Abstract

Globally, 2014 and 2015 were the two warmest years on record. At odds with these global records, eastern Canada experienced pronounced annual cold anomalies in both 2014 and 2015, especially during the 2013/14 and 2014/15 winters. This study sought to contextualize these cold winters within a larger climate context in Toronto, Ontario, Canada. Toronto winter temperatures (maximum Tmax, minimum Tmin, and mean Tmean) for the 2013/14 and 2014/15 seasons were ranked among all winters for three periods: 1840/41–2015 (175 winters), 1955/56–2015 (60 winters), and 1985/86–2015 (30 winters), and the average warming trend for each temperature metric during these three periods was analyzed using the Mann–Kendall test and Thiel–Sen slope estimation. The winters of 2013/14 and 2014/15 were the 34th and 36th coldest winters in Toronto since record-keeping began in 1840; however these events are much rarer, relatively, over shorter periods of history. Overall, Toronto winter temperatures have warmed considerably since winter 1840/41. The Mann–Kendall analysis showed statistically significant monotonic trends in winter Tmax, Tmin, and Tmean over the last 175 and 60 years. These trends notwithstanding, there has been no clear signal in Toronto winter temperature since 1985/86. However, there was a statistically significant increase in the diurnal temperature range in that period, indicating an expansion of winter extremes. It is proposed that the possible saturation of urban heat island–related warming in Toronto may partially explain this increase in variation. Also, anomalies in the position of the polar jet stream over Toronto during these cold events are identified. No direct influence of major teleconnections on Toronto winter temperature is found.

1. Introduction

Globally, 2014 and 2015 were the two hottest years ever recorded. According to the U.S. National Oceanic and Atmospheric Administration’s National Centers for Environmental Information (NOAA NCEI 2015a, 2016), both years set global records for the highest annual mean temperature. Winter 2013/14 was the eighth-hottest winter, globally. Likewise the seasonal mean temperature for winter 2014/15 was the highest since records began in 1880 (NOAA NCEI 2015b). Notwithstanding these global trends, 2014 and 2015 were only the 22nd and 12th warmest years since 1880, as calculated from the Global Historical Climatology Network (GHCN) gridded dataset (Smith et al. 2008; Lawrimore et al. 2011; Huang et al. 2015). In both years, Canada’s annual temperature anomaly was split between western and eastern Canada, with positive anomalies in the west and negative anomalies in the east. Part of this dichotomy between national and global temperatures can be attributed to the anomalously cold winters in 2013/14 and 2014/15.

Winter 2013/14 saw negative anomalies from the 1981–2010 average temperature throughout almost all of Canada—with the largest anomalies extending throughout most of Saskatchewan to Ontario and southeastern Quebec and the maritimes—as well as the eastern United States (see Fig. 1, top). The overland national mean temperature Tmean anomaly during winter 2013/14 was 1.3°C below the 1981–2010 baseline average, based on the GHCN data. In the Great Lakes region, lake ice posed challenges for the shipping industry as well as local water utilities, and ice shoves caused flooding in areas along the Niagara River (Environment Canada 2014). A December 2013 ice storm in Toronto, Ontario, caused 600 000 households to lose power for up to 10 days and had a severe impact on the city’s tree canopy, with an estimated 50%–80% canopy loss on some streets and an overall canopy loss of 20% across the city (Environment Canada 2014).

Fig. 1.

Departures from 1981–2010 average baseline temperature during the winters of (top) 2013/14 and (bottom) 2014/15. Data are from GHCN–Monthly version 3.3.0 (Lawrimore et al. 2011) and ERSST.v4 (Huang et al. 2015).

Fig. 1.

Departures from 1981–2010 average baseline temperature during the winters of (top) 2013/14 and (bottom) 2014/15. Data are from GHCN–Monthly version 3.3.0 (Lawrimore et al. 2011) and ERSST.v4 (Huang et al. 2015).

Winter 2014/15 was marked by the stark contrast of positive temperature anomalies along the Pacific coast (between 1.8° and 3.0°C) and negative anomalies in the east, with the largest negative anomalies in the Great Lakes and St. Lawrence River regions (between −2.3° and −2.8°C). Indeed, that season saw cold anomalies from the 1981–2010 GHCN baseline across the northeastern United States and much of eastern Canada (see Fig. 1, bottom), with an overall temperature anomaly of −0.29°C. In Toronto, the cold February weather in 2015 took its toll on city infrastructure, with 48 water main breaks by mid-February and over 1000 reports of frozen pipes in households across Toronto (Environment Canada 2015). According to Environment Canada (2015), February also saw agricultural impacts, including delayed maple tapping and damaged grape crop at vineyards in the Great Lakes region. Winter 2014/15 also saw increased lake ice in February that had a direct negative impact on the shipping industry, delaying shipments and deliveries by several weeks in some cases (Environment Canada 2015).

The recent cold winters notwithstanding, Environment and Climate Change Canada estimates that the national average winter temperature in Canada warmed by 3.0°C over the 68-yr period from 1948 to 2015 (Environment and Climate Change Canada 2015). In the city of Toronto, studies have also found that winters are becoming less cold (minimum temperature Tmin is increasing), and record extreme cold temperatures were shown to have decreased between 1971 and 2000 (Allen et al. 2015). Likewise, freeze–thaw cycles in the city have contracted (Ho and Gough 2006), leading to shorter winters.

In spite of the overwhelming evidence in support of the theories of climate change and global warming, anomalous (relatively extreme) cold events like those experienced in winters 2013/14 and 2014/15 often provoke a dialogue regarding the veracity of these theories. Therefore, this paper seeks to examine how our lived climatological experience influences the relative rarity of cold events like those that affected Toronto in 2013/14 and 2014/15. Concretely, this study will respond to the following research questions: First, how do the cold Toronto winters of 2013/14 and 2014/15 rank in a long-term climatological context, and how does this ranking change when we consider shorter periods of recent history? Second, how has winter temperature in Toronto evolved since the start of meteorological record-keeping at the University of Toronto in 1840? Finally, this paper will explore the climatological factors that influence winter temperature in Toronto, and explore possible causes for weather anomalies within the context of global climate change.

Toronto is characterized by a humid continental climate. Because of the city’s geographical location near the typical latitudinal limits of the northern polar vortex, its climate is heavily influenced by the interaction of polar and subtropical air masses (Gough and Sokappadu 2016). Additionally, Toronto’s climate is affected by natural and anthropogenic phenomena. The highly urbanized center of Toronto is affected by an urban heat island (UHI), which causes city temperatures to warm above those of surrounding rural areas. Toronto’s urban heat island causes nighttime temperatures to be approximately 3°C warmer than those of surrounding rural areas (Gough and Rozanov 2001). For most of the year, however, the effect of the UHI on maximum temperatures is regulated by the effect of Lake Ontario, one of the Laurentian Great Lakes. This lake breeze is strongest in the spring and summer months, as cool air from over Lake Ontario mitigates hot afternoon temperatures (Mohsin and Gough 2012). During the winter season, however, the influence of the lake on temperatures in downtown Toronto is diminished; indeed, Mohsin and Gough (2010) found no evidence of a lake effect during winter months in Toronto. During cold months, lake ice forms on the surface of Lake Ontario. According to those authors, this ice—which generally begins to form at the end of December and peaks in February—acts to insulate the lake, decreasing its thermal inertia. Without the effect of the lake, there is more warming in Toronto maximum temperature Tmax during cold months (Mohsin and Gough 2010). Finally, Toronto occasionally receives lake effect snow from over Lake Ontario when winds blow from the south or south east, but the city is shielded against the brunt of southern Ontario’s lake effect snow by the Niagara Escarpment to the south and west, and the Oak Ridges Moraine to the north (Mohsin and Gough 2010).

Understanding the history of winter weather in Toronto can allow us to better understand the relative rarity of anomalous cold events, such as those that occurred 2013/14 and 2014/15. Understanding extreme weather is the first step toward minimizing the social, economic, and physical impacts of extreme events.

2. Methodology

Daily data for the Toronto climate station (43°40′N, 79°24′W) were obtained from Environment and Climate Change Canada for the period from 1 March 1840 to 28 February 2015—approximately 175 years of data (online at http://climate.weather.gc.ca/). The Toronto climate station is located on the downtown campus of the University of Toronto and was selected because it has the longest continuous climate record in Canada and is centrally located within the area that was affected by the two cold winters in this study. From the daily data, averaged monthly, seasonal, and annual time series were produced for Tmax, Tmin, and Tmean. As is customary in the climate sciences, this study used the definition of winter as the period from 1 December of a given year, to the last day of February of the following year (i.e., DJF). As such, the period of winters analyzed includes all winters between winter 1840/41 and winter 2014/15.

We sought to describe how lived experience influences the relative rarity of an anomalous cold event. To this end, we ranked winter temperature (Tmax, Tmin, and Tmean) during these two cold winters among all winters in three time periods: 1841–2015 (175 winters), corresponding to the extent of the historical record; 1956–2015 (60 winters), approximately corresponding to the lived experience of baby boomers; and 1986–2015 (30 winters), a period that roughly corresponds to the lived experience of generation Y (millennials), and the shortest period of analysis that is typically used in climate studies. Return intervals (RI) were generated to quantify the relative rarity of these cold events. These intervals were calculated by the equation T = (n + 1)/m, where n is the total number of observations (175, 60, or 30 yr), and m is the respective rank of each cold event.

We used locally weighted scatterplot smoothing (LOESS) curves to visualize the evolution of each temperature variable since winter 1840/41. Subsequently, the nonparametric Mann–Kendall test was used to detect any significant trends in winter temperature (Tmax, Tmin, and Tmean) over the same periods, so as to quantify this evolution. The nonparametric Mann–Kendall test was chosen following a visual analysis of quantile–quantile (qq) plots and histograms for each data series and time period. Although the full dataset (1840–2015) was normally distributed, the smaller 60- and 30-yr subsets did not follow a normal distribution and, therefore, did not meet the assumptions of parametric tests. The Mann–Kendall test was chosen for all data so as to ensure easy trend comparison between the different time periods analyzed.

Before performing the trend analysis, each of the datasets was tested for autocorrelation at annual lags of 1, 2, and 3 winters using the Durbin–Watson test. In cases where the data were found to show autocorrelation, the time series was subjected to prewhitening as described by Yue et al. (2002). The same authors, however, warn that the prewhitening procedure can, in some cases, mask or eliminate trends in a time series (Yue and Wang 2002). For this reason, the results of the Mann–Kendall test after prewhitening and the Mann–Kendall test without performing prewhitening were compared; if there was no difference in the significance level of the detected trend, then the results of the Mann–Kendall test without prewhitening were preferred. Finally, the Thiel–Sen slope approximation was used, as per Mohsin and Gough (2010, 2012), to determine the magnitude of any detected trends.

Finally, we tracked how the variation in winter temperature changed over each of the time periods we analyzed by examining the standard deviation of each subset.

After performing the above tests, and undertaking a preliminary interpretation of the results, we deemed that further analysis was required. We performed a detailed trend analysis of the average winter diurnal temperature range (DTR) as a measure of variance in winter temperatures. Likewise, we studied monthly winter temperature (Tmax, Tmin, and Tmean) for our three time periods. Finally, to quantify the impact that the anomalously cold winters of 2013/14 and 2014/15 have on the period trend analysis, we reperformed the Mann–Kendall test on subsets of the data that exclude the cold winters.

a. A note on data quality

The Toronto weather station has moved a number of times since it was established in 1840. To account for this, and as part of a Canada-wide effort to ensure the quality of archived hydrometeorological data, the Toronto climate station dataset was tested for inhomogeneity and subsequently homogenized by Vincent et al. (2012) and provided as monthly time series. We also downloaded this homogenized data for the period from March 1840 to December 2015, and generated averaged seasonal and annual time series.

We performed all of our analyses using both the original archival data available from Environment and Climate Change Canada, as well as the homogenized data. In our analysis, the homogenized data did not significantly change any of the tests performed.

The homogenized set only affected the ranks of the seasons under study due to the precision lost by using monthly data instead of daily values. Both the original daily archival data and the monthly homogenized data provide temperature values measured to the instrumental limit of 0.1°C; however, when we allow for a standard error of , we can have more confidence in the precision of the daily monthly and seasonal means generated form the daily values, as opposed to the monthly data. As such, the winters that are reported herein as ranked 33, 34, 35, and 36—winters 1993/94, 2013/14, 1848/49, and 2014/15, respectively—were all ranked 33 (Tmean = −5.5°C) when using the homogenized monthly data. In this sense, the use of archival data versus homogenized data had no significant effect to the rankings reported herein, but the daily data provided better precision.

Additionally, the homogenized time series did not improve autocorrelation in any of the time series that we prepared for the trend analysis. In fact, the homogenized data actually introduced autocorrelation in Tmax for the period from 1985 to 2015. The trends detected by the Mann–Kendall test also did not differ significantly between the original archival data and the homogenized data by any more than 0.02°C. Whether these differences were the product of the homogenization process or were simply introduced rounding errors is not clear.

Given the above observations, this study preferred the original Environment and Climate Change Canada archival data. The results of the analysis reported herein are generated from this original data. We took great care to ensure that our selection of source data did not introduce inhomogeneities or mask real observed trends.

b. On the drivers of winter temperature variability

To study the influence of major teleconnections on winter temperature during the past 30 years, we identified individual winter seasons that were affected by any one of El Niño–Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO), and the Arctic Oscillation (AO). The oceanic Niño index is provided by NOAA in overlapping 3-month periods. Null (2015) provides a curated list of ENSO events, defined as five continuous overlapping periods with an ENSO index of magnitude 0.5 or greater. When the ENSO index is positive, the event is called an El Niño event. When the index is negative, the event is called La Niña. The NAO and AO indices are also provided by NOAA in monthly datasets. We averaged these indices over the 3-month winter season and identified years in which the seasonal average index was positive (above 0.5) or negative (below −0.5). We used the Mann–Whitney U test to compare the average winter temperature during positive or negative phases of the above oscillations and during years that were not affected by the oscillation.

Finally, we performed an air mass analysis for Toronto, using the classification scheme set out in Sheridan (2002). Sheridan’s spatial synoptic classification system (SSC2) classifies daily meteorological conditions of a given weather station using six air mass types: dry polar (DP) air, originating from northern Canada or the Arctic region; dry moderate (DM), which is air that is modified upon passing through the Rocky Mountains; dry tropical (DT), from the deserts of the southwestern United States or Rocky Mountain chinook winds; moist polar (MP), which is generated over the North Atlantic or Hudson’s Bay; moist moderate (MM), when MP air becomes modified or occurs near a warm front; and moist tropical (MT, MT+, and MT++), originating over the Atlantic Ocean or the Gulf of Mexico. Finally, there is a transitional (TR) category for days when air masses change from one class to another. SSC2-classified air mass data are available for download from a database maintained by Sheridan, including data from 1950 to present at Toronto Pearson International Airport—an international airport on the west end of Toronto in the neighboring city of Mississauga, 20 km from the Toronto city center. For this purposes of this study, we have binned the climatological distributions of polar (DP and MP), moderate (DM and MM), and tropical (DT, MT, MT+, and MT++) air masses for December 2013 and 2014, as well as January and February 2014 and 2015. We calculated the percentage of days during each month when air masses in Toronto were of one of the above binned classes. Note that the percentages do not sum to 100%, as we have omitted the transitional air masses. We also compare the results of our analysis with the climatological mean for the available dataset (winters 1950/51 through 2014/15), as well as winter 1993/94, which was the coldest winter in Canada in the past 30 years and showed similar characteristics to the winters of 2013/14 and 2014/15.

3. Results and discussion

a. Climate context of cold winters 2013/14 and 2014/15

In a historical context, winter 2013/14 was the 34th coldest winter since 1840, a rank only slightly colder than the winter of 2014/15, which was ranked as the 36th coldest winter on average (see Table 1). The mean temperature (Tmean = −5.5°C) of both of these seasons was below the seasonal average of −3.8°C, when compared to the 175 years of recorded history at Toronto. Indeed, the mean temperatures of both the winter seasons under study are among the top six coldest in both our 60- and 30-yr periods. Winter 2013/14 was the fifth coldest winter experienced in the last 60 years and the second coldest in the last 30 years (after winter 1993/94). Winter 2014/15 was the sixth coldest in the past 60 years and the third coldest in the past 30 years. The minimum temperature during winter 2014/15 was the lowest experienced in the past 30 years.

Table 1.

Cold rankings over three time periods for Tmin, Tmean, and Tmax (°C) for winters 2013/14 and 2014/15. We present three values: the ranking (rank), the RI (yr), and the average temperature Tavg (°C). Italicized numbers indicate that the winter was colder than the average for the time period. Asterisks indicate that the difference in the temperature exceeded one standard deviation.

Cold rankings over three time periods for Tmin, Tmean, and Tmax (°C) for winters 2013/14 and 2014/15. We present three values: the ranking (rank), the RI (yr), and the average temperature Tavg (°C). Italicized numbers indicate that the winter was colder than the average for the time period. Asterisks indicate that the difference in the temperature exceeded one standard deviation.
Cold rankings over three time periods for Tmin, Tmean, and Tmax (°C) for winters 2013/14 and 2014/15. We present three values: the ranking (rank), the RI (yr), and the average temperature Tavg (°C). Italicized numbers indicate that the winter was colder than the average for the time period. Asterisks indicate that the difference in the temperature exceeded one standard deviation.

When compared to the entire record, winters with a mean temperature similar to 2013/14 and 2014/15 are not exceptionally rare, with a return interval of 5.18 and 4.89 yr, respectively. Both winters were characterized by colder-than-usual maximum temperatures, with RI of 8.80 and 5.68 yr, respectively. It is interesting to note, however, that this relationship changes when we consider our lived experience. Within the last 60 years, a winter of the magnitude of winter 2013/14 had an RI of 12.2 yr for Tmin and Tmean and 20.3 yr for Tmax. Over the last 30 years, however, such temperatures occur only once every 15.5 yr. Winter 2014/15, on the other hand, showed a very clear cold (Tmin) bias, with the longest return intervals for Tmin, relative to the 60- and 30-yr periods. Of note is the fact that the Tmin return interval is actually shortest when we consider the 2014/15 minimum temperature within the full range of recorded history (4.51 yr).

When examined individually, the coldest months of each of the two winters differ (see Table 2). Winter 2013/14 was characterized by three cold months that were all well below the period average. January 2014 was the coldest month of that season. Winter 2014/15, on the other hand, saw temperatures that ranged from a mild December (153rd coldest out of 175 winters) to a bitterly cold February—the 5th coldest since the start of records in Toronto and coldest since 1934. February 2015 was over 4°C colder than any month during the two winter seasons under study.

Table 2.

As in Table 1, but for cold rankings over three time periods for December, January, and February Tmean for winters 2013/14 and 2014/15.

As in Table 1, but for cold rankings over three time periods for December, January, and February Tmean for winters 2013/14 and 2014/15.
As in Table 1, but for cold rankings over three time periods for December, January, and February Tmean for winters 2013/14 and 2014/15.

b. Evolution of Toronto winter temperature since 1840/41

Figure 2 shows density curves for winter Tmax and Tmin at Toronto during our three study periods. The graph illustrates that, as we reduce the amount of history under consideration, warmer temperatures consistently represent a higher proportion of winter weather. Indeed, the horizontal distance between the left side of the curves for 1841–2015 and 1956–2015 indicate that the coldest winter temperatures in Toronto have not been experienced in the last 60 years. The effect is more pronounced in Tmin than it is in Tmax, with a larger lateral shift in temperatures toward warmer weather. While the proportion of warm weather consistently increases as we reduce our period of study. The variability in temperatures increases as we shorten our study period, as illustrated by the width of the 1986–2015 density curve. Figure 2, thus shows that, as our lived experience with winter in Toronto decreases, so too does our experience of colder winter temperatures.

Fig. 2.

Toronto winter temperature density for Tmax and Tmin over three time periods.

Fig. 2.

Toronto winter temperature density for Tmax and Tmin over three time periods.

Table 3 summarizes the results of the Durbin–Watson test for serial autocorrelation on the three temperature variables and three time periods. Of all the datasets analyzed, autocorrelation was only found in the full 175-yr dataset, with significant correlation factors at lag 2 winters for both Tmin and Tmean. For those two series, data were first prewhitened [as per Yue et al. (2002)] before the trend analysis was performed.

Table 3.

Durbin–Watson test results. An asterisk indicates that data are serially correlated at the 5% level.

Durbin–Watson test results. An asterisk indicates that data are serially correlated at the 5% level.
Durbin–Watson test results. An asterisk indicates that data are serially correlated at the 5% level.

Figure 3, top, shows winter Tmax, Tmin, and Tmean, at Toronto from winter 1840/41 to winter 2014/15. The variables have also been fitted with LOESS curves (Fig. 3, bottom), which show approximately increasing trends for all three variables. Increases in Toronto winter temperatures, however, are clearly nonlinear; there is a noticeable lull in temperature change or perhaps even a decrease in temperature prior to approximately the 1930s, after which temperature appears to rise steadily.

Fig. 3.

Toronto winter Tmax, Tmin, and Tmean from 1841 to 2015: (top) time series and (bottom) LOESS curve.

Fig. 3.

Toronto winter Tmax, Tmin, and Tmean from 1841 to 2015: (top) time series and (bottom) LOESS curve.

The Mann–Kendall trend analysis—which is summarized in Table 4—showed consistent warming in Tmax, Tmin, and Tmean, in the 175- (at the 0.001 significance level) and 60-yr (at the 0.1 significance level) time periods. In the past 30 years, however, none of the temperature variables showed any significant trend, a result that is at odds with the global trends.

Table 4.

Summary of Mann–Kendall trend analysis and Theil–Sen estimator results. The table shows Kendall’s τ for each variable and time period. Values in parentheses indicate τ after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.1 level (one asterisk).

Summary of Mann–Kendall trend analysis and Theil–Sen estimator results. The table shows Kendall’s τ for each variable and time period. Values in parentheses indicate τ after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.1 level (one asterisk).
Summary of Mann–Kendall trend analysis and Theil–Sen estimator results. The table shows Kendall’s τ for each variable and time period. Values in parentheses indicate τ after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.1 level (one asterisk).

Winter Tmax at Toronto warmed the least of the three temperature variables. Over the 175-winter period, the maximum temperature increased by approximately 1.9°C. Warming accelerated within the last 60 years, which account for more than three-quarters of the total warming since 1840/41. In the last 30 years, however, we found no significant increase in winter maximum temperature. It is notable that over the three periods studied, the standard deviation (σ) for Tmax was between σ = 1.5° and 1.6°C.

The average minimum winter temperature showed the most pronounced warming of the three variables studied: Tmin increased by approximately 2.9°C between 1840/41 and 2014/15. Of this warming, approximately 62% (2.3°C) occurred in the last 60 years. The Mann–Kendall test detected a negative trend in winter minimum temperature in the past 30 years, but the trend was not statistically significant. It should be noted that the winter minimum temperature was the variable with the most variation. The standard deviation for Tmin decreased from σ = 2.2° to 1.8°C for the 175- and 60-yr periods, respectively. Over the last 30 years, however, there was more variation than over the last 50 years, with a standard deviation of σ = 1.9°C.

Finally, the winter mean temperature at Toronto warmed by approximately 1.9°C between 1840/41 and 2014/15. As with Tmax and Tmin, the last 60 years account for a greater portion of the total (approximately 62% of warming since 1840/41). The variance in mean temperature for each period was roughly constant at σ = 1.9°, 1.6°, and 1.7°C in the 175-, 60-, and 30-yr periods, respectively. In the last 30 years, the Mann–Kendall test showed a negative correlation between the winter mean temperature and time, but this trend was not statistically significant.

The averaged seasonal DTR provides a further measure of changes in temperature variation over the same time periods (see Table 5). The seasonal average DTR showed a significant decrease over the full 175-winter period, a result of the faster warming in Tmin when compared to Tmax. The Mann–Kendall test did not detect any significant trend in DTR in the last 60 years. It is notable, however, that the trend analysis did detect an increase in seasonal average DTR of approximately 0.02°C yr−1 over the past 30 years (significant at the 0.05 level). This result indicates that there has been greater variation (wider range), and more extreme temperatures over the last 30 winters, and suggests the reversal of the previous trend.

Table 5.

As in Table 4, but for DTR.

As in Table 4, but for DTR.
As in Table 4, but for DTR.

Figure 4, top, shows LOESS curves for Toronto winter Tmax, Tmin, and Tmean for the 30 years from 1986 to 2015. All three temperature variables exhibit a hump-like shape, indicating decreasing temperatures in approximately the past 15 years. For contrast, Fig. 4, bottom, shows LOESS curves for the 28-yr period from 1986 to 2013. By excluding the anomalously cold winters at the end of the period, the LOESS curve takes on a very different shape, showing an apparent increase in all three temperature variables in the last five years of the period. To quantify the impact that the anomalously cold winters of 2013/14 and 2014/15 have on the period trend analysis, we performed the Mann–Kendall trend analysis on data subsets that excluded those cold winters. Table 6 shows the difference in the results of the trend analysis when the cold winters are omitted. While the Mann–Kendall test did not detect any significant trends in the 28-yr period preceding 2013/14, the omission of the cold winters did eliminate the negative trends that were estimated in the original results. Likewise, there were considerable differences in period trends for Tmax and Tmin. The Thiel–Sen slope estimation also showed that there was more warming in the 28-yr period before the two cold winters than there was in the 30-yr period including them. By excluding the cold seasons, the period trend for Tmin between 1840/41 and 2012/13 was 1.2°C larger than the period trend calculated for the 175-winter period. In addition, the significance level of the trends for all three variables in the 58-yr period preceding 2013/14 increased from 0.1 (for 1956–2015) to 0.05.

Fig. 4.

LOESS curves for Toronto winter Tmax, Tmin, and Tmean (top) from 1986 to 2015 and (bottom) from 1986 to 2013.

Fig. 4.

LOESS curves for Toronto winter Tmax, Tmin, and Tmean (top) from 1986 to 2015 and (bottom) from 1986 to 2013.

Table 6.

Difference in trend analysis results by excluding winters 2013/14 and 2014/15. The table shows difference in slope (calculated by the Thiel–Sen slope estimation; °C yr−1) by excluding the winters of 2013/14 and 2014/15. Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.05 level (one asterisk).

Difference in trend analysis results by excluding winters 2013/14 and 2014/15. The table shows difference in slope (calculated by the Thiel–Sen slope estimation; °C yr−1) by excluding the winters of 2013/14 and 2014/15. Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.05 level (one asterisk).
Difference in trend analysis results by excluding winters 2013/14 and 2014/15. The table shows difference in slope (calculated by the Thiel–Sen slope estimation; °C yr−1) by excluding the winters of 2013/14 and 2014/15. Italicized values indicate statistically significant monotonic trend at the 0.001 level (two asterisks) and 0.05 level (one asterisk).

To diagnose which part of the past 30 years showed the greatest increase in variation, we split the period into two 15-yr periods from 1986 to 2000 and from 2001 to 2015. The standard deviation of all three temperature variables was higher in the past 15 years (1.8°C for Tmax, 2.2°C for Tmin, and 2.0°C for Tmean) than in the preceding 15-yr period (1.4°C for Tmax, 1.5°C for Tmin, and 1.5°C for Tmean).

The results of our analysis of winter monthly temperature are summarized in Table 7. Figure 5 shows LOESS curves for the three temperature variables for the winter months at the Toronto weather station. With the exception of the January maximum temperature, there were significant warming trends for all temperature variables in all three months for the 175-yr period from 1840/41 to 2014/15. However, unlike the trends detected in the average winter temperature datasets, there were no significant trends detected in the last 60 years in January or February temperatures. Only December showed a positive trend in all three variables at the 0.05 significance level.

Table 7.

Trend analysis for monthly temperatures for December, January, and February from 1840/41 to 2014/15. The table shows Kendall’s τ for maximum, minimum, and mean temperature during winter months. Please note that the December temperature was analyzed for the periods of 1840–2014, 1965–2014, and 1985–2014. Values in parentheses indicate result (if different) after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend (see footnotes for significance).

Trend analysis for monthly temperatures for December, January, and February from 1840/41 to 2014/15. The table shows Kendall’s τ for maximum, minimum, and mean temperature during winter months. Please note that the December temperature was analyzed for the periods of 1840–2014, 1965–2014, and 1985–2014. Values in parentheses indicate result (if different) after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend (see footnotes for significance).
Trend analysis for monthly temperatures for December, January, and February from 1840/41 to 2014/15. The table shows Kendall’s τ for maximum, minimum, and mean temperature during winter months. Please note that the December temperature was analyzed for the periods of 1840–2014, 1965–2014, and 1985–2014. Values in parentheses indicate result (if different) after prewhitening autocorrelated data. Slope indicates the Thiel–Sen slope estimation (°C yr−1). Italicized values indicate statistically significant monotonic trend (see footnotes for significance).
Fig. 5.

LOESS curves for Tmax, Tmin, and Tmean for (top) December 1840–2014, (middle) January 1841–2015, and (bottom) February 1841–2015.

Fig. 5.

LOESS curves for Tmax, Tmin, and Tmean for (top) December 1840–2014, (middle) January 1841–2015, and (bottom) February 1841–2015.

Over the 175-yr period, December saw the greatest increase in maximum temperature, by approximately 2.6°C. Meanwhile, minimum temperatures in both December and February increased by approximately 0.03°C yr−1 since 1840/41. January, the coldest month of the year, showed the smallest increase in Tmin at 0.02°C yr−1.

The rapid warming of December could be indicative of a temporal shift in winter weather. While there were no significant trends detected by the Mann–Kendall test, in the past 30 years, December was the only winter month that did not show a negative temperature change. A supplementary Mann–Kendall test revealed nonsignificant negative trends in the past 30 years in the mean temperature of both March (−0.02 °C yr−1) and April (−0.01 °C yr−1), as well as a nonsignificant positive trend in May mean temperature (0.01 °C yr−1). While potentially the result of noise in the dataset, these trends could be preliminary indicators of a months-long shift in the temporal patterns of winter weather in Toronto, and should be monitored in the future.

Within the past 30 years, the significant warming trends in Toronto winter temperature appear to have broken down. While this could simply be the case of a signal being lost in the noise of an increasingly varied seasonal weather series, it is possible that some external driver is influencing Toronto winter temperatures. This study examined some of the myriad factors that influence Toronto winter temperatures, including the urban heat island affect, major global circulation pattern teleconnections, and the polar jet stream.

c. Factors influencing Toronto winter temperature

1) Urbanization

The Toronto weather station is located in the heart of the city’s downtown core. Since the 1840s, this area has transformed from the small colonial settlement of York to one of North America’s largest cities. This process of growth and urbanization has had significant impacts on the local climate. Mohsin and Gough (2010, 2012) compared urban stations in Toronto with nearby rural stations with similar climatological characteristics. In their 2010 study, Mohsin and Gough determined that the mean annual and mean winter temperatures at Toronto began to increase after 1928. The trend was stronger for Tmin than for Tmax, which is consistent with the fact that the urban heat island disproportionately affects nighttime minimum temperatures. Indeed, between 1971 and 2000, the amount of warming at Toronto that could be attributed to the urban heat island ranged from 0.01° to 0.02°C decade−1 for the period from 1970 to 2000. A Mann–Kendall test and Thiel–Sen estimator slope approximation shows that total warming (Tmean) during that period was approximately 0.25°C decade−1 at the 0.1 significance level. The UHI, therefore, accounted for between 4% and 8% of total warming from 1970 to 2000.

In contrast, at Toronto Pearson abrupt increases in temperature started after 1948 (Mohsin and Gough 2010). From 1970 to 2000, a period during which the area was still undergoing rapid urbanization, the warming that could be attributed to the UHI was between 0.3° and 0.35°C decade−1 (Mohsin and Gough 2012). The total warming for the same period at Toronto Pearson—as calculated from Environment and Climate Change Canada archival weather data—was approximately 0.6°C decade−1 at 0.01 significance. At the urbanizing Toronto Pearson station, therefore, the UHI accounted for over 50% of total warming in the period.

In addition to the background warming from global climate change, urbanization has contributed to higher temperatures in Toronto since the late 1920s (Mohsin and Gough 2010). However, given that the area around the Toronto station is fully urbanized, and considering the slower warming at Toronto compared to Toronto Pearson, additional warming from the UHI at that station might have reached saturation (Mohsin and Gough 2010). This slow-down in urbanization-related warming may have contributed to the apparent disappearance of a significant monotonic trend in Toronto winter temperature over the past 30 years in Toronto; however, it does not explain the occurrence of the anomalously cool winters in 2013/14 and 2014/15.

2) Influence of major teleconnections on Toronto winter temperatures

Winter temperature in Canada is influenced by the world’s major climatological teleconnections, phenomena where changes in circulation patterns in one part of the globe will have an influence on other, physically separated parts (NOAA National Weather Service 2009). Three such teleconnections that are understood to have a direct influence on Toronto’s winter climate are ENSO, an oceanic circulation anomaly in the Pacific Ocean that has impacts on atmospheric circulation patterns around the world and affects Toronto during its positive phase; and the AO and the NAO, which are atmospheric modes related to sea level pressure patterns north of 50°N. ENSO is an oceanic circulation anomaly; however, its impacts in Toronto are manifested through the atmospheric changes that occur during an El Niño event. In this sense, while these phenomena are of distinct origin, we have opted to consider the atmospheric modes together with ENSO in this section.

When the NAO or the AO is negative, the polar jet stream is weakened and penetrates farther south, which could result in colder temperatures in the city of Toronto. On the other hand, during positive El Niño events, the Toronto region can experience warmer winter weather. During positive NAO or AO events, the polar jet stream remains farther north and the cold Arctic air does not penetrate as far south as Toronto.

We performed a Mann–Whitney (M–W) U test to compare the average winter temperature during the past 30 years in Toronto during positive (Table 8) or negative (Table 9) events for each oscillation and during years without the effects of the oscillation. The average temperature during winters affected by any one of the oscillations, in either their positive or negative phase, did not show any significant departure from the average temperatures from years where the oscillation was not active. Likewise, a Mann–Kendall test showed no statistically significant relationship between winter Tmean and AO or between Tmean and NAO events. This result was confirmed by a Spearman correlation test, which showed no statistical significance for either index when correlated with Tmean.

Table 8.

Comparison of temperature (°C) during positive ENSO, NAO, and AO events, compared to nonevent years.

Comparison of temperature (°C) during positive ENSO, NAO, and AO events, compared to nonevent years.
Comparison of temperature (°C) during positive ENSO, NAO, and AO events, compared to nonevent years.
Table 9.

As in Table 8, but for negative ENSO, NAO, and AO events.

As in Table 8, but for negative ENSO, NAO, and AO events.
As in Table 8, but for negative ENSO, NAO, and AO events.

When averaged over the winter season, there was no significant relationship between any of the three major oscillations and winter temperatures in Toronto in the past 30 years. Future research should focus on the combined influence of two or all of these indexes on winter temperature in Toronto and seek to identify if there is a lagged response of temperature in the city to teleconnection cycles.

3) Weakening of the jet stream due to Arctic warming

Cold polar air from the Arctic is constrained in the Arctic region along the polar front. The polar front and the polar front jet stream follow an undulating longitudinal Rossby wave pattern. The magnitude of these waves—the size of the undulations—controls how far south polar air can penetrate, influencing temperature in the midlatitudes. Recent research has focused on the nature of these undulations. According to Francis and Vavrus (2012), the warming of Arctic air masses in recent decades has resulted in a modification of the pressure differentials between cold polar air and warm air from over the continental United States. As these pressure differentials weaken, the polar jet stream also becomes weaker. With a weakened polar jet, the amplitude of the north–south Rossby wave pattern increases, and the jet stream begins to meander. This meandering jet stream shifts more slowly and can sometimes curve back toward itself, causing small droplets of cold air to form over temperate areas in the midlatitudes (Francis and Vavrus 2012). It should be noted that there has not yet been a statistically robust study of this phenomenon (Vihma 2014), with a number of studies providing somewhat contradictory evidence (see Francis and Vavrus 2012; Screen and Simmonds 2013; Barnes 2013). This paper, therefore, explores the possibility that a meandering jet stream may have affected Toronto winter temperature in the past decades.

Figure 6 shows the 500-mb heights (1 mb = 1 hPa) and anomalies for the polar jet stream during January 2014 and February 2015, the coldest months of each of the winter periods being studied. The cold anomalies illustrated in Fig. 1 correspond closely to the 500-mb height anomalies in Fig. 6. It should be noted that although we found no significant difference in Toronto winter temperatures during NAO events, a positive phase NAO is characterized by warm Arctic weather. During the winters of 2013/14 and 2014/15, the NAO was in a positive phase, which may have led to more warming in the Arctic and, therefore, an even weaker jet stream. This may, in part, explain the deeper penetration of cold Arctic air into the south during those seasons. In this sense, while a positive NAO shows no direct relationship with colder winter temperatures in Toronto, the NAO does affect the position of the polar jet and may indirectly influence temperatures.

Fig. 6.

The 500-mb heights and anomalies for the polar jet stream during (top) January 2014 (NOAA NCEI 2014) and (bottom) February 2015 (NOAA NCEI 2015b).

Fig. 6.

The 500-mb heights and anomalies for the polar jet stream during (top) January 2014 (NOAA NCEI 2014) and (bottom) February 2015 (NOAA NCEI 2015b).

The daily air mass origin for each day of winters 2013/14 and 2014/15 is displayed in Fig. 7 and the results of the air mass analysis are presented in Table 10. During winter 2013/14, on 60.2% of days (52 days) at Toronto Pearson, the prevailing air masses in Toronto were of polar origin. Of the 16 cold days that occurred in January, 10 days fell within the last fortnight of the month, and continued through February, which experienced a total of 20 polar air days (71.4%). Of these, 13 days occurred in the first two weeks of that month, indicating a prolonged period of nearly exclusively polar air in Toronto from mid-January to mid-February 2014. The number of days with polar air during February was 13.3% above the climatological (1950–2015) norm.

Fig. 7.

SCC2 air masses at Toronto Pearson during DJF (top) 1993/94, (middle) 2013/14, and (bottom) 2014/15.

Fig. 7.

SCC2 air masses at Toronto Pearson during DJF (top) 1993/94, (middle) 2013/14, and (bottom) 2014/15.

Table 10.

Monthly and winter air mass analysis for Toronto Pearson. This table presents the results of the binned air mass analysis, with values presented as the percentage of days that air in Toronto was of a particular binned class. Values marked in boldface (italics) showed an absolute change of greater than 10% above (below) the climatological norm (1950–2015).

Monthly and winter air mass analysis for Toronto Pearson. This table presents the results of the binned air mass analysis, with values presented as the percentage of days that air in Toronto was of a particular binned class. Values marked in boldface (italics) showed an absolute change of greater than 10% above (below) the climatological norm (1950–2015).
Monthly and winter air mass analysis for Toronto Pearson. This table presents the results of the binned air mass analysis, with values presented as the percentage of days that air in Toronto was of a particular binned class. Values marked in boldface (italics) showed an absolute change of greater than 10% above (below) the climatological norm (1950–2015).

Winter 2014/15 experienced 55 days during which the air over Toronto Pearson was of polar origin (61.1%). That season, the relatively mild December gave way to cold weather in January and February. In particular, February 2015 was characterized by polar air on a total of 23 days (82.1%), which represents an increase of 24.1% over the climatological norm. Indeed, winter 2014/15 was characterized by a prolonged cold event, with cold air in the city from 24 January through 2 March 2015. During this 38-day period, there were only three brief periods of mixing with transitional air masses, none of which exceeded three days in length. There was not a single day during February 2015 when the air mass was moderate or tropical—typically 28.4% and 1.5%, respectively—suggesting a prolonged period during which the polar jet was south of Toronto.

The coldest winter in Toronto in the past 30 years occurred between December 1993 and February 1994. According to Assel et al. (1996), the temperature during that season can also be explained by an anomalously low polar vortex. For the sake of comparison, we performed an air mass analysis for this winter as well. The analysis supported the hypothesis put forth by Assel et al. Winter 1993/94 showed an anomalously high number of days characterized by polar air masses in both January (74.2%) and February (67.9%), 18.8% and 9.8% above the climatological norm, respectively.

Interestingly, both December 1993 and December 2014 showed anomalously low proportions of days with polar air, with 10.4% and 11.6% more moderate air than the climatological norm. These warm Decembers may have incremented the susceptibility of Torontonians to the cold weather of the following months, and are likely to have influenced the measure of seasonal DTR, reported earlier in this study.

The air mass analysis shows that the prevailing air masses in Toronto during the coldest months of the cold winters in 2013/14 and 2014/15 were polar air masses, which indicates that the polar jet stream was south of the city for prolonged periods during those months. In fact, both cold events described herein occurred over the span of approximately four weeks, suggesting prolonged blocks in in the polar jet stream.

While this study has focused on winter temperature, further evidence suggests that anomalous cooling in Toronto may not be restricted to the winter months. Gough and Sokappadu (2016) studied Toronto’s cold 2014 summer in a climate context. The authors analyzed the binned climatological distribution of cool, moderate, and warm air masses during the summer season and identified the larger amplitude of Rossby waves along the polar front as a possible cause for anomalously cold weather in Toronto during summer 2014.

This study has found increasing variation in winter temperatures in the past 30 years in Toronto. Given the direct evidence linking the weakened or meandering jet stream to cold events during the 1993/94 winter (Assel et al. 1996), summer 2014 (Gough and Sokappadu 2016), and the cold winters of 2013/14 and 2014/15, it is possible that changes in the behavior of the jet stream are responsible not only for changes in winter temperature, but for temperature extremes in Toronto year-round. Further study is required to assess the long-term effects of jet stream shifts on Toronto temperatures, and to assess whether the spatial and temporal variation in jet stream behavior will continue to increase in a statistically significant manner.

4. Summary

Cold winter temperatures in Toronto, such as the anomalously cold winters of 2013/14 and 2014/15, have direct impacts on human health, infrastructure, shipping, and agriculture. This study sought to determine the climate context for these cold winters by ranking winter temperature over the last 175, 60, and 30 years. In the long-term climate record, winters 2013/14 and 2014/15 were among the top-36 coldest winters since records began at the Toronto weather station in 1840, with a return interval of approximately 5 yr. Likewise, these two seasons were among the top three coldest seasons in the last 30 years. To understand whether these winters represent one-time anomalies, or are part of a larger increase in winter extremes, we examined the evolution of Toronto winter Tmax, Tmin, and Tmean since winter 1840/41. Toronto has experienced significant warming in all three temperature variables when studied over the past 175 or 60 years. Over the last 30 years, however, there was no clear signal in winter temperature. The increasing variation and diurnal temperature range observed over the past 30 years, however, indicate that the frequency of extreme winter temperatures could be increasing.

The specific factors that generate noise and variation in Toronto’s winter temperature over the past 30 years are not known. Major atmospheric and oceanic oscillations such as El Niño–Southern Oscillation, the North Atlantic Oscillation, and the Arctic Oscillation have direct and indirect effects on weather in the city; however, this study found no significant difference in winter temperatures in Toronto in seasons where the mentioned oscillations were in either a positive or a negative phase. Further study is required to better understand how the interaction of these phenomena can affect Toronto’s winter temperature, and the temporal scale on which these effects will be apparent.

Other aspects of Toronto’s climatology, such as the lake effect and the urban heat island, have not been studied in great detail beyond the year 2000. Studies outside of Toronto have shown that lake effect snow in winter months is increasing due to climate warming (Burnett et al. 2003), however, the effect of climate change on the lake effect in Toronto has yet to be determined. Further detailed study is required.

The anomalously cold weather in winters 2013/14 and 2014/15 can be attributed to the anomalously high proportion of polar air over the city during those seasons. This influx of polar air may be due to shifts in the northern polar vortex and the strength of the polar jet stream. As the Arctic continues to warm, the behavior of the polar front may continue to be erratic and unpredictable. Given that Toronto is situated along the polar front, it is likely that the city can expect more cold winters, late winter weather, false springs, and even year-round temperature extremes as the once predictable behavior of the polar front changes due to accelerated Arctic warming.

Acknowledgments

Thank you to T. Mohsin for teaching skills necessary for effective trend analysis. Thank you to J. Jien, K. Leung, and M. Sheremata for additional feedback. Thank you to the editor and the anonymous reviewers who provided valuable feedback and suggestions for improvements to this paper.

REFERENCES

REFERENCES
Allen
,
S. M. J.
,
W. A.
Gough
, and
T.
Mohsin
,
2015
:
Changes in the frequency of extreme temperature records for Toronto, Ontario, Canada
.
Theor. Appl. Climatol.
,
119
,
481
491
, doi:.
Assel
,
R. A.
,
J. E.
Janowiak
,
S.
Young
, and
D.
Boyce
,
1996
:
Winter 1994 weather and ice conditions for the Laurentian Great Lakes
.
Bull. Amer. Meteor. Soc.
,
77
,
71
88
, doi:.
Barnes
,
E. A.
,
2013
:
Revisiting the evidence linking Arctic amplification to extreme weather in midlatitudes
.
Geophys. Res. Lett.
,
40
,
4728
4733
, doi:.
Burnett
,
A. W.
,
M. E.
Kirby
,
H. T.
Mullins
, and
W. P.
Patterson
,
2003
:
Increasing Great Lake–effect snowfall during the twentieth century: A regional response to global warming?
J. Climate
,
16
,
3535
3542
, doi:.
Environment and Climate Change Canada
,
2015
: Climate Trends and Variations Bulletin—Winter 2014–2015, 5 pp. [Available online at http://publications.gc.ca/pub?id=9.507871&sl=0.]
Environment Canada
,
2014
: Great Lakes Climate Quarterly—March 2014. [Available online at http://www.ec.gc.ca/eau-water/?n=B6344EC5-1.]
Environment Canada
,
2015
: Great Lakes Climate Quarterly—March 2015. [Available online at http://www.ec.gc.ca/eau-water/?n=F5329B03-1.]
Francis
,
J. A.
, and
S. J.
Vavrus
,
2012
:
Evidence linking Arctic amplification to extreme weather in mid-latitudes
.
Geophys. Res. Lett.
,
39
, L06801, doi:.
Gough
,
W. A.
, and
Y.
Rozanov
,
2001
:
Aspects of Toronto’s climate: Heat island and lake breeze
.
Can. Meteor. Oceanogr. Soc. Bull.
,
29
,
67
71
.
Gough
,
W. A.
, and
S.
Sokappadu
,
2016
:
Climate context of the cold summer of 2014 in Toronto, ON, Canada
.
Theor. Appl. Climatol.
,
126
,
183
189
, doi:.
Ho
,
E.
, and
W. A.
Gough
,
2006
:
Freeze thaw cycles in Toronto, Canada in a changing climate
.
Theor. Appl. Climatol.
,
83
,
203
210
, doi:.
Huang
,
B.
, and Coauthors
,
2015
: Extended Reconstructed Sea Surface Temperature (ERSST), version 4. NOAA National Centers for Environmental Information, accessed 17 March 2017, doi:.
Lawrimore
,
J. H.
,
M. J.
Menne
,
B. E.
Gleason
,
C. N.
Williams
,
D. B.
Wuertz
,
R. S.
Vose
, and
J.
Rennie
,
2011
:
An overview of the Global Historical Climatology Network monthly mean temperature data set, version 3
.
J. Geophys. Res.
,
116
, D19121, doi:.
Mohsin
,
T.
, and
W. A.
Gough
,
2010
:
Trend analysis of long-term temperature time series in the Greater Toronto area (GTA)
.
Theor. Appl. Climatol.
,
101
,
311
327
, doi:.
Mohsin
,
T.
, and
W. A.
Gough
,
2012
:
Characterization and estimation of urban heat island at Toronto: Impact of the choice of rural sites
.
Theor. Appl. Climatol.
,
108
,
105
117
, doi:.
NOAA NCEI
,
2014
: State of the Climate: Synoptic Discussion for January 2014. NOAA National Centers for Environmental Information. [Available online at https://www.ncdc.noaa.gov/sotc/synoptic/201401.]
NOAA NCEI
,
2015a
: State of the Climate: Global Analysis for Annual 2014. NOAA National Centers for Environmental Information. [Available online at https://www.ncdc.noaa.gov/sotc/global/201413.]
NOAA NCEI
,
2015b
: State of the Climate: Synoptic Discussion for February 2015. NOAA National Centers for Environmental Information. [Available online at https://www.ncdc.noaa.gov/sotc/synoptic/201502.]
NOAA NCEI
,
2016
: State of the Climate: Global Analysis for Annual 2015. NOAA National Centers for Environmental Information. [Available online at https://www.ncdc.noaa.gov/sotc/global/201513.]
NOAA National Weather Service
,
2009
: NOAA National Weather Service Glossary. [Available online at http://w1.weather.gov/glossary.]
Null
,
J.
,
2015
: El Niño and La Niña years and intensities based on Oceanic Niño Index (ONI). [Available online at http://ggweather.com/enso/oni.htm.]
Screen
,
J. A.
, and
I.
Simmonds
,
2013
:
Exploring links between Arctic amplification and mid-latitude weather
.
Geophys. Res. Lett.
,
40
,
959
964
, doi:.
Sheridan
,
S. C.
,
2002
:
The redevelopment of a weather-type classification scheme for North America
.
Int. J. Climatol.
,
22
,
51
68
, doi:.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
, doi:.
Vihma
,
T.
,
2014
:
Effects of Arctic sea ice decline on weather and climate: A review
.
Surv. Geophys.
,
35
,
1175
1214
, doi:.
Vincent
,
L. A.
,
X. L.
Wang
,
E. J.
Milewska
,
H.
Wan
,
F.
Yang
, and
V.
Swail
,
2012
:
A second generation of homogenized Canadian monthly surface air temperature for climate trend analysis
.
J. Geophys. Res.
,
117
, D18110, doi:.
Yue
,
S.
, and
C. Y.
Wang
,
2002
:
Applicability of prewhitening to eliminate the influence of serial correlation on the Mann–Kendall test
.
Water Resour. Res.
,
38
, 1068, doi:.
Yue
,
S.
,
P.
Pilon
,
B.
Phinney
, and
G.
Cavadias
,
2002
:
The influence of autocorrelation on the ability to detect trend in hydrological series
.
Hydrol. Processes
,
16
,
1807
1829
, doi:.

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).