The ratio of snow to total precipitation (S/P) is a hydrologic indicator that is sensitive to climate variability and can be used to detect and monitor hydrologic responses to climatic change. Changes in S/P ratio over time could influence the magnitude and timing of spring runoff and recession to summer baseflow. The S/P ratio for 21 U.S. Historical Climatology Network sites in New England was examined. Eleven out of twenty-one sites in New England had significant decreasing annual S/P ratios from 1949 to 2000. Annual trends in S/P are predominantly a result of decreasing snowfall, and to a lesser extent, increasing rainfall. The most consistent trends were in northernmost New England where all four sites had decreasing ratios, and in the coastal and near-coastal areas where five out of eight sites had significantly decreasing ratios. The four sites in northernmost New England, which had the strongest and most coherent trends, showed an average decrease in annual S/P ratio from about 0.30 in 1949 to 0.23 in 2000. Trends in winter S/P ratio were less geographically consistent. Seven out of 21 sites had significantly decreasing winter S/P ratios. Most northern New England and coastal to near-coastal sites had statistically significant trends (p < 0.05) or weak, but not significant trends (p < 0.2). When trends in S/P were analyzed on a monthly basis for the northernmost sites, it was evident that decreasing S/P trends were significant for March and December only. Significant correlations were observed between winter S/P ratios in northern New England and the timing of spring runoff, the North Atlantic Oscillation (NAO) index, and the Pacific–North American (PNA) index. Significant correlations were observed between winter S/P ratios averaged for all of New England and the NAO and PNA.
Surface air temperature has increased globally during the twentieth century, with substantial increases taking place in most northern temperate regions (Houghton et al. 2001). Biological responses to recent warming trends throughout northern temperate latitudes are well documented in recent reviews (Walther et al. 2002; Root et al. 2003; Parmesan and Yohe 2003). In New England, phenological changes consistent with earlier spring warming have been reported including springtime advances in timing of lilac blooming (Schwartz and Reiter 2000), bird migration (Dunn and Winkler 1999), and anadromous fish migration (Huntington et al. 2003). There is also evidence for a lengthening of the growing season in New England from the late 1700s/early 1800s to the 1930s and 1940s (Baron and Smith 1996) and in recent decades (Cooter and Le Duc 1995).
The warming trend over most of New England over the past 70 yr is well documented (Keim et al. 2003). Hydrologic variables in New England have shown temporally coherent trends that are consistent with winter and spring warming. For example, dates of spring lake ice-out (Hodgkins et al. 2002), river ice-out (Dudley and Hodgkins 2002), and snowmelt-driven spring runoff (Hodgkins et al. 2003) are all becoming earlier and are correlated with warming surface air temperatures. There is also evidence for increases in late-winter snow density (Dudley and Hodgkins 2002) and decreases in river ice thickness (Huntington et al. 2003) that is consistent with climate warming in recent decades in Maine. Similarly, and more broadly, there have been systematic decreases in April snow-cover extent in North America (Brown 2000) and an advance in the timing of high spring flows in western North America (Leith and Whitfield 1998; Cayan et al. 2001) and Siberia (Yang et al. 2002).
There are also human dimensions to climate warming that are directly related to changes in snowfall in New England. For example, Hamilton et al. (2003) report that part of the twentieth-century migration of New Hampshire's ski industry from throughout the state to the northern mountainous region, and consolidation into far fewer operations capable of large-scale snowmaking, is an adaptation to warmer and less snowy winters of recent decades. Rising sea surface temperatures during the last quarter of the twentieth century (Houghton et al. 2001) have been implicated as one possible cause of the observed decline in abundance of winter flounder (Pseudopleuronectes americanus) in New England over that period (Shea et al. 2001). The decline in maple syrup production in Vermont, and the increase in Quebec, Canada, in recent decades is thought to be related to climate variability (due to systematic changes in precipitation and temperature) and socioeconomic factors (Shea et al. 2001).
Karl et al. (1993) found that the annual ratio of snow to total precipitation (S/P) had decreased significantly (p < 0.01) over Canada south of 55°N latitude during the period 1980–90, compared with the previous three decades (1950–79). For the period 1950–90 there was also a decreasing trend (0.7% decade−1) in the annual S/P ratio in southern Canada. Karl et al. (1993) found a decreasing trend in annual S/P ratio over the contiguous United States for the same period but it was not statistically significant.
Continued climate forcing from increasing atmospheric burden of radiatively active gases during the twenty-first century is projected using general circulation models (GCMs) to result in climate warming and possible changes in precipitation (McCarthy et al. 2001). Hydrologic responses to changes in climate for specified climate forcing are estimated from hydrologic models “driven” by GCM projections at the appropriate spatial and temporal scale (Chalecki and Gleick 1999; Arnell et al. 2001). Predicted hydrologic responses to climate changes associated with midrange emission scenarios suggest a likelihood for reduced snow cover, earlier snowmelt, and changes in the seasonal pattern of runoff (McCarthy et al. 2001). Coherent trends in multiple hydrologic indicator variables, as have been documented in the New England region, strengthen confidence in the predictions of hydrologic models. Shifts from nival to pluvial precipitation regimes have been predicted, but have received far less attention than changes in other hydrologic variables. Ongoing changes in the S/P ratio could be quite important for interpretation of other hydrologic trends because they could influence timing of spring runoff, timing of river ice-out, and late-winter snow density.
This paper examines the ratio of snow to total precipitation using daily snow water equivalent data for 21 stations in New England and relates these time series to other hydroclimatological time series within the region, including the North Atlantic Oscillation index, the trough axis index, and the Pacific–North American index. It also evaluates the S/P ratio as a hydrologic indicator of climate variability for use in detection and monitoring of hydrologic response to climatic change. We are interested in the detection of trends in long-term hydrologic data to support a better quantitative understanding of hydrologic response to climate change. Such analysis can be used to improve hydrologic model development by constraining model sensitivity and providing a qualitative basis for model validation.
Temperature and precipitation data were obtained from the U.S. Historical Climatology Network (USHCN; Karl et al. 1990). USHCN data have been subjected to quality control and homogeneity testing and adjustment procedures for bias originating from changes in time of observation (Karl et al. 1986), station moves (Karl and Williams 1987), and urban warming (Karl et al. 1988). Keim et al. (2003) showed that the USHCN dataset is more reliable for long-term trend analysis than the larger dataset of the National Climate Data Center's Climate Divisions.
We used temperature and precipitation data from all USHCN stations in New England where continuous daily records were available to calculate annual or winter (here defined as December through March) composite records that met specific data completeness criteria. Any year having four or more days of missing precipitation or snowfall data was considered incomplete and the entire year was removed from the analysis. There were many years (average of 9.6 yr per site) that contained between one and three missing days per year so that the number of years meeting the criteria for the analysis could be substantially increased by including these years. We concluded that the potential bias associated with eliminating these years outweighed the potential bias associated with not including precipitation on those days. Several USHCN sites in southern New England, most notably in western Massachusetts and Connecticut, had insufficient data for analysis.
Any site containing less than 50% complete records within any 10-yr period during 1949 through 2000 was excluded from the analysis. This completeness criterion was used to control for possible temporal bias that could be introduced by having excessive missing data in any part of the record. Record completeness was defined as the total number of years having complete annual or winter records divided by the total number of years spanned by the record for each individual site. For the 21 sites included in this study, data completeness averaged 87% for the annual analysis and 93% for the winter analysis (Table 1). Time series for most USHCN sites in New England begin between 1948 and 1955. The sites at Presque Isle, Orono, and Ripogenus Dam, Maine, did not have any complete annual data after the mid-1990s.
Following Bradbury et al. (2002b), March was included as a winter month because at many sites in New England, March receives as much (or more) snowfall as December. As winter progresses, the longitudinal position of the East Coast trough migrates eastward resulting in cooler temperatures and more snowfall in March over New England than if the trough were stationary throughout the winter (Bradbury et al. 2002b).
Where snowfall was recorded for a given day, we used the independently recorded daily total liquid equivalent precipitation amount as the snow water equivalent (SWE) for the day. There can be positive or negative biases using this procedure because on days when precipitation falls as both snow and rain, it is recorded entirely as snow or entirely as rain depending on whether there was measurable snow at the time of measurement. There are also differences among observers in terms of the number of times that snowfall is recorded on any given day. Under mixed-precipitation conditions, more frequent observation may result in more days with snowfall recorded; under such circumstances, our algorithm would result in a bias toward more snowfall recorded than actually fell. We do not know how large the net effect of such bias could be, nor whether there could be systematic bias owing to changes in the frequency of mixed-precipitation events.
If there was a bias toward an increasing frequency of mixed-precipitation events over time, this would lead to an overestimation of snowfall in more recent times. This type of bias, if present, would result in a more conservative test of the hypothesis that the ratio of solid-to-total precipitation has decreased over time. On the other hand, for unshielded gauges, undercatch increases with increasing wind speed and is greater for solid than for liquid precipitation (Doesken and Judson 1996). A temporal trend toward decreasing S/P ratio could be exaggerated because gauge undercatch bias could record proportionately less snow under windier conditions. The bias can be accentuated under conditions of heavier wet snow that tends to accumulate on the top edges of the cylindrical gauges reducing the effective surface area (Doesken and Judson 1996). These biases may balance one another to some extent although no attempt was made in this analysis to quantify potential errors of adjusting for net biases because of insufficient data on undercatch at each site.
Time series were compiled for SWE, total precipitation, and the ratio of SWE to total precipitation for annual, winter (December through March) and monthly (October through April) periods. Time series also were developed for the timing of spring runoff using the average winter/spring seasonal center-of-volume date (WSCV) for eight unregulated river streamflow gauging stations in northern and mountainous areas of Maine and New Hampshire (USGS station identification numbers 1013500, 1014000, 1031500, 1047000, 1055000, 1057000, 1064500, 1076500) described by Hodgkins et al. (2003). The WSCV is defined as the Julian date (sequential day of year) on which 50% of the total runoff volume that occurs from 1 January to 31 May has passed the station. This variable has been shown to be sensitive to late-winter/early-spring air temperature, particularly in this region of New England (Dudley and Hodgkins 2002; Hodgkins et al. 2003).
The North Atlantic Oscillation (NAO) index is a measure of the difference in sea level pressure between Lisbon, Portugal, and Stykkisholmur, Iceland (http://www.cgd.ucar.edu/~jhurrell/nao.html). The NAO has been shown to be related to low-frequency climate variations over the North Atlantic (Hurrell et al. 2003), snowfall over New England (Hartley and Keables 1998), New England regional winter streamflow (Bradbury et al. 2002a), and temperature Bradbury et al. (2002b). The NAO (also known as the Northern Hemisphere annular mode, or NAM) has also been shown to modulate high-frequency (daily) winter climatic variation in high-latitude continental regions (Thompson and Wallace 2001; Wettstein and Mearns 2002). Annual and winter (December through March) NAO data were obtained from Hurrell et al. (2003; http://www.cgd.ucar.edu/~jhurrell/nao.html). The trough axis index (TAI) is a measure of the monthly mean longitudinal position of the East Coast trough (between 40° and 47.5°N) and has been shown to be related to winter climate variability in New England (Bradbury et al. 2002b). The Pacific–North America Index (PNA; Wallace and Gutzler 1981) is another teleconnection that has been evaluated in relation to climate variability in New England (Hartley and Keables 1998; Bradbury et al. 2002b). TAI data were obtained from J. Bradbury (2003, University of Massachusetts, personal communication) and PNA data were obtained from the University of Washington (http://tao.atmos.washington.edu/data_sets/pna/) where the record is maintained based on the approach of Wallace and Gutzler (1981). Multivariate El Niño–Southern Oscillation (ENSO) and the Pacific decadal oscillation (PDO) indices were obtained for correlation with the winter S/P ratio because they also have been related to hydroclimate in New England (Bradbury et al. 2003). PDO data were obtained from the University of Washington (http://tao.atmos.washington.edu/pdo/) (Mantua et al. 1997). Multivariate ENSO index (MEI) data were obtained from the National Oceanic and Atmospheric Administration (http://www.cdc.noaa.gov/~kew/MEI/mei.html#ref&_wt1) (Wolter and Timlin 1998).
Temporal trend tests were conducted on individual sites, 21-site aggregate (New England region); and 4-site aggregate (northern Maine and northwestern New Hampshire) time series data with a nonparametric test for monotonic trend based on Kendall's tau statistic (Helsel and Hirsch 1992). Using this test, no assumptions of normality of the distribution are required, and serial correlation is assumed to be negligible. The Durbin–Watson statistic was calculated to test for serial correlation. There was no temporal bias in the distribution of missing values. Time series for S/P ratio were plotted with a locally weighted scatterplot smooth (LOWESS; Helsel and Hirsch 1992) curve, with a weighting function of 66% for the period of record, for graphical interpretation of the trend. Correlation analysis (Pearson's r with Fisher's p-value for significance of the Pearson's r) was used to determine relations between the annual and winter S/P ratio and hydroclimatological variables and teleconnections. Trends or correlations with p < 0.05 were considered statistically significant. Trends with p in the range of 0.05–0.20 were considered weak, but not significant.
3. Results and discussion
a. Time series analysis
Eleven out of twenty-one sites had significant (p-values < 0.05) decreasing annual S/P ratios (Table 1, Fig. 1). The areas showing the most coherent trends in annual S/P ratio were northwestern Maine and northernmost New Hampshire (northern New England) where all sites had significant decreasing annual trends. Five out of eight of the coastal and near-coastal sites also had significant decreasing annual S/P ratio trends. Three additional sites had weak, but not significant (p-values between 0.05 and 0.20) trends toward decreasing annual S/P ratio. Seven sites had significant decreasing trends in winter S/P ratio and an additional four sites had weak trends toward decreasing winter S/P ratio. None of the sites had significant, or even weak, trends toward increasing winter S/P ratio. The tendency for the northernmost sites to show stronger annual than winter trends indicates that part of the trend in S/P ratio may be driven by increases in nonwinter precipitation. When the data were aggregated, the entire New England region and the northernmost region had significant decreasing trends in S/P ratio for annual and winter periods over the period 1949–2000 (Table 2).
Data from the four sites in northern New England that showed the strongest geographically coherent trends in decreasing annual S/P ratio were aggregated. The S/P ratio of the aggregate data decreased from about 0.30 to 0.23 from 1949 through 2000 (Fig. 2). Based on the aggregate LOWESS curve, most of the decrease in S/P occurred after 1975. Several other hydrologic and climatic time series including: annual S/P in Canada south of 55°N (Karl et al. 1993), timing of lake ice-out (Hodgkins et al. 2002), timing of streamflow (Hodgkins et al. 2003), river ice thickness (Huntington et al. 2003), and surface air temperature (Houghton et al. 2001) also indicate that the majority of change occurs after 1975. This coherence in multiple hydrologic indicators suggests sensitivity to a common climate forcing.
One of the northern New England sites, Presque Isle, Maine, had a significant decreasing trend in winter S/P ratio, and two of the remaining sites in that region had weak decreasing trends in winter S/P ratio (Table 1). The S/P ratio of the aggregate data for the four sites in northern New England decreased from about 0.73 to 0.69 from 1949 to 1998 (Fig. 2).
We analyzed S/P ratio for trends in individual months over the period 1949 through 2000 to determine which months had the strongest trends. For the four sites in northern New England, March and December had significant decreasing trends in S/P (Fig. 3). Both months had similar significance levels but the rate of decrease in S/P was greater in December than March. This suggests that the annual and winter trends are driven in large part by changes in S/P near the beginning and end of the winter season when temperatures are more frequently near freezing. During the coldest months, January and February, temperatures are usually well below freezing so small increases in temperature are not likely to result in a shift from snow to rain. We expected to find the most significant effects at the beginning and end of the winter season based on previous studies of changes in growing-season length (Cooter and LeDuc 1995; Easterling 2002) and model predictions for responses to climate warming (McCarthy et al. 2001). We did not observe significant trends in S/P ratio for the months of November and April which represent the very beginning and ending of the snow season; however, during these months typically less than 50% of the precipitation falls as snow.
Five out of eight coastal or near-coastal sites showed significant decreasing winter S/P ratio trends and one additional site (Blue Hill, MA) showed a weak, but not significant, decreasing trend (Table 1, Fig. 1). Sites in northern Vermont showed no significant trends in annual or winter S/P ratio, though the site at St. Johnsbury showed a weak decreasing trend in annual S/P ratio. Three out of four sites in the upper Connecticut River valley showed weak, or for Keene, New Hampshire, significant, decreasing annual S/P ratio trends. Only one of these sites (Keene, NH) showed a significant decreasing winter S/P ratio trend.
Comparisons with other recent studies showing hydrologic responses to climate variability indicated some consistent geographic patterns in responses within New England. Northern New England had the most consistent trends in annual S/P ratio and had the most consistently significant trends in earlier (by 1 to 2 weeks during the twentieth century) high spring flows (Hodgkins et al. 2003). In northern New England, median seasonal maximum snow depths average 71 to >81 cm compared with 25 to 50 cm in southern and coastal regions, and 50 to 60 cm over other inland regions of New England (Cember and Wilks 1993). These northern New England sites had the strongest trends toward decreasing S/P ratio and earlier high spring flows. This region has substantially greater snow accumulation than in more southerly regions; thus warming would have a greater impact on snowmelt- (and rain on snow)-driven runoff.
Most of the unregulated rivers in northern New England that have been studied have also shown significant trends toward earlier river ice-out and fewer total ice-affected flow days (Hodgkins et al. 2004, manuscript submitted to Climatic Change). These trends revealed a geographic pattern similar to that observed for S/P ratio, such that the strongest (lowest p values) trends occurred in northern New England. Ice thickness on 1 March on the Piscataquis River in central Maine has also decreased during the twentieth century (Huntington et al. 2003). These trends in river ice phenology and in the timing of high spring flows were significantly correlated with late-winter/early-spring temperatures. These earlier papers did not attempt to correlate the observed changes in hydrologic variables to the S/P ratio. The S/P ratio time series is substantially shorter than that available for most of the other hydrologic variables. However, we expected to observe significant correlations between S/P ratio and hydrologic variables that were consistently correlated with late-winter/early-spring air temperatures for two reasons. First, most of the observed changes in hydrologic and climatic variables during the twentieth century have occurred during recent decades for which S/P ratio data are available. Second, analyses of snow cover over the Northern Hemisphere (Brown 2000); snowfall in parts of North America (Karl et al. 1993); frost over New England (Cooter and LeDuc 1995; Easterling 2002); and snow density in Maine (Dudley and Hodgkins 2002) are all consistent with decreasing S/P ratio during the last 50 yr.
Decreasing S/P ratio could be explained by snowfall decreases that were proportionately larger than decreases in rainfall; by constant snowfall and increasing rainfall; or increases in both, but larger increases in rainfall than snowfall. We tested for temporal trends in total annual and total winter rainfall, and we tested for temporal trends in total snow water equivalent to determine which could best explain the observed trends in S/P ratio. For the New England region, neither total annual rainfall nor winter rainfall increased significantly (1949–2000) (Table 2). Annual rainfall showed a weak (p = 0.10) but insignificant increasing trend. The four northern Maine and northernmost New Hampshire sites did not show any significant trends in annual or winter rainfall amount. Total annual SWE over the entire New England region and northern New England decreased significantly (Table 2). The observed weak trends (p < 0.2) toward decreasing winter SWE over New England and northern New England were not significant.
Based on these trends in nival to pluvial precipitation for the New England region, annual trends in S/P are predominantly a result of decreasing snowfall, and to a lesser extent, increasing rainfall (particularly during the nonwinter period). It has been suggested that climate warming may result in both increased precipitation and increased snowfall in many northern temperate latitude areas (McCarthy et al. 2001). The snowfall data for the 21 USHCN sites analyzed in this study and in the analyses by Hamilton et al. (2003) do not support an increase in snowfall during the latter half of the twentieth century in New England; on the contrary, they indicate decreasing snowfall.
Others have reported increases in precipitation for the northeastern United States and for the southeastern Canadian provinces during the twentieth century (Dai et al. 1997; Karl and Knight 1998; Groisman et al. 2001). However, the period 1976–99 showed decreasing precipitation while the period 1946–75 showed increasing precipitation over this region (Houghton et al. 2001). These opposing trends during the third and fourth quarters of the twentieth century may explain why we did not find stronger trends in S/P at many sites.
In southern Canada, the decreasing trend in annual S/P ratio for the period 1950–90 reported by Karl et al. (1993) may have been driven primarily by a decrease in snow because total precipitation increased only slightly over the period, but neither trend was statistically significant. Karl et al. (1993) related their results to unprecedented warming trends but they did not investigate trends in winter S/P ratio, so it is not clear how important increases in precipitation outside of the winter period may have been in forcing the trend they observed. More recent analyses (Houghton et al. 2001) indicate increases in precipitation over this region occurred during all seasons, when evaluated on a century scale.
Karl et al. (1993) assumed a temporally and spatially stationary ratio of snow depth to SWE for converting monthly cumulative snow depth into SWE prior to their time series analysis. This assumption could introduce a spurious trend if there had been systematic change in the density of the falling snow over time. Dudley and Hodgkins (2002) have shown evidence for statistically significant increasing snow density for the cumulative snowpack in late winter/early spring for coastal and eastern Maine over the period 1950–2000. Part of the increase in snow density observed by Dudley and Hodgkins (2002) could be a result of a systematic increase in the density of snow falling during late winter/early spring over time. The density of falling snow increases with increasing air temperature (Dube 2003); therefore, it is likely that increases in air temperature have resulted in increases in the density of falling snow. This increase in snowpack SWE in late winter/early spring, or ripening of the snowpack, is consistent with warmer spring temperatures, more frequent rain on snow, and/or denser (wetter) snow. Such a trend, if present at the sites analyzed by Karl et al. (1993), would result in a bias toward decreasing S/P ratio.
Trends in surface air temperature for New England and for northern New England for annual and winter periods were not significant (Table 2). This lack of significance is a result of cooling trends during the third quarter of the twentieth century in spite of substantial warming trends during the fourth quarter of the twentieth century (Houghton et al. 2001). There was no significant trend in annual NAO index, but for the winter period, the NAO index increased significantly for the period 1948–2001 (Table 2, Fig. 4). There was no significant trend in the TAI for the period 1949–97. The PNA increased significantly over the period 1948–2001 (Table 2). The average WSCV date for eight unregulated rivers in northern New England decreased significantly (indicating earlier spring runoff in recent years) during the period 1948–99.
b. Correlation analysis
The annual and winter S/P ratios at the four northern New England sites were significantly and positively correlated with timing of spring runoff (WSCV date) at eight northern New England streamflow gauging stations (Table 3). The S/P ratio has decreased as the WSCV date has occurred earlier in the spring. This correlation was expected because both variables would likely be responsive to increasing surface air temperature and Hodgkins et al. (2003) have reported significant trends in WSCV date and average March + April surface air temperatures for this region. For New England, winter S/P ratio was not correlated with surface air temperature, but annual S/P ratio was significantly and negatively correlated with surface air temperature (Table 3). For northern New England, the reverse was true; that is, winter S/P ratio was significantly correlated with temperature and annual S/P ratio was not.
We expected that warmer air temperature would be associated with a decrease in S/P ratio as more precipitation occurred as rain versus snow. Our results generally are consistent with this hypothesis, particularly for the northern sites. Together these results indicate that decreases in S/P ratio are variable across the region but are likely associated with decreases in snowfall and, to a lesser extent, increases in total annual rainfall over the New England region. Increasing spring surface air temperatures are likely driving these trends in the ratio of nival-to-pluvial precipitation.
Winter S/P ratios for New England, and for northern New England, were significantly and negatively correlated with winter NAO (Table 3). The annual S/P ratio for northern New England was also significantly and negatively correlated with annual NAO, while annual S/P ratio for all of New England was weakly correlated with NAO. The inverse relationship is consistent with the weak positive correlation between northern/inland New England streamflow and the NAO index that was strongest in winter reported by Bradbury et al. (2002a). During positive NAO winters, a decreased S/P ratio would favor more rainfall as opposed to snow, which would lead to more winter streamflow and an earlier WSCV. These results help to explain why an NAO– streamflow link is observed, but no significant NAO– precipitation link has been documented for this region.
The NAO modulates the circulation pattern over the middle and high latitudes and regulates the number and intensity of significant weather events at middle and high latitudes. In the negative phase of the NAO, high pressure often builds over Greenland, which produces a blocking pattern that favors cold air advection and increased snowfall in northern New England (Bradbury et al. 2003). The positive phase of the NAO features a strong polar vortex, with the midlatitude jet stream shifted to the north of its normal position. In the positive phase, there is an increase in the occurrence of extreme warm days over much of the contiguous United States and a decrease in the occurrence of snowstorms affecting the Northeast. This out-of-phase relationship between the NAO index and snowfall over New England has been observed (Hartley and Keables 1998; Thompson and Wallace 2001; Hamilton et al. 2003). The persistence of high index polarity (positive phase) NAM during the 1980s and 1990s resulted in less cold arctic air spilling into midlatitude regions and fewer large snowstorms (“nor'easters”) during recent decades compared with earlier in the twentieth century (Davis and Dolan 1993; Thompson and Wallace 2001). A significant decrease in snowfall has been reported for two sites (Keene and Berlin) in southern and northern New Hampshire, respectively, during the latter half of the twentieth century (Hamilton et al. 2003). This would suggest that the observed trends toward decreased S/P ratio during the latter half of the twentieth century are associated with this hemispheric-scale pattern of climate variability that has resulted in warmer winter temperatures and decreased snowfall. The association between atmospheric circulation and S/P ratio is evident when comparing the inverse relations between these time series (Figs. 2 and 4).
There were weak, but not significant, positive correlations between winter S/P ratio and TAI for all of New England and for northern New England (Table 3). We expected to observe stronger correlations between S/P and TAI because the TAI is significantly correlated with winter precipitation, particularly over more northern latitudes in New England (Bradbury et al. 2002b). As the East Coast pressure trough (TAI) shifts eastward, there is less winter precipitation over the New England region. Bradbury et al. (2002a) have reported a significant correlation between TAI and NAO such that negative NAO conditions accompany an eastward-displaced trough.
The PNA was significantly and negatively correlated with the winter S/P ratio over New England (but not northern New England). This correlation contrasts with the reports that PNA was not significantly correlated with northeastern climate variability (Bradbury et al. 2002a; Hartley and Keables 1998). Neither the PDO nor multivariate ENSO indices were significantly correlated with winter S/P ratio for New England (Table 3).
This time series analysis supports a decreasing trend in annual and winter S/P ratio from 1949 to 2000 over substantial parts of New England. Eleven out of 21 sites showed significant decreasing trends in annual S/P ratios and 7 out of 21 sites had significantly decreasing winter S/P ratios. The geographic patterns in the annual and winter decreasing S/P trends were similar to and consistent with other hydroclimatological analyses in New England of trends in timing of spring runoff, river ice dynamics, and river ice thickness. The strongest trends were in the northernmost, coastal, and near-coastal regions. High interannual variability, short periods of record, and lack of completeness in existing records make it difficult to make more definitive conclusions. Longer-term and more complete records have shown more definitive trends in lake ice-out (Hodgkins et al. 2002), timing of spring flow (Dudley and Hodgkins 2002; Hodgkins et al. 2003), and river ice-out (Hodgkins et al. 2004, manuscript submitted to Climatic Change) in this region. Hydroclimatological time series that begin around 1950 are anchored during the period of regional cooling from 1946 to 1975 (Houghton et al. 2001) that also included several years of exceptionally high snowfall in Maine (Dudley and Hodgkins 2002). Anchoring this S/P ratio time series analysis at the beginning of this period of cooling may mask a stronger longer-term trend, but this cannot be verified without data prior to 1950. The National Climatic Data Center plans to release TD3206 Data Recovery Program files that will permit an extension of available digital liquid precipitation and snow records back into the first half of the twentieth century for certain sites. Once these records are released it may be possible to provide a longer-term analysis of temporal trends in S/P ratio that will be of comparable duration to other hydrologic indicator variables that have been investigated in New England.
The S/P ratio is a sensitive hydrologic indicator variable that can be used to detect and monitor hydrologic response to climate variability in the future. Significant trends in S/P ratio lend strength and confidence to hydrologic model prediction because they demonstrate that empirical observations are consistent with model predictions. The correlations between S/P ratio and precipitation and between S/P ratio and temperature are consistent with sensitivity to past and future climatic change. There is also an association between the S/P ratio and hemispheric-scale (i.e., NAO and PNA) and, potentially, regional-scale (i.e., TAI) atmospheric circulation patterns that indicates that part of the variation and trend in S/P ratio is likely attributable to variability in these indices. It remains uncertain whether ongoing climate warming influences the variability in NAO (Visbeck et al. 2001), PNA, or TAI. Longer-term changes in S/P ratio over time could be quite important to the extent that they influence the magnitude and timing of spring runoff and recession to summer baseflow.
We thank Mark Ayers and Jamie Shanley, both from the U.S. Geological Survey for helpful reviews of earlier drafts of this manuscript. We thank James Bradbury of the University of Massachusetts for insightful comments regarding atmospheric circulation. We are grateful to Thomas Hawley from the U.S. National Weather Service in Gray, Maine, for helpful insights into the methodology of snow measurement. We thank Pavel Groisman and David Easterling from NOAA for helpful information about previous related studies and the USHCN data structure. Ivan Dube of the Meteorological Service of Canada provided insights into the climatic regulation of the density of falling snow. This study was funded by the U.S. Geological Survey, Office of Surface Water.
Corresponding author address: Thomas G. Huntington, U.S. Geological Survey, 196 Whitten Rd., Augusta, ME 04330. Email: firstname.lastname@example.org