Variations in temperature‐related extreme events (1975–2014) in Ny‐Ålesund, Svalbard

We present a comprehensive analysis of temperature‐related extreme events in Ny‐Ålesund (78.9°N, 11.9°E) using data from three meteorological stations. The results show that annual mean temperatures in Ny‐Ålesund increase at a rate that is four times faster than the global mean from 1975 to 2014 with no ‘hiatus’ in recent decades. The annual diurnal temperature range shows a negative trend as minimum daily temperature increases at a faster rate than maximum daily temperature. A negative trend in cold extremes and a positive trend in warm extremes are observed. This asymmetry hints at potential changes in the probability distribution of temperatures in Ny‐Ålesund.


Introduction
Extreme events, such as heat waves, droughts, floods, and hurricanes, interact with exposed and vulnerable human and natural systems to produce disasters (IPCC, 2013). Overwhelming evidence indicates that climate extremes have frequently occurred on both regional and global scales in recent years (Osborn et al., 2000;Alexander et al., 2006;Donat et al., 2013;Chen and Sun, 2015). The intensification of climate extremes is causing dangerous changes in the economy and ecosystems, and to society and human health (Handmer et al., 2012). Enhanced warming is observed at high latitudes in response to increased CO 2 relative to the warming trends at mid-and low-latitudes. Polar amplification may cause an increased probability of extreme weather events in mid-latitudes (Francis and Vavrus, 2012). Therefore, the variations in mean climate and extreme events at high latitudes should be investigated.
Recent studies indicate that climate extremes have changed at mid-high latitudes. Cold/warm extreme events show a positive trend in North America since the late 1960s (CCSP, 2008a;Peterson et al., 2008;Zhang et al., 2010). In European and Mediterranean countries, a decrease in large-scale occurrences of cold extremes and an increase in warm/hot extremes are generally consistent with global trends of temperatures and their extremes (Moberg et al., 2006;Della-Marta et al., 2007). A significant positive trend in the frequency of heat waves and a negative trend in the number of cold spells have also been observed in Australia and the majority of Asia since the middle of the 20th century (Chambers and Griffiths, 2008;Choi et al., 2009;You et al., 2010;Wang et al., 2013;Chen and Sun, 2014).
However, few studies have examined extreme events in polar regions. Tuomenvirta et al. (2000) indicated that mean maximum and minimum temperatures in western coastal Greenland show a decreasing trend from 1950 to 1995 and opposite trends over the Nordic Seas and Fenno-Scandia. These results indicate considerable regional differences for extreme events over the Arctic. A high frequency in warm days is observed at Svalbard Airport from 1975 to 2010 with a negative trend in cold nights (Bednorz, 2011;Bednorz and Kolendowicz, 2013;Tomczyk and Bednorz, 2014).
Ny-Ålesund (78.9 ∘ N, 11.9 ∘ E), which is located on the west coast of Svalbard (Spitsbergen), is one of the northernmost archipelagos in the Arctic. Although Ny-Ålesund has its own weather features, it provides valuable evidence for climate change in the general Arctic. Several studies have analysed climate change in Ny-Ålesund and identified a significant warming in recent decades (Førland and Hanssen-Bauer, 2000;Førland et al., 2011;Maturilli et al., 2013Maturilli et al., , 2015. However, extreme events and their changes have been given minimal attention. In this study, we investigate changes in temperature and its related extremes in Ny-Ålesund based on data from three meteorological stations. This investigation may advance our understanding of intensity and frequency of climate extremes in the Arctic and the influence of global warming on extreme events.

Climate extreme indices
The 14 temperature-related extreme indices proposed by the Expert Team on Climate Change Detection and Indices (ETCCDI) are analysed in this study ( Table 1). The ETCCDI climate indices are very popular and extensively utilized in climate research and related fields (e.g. Zhang et al., 2011;Sillmann et al., 2013) due to their robustness and fairly straightforward calculation and interpretation. The indices are primarily selected for the assessment of changing climates, including changes in intensity, frequency and duration of temperature-related events. The indices are calculated using a FORTRAN package, as documented at the ETCCDI climate change indices website (http://etccdi.pacificclimate.org/software.shtml). All raw data from the three stations are resampled to a common daily resolution using an arithmetic mean. Note that we use the Sen's slope method (Sen, 1968) to compute the linear trend of the considered variables and the non-parametric Mann-Kendall approach for the significance test. Regarding probability density distribution (PDF), we employ the Kolmogorov-Smirnov two-sample test to determine if two samples are from the same distribution.

Results
The annual mean air temperature in Ny-Ålesund shows a significant warming trend from 1975 to 2014 (∼0.79 ∘ C decade −1 ; Table S1), which is four times larger than the trend for global mean temperature (∼0.17 ∘ C decade −1 ) (Figure 1(a)). Temperatures during the winter season exhibit an enhanced warming (1.75 ∘ C decade −1 ) with greater inter-annual variability (3.34 ∘ C year −1 ). The observations from the three stations show reasonable consistency (Figure 1). Based on an 11-year sliding linear trend, the magnitude of the warming trend in Ny-Ålesund is gradually decreased (Figure 1(b)), which is consistent with the global mean. The results show a continuous warming trend in the winter temperatures in Ny-Ålesund since the mid-1990s, whereas the trend of global mean temperature is weakened and turns to a negative phase in recent years (Figure 1(d)). The global mean summer temperatures exhibit a 'hiatus' since the late 1990s, whereas summer temperatures in Ny-Ålesund exhibit a negative trend in recent years (Figure 1(f)). These results   indicate different temperature behaviours between the Ny-Ålesund and the globe. Thus, the variations in temperature-related extremes in Ny-Ålesund should be examined. Figure 2 illustrates the time series of mean maximum and minimum daily temperatures in Ny-Ålesund. Both time series show a significant positive trend in the past 40 years (Figure 2(a)-(c)), especially for the winter season since mid-1990s (Table S1). The annual diurnal temperature range (DTR) exhibits a decreasing trend (0.012 ∘ C year −1 ) during 1975-2014 (Figure 1(g)) as the minimum daily temperature increases at a faster rate than the maximum daily temperature (Table S1). The negative trend in annual DTR since the mid-1980s is contradictory to the DTR behaviour over global land surfaces (Rohde et al., 2013). Note that the summer DTR shows a positive trend (0.02 ∘ C year −1 ) due to the rapid increase in the maximum daily temperature.  In response to the increased temperatures, a significant negative trend in the annual occurrence of cold days (∼2.7% decade −1 ) and nights (∼3.5% decade −1 ) is observed from 1975 to 2014 (Figure 3(b) and (d); Table S2). Although the annual occurrence of warm days and nights shows a long-term upward trend in the past 40 years (Figure 3(a) and (c)), their magnitudes (1.3-1.8% decade −1 ) are relatively small. The annual occurrence of warm nights does not exhibit a significant trend from 1975 to 2000 and shows a negative trend in recent years (Figure 3(c)). The temperatures on the coldest days (nights) significantly increase during 1975-2014 (Figure 3(f) and (g)) with a significant positive trend of 2.3 (2.5) ∘ C decade −1 (Table S2). However, the temperatures on the hottest days and nights reveal a large magnitude of inter-annual variations with an insignificant long-term upward trend (Figure 3(e) and (h)). This result may be attributed to a small warming trend in summer temperatures. In addition, the local SST and large-scale circulation patterns are likely to influence the absolute temperature indices (e.g. temperatures on the hottest days) based on previous studies (Scaife et al., 2008;Alexander et al., 2009).
The cold spell duration exhibits a significant negative trend of ∼8 days decade −1 from 1975 to 2014 with a larger decreasing trend from 1994 to 2010 (Figure 3(j); Table S2). The significant decline since the early 1990s is similar to the trend in cold spells at middle and high latitudes Chen and Sun, 2015). Although the warm spell duration increased over the past 40 years, minimal change and a downward trend since the early 2000s is observed (Figure 3(i)). The annual occurrence of frost days and ice days in Ny-Ålesund shows a significant negative trend (3-9 days decade −1 ) during the study period, especially since the early 1990s (Figure 3(k) and (l)). The growing season length, which affects hydrologic factors and biologic territories (Logan et al., 2003;CCSP, 2008b), increases in Ny-Ålesund from 1975 to 2014 (Figure 3(m)) with a long-term trend of ∼4 days decade −1 . This result is similar to the change in the growing season length in the eastern and central Tibetan Plateau (Liu et al., 2006). The temperature-related extremes in Ny-Ålesund show a negative trend in cold extremes and a positive trend in warm extremes. The negative trend in extreme minimum temperatures is larger than the positive trend in extreme maximum temperatures. This asymmetry hints at potential changes in the shape and/or scale of the distribution of temperatures in Ny-Ålesund. According to the probability distribution function, it is likely (at 99% confidence level) that there is a decreased frequency in the cold tails of temperatures after mid-1990s and a decreased frequency of temperature above freezing point (Figure 4). The maximum and minimum temperatures indicate a similar shift. However, additional observations are required to accurately assess potential changes in the shape of the distribution of temperature.

Summary
In this study, we present a comprehensive picture of changes in temperature-related extremes from 1975 to 2014 based on the station data in Ny-Ålesund. The annual warming trend in Ny-Ålesund during the past 40 years is four times larger than the global mean. The continuous warming since the mid-1990s is inconsistent with the 'hiatus' in global warming trend in terms of annual mean temperature. The annual DTR exhibits a negative trend from 1975 to 2014, whereas the summer DTR shows a positive trend. In response to the increased temperature, the temperature-related extremes in Ny-Ålesund show a negative (positive) trend in cold (warm) extremes. Specifically, Ny-Ålesund experiences a significant increase (decrease) in the annual occurrence of warm days (cold nights). The temperatures on the coldest days (nights) exhibit a significant positive trend of 2.3 (2.5) ∘ C decade −1 , whereas the trend of the temperatures on the hottest days (nights) is moderate. The annual occurrence of cold spells, frost days, and ice days shows a long-term negative trend from 1975 to 2014, whereas the annual occurrence of warm spells shows a positive trend. In addition, the growing season length is prolonged in Ny-Ålesund with a trend of ∼4 days decade −1 .

Supporting information
The following supporting information is available:  Table S1. Linear trends ( ∘ C year −1 ) of the maximum temperature, minimum temperature, diurnal temperature range and mean temperature at the three stations. Table S2. Linear trends of temperature-related extremes (Table 1) at the three stations. Units are % year −1 for TX90p, TX10p, TN90p, and TN10p, ∘ C year −1 for TXx, TXn, TNx, and TNn, and days year −1 for the remaining indices.