rain occurred at different observation times; these days are counted as both snowfall and rainfall days and also counted as mixed-phase precipitation days. Wet days are defined as any day during which any form of precipitation was recorded for at least one of the four observations. Each category of precipitation type is extracted from the original dataset independently.

The number of days falling into each category of snow, rain, mixed phase, and wet are totaled for each month and then grouped into three seasons: winter [December–February (DJF)], spring [March–June (MAMJ) to include all snowfall events in late spring], and fall [September–November (SON)]. If there is one day that has missing weather records, that season is considered missing. Among these 223 original stations, 80 stations are retained for analyses based on the criteria that they have quality data starting no later than 1940–41. The number of missing years for snowfall days ranges from 1–14 for winter, 2–16 for spring, and 2–14 for fall. The number of missing years for rainfall days ranges from 2–15 for winter and spring and 2–16 for fall during the 55 years of the study period.

Daily air temperature—measured at 2 m AGL—is averaged from the same four observation times. If there is one observation missing, the day is considered as missing and, if more than 10% of observations are missing, the monthly and, thus, seasonal mean temperature is considered as missing.

Correlation analysis is applied to each station between each category of precipitation days and air temperature for each season and the statistical significance level is determined based on sample size. The rate of change is determined based on simple linear regression analysis (von Storch and Zwiers 1999) with precipitation days as a dependent variable and air temperature an independent variable. The coefficient of the slope represents the rate of precipitation day change per each degree of air temperature increase for the study time period. A scatterplot of correlation coefficients/rates of change versus seasonal air temperature is produced for all stations to reveal changes in the relationships based on air temperature conditions across the study region.

To further examine the possible shift from solid to liquid precipitation associated with air temperature change during spring and fall, days are classified into two groups based on the daily mean air temperature for each season. Since the critical temperature range for a possible switch from solid to liquid form is from −5°C to +5°C, group 1 only includes days in which mean daily air temperature falls within that range in order to study the relationships between both snowfall and rainfall days and air temperature. Group 2 only includes days on which mean daily air temperature is higher than 5°C to study the relationship between changes in rainfall days and air temperature. Since the total number of days for each group in each season varies from one year to the next, the extracted number of snowfall/rainfall days from each group is divided by the total days in the group of the corresponding season and multiplied by 100 to standardize the data. Therefore, the snowfall/rainfall days derived for each group is actually the percentage of days on which snowfall/rainfall occurred for each group in each season. Then the snowfall/rainfall days are correlated with mean seasonal air temperatures averaged from the corresponding group. In addition, analyses are applied to combined spring and fall seasons for each group to test if results are sensitive to the sample sizes and further confirm the findings.

3. Results

Seasonal mean total snowfall and rainfall days are shown in Fig. 1. Snowfall days increase from south to north in accordance with decreasing distance to the Arctic Ocean in combination with air temperature distributions. Snowfall days range from 20 to 65 days for the winter three months (Fig. 1a), 10 to 70 days for the four spring months (Fig. 1b), and 10 to 55 days for the three autumn months (Fig. 1c). The number of rainfall days increases toward the west with the largest number occurring over European Russia. During winter, there are very few rainfall days—fewer than 2 days over Siberia and 2–6 over European Russia. Rainfall increases quickly toward the westernmost stations in the study region (Fig. 1d). During spring, rainfall days range from 12 to 28 days, the highest number occurring over European Russia (Fig. 1e). During autumn, rainfall days range from 6 to 36 days (Fig. 1f). Mixed-phase precipitation days range from 1 to 3 days in Siberia and are mostly 6–9 days over European Russia during winter (Fig. 2a). During spring, the mixed-phase days increase to 3–11 days in Siberia and 4–13 days over European Russia (Fig. 2b). There were 3–9 mixed-phase days over Siberia and 3–12 over European Russia during autumn (Fig. 2c). Wet day distributions generally follow that of snowfall days. In winter, the number of wet days is very similar to that of snow days over Siberia, but it reaches 70 over northern European Russia (Fig. 2d). During spring, wet days range from 20 to 80 days (Fig. 2e) and, during autumn, 20 to 70 wet days (Fig. 2f).

Fig. 1.

Mean snowfall and rainfall day distribution over the study region for winter (DJF), spring (MAMJ), and autumn (SON) during the time period 1936–90. Circles indicate weather station locations.

Fig. 1.

Mean snowfall and rainfall day distribution over the study region for winter (DJF), spring (MAMJ), and autumn (SON) during the time period 1936–90. Circles indicate weather station locations.

Fig. 2.

As in Fig. 2 but for mixed-phase days and wet days.

Fig. 2.

As in Fig. 2 but for mixed-phase days and wet days.

The correlation coefficients between snowfall days and surface air temperature have positive values for 58 stations during winter (Fig. 3a). Nineteen stations are statistically significant at the 95% confidence level and above, and most of them are located over the northern portion of the study region. Negative values are found at the rest of the 22 stations located on the southern and western edges of the study region with only four stations showing statistical significance during the winter season. This suggests that more snowfall days are expected over large areas of northern Eurasia but probably fewer snowfall days over the southern edge of northern Eurasia under warming air temperatures during the winter season.

Fig. 3.

Distribution of correlation coefficients between snowfall/rainfall days and surface air temperature. Red: positive correlation; blue: negative correlation. Solid large filled circle: statistically significant at the 99% confidence level and above; solid small filled circles: statistically significant at the 95% confidence level and above.

Fig. 3.

Distribution of correlation coefficients between snowfall/rainfall days and surface air temperature. Red: positive correlation; blue: negative correlation. Solid large filled circle: statistically significant at the 99% confidence level and above; solid small filled circles: statistically significant at the 95% confidence level and above.

For the spring season, negative correlation coefficient values are found at 76 stations, and 51 of them are statistically significant at the 95% confidence level and above (Fig. 3b). For the autumn season, 79 stations show negative correlation coefficients and 51 of them are statistically significant at the 95% confidence level and above (Fig. 3c). No significant positive correlation stations are found for both seasons. This indicates that snowfall days decrease as air temperature increases during both spring and autumn seasons over the study region.

The relationship between rainfall days and air temperature is predominantly positive (60 stations) during the winter season, and 29 stations, mostly located over European Russia, are statistically significant at the 95% confidence level and above (Fig. 3d). There is only one station showing a significant negative correlation. It should be noted that, if the number of statistically significant stations is less than four (5% of total stations), this could be due to random chance. Seven stations over central and eastern Siberia have almost no rainfall days, so they are not statistically meaningful and correlation is not performed (marked by small open black circles). This suggests that rainfall days increase as air temperature increases; this is most evident over European Russia during the winter season (since there are very few rainfall days over Siberia). During spring, there are 26 positive and 54 negative correlation stations. The only three statistically significant positive correlation stations are all located over northern European Russia, while there are 10 negative correlation stations over the southwestern portion of the region (Fig. 3e). For the autumn season, correlation is positive at most stations (70 stations), although only 17 are statistically significant and are located mostly over the higher-latitude regions of the region (Fig. 3f).

The scatterplot of correlation coefficients versus station air temperatures at all stations and for all three seasons is shown in Fig. 4a. It seems that no significant positive correlation between snowfall days and air temperature occurs when mean air temperature is about −8°C and above. There are a few incidences of negative correlation occurring at stations where the air is colder than about −8°C, but only three cases (one station during winter and two stations during autumn) that are statistically significant. This implies that, if mean seasonal air temperature remains below −8°C, snowfall days are likely to continue to increase as air temperature increases but may quickly switch to no change or decreasing if the mean air temperature approaches −8°C.

Fig. 4.

Scatterplot of correlation coefficients/rates of change vs station surface air temperatures. Blue star: winter statistically significant stations; blue triangle: winter nonstatistically significant stations; green plus: spring statistically significant stations; green diamond: spring nonstatistically significant stations; orange cross: autumn statistically significant stations; orange square: autumn nonstatistically significant stations. The ordinate is days per degree in a given season.

Fig. 4.

Scatterplot of correlation coefficients/rates of change vs station surface air temperatures. Blue star: winter statistically significant stations; blue triangle: winter nonstatistically significant stations; green plus: spring statistically significant stations; green diamond: spring nonstatistically significant stations; orange cross: autumn statistically significant stations; orange square: autumn nonstatistically significant stations. The ordinate is days per degree in a given season.

For spring and autumn, the correlation coefficients decrease (or become larger negative values) as air temperature increases, suggesting that higher air temperatures are likely associated with more significant negative correlation with snowfall days. The rate of snowfall day change with each 1°C increase is mostly around 1–2 days during winter and snowfall decrease is around 1–3 days during spring and 1–4 days during autumn (Fig. 4b).

For rainfall days, the correlation coefficient evidently increases as air temperature increases during winter (Fig. 4c). During spring, there is a change from positive to negative correlations when the spring air temperature increases to above the freezing point. This is consistent with the common assumption [as was clearly stated in the IPCC report (Alley et al. 2007)] that snowfall will change to rainfall when air temperature rises above 0°C. For autumn, no significant positive correlations occur at stations with higher seasonal air temperatures of about 8°C or above. All of these positive and negative signs are consistent with the geographical distribution of the stations shown in Fig. 3.

The rate of rainfall day change increases quickly as air temperature increases during winter from near 0 to 2.5 days per 1°C temperature increase (Fig. 4d). For autumn, it increases from 0.5 to 2.5 days per 1°C temperature increase. For spring, the rate of rainfall day decrease is very small, less than 0.5 days per 1°C temperature increase, but the magnitude seems to increase slightly as air temperature increases. Once the rate of change is divided by the mean number of rainfall days, the trend disappears (not shown).

For mixed-phase precipitation days, results are quite similar to rainfall days but less significant. The winter season shows a prevailing positive correlation with air temperature; 61 stations show positive correlation with 23 statistically significant (Fig. 5a). Spring shows prevailing negative correlation; 62 stations show negative correlation, but only 7 are statistically significant (Fig. 5b). Autumn shows mixed positive and negative patterns with only 3 significant positive and 3 significant negative correlations (Fig. 5c).

Fig. 5.

As in Fig. 2 but for mixed-phase precipitation days and wet days.

Fig. 5.

As in Fig. 2 but for mixed-phase precipitation days and wet days.

For wet days, the winter season shows predominantly positive correlations (61 stations) with 29 stations being statistically significant (Fig. 5d); the spring season shows prevailing negative correlations (71 stations) with 41 being statistically significant (Fig. 5e). Autumn also shows a majority of stations with negative correlations (68) with 16 statistically significant stations (Fig. 5f). This suggests that the autumn season reduction in wet days is mostly due to the reduction in snowfall days. In other words, the decrease in snowfall days associated with air temperature is faster than the increase in rainfall days during the autumn season, as reflected in Figs. 4b and 4d. The correlation between rainfall days and snowfall days shows a mixture of positive and negative correlations with only 7 stations located over higher latitudes showing a statistically negative correlation at a 95% confidence level, suggesting that the reduction in snowfall days is not necessarily related to the increase in rainfall days in general for the study region during the autumn season.

The rate of change for wet days is between 1 and 4 days for winter; most stations have values of 1–2 days per 1°C temperature increase. This translates into about 3% of total wet days. For spring this ranges from −1 to −5 days per 1°C increase in air temperature, translating to about −7% wet days. For the autumn the rate ranges from −1 to −3 days with most stations having between −1 and −2 days corresponding to about −5% of wet days per 1°C increase in air temperature.

The correlation results for snowfall and rainfall days during spring and autumn using the group 1 data (days with mean air temperature between −5°C and 5°C) are shown in Fig. 6. The negative correlation between snowfall days and air temperature is evident in both seasons. Compared to Figs. 3b and 3c, which use entire days of the seasons, the largest improvement in the snowfall day and temperature correlation is at northern Siberian stations. During spring, there are 75 negative correlation stations with 43 significant at a 95% confidence level and above (Fig. 6a). It appears that the statistically significant stations changed from central European Russia to northern Siberia (Fig. 6a). For autumn, there are 78 negative correlation stations and 63 are statistically significant at a 95% confidence level and above (Fig. 6b). Again, the northern central Siberian stations become statistically significant. This reinforces the trend of snowfall days decreasing with air temperature when the daily temperatures are near the freezing point (definitely higher than −8°C).

Fig. 6.

Distribution of correlation coefficients between snowfall/rainfall days and surface air temperature for group 1 days (in which mean daily air temperature falls between −5°C and 5°C) for spring and autumn seasons.

Fig. 6.

Distribution of correlation coefficients between snowfall/rainfall days and surface air temperature for group 1 days (in which mean daily air temperature falls between −5°C and 5°C) for spring and autumn seasons.

The correlation between rainfall days and air temperature during spring using this group is opposite to that of using all spring days. All stations, except one, have a positive correlation between rainfall days and air temperature, with 57 stations statistically significant at the 95% confidence level and above (Fig. 6c). The positive correlation between rainfall and air temperature during autumn is more significant and geographically dominant compared to that of using all autumn days. All stations show positive correlation and 57 are statistically significant at the 95% confidence level and above (Fig. 6d). Results from combining the two seasons are very similar. For snowfall days, there are 76 negative correlation stations with 44 that are statistically significant at a 95% confidence level and above; for rainfall days, there are 79 positive correlation stations with 57 that are statistically significant at 95% and above. These results suggest that as air temperature increases snowfall days decrease and rainfall days increase across the entire study region when daily air temperature is around the freezing point of −5°C to 5°C.

Results for group 2 data (days with mean air temperature higher than 5°C) show that rainfall days have a predominantly negative correlation with air temperature for both seasons (Fig. 7). During spring all stations show negative correlation and 64 are statistically significant at the 95% confidence level and above (Fig. 7a). During autumn 71 stations have a negative correlation and 27 are statistically significant at the 95% confidence level and above (Fig. 7b). The stations with significant negative correlations are mostly located in the western and some southern parts of the study region. Thus, the decreases in autumn rainfall days are not as widespread as spring rainfall days over the study region for higher air temperature days. Results from the combined spring and autumn seasons are the same as that of spring season. The results from this group 2 data indicate that for days when air temperature is higher, rainfall days switch to decreasing as air temperature increases. Please keep in mind that these results are probably related to the continental nature of the region that receives its water vapor from remote oceans, and interpolation to other regions requires more analyses.

Fig. 7.

Distribution of correlation coefficients between rainfall days and air temperature for group 2 days (in which the mean daily air temperature is higher than 5°C) for spring and autumn seasons.

Fig. 7.

Distribution of correlation coefficients between rainfall days and air temperature for group 2 days (in which the mean daily air temperature is higher than 5°C) for spring and autumn seasons.

To further determine if there is a threshold temperature at which correlation between rainfall days and air temperature changes, the relationship between the two at monthly scale is examined for both spring and autumn seasons. Owing to the very large temperature range during these seasons, the study at monthly scale is another way to look for more specific information and reduce the range of air temperatures. The monthly correlation results are shown in Fig. 8. For the spring months, all significant correlations are positive in March. April and May have both significant positive and negative correlations. All stations except one have negative correlations in June (Fig. 8a). The switch from positive to negative correlation seems to occur around 6°C, after which significant negative correlations occur, except for two cases with positive correlation at air temperature around 7°C. The threshold of 6°C is a little clearer in autumn. All significant correlations in November and October are positive and there are both positive and negative correlations in September (Fig. 8b). The switch between significant positive and negative correlation occurs at 6°C. This suggests that rainfall days increase as air temperature increases until the air temperature reaches 6°C and then decrease as air temperature increases for this study region.

Fig. 8.

Scatterplot of correlation coefficients between monthly rainfall days and air temperature vs monthly mean air temperature for (a) spring (green: March; blue: April; red: May; orange: June), and (b) autumn (green: September; blue: October; red: November). Significant correlations indicated by the cross, star, and ×. Nonsignificant correlations indicated by the triangle, square, and diamond.

Fig. 8.

Scatterplot of correlation coefficients between monthly rainfall days and air temperature vs monthly mean air temperature for (a) spring (green: March; blue: April; red: May; orange: June), and (b) autumn (green: September; blue: October; red: November). Significant correlations indicated by the cross, star, and ×. Nonsignificant correlations indicated by the triangle, square, and diamond.

4. Summary and discussion

The purpose of this study is to understand potential changes in frequency of snowfall, rainfall, and wet days under a warming climate by examining their relationship with air temperature for 80 stations over northern Eurasia during the available continuous data record period from 1936 to 1990. If historical trends continue into the future as the climate warms we can expect more wet days in winter, while spring and autumn would have fewer wet days as air temperature increases over northern Eurasia. The rate of change ranges from 1 to 4 wet days (about 3% on average) increase with each 1°C temperature increase during winter, but 1 to 5 (about 7% average) and 1 to 3 days (about 5% average) reduction in wet days for spring and autumn for each 1°C temperature increase, respectively. These changes in wet days are predominantly due to changes in snowfall days. Increasing air temperature would bring more precipitation days of all types during winter, but decreasing snowfall days and possibly rainfall days during spring. The increases in rainfall days during autumn have not compensated for losses in snowfall days and thus fewer precipitation days occur as the climate becomes warmer. Also, the data show that the number of snowfall days increases significantly with increasing air temperature in northern parts of the study region; the increase becomes insignificant as you move south, and changes to decreasing snowfall days at the southern marginal stations where mean air temperature is −8°C and above. This suggests that increases in snowfall days could suddenly switch to no change or decreases if a station’s winter mean air temperature approaches this threshold or “tipping point” during the winter season. Afterward, large increases in rainfall days would still prevail thus wetter winters will last for a while until the mean monthly temperature reaches about 6°C. This also applies to the spring and autumn seasons. Rainfall days increase with air temperature until the station mean monthly air temperature reaches 6°C and above, then rainfall days switch to decreasing with increasing air temperature.

The relationship between frequency of snowfall and air temperature revealed in this study is, in general, consistent with the relationship between a water-equivalent amount of snowfall and air temperature found in Canada (Davis et al. 1999). Davis et al. revealed that increasing snowfall amounts with increasing air temperature switches to decreasing when the mean January air temperature goes above about −7°C at stations over eastern Canada and the west coast, but the relationships are more complex at stations located in the lee of the Rockies (where threshold temperatures are lower). In this study, over northern Eurasia, the highest winter-mean air temperature at stations with a significant positive correlation is −7.8°C and the lowest winter mean at stations with a significant negative correlation is −5.1°C. The cutoff at −8°C is based more on where the positive correlation ends than where the negative correlation starts. Thus the threshold temperature of transition could be in a range between −8°C and −6°C and could be defined more precisely if there were more statistically significant stations falling into this range. The shift in direction of relationship between snowfall days and air temperature also explains the observed differences between rainfall days and rain-on-snow days over European Russia during the same study time period (Ye et al. 2008). They found that the number of rain-on-snow days is very close to the rainfall days for stations where mean air temperature is below −8°C but rain-on-snow days dropped to about half of the rainfall days at stations where the air temperature is higher. This is apparently related to decreasing snowfall days—thus fewer snow cover days for locations where mean winter air temperature is −8°C and higher.

The opposite relationship between rainfall days and air temperature during spring and autumn is likely due to the different proportion of days that have a daily air temperature below 6°C versus above 6°C. Based on group 1 and group 2 data, the majority of spring days fall into group 2 (60%–90% of all spring days) except for a few stations over northern European Russia. Thus, the overall relationship with air temperature is predominantly negative and only a few stations over northern European Russia have positive correlations as shown in Fig. 3e. For autumn, there are about an equal number of days in these two groups except for a few of the southernmost stations, thus the positive relationship prevails in autumn. This is also reflected in the fact that mean seasonal air temperatures during spring are warmer than in autumn at a majority of stations over the study region.

To have a better idea of how many degrees of warming it would take for decreasing winter snowfall days to occur in various parts of the study region, the mean winter air temperature map averaged over 30 years (1960–89) is shown in Fig. 9. The −8°C isotherm is found over southwestern European Russia and the southern edge of western Siberia. The Alley et al. (2007) projection indicates possible 4°–6°C increases in surface air temperature over the study region by 2090–99 (Trenberth et al. 2007). Taking into consideration the approximate 1°C warming during the most recent decade over the Arctic region (Trenberth et al. 2007), decreasing snowfall will be present over the majority of European Russia and southwestern Siberia (current −13°C and −15°C isotherm locations).

Fig. 9.

Mean surface air temperature during the winter season averaged from 1960/61 to 1989/90.

Fig. 9.

Mean surface air temperature during the winter season averaged from 1960/61 to 1989/90.

Increasing precipitation frequency during winter seems to be consistent with analyses of evidence for an ongoing intensification of the global hydrologic cycle (Dai 2006; Dirmeyer and Brubaker 2006; Held and Soden 2006; Huntington 2006; Trenberth et al. 2007), and decreasing rainfall days in a warmer climate is also consistent with the decreasing precipitation trends found in lower-latitude regions, especially during warm seasons (Dai et al. 1997,2004; Trenberth et al. 2007).

It is worth noting that changes in wet days are not necessarily related to amounts of total precipitation. Studies have found that, in China and Japan, although the frequency of precipitation days has decreased, the mean precipitation amount has increased owing to increases in high-intensity precipitation days (Easterling et al. 2000; Liu et al. 2005). Most studies have found that high-intensity precipitation days have increased in association with the warming climate (Groisman and Karl 1999; Groisman et al. 2005; Karl and Knight 1998). A study over Switzerland found that winter precipitation frequency and strength increased during 1901–2000 with the strongest signal over northern and western Switzerland (Schmidli and Frei 2005). However, a study of precipitation in Italy found general decreases in wet days even during winter but more frequent heavy precipitation days (Brunetti et al. 2004). The results from this study show that the differences in the trends of the Swiss and Italian studies are very likely due to the fact that Italy has a warmer climate than Switzerland. A recent study based on global climate model output under anthropogenic forcing scenarios suggests increases in precipitation frequency, amount, and intensity during winter seasons over four major river basins: the Volga, Ob, Yenisei, and Lena (Khon et al. 2007). This is consistent with the idea that the frequency of all types of precipitation might be expected to increase with increasing air temperatures in cold regions.

The threshold of −8°C for a switch in the change direction of snowfall days and −6°C for a switch in the change direction of rainfall days is based on the northern Eurasian station data during the study period 1936–89. Future study is needed to confirm the findings and provide more robust statistics about these relationships by using updated data sources and expanded study into other geographical regions. This knowledge will help us to better understand the nature of precipitation characteristics and to predict future changes in hydrological cycles under a warming climate in many regions of the world.

Acknowledgments

This research is supported by the NOAA Data and Climate Change Detection Program Grant NA05OAR4311112. The author would like express her appreciation to the CDIAC for providing Russian synoptic station data, Argyl Houser for proofreading the manuscript, and the two anonymous reviewers for their insightful comments that significantly enhanced the quality of this manuscript.

REFERENCES

REFERENCES
Alley
,
R. B.
, and
Coauthors
,
2007
:
Summary for policymakers. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 1–18
.
Brown
,
R. D.
, and
R. O.
Braaten
,
1998
:
Spatial and temporal variability of Canadian monthly snow depths, 1946–1995.
Atmos.–Ocean
,
36
,
37
54
.
Brunetti
,
M.
,
M.
Maugeri
, and
F.
Monti
,
2004
:
Changes in daily precipitation frequency and distribution in Italy over the last 120 years.
J. Geophys. Res.
,
109
.
D05102, doi:10.1029/2003JD004296
.
Dai
,
A.
,
2006
:
Recent climatology, variability, and trends in global surface humidity.
J. Climate
,
19
,
3589
3606
.
Dai
,
A.
,
I. Y.
Fung
, and
A. D.
Delgenio
,
1997
:
Surface observed global land precipitation variations during 1900–1988.
J. Climate
,
10
,
2943
2962
.
Dai
,
A.
,
P. J.
Lamb
,
K. E.
Trenberth
,
M.
Hulme
,
P. D.
Jones
, and
P.
Xie
,
2004
:
The recent Sahel drought is real.
Int. J. Climatol.
,
24
,
1323
1331
.
Davis
,
R. E.
,
M. B.
Lowit
,
P. C.
Knappenberger
, and
D. R.
Legates
,
1999
:
A climatology of snowfall-temperature relationships in Canada.
J. Geophys. Res.
,
104
,
D10
.
11985
11994
.
Dirmeyer
,
P. A.
, and
K. L.
Brubaker
,
2006
:
Evidence for trends in the Northern Hemisphere water cycle.
Geophys. Res. Lett.
,
33
.
L14712, doi:10.1029/2006GL026359
.
Easterling
,
D. R.
,
J. L.
Evans
,
P.
Ya Groisman
,
T. R.
Karl
,
K. E.
Kunkel
, and
P.
Amebenje
,
2000
:
Observed variability and trends in extreme climate events: A brief review.
Bull. Amer. Meteor. Soc.
,
81
,
417
425
.
Frei
,
A.
,
J.
Miller
, and
D.
Robinson
,
2003
:
Improved simulations of snow extent in the second phase of the Atmospheric Model Intercomparison Project (AMIP-2).
J. Geophys. Res.
,
108
.
4369, doi:10.1029/2002JD003030
.
Giorgi
,
F.
,
X.
Bi
, and
J.
Pal
,
2004
:
Mean, interannual variability and trends in a regional climate change experiment over Europe. II: Climate change scenarios (2071–2100).
Climate Dyn.
,
23
,
839
858
.
Groisman
,
P. Ya
, and
T. R.
Karl
,
1999
:
Changes in the probability of heavy precipitation: important indicators of climatic change.
Climatic Change
,
42
,
243
283
.
Groisman
,
P. Ya
,
R. W.
Knight
,
D. R.
Easterling
,
T. R.
Karl
,
G. C.
Hegerl
, and
V. N.
Razuvaev
,
2005
:
Trends in intense precipitation in the climate record.
J. Climate
,
18
,
1236
1251
.
Groisman
,
P. Ya
,
R. W.
Knight
,
V. N.
Razuvaev
, and
O. N.
Bulygina
,
2006
:
State of the ground: Climatology and changes during the past 69 years over northern Eurasia for a rarely used measure of snow cover and frozen land.
J. Climate
,
19
,
4933
4955
.
Held
,
I. M.
, and
B. J.
Soden
,
2006
:
Robust responses of the hydrological cycle to global warming.
J. Climate
,
19
,
5686
5699
.
Huntington
,
T. G.
,
2006
:
Evidence for intensification of global water cycle: Review and synthesis.
J. Hydrol.
,
319
,
83
95
.
Huntington
,
T. G.
,
G. A.
Hodgkins
,
B. D.
Keim
, and
R. W.
Dudley
,
2004
:
Changes in the proportion of precipitation occurring as snow in New England (1949–2000).
J. Climate
,
17
,
2626
2636
.
Karl
,
T. R.
, and
R. W.
Knight
,
1998
:
Secular trends of precipitation, frequency, and intensity in the United States.
Bull. Amer. Meteor. Soc.
,
79
,
231
241
.
Khon
,
V. C.
,
I. I.
Mokhov
,
E.
Roeckner
, and
V. A.
Semenov
,
2007
:
Regional changes of precipitation characteristics in northern Eurasia from simulations with global climate model.
Global Planet. Change
,
57
,
118
123
.
Liu
,
B.
,
M.
Xu
,
M.
Henderson
, and
Q.
Ye
,
2005
:
Observed trends of precipitation amount, frequency, and intensity in China, 1960-2000.
J. Geophys. Res.
,
110
.
D08103, doi:10.1029/2004JD004864
.
Razuvaev
,
V. N.
,
E. B.
Apasova
, and
R. A.
Martuganov
,
1995
:
Six- and three-hourly meteorological observations from 223 U.S.S.R. stations. Carbon Dioxide Information Analysis Center Rep. NDP-048/R1, Oak Ridge National Laboratory, Oak Ridge, Tennessee, 124 pp
.
Roesch
,
A.
, and
E.
Roechner
,
2006
:
Assessment of snow cover and surface albedo in the ECHAM5 general circulation model.
J. Climate
,
19
,
3828
3843
.
Schmidli
,
J.
, and
C.
Frei
,
2005
:
Trends of heavy precipitation and wet and dry spells in Switzerland during the 20th century.
Int. J. Climatol.
,
25
,
753
771
.
Semenov
,
V. A.
, and
L.
Bengtsson
,
2002
:
Secular trends in daily precipitation characteristics: Greenhouse gas simulation with a coupled AOGCM.
Climate Dyn.
,
19
,
123
140
.
Serreze
,
M. C.
, and
Coauthors
,
2000
:
Observational evidence of recent change in the northern high-latitude environment.
Climatic Change
,
6
,
159
208
.
Trenberth
,
K. E.
, and
Coauthors
,
2007
:
Observations: Surface and atmospheric climate change.
Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336
.
Von Storch
,
H.
, and
F. W.
Zwiers
,
1999
:
Statistical Analysis in Climate Research.
Cambridge University Press, 484 pp
.
Ye
,
H.
,
2001a
:
Characteristics of winter precipitation variation over northern Eurasia and their connections to sea surface temperatures over the Atlantic and Pacific Oceans.
J. Climate
,
14
,
3140
3155
.
Ye
,
H.
,
2001b
:
Quasi-biennial and quasi-decadal variations in snow accumulation over northern central Eurasia and their connections to Atlantic and Pacific Oceans and Atmospheric circulation.
J. Climate
,
14
,
4573
4584
.
Ye
,
H.
, and
J. R.
Mather
,
1997
:
Polar snow cover changes and global warming.
Int. J. Climatol.
,
17
,
155
162
.
Ye
,
H.
,
H.
Cho
, and
P.
Gustafson
,
1998
:
The changes of Russian winter snow accumulation during 1936-1983 and its spatial patterns.
J. Climate
,
11
,
856
863
.
Ye
,
H.
,
D.
Yang
, and
D.
Robinson
,
2008
:
Winter rain-on-snow and its association with air temperature in Northern Eurasia.
Hydrol. Process.
,
22
,
2728
2736
.

Footnotes

Corresponding author address: Hengchun Ye, Department of Geography and Urban Analysis, California State University, Los Angeles, Los Angeles, CA 90032-8222. Email: hye2@calstatela.edu