- Email: [email protected]
Variability of observed temperature-derived climate indices in the Arctic
Global and Planetary Change 69 (2009) 214–224
Contents lists available at ScienceDirect
Global and Planetary Change j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g l o p l a c h a
Variability of observed temperature-derived climate indices in the Arctic Heidrun Matthes ⁎, Annette Rinke, Klaus Dethloff Alfred Wegener Institute for Polar and Marine Research, Potsdam, Germany
a r t i c l e
i n f o
Article history: Received 18 June 2009 Accepted 10 October 2009 Available online 21 October 2009 Keywords: Arctic temperature climate extremes
a b s t r a c t Arctic temperature is analyzed in view of its extremes based on climate indices derived from daily mean, maximum, and minimum temperature. This analysis is done for the pan Arctic domain and region-specific for the eastern and western Russian Arctic. The variability of temperature-related indices over the last four decades is presented, in which the spatial distribution and regional differences as well as its temporal trends are discussed. The analysis is based on ERA40 data and station data in the Russian Arctic. Station-based results for 1958–2008 show a significant decrease of frost days (− 0.8 days/decade) over the eastern Russian Arctic in spring. The trends in warm (cold) spell days are not statistically significant; except in western Russian Arctic in summer for cold spell days (− 2 days/decade). The inter-annual variability of the indices is large and shows pronounced decadal variations. ERA40 data generally reproduce well the inter-annual variability in the climate indices as seen in observations, but they show some deficiencies in the magnitudes and trends of the indices. © 2009 Elsevier B.V. All rights reserved.
1. Introduction The Arctic is under rapid transition and observations show a consistent picture of surface warming and reduction in all components of the cryosphere (IPCC, 2007). Studies on trends of seasonal mean temperature indicate warming trends over most of the Arctic, especially pronounced in winter and spring, but depending on the period analyzed (Serreze and Francis, 2006). The ACIA (2005) report presents the Arctic warming for annual trends as well. It is also suggested that the expected future warming may be associated with a greater frequency of extremely warm temperatures (recently, e.g., Rinke and Dethloff, 2008). The understanding of the climate variability of extremes on a regional level is of increasing importance as it directly impacts the residents and is needed for investigations of ecological and societal changes. In recent years, specific months (e.g., record warm March 1996 in Northern Alaska, record warm spring and summer 2007 in the Arctic) were characterized by record temperatures, and attracted great attention in the public. However, it is apparent that the patterns and variability of extreme events have received little attention in the Arctic assessments. Thus, the question has arisen whether significant changes of extreme temperature are already present in the Arctic. Global analyses of temperature and precipitation extremes have been conducted by various authors using land-based gridded daily data sets. For example, Kiktev et al. (2003) show trends in the annual number of frost days for the northern
⁎ Corresponding author. Alfred Wegener Institute for Polar and Marine Research Telegrafenberg A43, D-14473 Potsdam, Germany. Tel.: +49 331 2882101; fax: +49 331 2882178. E-mail address: [email protected] (H. Matthes). 0921-8181/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.gloplacha.2009.10.004
hemisphere, indicating a significant decrease in frost days of up to 4 days/decade for almost all mid-latitudes. For high latitudes, trends are mainly insignificant or could not be calculated due to a lack of data. The authors calculate a warming over Scandinavia and Alaska associated with a decreasing trend in frost days of up to 2 days/ decade; a cooling is calculated over Eurasia south of Novaya Zemlya. Alexander et al. (2006) provide more complete information for the Arctic. They calculate that the number of frost days and the cold spell days decreased over most parts of the Arctic. However, most trends over the high latitudes are not significant. Our analysis aims at providing more regional results for temperature-based climate indices over the Arctic and is based on the European Centre for Medium-Range Weather Forecasts reanalysis (ERA40) and on Russian station data, covering 1958–2008. The detailed regional analysis for the Russian Arctic has its background in the EU project CARBO—North (http://www.carbonorth.net/) which aims at quantifying the carbon budget in Northern Russia. For this purpose, past, recent and future climate changes are investigated in the Russian Arctic. The paper continues in Section 2 with a description of the applied data sets and the performed data analysis. In Section 3, the spatial and temporal variability of selected temperature-related climate indices and their trends are presented. The results are summarized in Section 4. 2. Data and analysis 2.1. Daily temperature data sets The data set for the pan Arctic analysis is the ERA40 reanalysis data set which covers the 44-year-long period from 1958 to 2001
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
(Uppala et al., 2005). For the comparison with the station data, the ERA40 data have been interpolated on the station locations. The indices have then been calculated from the interpolated time series after removing those days which had missing values in the station data. The station data set used for the analysis over the Russian Arctic is called “Global Summary of the Day” (GSOD; www.ncdc.noaa.gov/oa/ mpp/freedata.html). From this, a total of 644 stations (all located in the Russian Arctic and north of 60 N) have been used, covering 1958– 2008. Compared to ERA40, here the last recent 7 years 2002–2008 are included. It is the case that almost no station has full data coverage over the whole 51-year-long period. The data set has been split for the analysis into two separate parts, GSOD east (324 stations) and GSOD west (320 stations), indicating that they include all stations located east and west of Ural mountains, respectively (Fig. 1). Therewith, the analysis has been done regionally for the two different climate regions of east and west part of Russian Arctic. From both data sets, daily minimum, maximum and mean temperatures have been analyzed.
2.2. Climate indices and trend analysis The classification of extreme temperature is based on climate indices (Peterson et al., 2001) and shown in Table 1. Here in the paper, selected temperature-related indices are discussed which are important for the living conditions in the Arctic. In addition to temperature indices (warm and cold spell days), indices illustrating vegetation conditions (growing season length, growing degree days) and frost conditions (frost days) are studied. A basis for calculating the indices has been defined to take the occurrence of missing data values into account. The data of a station were included in the analysis of a season if the station had 10 or less missing values in the input data within that season, except for indices that count consecutive days (like cold spells). For those indices, only seasons with full data coverage were included. The presented times series show a station mean for a specific season, if at least 10% of the
215
Table 1 Classification of extreme temperature based on climate indices. TX90 is the 90th percentile calculated from a five-day gliding mean of the daily maximum temperature of the reference period 1961–1990 and TN10 is the 10th percentile calculated from a five-day gliding mean of the daily minimum temperature of the reference period. t_mean is used as denotation for the daily mean temperature, t_min and t_max for daily minimum and maximum temperature respectively. Frost days Cold spell days
FD CSDI
Cold nights Warm spell days
TN10p WSDI
Warm day-times Growing degree days
TX90p GD4
Growing season length
GSL
# of days per season with t_min < 0 °C # of at least 6 consecutive days with t_min < TN10 # of days per season with t_min < TN10 # of at least 6 consecutive days with t_max > TX90 # of days per season with t_max > TX90 Sum of daily temperature means per season for days with t_mean > 4 °C # of days between the first occurrence of 6 consecutive days with t_mean > 5 °C and the first occurrence after the 1st July of at least six consecutive days with t_mean < 5 °C
stations deliver data for that season. Counters (like frost days) were normalized on the maximum number of days in the season, accounting for the existence of missing values. To show the variability of the inter-annual variability within the calculated time series, the standard deviation development relative to an 11-year gliding mean (and assigned to the center of the 11-year window) is included in the figures. The standard deviation is calculated if at least 6 years within the 11-year window have data. Furthermore, trends in the climate indices have been analyzed. They were calculated using a linear regression with the least squares method. To gain information on the significance of the obtained trends, a bootstrapping approach was used as described in Kiktev et al. (2003), except for the moving block re-sampling. To account for spatial correlation in the gridded ERA40 data, the same re-sampling sequence was used for each grid point (see again Kiktev et al., 2003). All trends marked significant are significant at the 95%-level. Significance was assigned to a “zero trend”, if the sum of squares of deviations from the mean were less than 5% of the mean. As the time series in the station data contain missing values, it is necessary to make sure that a shown trend is evenly based on values from the entire period. Therefore, the whole 51-year period has been divided in four 11-year-long intervals and a 7-year-long interval for the rest (1958–1968, 1969–1979, 1980–1990, 1991–2001, and 2002–2008). And thus, a trend for a certain period is only shown, if at least five years (respectively three years in the last period) of data are contained for each of these intervals within the trend's period. Additionally, the spatial distribution of stations delivering values to a certain index could have changed throughout the analyzed period and thereby influenced the calculated trends. To asses this problem, the spatial distribution of missing values over time was analyzed for each index. The missing values were found to be rather evenly distributed over the whole area taken into account. 3. Results 3.1. Frost days
Fig. 1. Map of station distribution of the GSOD data set. In total, daily data from these 644 stations have been analyzed (324 stations in eastern part plotted black, 320 stations in western part plotted blue). Colors refer to the orographic height [m].
Fig. 2 shows the 44-year seasonal mean frost day patterns for the transition seasons spring (March–May) and autumn (September– November), based on ERA40. As expected, the maximum numbers of frost days occur over the central Arctic and Greenland. In autumn the numbers of frost days are still moderate over Siberia (~60–80 days) and Alaska/Canadian Archipelago (50–70 days) due to the transition from summer. But then in spring, the large amount of frost days over most parts of the Arctic continues from winter. It is of interest to investigate if the number of frost days changed within the period
216
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 2. Seasonal patterns and trends of frost days in spring and autumn, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
because its changes in spring and autumn could influence the growing season length. Therefore, Fig. 2 further presents the spatial pattern of the trends of frost days in spring and autumn, based on ERA40. It indicates that most of the trends are small (within ± 2 days/decade) and statistically not significant. The largest and significant negative trends (of up to − 4 days/decade) are located in the northern North Atlantic, Kara-, Laptev- and Beaufort Seas and the Canadian Archipelago which are associated with the northward movement of the sea ice boundary due to the warming. Further significant negative trends (−1 to −3 days/decade) are found over the West Siberian Plain and Alaska during spring. This finding of less frequent frost days is in agreement with earlier studies over the northern hemisphere (Frich et al., 2002; Kiktev et al., 2003; Klein Tank and Können, 2003). Significant positive trends, indicating an increase of frost days, have been calculated over two areas: (i) over the Labrador Sea, north of Greenland and over Iceland in spring (up to 3 days/decade) and (ii) over a small area near the mouth of Lena River in autumn (up to 1 day/ decade). Earlier studies also report on negligible or positive trends in cold areas with abundant snow (e.g., Moberg et al., 2005). Over inner Greenland in both seasons, areas with nearly constant time series of frost days are found, the slopes of the least squares lines calculated for those were zero and found significant. The detailed regional analysis of the changes of frost days over the Russian Arctic is presented in Fig. 3 and Table 2. In the figure, a remarkable year-to-year variability of the frost days is obvious in the GSOD observations. This variability is larger in western than in eastern part of Russia; this is seen for both seasons. The inter-annual amplitude (maximum minus minimum of occurred frost days during the 51year-long period) is 24 (8) days in spring, and 34 (11) days in autumn in the western (eastern) part. This east–west difference is because the western part is under the strong influence of the Atlantic storm tracks which brings variable weather conditions, while the eastern part is influenced by a strong continental climate associated with higher numbers of frost days and lower variability. Further, in both regions, the variability is larger in autumn than in spring. This can be partly explained by an analysis of the minimum temperature itself, on which frost days are based. For both eastern and western Russian Arctic,
autumn is warmer than spring. The variability of minimum temperature within one season (i.e. the fluctuation around the corresponding seasonal mean value) is in autumn and spring of similar magnitude, but in colder spring, this variability does not translate into a variability of frost days because the mean minimum temperature is too far below 0 °C. Based on the calculated standard deviations (see the bars in Fig. 3), a clear decadal variation in the inter-annual variability of frost days becomes obvious. Considering the calculated trends (Table 2), it becomes clear that no station-based trend is significant in western Russia. The numbers show positive and negative trends depending on the considered time period. The long-term station trends are positive, indicating an increase of frost days, while the short-term trends (which consider only the recent 3 decades) are negative, indicating a
Fig. 3. Year-to-year variability and trends of frost days in spring and autumn, 1958– 2008, based on station data (GSOD west: light green lines, GSOD east: dark green lines) and ERA40 (light and dark red lines). The left y-axis is for the frost days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details. Upper panel: spring, lower panel: autumn. The corresponding slopes of the trends and their significances are given in Table 2; significant trends are plotted as solid lines.
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224 Table 2 Seasonal trends of frost days [days/decade] in spring and autumn, covering different time periods. The numbers are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.52 1.31 − 0.40 − 3.31
− 0.05 − 2.06 − 4.62
− 0.69⁎ − 0.49 − 0.73 − 1.73⁎
− 1.04⁎ − 1.60⁎ − 2.77⁎
Autumn 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.62 1.55 − 0.43 − 0.54
0.20 − 2.10 − 1.24
− 0.13 0.50 0.43 0.66
− 0.25 − 0.52 0.02
decrease of frost days. On the contrary, some spring trends in eastern part are significant. A decrease in frost days is calculated here. The strongest trend appears in the 22-year-long period 1980–2001 with
217
−1.73 days/decade. For the whole 51 years 1958–2008, the trend of −0.69 days/decade is also statistically significant. In autumn, the trends are positive (increase of frost days), but are non-significant. It is obvious from Fig. 3 that this positive trend has been broken down in recent years due to the occurrence of relative low frost days since 2003. The comparison of the ERA40 and station data shows that the number of frost days is generally underestimated by the reanalysis in both seasons and for both regions, particularly in the eastern region. However, the general shape of observed variability is well reproduced by ERA40. The trend calculations show that the ERA40-based trend magnitudes do not agree well with those from observations. However, the sign of the trend is correctly reproduced for most of the cases. Generally, ERA40 shows stronger decrease of frost days than observed. 3.2. Cold spells The occurrence of cold spells is of importance in all seasons, and thus Fig. 4 presents the four seasonal mean patterns, based on ERA40. The figure shows that, in the mean, cold spells occur all over the year with a similar frequency except of a clear minimum in summer, although their
Fig. 4. Seasonal patterns and trends of cold spell days, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
218
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
occurrence is regionally non-uniformly distributed in all seasons. Generally, the spatial distribution of cold spells in autumn is similar to that of spring, but less distinctive. In winter and both transition seasons, a maximum occurrence of cold spells (6–9 days) is found over the Baffin Bay and around Novaya Zemlya. Another prominent area of many cold spells is found over central and eastern Siberia, particularly expressed in spring (7–9 days). Partly, these areas are spatially correlated with areas characterized by maximum cold nights (Baffin Bay in winter and spring, central Siberia in spring, not shown), because many cold nights may lead to many cold spell days, as cold spell days are per definition consecutive cold nights. An analysis of the circulation patterns in years with many cold spell days over the Baffin Bay shows a more prominent high over the Arctic Ocean and a more pronounced Icelandic Low than average (not shown), leading to cold north winds over Baffin Island and the Baffin Bay causing the increase in cold nights. In years with many cold spell days over central Siberia in spring, the Arctic High is strong and largely extended towards the pole and the Russian coast, leading to stable, cold conditions over the area. Over northern Europe and western Russia relatively few cold spells occur. Here cold nights are often separated by warmer days due to the frequent transport of relatively mild air from the ocean. The lowest numbers of cold spell days occur in summer. Here, over eastern Siberia, Alaska and parts of northwestern Canada a strong anti-correlation with cold nights exists, i.e. those areas are characterized by maximum cold nights (up to 15 days), but minimum cold spells (0–3 days), indicating that the cold nights appear separately. The trends of cold spell days are generally small and in the order of ± 2 day/decade (Fig. 4). Significant trends in cold spells are partly calculated in regions where frost days show no trend at all (e.g., in spring over European Russia and northern Canada; in autumn over parts of eastern Siberia), indicating a change in temperature that is too far away from the absolute threshold of 0 ºC to be seen in frost days. In winter and transition seasons, significant negative trends in cold spells are mainly found over the Greenland Sea, Canadian Archipelago/Alaska and around Novaya Zemlya, which are the same regions characterized by a decrease of frost days. The strongest decrease of cold spells occurs over the Fram Strait in winter and spring (up to −4 days/decade). Significant positive trends (1–2 days/decade) are mainly found over parts of the Labrador Sea in all seasons and over the Canadian Basin in summer. In those places where the maximum number of cold spells appear (Baffin Bay, East Siberian mountains; see above), their occurrence has not changed over the last 4 decades. To analyze in how many events the cold spell days occurred, the mean number of cold spell day events has been calculated, excluding years with no cold spell days (not shown). It shows that few cold spell days occur in many events over the Atlantic storm track area, while many cold spell days appear in very few events (practically only one) over Canada, Alaska and Siberia. The regional analysis of cold spell days for the Russian Arctic (Table 3, Fig. 5) based on the GSOD station observations shows that almost all calculated trends are statistically not significant. The only exception is found in western part in summer; here significant negative trends of −2.18 days/decade for 1980–2001 and −1.19 days/decade for 1969– 2001 are calculated. It is also obvious that strong (as it is for summertime) cold spell events of 5–10 days, which occurred regularly in the 60 s–80 s, do not continue to appear after the mid 80s. A similar decreased variation of the inter-annual variability is seen in spring and in winter, whereas in autumn such a decrease of variability cannot be confirmed. In eastern part, the many missing values in the time series prevent the calculation of almost all trends. For the longer records, the cold spells appear to decrease in all seasons. The calculated variability (see the bars in Fig. 5) shows, contrary to the frost days, that the differences between the eastern and western stations are not so large. This can be explained by the daily minimum temperature itself (not shown). While its variability is for east and west of similar magnitude, the magnitudes are different in the transition seasons. Here, the min-
imum temperatures can exceed 0 °C (the threshold used for the frost days) in the western Russian Arctic, while that does never happen for the eastern Russian Arctic. Because the cold spell days-definition uses a relative threshold which depends on the temperature distribution of each station, the variability in temperature is translated into the variability of cold spell days equally for east and west. To gain information on the frequency of cold spell events, time series, trend and variability have also been calculated for the number of cold spell events (not shown). Obtained results are quite similar to what is presented for cold spell days. This similarity indicates that cold spell days often occur as multiple separate events. Fig. 5 further presents the comparison of ERA40 and station observations. Generally, the ERA40 data well reproduce both magnitude and variability seen in the station data. 3.3. Warm spells The occurrence of warm spells is of importance in all seasons, but based on ERA40 it is obvious that the maximum numbers occur in spring and autumn (Fig. 6). In spring, the highest numbers of warm spell days are located over the Arctic Ocean (6–9 days), indicating that the occurring warm day-times (for definition see Table 1) mostly arise in a row (the number of warm day-times is only 8–11 days; an average of 1 to 1.2 events is calculated; not shown). The absolute numbers of warm day-times peak in spring along the Atlantic storm track and reach a maximum of 10–17 days (not shown). But there, only few warm spells (4–6 days) occur because the often passing and changing storms bring these warm day-times not on at least 6 consecutive days. This is confirmed by the analysis of the average number of warm spell day events (years with no warm spells not taken into account), where a clear maximum along the Atlantic storm track is found (not shown). An analysis of the circulation patterns associated with many warm spell days over the Arctic Ocean in spring shows that the Arctic high is shifted towards Canada and Alaska, leaving most of the Arctic Ocean under the influence of pronounced low pressure and North Atlantic storms transport warm air into the Arctic. In winter and summer, less distinct regional differences are obvious and the mean number of warm spells is low (2–7 days). This happens although the number of warm day-times is largest (10–20 days) in both seasons (not shown), indicating that warm day-times often occur separated by colder days. In autumn, the large occurrence of warm spells (6–8 days) over the northern North Atlantic correlates with the occurrence of many warm day-times (14–17 days) in this region. Fig. 6 includes the trends over the 44 years. In winter, positive and negative trends in warm spell days occur; they are small, very local and mostly not significant over the whole Arctic. The small area east of Greenland which shows a significant positive trend corresponds to a decrease in cold spell days (compare to Fig. 4). In spring, pronounced significant positive trends occur over the Barents- and Kara Seas, Taimyr Peninsula and the West Siberian Plain with up to 3 days/ decade. Only over a small part of this area (west of Novaya Zemlya), an anti-correlation with the trends in cold spell days is found. Also, these positive trends only cover a small part of the area showing maximum occurrences of warm spell days. In summer, almost all significant trends show an increase in warm spell days; the strongest trends are located over sea, in particular south and west of Greenland with up to 6 days/decade and over the central Arctic Ocean with up to 4 days/ decade. In autumn, significant positive trends are found over northern Greenland and parts of its surrounding seas, around Svalbard and over the Chukchi- and East Siberian Seas. Significant negative trends are calculated over parts of central and eastern Siberia. In all seasons except winter, a connection between trends in warm day-times and warm spell days is found, i.e. both indices show the same trend magnitudes and similar locations of significance, indicating that the increase in warm day-times led to the increase in warm spell days. In winter, the trends in warm spells are smaller and less significant than
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
219
Table 3 Seasonal trends of cold spell days [days/decade] and number of cold spell events [events/decade], covering different time periods. The left column always refers to cold spell days, the right column to the number of events. The slopes are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Winter 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.96 − 0.47
− 0.12 − 0.04
− 1.35 − 0.64
− 0.15 − 0.07
− 0.64 0.98
− 0.05 0.13
− 0.32 1.34
0.05 0.14
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.58 − 0.80 − 0.67 − 1.88
− 0.06 − 0.09 − 0.08 − 0.19
− 0.86 − 0.83 − 2.19
− 0.09 − 0.10 − 0.22
− 0.17 0.40
− 0.04 0.04
− 0.07 0.68
− 0.05 0.05
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.35 − 0.47 − 1.19 − 2.18⁎
− 0.04 − 0.06 − 0.14⁎ − 0.27⁎
− 0.57 − 1.50⁎ − 2.22⁎
− 0.07 − 0.18⁎ − 0.29⁎
0.24 0.34
0.02 0.04
0.13 − 0.06
0.00 0.00
Autumn 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
1.20 1.69
0.11 0.14
1.10 1.72
0.10 0.15
− 0.88 − 1.28
− 0.09 − 0.11
− 0.58 − 1.08
− 0.06 − 0.08
Fig. 5. Year-to-year variability and trends of cold spell days, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 3; significant trends are plotted as solid lines. The left y-axis is for the cold spell days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details.
220
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 6. Seasonal patterns and trends of warm spell days, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
in warm day-times, indicating that the increased number of warm day-times occurs separated from each other. Like for cold spells, the regional analysis of warm spells for the Russian Arctic (Table 4, Fig. 7) based on the GSOD stations shows that the calculated trends are statistically not significant. But contrary to cold spells which tend to decrease, the warm spells tend to increase. In the western part, the warm spells have increased by 0.2 (0.6) days/decade in spring (summer), but have remained unchanged in autumn, considering the full 51-year-long record. In the eastern part, the many missing values in the time series prevent the calculation of most trends. But from the few trends calculated, the trend magnitudes in east appear very similar to those from west. A strong increase in the variability of warm spell days is found (about a doubling over the whole period) in the winter time series for the western and eastern stations, a behavior that is not or only less pronounced found in the other seasons. This is partly in line with the results derived from cold spell days, where the analysis of the winter season shows a decrease in variability for Western Russia. But the comparison with other seasons and Eastern Russia also demonstrates that this correlation is not mandatory.
Fig. 7 additionally presents the comparison of ERA40 and station data. Generally, both the magnitude and variability of the warm spell days calculated based on ERA40 agree well with the station observations. 3.4. Growing degree days Growing degree days describe the intensity of the growing season. Fig. 8 shows their regional distribution over the Arctic, calculated from ERA40 data for summer (as the summer is in the Arctic the only relevant season for it). The figure reflects the north– south temperature gradient and distinguishes the high mountain ranges in Alaska and eastern Siberia. The areas with high mountains can easily be identified by up to 300 °C lower values compared with regions of same latitude (see e.g. the southern Chugach Mountains, Alaska). The ERA40 trend analysis shows that all significant trends are positive and in the range of 5 °C/decade to 30 °C/decade. They are located over western Greenland, the mouth of the river Ob, the Mackenzie Mountains
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
221
Table 4 Seasonal trends of warm spell days [days/decade] and number of cold spell events [events/decade], covering different time periods. The left column always refers to warm spell days, the right column to the number of events. The slopes are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Winter 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.71 1.31
0.08 0.17
0.77 1.22
0.06 0.16
0.51 − 0.10
0.07 − 0.03
0.83 − 0.18
0.09 − 0.08
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.25 0.19 0.76 0.66
0.01 0.00 0.06 0.08
0.23 0.83 0.75
0.00 0.06 0.07
0.73 1.02
0.07 0.08
1.08 1.45
0.08 0.09
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.64 0.18 0.41 1.15
0.06 0.00 0.02 0.10
0.27 0.33 1.02
0.02 0.02 0.07
− 0.02 0.06
0.00 0.02
0.45 0.31
0.07 0.11
Autumn 51 yrs (1958–2008) 44 yrs (1958 –2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.12 0.22 0.00 1.72
0.00 − 0.01 − 0.06 0.10
0.30 0.08 1.85
− 0.01 − 0.05 0.11
0.21 − 0.37
0.02 − 0.05
0.22 − 0.43
0.02 − 0.03
and the Brooks Range, the southern Alaska Range, over Iceland and parts of central Siberia. Most of those areas are high mountain ranges and/or are characterized by lower growing degree days, except in East Siberia.
The analysis of the growing degree days for the Russian Arctic based on GSOD station data shows positive trends for both the eastern and the western part (Fig. 9, Table 5). The trend magnitudes increase
Fig. 7. Year-to-year variability and trends of warm spell days, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 4; significant trends are plotted as solid lines. The left y-axis is for the warm spell days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details.
222
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224 Table 5 Seasonal trends of growing degree days [° C/decade] in summer, covering different time periods. The numbers are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
GSOD east
GSOD
ERA40
GSOD
ERA40
37.36⁎ 19.15 61.92⁎ 104.95⁎
19.06 63.30⁎ 108.52⁎
84.65⁎ 97.19⁎
74.40⁎ 100.86⁎
Again, the interpolated ERA40 magnitude and variability fit well with the station data.
3.5. Growing season length
Fig. 8. Seasonal patterns and trends of growing degree days for summer, based on ERA40 data, 1958–2001. The colors show the seasonal mean [° C] and the black isolines show the seasonal trend [° C/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 °C/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
with decreasing length of the analyzed period, indicating that the warming increased over the last two decades. Similar results were found by Førland et al. (2004) who analyzed station records over western Greenland, the North Sea and Scandinavia. Although they use a slightly different definition of growing degree days, they find that most analyzed stations show an increase of growing degree days over l976–2000, but a clearly larger increase is found for the period 1990– 2002. The recent data after 2001 for western Russia does not confirm the further increase in growing degree days (Fig. 9). The figure further displays that the inter-annual variability in the western part is larger than in the eastern part. Moreover, a pronounced decadal variability in the growing degree days is seen in west. In addition, an increase in the inter-annual variability by 30 °C is evident for the eastern stations if one compares the variability magnitude of 100 °C within the period 1995–2002 with that before 1995 (70 °C).
Fig. 10 shows the spatial distribution of the growing season length. The geographical pattern agrees with that of the growing degree days. Along the Arctic coast, the growing season is less than 10 days, but in the inland the higher summer temperatures allow a longer season. Highest values (up to 180 days) are found over western Russia, lowest values (20–40 days) occur over the Taimyr Peninsula and parts of the Ellesmere and Banks Islands. The presented ERA40 data agree well with satellite observations which determine the days of thaw and freeze and calculate the growing season length as the number of days in between (Smith et al., 2004; McDonald et al., 2004). Therefore the determined growing season length is systematically longer than our calculations, but the general spatial distributions are very similar. The conclusions from the ERA40 trends calculated for the growing season length agree with those for the growing degree days, indicating a slight cooling over the east Siberian mountains and parts of the Canadian Archipelago, and a warming mostly everywhere else. The calculated trends in the growing season length are significantly positive (up to 6 days/decade) in the southern West Siberian Plain, western European Russia, northern and southern Alaska, and parts of northwestern Canada. Significant negative trends occur sparsely over parts of eastern Siberia. The calculated trend patterns can be found similarly in Smith et al. (2004), though there they are based on a much shorter time period (1988–2002). The only region with an obvious difference in the trend of the growing season length is the Taimyr Peninsula, where Smith et al. (2004) calculated a decrease while here an increase is found. This is more consistent with Euskirchen et al. (2006), who calculate an increase from their terrestrial ecosystem model. Differences in the trend slopes are due to our usage of periods of warm/cold days for calculation. Those change more slowly than days of freeze/thaw as used by Smith et al. (2004) and therefore may cause our smaller trend slopes for positive as well as negative trends. The results calculated from ERA40 data are also in accordance with Groisman et al. (2003), who determine the growing season length from a station data set using 10 °C as a threshold and find
Fig. 9. Year-to-year variability and trends of growing degree days for summer, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 5; significant trends are plotted as solid lines. The left y-axis is for the growing degree days [° C] (dots), and the right y-axis is for the inter-annual variability [° C] (bars). See text (Section 2.2) for details.
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 10. Seasonal patterns and trends of growing season length, based on ERA40 data, 1958–2001. The colors show the seasonal mean [° C] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines, the line of the zero trend is shown white. The isoline spacing is 2 days/ decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
223
cold spell days shows an overall increase in lasting warm periods and decrease in lasting cold periods. This is associated with a shift in both daily maximum and minimum temperatures towards warmer temperatures as well as atmospheric circulation changes leading to more (less) stable conditions for warm (cold) spells. Additionally, a decrease in the variability of cold spells and a momentary increase in the variability of warm spells is detected, which is particularly pronounced in winter, and is expressed by the occurrence of fewer and shorter cold spell days and coevally longer warm spell events. The comparison between station data and interpolated ERA40 data shows good agreement for all calculated indices, though the model performs better on indices using relative thresholds (cold and warm spell days) than on indices using absolute thresholds (frost days). This is due to a better reproduction of the temperature variability in comparison to absolute temperature values. The growing degree days show positive trends over most parts of the Arctic and are significant in some areas like northern Alaska and northeastern Canada, based on ERA40. The regional analysis of Russia station data even shows an increase of trend magnitudes with time, though some calculated trends are not significant. Similar results are obtained from the analysis of the growing season length, which increases over most of the Arctic except eastern Siberia and parts of the Canadian Archipelago. Finally, we conclude that the warming in annual and seasonal mean temperature found e.g. by Serreze and Francis (2006) can also be partly confirmed in temperature-related climate indices. Additionally, it is indicated that these indices are characterized by a large interannual variability in the Arctic which in turn shows a pronounced decadal variability. Thus, we see high changeability even in our relatively short period of analysis, demonstrating that a combination of high natural variability in the Arctic and an increased radiative forcing due to higher greenhouse gas concentrations is an important factor when analyzing trends. Acknowledgements
the most prominent positive trends over Alaska (ca. 3 days/decade) and western Russia (ca. 2 days/decade). 4. Summary and conclusions In this study, an analysis of climate indices based on daily temperature data for 1958–2008 is presented for the pan Arctic domain with a special focus on the Russian Arctic. Data from the ERA40 reanalysis as well as station data have been used to gain information spatially and on a more regional scale. The focus is on the changes in temperaturerelated indices expressing temperature extremes, i.e. frost days, warm and cold spell days, growing degree days and growing season length. The spatial analysis of frost days clearly signs to a warming along the sea ice boundaries in both transition seasons. Also, few areas over land with increasing frost days were found. The high inter-annual variability in frost days is emphasized; it causes in fact only few significant trends. Distinct regional differences in the variability as well as in the amount of frost days are demonstrated in the comparison of eastern and western Russian stations and are attributed to the different regional climate conditions. The calculation of cold and warm spells shows that their spatial distribution is strongly dependent on the predominant atmospheric circulation patterns in the different years. A significant decrease of cold spells is calculated over marine areas (up to 6 days/decade), associated with the shifted sea ice boundary, in winter and transition seasons. A negative trend in cold spells is also found for the Russian stations, though it is not statistically significant. Examination of trends in warm spell days based on ERA40 data indicates a positive and significant trend in summer and the transition seasons. Regional station analysis provides positive, but non-significant trends. The investigation of both warm and
This research has been funded by the European Union project CARBO—North. We are grateful to I. Hebestadt for programing support. We thank two anonymous reviewers which helped to improve the manuscript. References ACIA, 2005. Arctic Climate Impact Assessment. Cambridge University Press. 1042p. Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank, A., Haylock, M., Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A., Ambenje, P., Kumar, K.Rupa, Revadekar, J., Griffiths, G., Vincent, L., Stephenson, David B., Burn, J., Aguliar, E., Brunet, M., Taylor, M., New, M., Zhai, P., Rusticucci, M., Vazquez-Aguirre, J.L., 2006. Global observed changes in daily climate extremes of temperature and precipitation. J. Geophys. Res. 111, D05109. Euskirchen, E.S., McGuire, A.D., Kicklighter, D.W., Zhuang, Q., Clein, J.S., Dargaville, R., Dye, D.G., Kimball, J.S., McDonald, K.C., Melillo, J.M., Romanovsky, V.E., Smith, N.V., 2006. Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Glob. Chang. Biol. 12, 731–750. Frich, P., Alexander, L.V., Della-Marta, P., Gleason, D., Haylock, M., Klein Tank, A.M.G., Peterson, T., 2002. Observed coherent changes in climatic extremes during the second half of the twentieth century. Clim. Res. 19, 193–212. Førland, E.J., Skaugen, T.E., Benestad, R.E., Hanssen-Bauer, I., Tveito, O.E., 2004. Variations in thermal growing, heating and freezing indices in the Nordic Arctic, 1900–2050. Arct. antarct. alp. res. 36 (3), 347–356. Groisman, P.Ya., Sun, B., Vose, R.S., Lawrimore, J.H., Whitfield, P.H., Førland, E.J., Hanssen-Bauer, I., Serreze, M.C., Razuvaev, V.N., Alekseev, G.V., 2003. Contemporary Climate Changes in High Latitudes of the Northern Hemisphere: Daily Time Resolution. International Symposium on Climate Change, Beijing, China, 31 March– 3 April, 2003, 1172. World Meteorological Organisation Publication, pp. 51–55. IPCC, 2007. IPCC, Climate Change 2007: The Physical Scientific Basis. In: Solomon, S., Qin, D., Manning, M. (Eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge. United Kingdom and New York. Kiktev, D., Sexton, D.M.H., Alexander, L., Folland, C.K., 2003. Comparison of modelled and observed trends in indices of daily climate extremes. J. climate 16, 3560–3571.
224
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Klein Tank, A.M.G., Können, G.P., 2003. Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99. J. Climate 16, 3665–3680. McDonald, K.C., Kimball, J.S., Njoku, E., Zimmermann, R., Zhao, M., 2004. Variability in springtime thaw in the terrestrial high latitudes: monitoring a major control on the biospheric assimilation of atmospheric CO2 with spaceborne microwave remote sensing. Earth interact. 8 (20), 1–23. Moberg, A., Sonechkin, D.M., Holmgren, K., Datsenko, N.M., Karlèn, W., 2005. Highly variable northern hemisphere temperatures from low- and high-resolution proxy data. Nature 433, 613–617. Peterson, T.C., Folland, C., Gruza, G., Hogg, W., Mokssit, A., Plummer, N., 2001. Report on the Activities of the Working Group on Climate Change Detection and Related Rapporteurs 1998–2001. WMO, Rep. WCDMP-47, WMO-TD 1071, Geneve. Rinke, A., Dethloff, K., 2008. Simulated circum-Arctic climate changes by the end of the 21st century. Glob. Planet. Change 62, 173–186.
Serreze, M.C., Francis, J.A., 2006. The arctic amplification debate. Clim. change 76, 241–264. Smith, N.V., Saatchi, S.S., Randerson, J.T., 2004. Trends in high northern latitude soil freeze and thaw cycles from 1988 to 2002. J. Geophys. Res. 109, D12101. Uppala, S.M., Kålberg, P.W., Simmons, A.J., Andrae, U., Bechtold, V., Da Costa, Fiorino, M., Gibson, J.K., Haseler, J., Hernandez, A., Kelly, G.A., Li, X., Onogi, K., Saarinen, S., Sokka, N., Allan, R.P., Andersson, E., Arpe, K., Balmaseda, M.A., Beljaars, A.C.M., Berg, Van De, L., Bidlot, J., Bormann, N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, M., Fuentes, M., Hagemann, S., Hólm, E., Hoskins, B.J., Isaksen, L., Janssen, P.A.E.M., Jenne, R., McNally, A.P., Mahfouf, J.-F., Morcrette, J.-J., Rayner, N.A., Saunders, R.W., Simon, P., Sterl, A., Trenberth, K.E., Untch, A., Vasiljevic, D., Viterbo, P., Woollen, J., 2005. The ERA-40 re-analysis. Quart. J. Roy. Meteor. Soc. 130, 2961–3012.
Contents lists available at ScienceDirect
Global and Planetary Change j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g l o p l a c h a
Variability of observed temperature-derived climate indices in the Arctic Heidrun Matthes ⁎, Annette Rinke, Klaus Dethloff Alfred Wegener Institute for Polar and Marine Research, Potsdam, Germany
a r t i c l e
i n f o
Article history: Received 18 June 2009 Accepted 10 October 2009 Available online 21 October 2009 Keywords: Arctic temperature climate extremes
a b s t r a c t Arctic temperature is analyzed in view of its extremes based on climate indices derived from daily mean, maximum, and minimum temperature. This analysis is done for the pan Arctic domain and region-specific for the eastern and western Russian Arctic. The variability of temperature-related indices over the last four decades is presented, in which the spatial distribution and regional differences as well as its temporal trends are discussed. The analysis is based on ERA40 data and station data in the Russian Arctic. Station-based results for 1958–2008 show a significant decrease of frost days (− 0.8 days/decade) over the eastern Russian Arctic in spring. The trends in warm (cold) spell days are not statistically significant; except in western Russian Arctic in summer for cold spell days (− 2 days/decade). The inter-annual variability of the indices is large and shows pronounced decadal variations. ERA40 data generally reproduce well the inter-annual variability in the climate indices as seen in observations, but they show some deficiencies in the magnitudes and trends of the indices. © 2009 Elsevier B.V. All rights reserved.
1. Introduction The Arctic is under rapid transition and observations show a consistent picture of surface warming and reduction in all components of the cryosphere (IPCC, 2007). Studies on trends of seasonal mean temperature indicate warming trends over most of the Arctic, especially pronounced in winter and spring, but depending on the period analyzed (Serreze and Francis, 2006). The ACIA (2005) report presents the Arctic warming for annual trends as well. It is also suggested that the expected future warming may be associated with a greater frequency of extremely warm temperatures (recently, e.g., Rinke and Dethloff, 2008). The understanding of the climate variability of extremes on a regional level is of increasing importance as it directly impacts the residents and is needed for investigations of ecological and societal changes. In recent years, specific months (e.g., record warm March 1996 in Northern Alaska, record warm spring and summer 2007 in the Arctic) were characterized by record temperatures, and attracted great attention in the public. However, it is apparent that the patterns and variability of extreme events have received little attention in the Arctic assessments. Thus, the question has arisen whether significant changes of extreme temperature are already present in the Arctic. Global analyses of temperature and precipitation extremes have been conducted by various authors using land-based gridded daily data sets. For example, Kiktev et al. (2003) show trends in the annual number of frost days for the northern
⁎ Corresponding author. Alfred Wegener Institute for Polar and Marine Research Telegrafenberg A43, D-14473 Potsdam, Germany. Tel.: +49 331 2882101; fax: +49 331 2882178. E-mail address: [email protected] (H. Matthes). 0921-8181/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.gloplacha.2009.10.004
hemisphere, indicating a significant decrease in frost days of up to 4 days/decade for almost all mid-latitudes. For high latitudes, trends are mainly insignificant or could not be calculated due to a lack of data. The authors calculate a warming over Scandinavia and Alaska associated with a decreasing trend in frost days of up to 2 days/ decade; a cooling is calculated over Eurasia south of Novaya Zemlya. Alexander et al. (2006) provide more complete information for the Arctic. They calculate that the number of frost days and the cold spell days decreased over most parts of the Arctic. However, most trends over the high latitudes are not significant. Our analysis aims at providing more regional results for temperature-based climate indices over the Arctic and is based on the European Centre for Medium-Range Weather Forecasts reanalysis (ERA40) and on Russian station data, covering 1958–2008. The detailed regional analysis for the Russian Arctic has its background in the EU project CARBO—North (http://www.carbonorth.net/) which aims at quantifying the carbon budget in Northern Russia. For this purpose, past, recent and future climate changes are investigated in the Russian Arctic. The paper continues in Section 2 with a description of the applied data sets and the performed data analysis. In Section 3, the spatial and temporal variability of selected temperature-related climate indices and their trends are presented. The results are summarized in Section 4. 2. Data and analysis 2.1. Daily temperature data sets The data set for the pan Arctic analysis is the ERA40 reanalysis data set which covers the 44-year-long period from 1958 to 2001
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
(Uppala et al., 2005). For the comparison with the station data, the ERA40 data have been interpolated on the station locations. The indices have then been calculated from the interpolated time series after removing those days which had missing values in the station data. The station data set used for the analysis over the Russian Arctic is called “Global Summary of the Day” (GSOD; www.ncdc.noaa.gov/oa/ mpp/freedata.html). From this, a total of 644 stations (all located in the Russian Arctic and north of 60 N) have been used, covering 1958– 2008. Compared to ERA40, here the last recent 7 years 2002–2008 are included. It is the case that almost no station has full data coverage over the whole 51-year-long period. The data set has been split for the analysis into two separate parts, GSOD east (324 stations) and GSOD west (320 stations), indicating that they include all stations located east and west of Ural mountains, respectively (Fig. 1). Therewith, the analysis has been done regionally for the two different climate regions of east and west part of Russian Arctic. From both data sets, daily minimum, maximum and mean temperatures have been analyzed.
2.2. Climate indices and trend analysis The classification of extreme temperature is based on climate indices (Peterson et al., 2001) and shown in Table 1. Here in the paper, selected temperature-related indices are discussed which are important for the living conditions in the Arctic. In addition to temperature indices (warm and cold spell days), indices illustrating vegetation conditions (growing season length, growing degree days) and frost conditions (frost days) are studied. A basis for calculating the indices has been defined to take the occurrence of missing data values into account. The data of a station were included in the analysis of a season if the station had 10 or less missing values in the input data within that season, except for indices that count consecutive days (like cold spells). For those indices, only seasons with full data coverage were included. The presented times series show a station mean for a specific season, if at least 10% of the
215
Table 1 Classification of extreme temperature based on climate indices. TX90 is the 90th percentile calculated from a five-day gliding mean of the daily maximum temperature of the reference period 1961–1990 and TN10 is the 10th percentile calculated from a five-day gliding mean of the daily minimum temperature of the reference period. t_mean is used as denotation for the daily mean temperature, t_min and t_max for daily minimum and maximum temperature respectively. Frost days Cold spell days
FD CSDI
Cold nights Warm spell days
TN10p WSDI
Warm day-times Growing degree days
TX90p GD4
Growing season length
GSL
# of days per season with t_min < 0 °C # of at least 6 consecutive days with t_min < TN10 # of days per season with t_min < TN10 # of at least 6 consecutive days with t_max > TX90 # of days per season with t_max > TX90 Sum of daily temperature means per season for days with t_mean > 4 °C # of days between the first occurrence of 6 consecutive days with t_mean > 5 °C and the first occurrence after the 1st July of at least six consecutive days with t_mean < 5 °C
stations deliver data for that season. Counters (like frost days) were normalized on the maximum number of days in the season, accounting for the existence of missing values. To show the variability of the inter-annual variability within the calculated time series, the standard deviation development relative to an 11-year gliding mean (and assigned to the center of the 11-year window) is included in the figures. The standard deviation is calculated if at least 6 years within the 11-year window have data. Furthermore, trends in the climate indices have been analyzed. They were calculated using a linear regression with the least squares method. To gain information on the significance of the obtained trends, a bootstrapping approach was used as described in Kiktev et al. (2003), except for the moving block re-sampling. To account for spatial correlation in the gridded ERA40 data, the same re-sampling sequence was used for each grid point (see again Kiktev et al., 2003). All trends marked significant are significant at the 95%-level. Significance was assigned to a “zero trend”, if the sum of squares of deviations from the mean were less than 5% of the mean. As the time series in the station data contain missing values, it is necessary to make sure that a shown trend is evenly based on values from the entire period. Therefore, the whole 51-year period has been divided in four 11-year-long intervals and a 7-year-long interval for the rest (1958–1968, 1969–1979, 1980–1990, 1991–2001, and 2002–2008). And thus, a trend for a certain period is only shown, if at least five years (respectively three years in the last period) of data are contained for each of these intervals within the trend's period. Additionally, the spatial distribution of stations delivering values to a certain index could have changed throughout the analyzed period and thereby influenced the calculated trends. To asses this problem, the spatial distribution of missing values over time was analyzed for each index. The missing values were found to be rather evenly distributed over the whole area taken into account. 3. Results 3.1. Frost days
Fig. 1. Map of station distribution of the GSOD data set. In total, daily data from these 644 stations have been analyzed (324 stations in eastern part plotted black, 320 stations in western part plotted blue). Colors refer to the orographic height [m].
Fig. 2 shows the 44-year seasonal mean frost day patterns for the transition seasons spring (March–May) and autumn (September– November), based on ERA40. As expected, the maximum numbers of frost days occur over the central Arctic and Greenland. In autumn the numbers of frost days are still moderate over Siberia (~60–80 days) and Alaska/Canadian Archipelago (50–70 days) due to the transition from summer. But then in spring, the large amount of frost days over most parts of the Arctic continues from winter. It is of interest to investigate if the number of frost days changed within the period
216
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 2. Seasonal patterns and trends of frost days in spring and autumn, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
because its changes in spring and autumn could influence the growing season length. Therefore, Fig. 2 further presents the spatial pattern of the trends of frost days in spring and autumn, based on ERA40. It indicates that most of the trends are small (within ± 2 days/decade) and statistically not significant. The largest and significant negative trends (of up to − 4 days/decade) are located in the northern North Atlantic, Kara-, Laptev- and Beaufort Seas and the Canadian Archipelago which are associated with the northward movement of the sea ice boundary due to the warming. Further significant negative trends (−1 to −3 days/decade) are found over the West Siberian Plain and Alaska during spring. This finding of less frequent frost days is in agreement with earlier studies over the northern hemisphere (Frich et al., 2002; Kiktev et al., 2003; Klein Tank and Können, 2003). Significant positive trends, indicating an increase of frost days, have been calculated over two areas: (i) over the Labrador Sea, north of Greenland and over Iceland in spring (up to 3 days/decade) and (ii) over a small area near the mouth of Lena River in autumn (up to 1 day/ decade). Earlier studies also report on negligible or positive trends in cold areas with abundant snow (e.g., Moberg et al., 2005). Over inner Greenland in both seasons, areas with nearly constant time series of frost days are found, the slopes of the least squares lines calculated for those were zero and found significant. The detailed regional analysis of the changes of frost days over the Russian Arctic is presented in Fig. 3 and Table 2. In the figure, a remarkable year-to-year variability of the frost days is obvious in the GSOD observations. This variability is larger in western than in eastern part of Russia; this is seen for both seasons. The inter-annual amplitude (maximum minus minimum of occurred frost days during the 51year-long period) is 24 (8) days in spring, and 34 (11) days in autumn in the western (eastern) part. This east–west difference is because the western part is under the strong influence of the Atlantic storm tracks which brings variable weather conditions, while the eastern part is influenced by a strong continental climate associated with higher numbers of frost days and lower variability. Further, in both regions, the variability is larger in autumn than in spring. This can be partly explained by an analysis of the minimum temperature itself, on which frost days are based. For both eastern and western Russian Arctic,
autumn is warmer than spring. The variability of minimum temperature within one season (i.e. the fluctuation around the corresponding seasonal mean value) is in autumn and spring of similar magnitude, but in colder spring, this variability does not translate into a variability of frost days because the mean minimum temperature is too far below 0 °C. Based on the calculated standard deviations (see the bars in Fig. 3), a clear decadal variation in the inter-annual variability of frost days becomes obvious. Considering the calculated trends (Table 2), it becomes clear that no station-based trend is significant in western Russia. The numbers show positive and negative trends depending on the considered time period. The long-term station trends are positive, indicating an increase of frost days, while the short-term trends (which consider only the recent 3 decades) are negative, indicating a
Fig. 3. Year-to-year variability and trends of frost days in spring and autumn, 1958– 2008, based on station data (GSOD west: light green lines, GSOD east: dark green lines) and ERA40 (light and dark red lines). The left y-axis is for the frost days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details. Upper panel: spring, lower panel: autumn. The corresponding slopes of the trends and their significances are given in Table 2; significant trends are plotted as solid lines.
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224 Table 2 Seasonal trends of frost days [days/decade] in spring and autumn, covering different time periods. The numbers are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.52 1.31 − 0.40 − 3.31
− 0.05 − 2.06 − 4.62
− 0.69⁎ − 0.49 − 0.73 − 1.73⁎
− 1.04⁎ − 1.60⁎ − 2.77⁎
Autumn 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.62 1.55 − 0.43 − 0.54
0.20 − 2.10 − 1.24
− 0.13 0.50 0.43 0.66
− 0.25 − 0.52 0.02
decrease of frost days. On the contrary, some spring trends in eastern part are significant. A decrease in frost days is calculated here. The strongest trend appears in the 22-year-long period 1980–2001 with
217
−1.73 days/decade. For the whole 51 years 1958–2008, the trend of −0.69 days/decade is also statistically significant. In autumn, the trends are positive (increase of frost days), but are non-significant. It is obvious from Fig. 3 that this positive trend has been broken down in recent years due to the occurrence of relative low frost days since 2003. The comparison of the ERA40 and station data shows that the number of frost days is generally underestimated by the reanalysis in both seasons and for both regions, particularly in the eastern region. However, the general shape of observed variability is well reproduced by ERA40. The trend calculations show that the ERA40-based trend magnitudes do not agree well with those from observations. However, the sign of the trend is correctly reproduced for most of the cases. Generally, ERA40 shows stronger decrease of frost days than observed. 3.2. Cold spells The occurrence of cold spells is of importance in all seasons, and thus Fig. 4 presents the four seasonal mean patterns, based on ERA40. The figure shows that, in the mean, cold spells occur all over the year with a similar frequency except of a clear minimum in summer, although their
Fig. 4. Seasonal patterns and trends of cold spell days, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
218
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
occurrence is regionally non-uniformly distributed in all seasons. Generally, the spatial distribution of cold spells in autumn is similar to that of spring, but less distinctive. In winter and both transition seasons, a maximum occurrence of cold spells (6–9 days) is found over the Baffin Bay and around Novaya Zemlya. Another prominent area of many cold spells is found over central and eastern Siberia, particularly expressed in spring (7–9 days). Partly, these areas are spatially correlated with areas characterized by maximum cold nights (Baffin Bay in winter and spring, central Siberia in spring, not shown), because many cold nights may lead to many cold spell days, as cold spell days are per definition consecutive cold nights. An analysis of the circulation patterns in years with many cold spell days over the Baffin Bay shows a more prominent high over the Arctic Ocean and a more pronounced Icelandic Low than average (not shown), leading to cold north winds over Baffin Island and the Baffin Bay causing the increase in cold nights. In years with many cold spell days over central Siberia in spring, the Arctic High is strong and largely extended towards the pole and the Russian coast, leading to stable, cold conditions over the area. Over northern Europe and western Russia relatively few cold spells occur. Here cold nights are often separated by warmer days due to the frequent transport of relatively mild air from the ocean. The lowest numbers of cold spell days occur in summer. Here, over eastern Siberia, Alaska and parts of northwestern Canada a strong anti-correlation with cold nights exists, i.e. those areas are characterized by maximum cold nights (up to 15 days), but minimum cold spells (0–3 days), indicating that the cold nights appear separately. The trends of cold spell days are generally small and in the order of ± 2 day/decade (Fig. 4). Significant trends in cold spells are partly calculated in regions where frost days show no trend at all (e.g., in spring over European Russia and northern Canada; in autumn over parts of eastern Siberia), indicating a change in temperature that is too far away from the absolute threshold of 0 ºC to be seen in frost days. In winter and transition seasons, significant negative trends in cold spells are mainly found over the Greenland Sea, Canadian Archipelago/Alaska and around Novaya Zemlya, which are the same regions characterized by a decrease of frost days. The strongest decrease of cold spells occurs over the Fram Strait in winter and spring (up to −4 days/decade). Significant positive trends (1–2 days/decade) are mainly found over parts of the Labrador Sea in all seasons and over the Canadian Basin in summer. In those places where the maximum number of cold spells appear (Baffin Bay, East Siberian mountains; see above), their occurrence has not changed over the last 4 decades. To analyze in how many events the cold spell days occurred, the mean number of cold spell day events has been calculated, excluding years with no cold spell days (not shown). It shows that few cold spell days occur in many events over the Atlantic storm track area, while many cold spell days appear in very few events (practically only one) over Canada, Alaska and Siberia. The regional analysis of cold spell days for the Russian Arctic (Table 3, Fig. 5) based on the GSOD station observations shows that almost all calculated trends are statistically not significant. The only exception is found in western part in summer; here significant negative trends of −2.18 days/decade for 1980–2001 and −1.19 days/decade for 1969– 2001 are calculated. It is also obvious that strong (as it is for summertime) cold spell events of 5–10 days, which occurred regularly in the 60 s–80 s, do not continue to appear after the mid 80s. A similar decreased variation of the inter-annual variability is seen in spring and in winter, whereas in autumn such a decrease of variability cannot be confirmed. In eastern part, the many missing values in the time series prevent the calculation of almost all trends. For the longer records, the cold spells appear to decrease in all seasons. The calculated variability (see the bars in Fig. 5) shows, contrary to the frost days, that the differences between the eastern and western stations are not so large. This can be explained by the daily minimum temperature itself (not shown). While its variability is for east and west of similar magnitude, the magnitudes are different in the transition seasons. Here, the min-
imum temperatures can exceed 0 °C (the threshold used for the frost days) in the western Russian Arctic, while that does never happen for the eastern Russian Arctic. Because the cold spell days-definition uses a relative threshold which depends on the temperature distribution of each station, the variability in temperature is translated into the variability of cold spell days equally for east and west. To gain information on the frequency of cold spell events, time series, trend and variability have also been calculated for the number of cold spell events (not shown). Obtained results are quite similar to what is presented for cold spell days. This similarity indicates that cold spell days often occur as multiple separate events. Fig. 5 further presents the comparison of ERA40 and station observations. Generally, the ERA40 data well reproduce both magnitude and variability seen in the station data. 3.3. Warm spells The occurrence of warm spells is of importance in all seasons, but based on ERA40 it is obvious that the maximum numbers occur in spring and autumn (Fig. 6). In spring, the highest numbers of warm spell days are located over the Arctic Ocean (6–9 days), indicating that the occurring warm day-times (for definition see Table 1) mostly arise in a row (the number of warm day-times is only 8–11 days; an average of 1 to 1.2 events is calculated; not shown). The absolute numbers of warm day-times peak in spring along the Atlantic storm track and reach a maximum of 10–17 days (not shown). But there, only few warm spells (4–6 days) occur because the often passing and changing storms bring these warm day-times not on at least 6 consecutive days. This is confirmed by the analysis of the average number of warm spell day events (years with no warm spells not taken into account), where a clear maximum along the Atlantic storm track is found (not shown). An analysis of the circulation patterns associated with many warm spell days over the Arctic Ocean in spring shows that the Arctic high is shifted towards Canada and Alaska, leaving most of the Arctic Ocean under the influence of pronounced low pressure and North Atlantic storms transport warm air into the Arctic. In winter and summer, less distinct regional differences are obvious and the mean number of warm spells is low (2–7 days). This happens although the number of warm day-times is largest (10–20 days) in both seasons (not shown), indicating that warm day-times often occur separated by colder days. In autumn, the large occurrence of warm spells (6–8 days) over the northern North Atlantic correlates with the occurrence of many warm day-times (14–17 days) in this region. Fig. 6 includes the trends over the 44 years. In winter, positive and negative trends in warm spell days occur; they are small, very local and mostly not significant over the whole Arctic. The small area east of Greenland which shows a significant positive trend corresponds to a decrease in cold spell days (compare to Fig. 4). In spring, pronounced significant positive trends occur over the Barents- and Kara Seas, Taimyr Peninsula and the West Siberian Plain with up to 3 days/ decade. Only over a small part of this area (west of Novaya Zemlya), an anti-correlation with the trends in cold spell days is found. Also, these positive trends only cover a small part of the area showing maximum occurrences of warm spell days. In summer, almost all significant trends show an increase in warm spell days; the strongest trends are located over sea, in particular south and west of Greenland with up to 6 days/decade and over the central Arctic Ocean with up to 4 days/ decade. In autumn, significant positive trends are found over northern Greenland and parts of its surrounding seas, around Svalbard and over the Chukchi- and East Siberian Seas. Significant negative trends are calculated over parts of central and eastern Siberia. In all seasons except winter, a connection between trends in warm day-times and warm spell days is found, i.e. both indices show the same trend magnitudes and similar locations of significance, indicating that the increase in warm day-times led to the increase in warm spell days. In winter, the trends in warm spells are smaller and less significant than
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
219
Table 3 Seasonal trends of cold spell days [days/decade] and number of cold spell events [events/decade], covering different time periods. The left column always refers to cold spell days, the right column to the number of events. The slopes are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Winter 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.96 − 0.47
− 0.12 − 0.04
− 1.35 − 0.64
− 0.15 − 0.07
− 0.64 0.98
− 0.05 0.13
− 0.32 1.34
0.05 0.14
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.58 − 0.80 − 0.67 − 1.88
− 0.06 − 0.09 − 0.08 − 0.19
− 0.86 − 0.83 − 2.19
− 0.09 − 0.10 − 0.22
− 0.17 0.40
− 0.04 0.04
− 0.07 0.68
− 0.05 0.05
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
− 0.35 − 0.47 − 1.19 − 2.18⁎
− 0.04 − 0.06 − 0.14⁎ − 0.27⁎
− 0.57 − 1.50⁎ − 2.22⁎
− 0.07 − 0.18⁎ − 0.29⁎
0.24 0.34
0.02 0.04
0.13 − 0.06
0.00 0.00
Autumn 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
1.20 1.69
0.11 0.14
1.10 1.72
0.10 0.15
− 0.88 − 1.28
− 0.09 − 0.11
− 0.58 − 1.08
− 0.06 − 0.08
Fig. 5. Year-to-year variability and trends of cold spell days, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 3; significant trends are plotted as solid lines. The left y-axis is for the cold spell days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details.
220
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 6. Seasonal patterns and trends of warm spell days, based on ERA40 data, 1958–2001. The colors show the seasonal mean [days] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 day/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
in warm day-times, indicating that the increased number of warm day-times occurs separated from each other. Like for cold spells, the regional analysis of warm spells for the Russian Arctic (Table 4, Fig. 7) based on the GSOD stations shows that the calculated trends are statistically not significant. But contrary to cold spells which tend to decrease, the warm spells tend to increase. In the western part, the warm spells have increased by 0.2 (0.6) days/decade in spring (summer), but have remained unchanged in autumn, considering the full 51-year-long record. In the eastern part, the many missing values in the time series prevent the calculation of most trends. But from the few trends calculated, the trend magnitudes in east appear very similar to those from west. A strong increase in the variability of warm spell days is found (about a doubling over the whole period) in the winter time series for the western and eastern stations, a behavior that is not or only less pronounced found in the other seasons. This is partly in line with the results derived from cold spell days, where the analysis of the winter season shows a decrease in variability for Western Russia. But the comparison with other seasons and Eastern Russia also demonstrates that this correlation is not mandatory.
Fig. 7 additionally presents the comparison of ERA40 and station data. Generally, both the magnitude and variability of the warm spell days calculated based on ERA40 agree well with the station observations. 3.4. Growing degree days Growing degree days describe the intensity of the growing season. Fig. 8 shows their regional distribution over the Arctic, calculated from ERA40 data for summer (as the summer is in the Arctic the only relevant season for it). The figure reflects the north– south temperature gradient and distinguishes the high mountain ranges in Alaska and eastern Siberia. The areas with high mountains can easily be identified by up to 300 °C lower values compared with regions of same latitude (see e.g. the southern Chugach Mountains, Alaska). The ERA40 trend analysis shows that all significant trends are positive and in the range of 5 °C/decade to 30 °C/decade. They are located over western Greenland, the mouth of the river Ob, the Mackenzie Mountains
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
221
Table 4 Seasonal trends of warm spell days [days/decade] and number of cold spell events [events/decade], covering different time periods. The left column always refers to warm spell days, the right column to the number of events. The slopes are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
GSOD east
GSOD
ERA40
GSOD
ERA40
Winter 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.71 1.31
0.08 0.17
0.77 1.22
0.06 0.16
0.51 − 0.10
0.07 − 0.03
0.83 − 0.18
0.09 − 0.08
Spring 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.25 0.19 0.76 0.66
0.01 0.00 0.06 0.08
0.23 0.83 0.75
0.00 0.06 0.07
0.73 1.02
0.07 0.08
1.08 1.45
0.08 0.09
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.64 0.18 0.41 1.15
0.06 0.00 0.02 0.10
0.27 0.33 1.02
0.02 0.02 0.07
− 0.02 0.06
0.00 0.02
0.45 0.31
0.07 0.11
Autumn 51 yrs (1958–2008) 44 yrs (1958 –2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
0.12 0.22 0.00 1.72
0.00 − 0.01 − 0.06 0.10
0.30 0.08 1.85
− 0.01 − 0.05 0.11
0.21 − 0.37
0.02 − 0.05
0.22 − 0.43
0.02 − 0.03
and the Brooks Range, the southern Alaska Range, over Iceland and parts of central Siberia. Most of those areas are high mountain ranges and/or are characterized by lower growing degree days, except in East Siberia.
The analysis of the growing degree days for the Russian Arctic based on GSOD station data shows positive trends for both the eastern and the western part (Fig. 9, Table 5). The trend magnitudes increase
Fig. 7. Year-to-year variability and trends of warm spell days, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 4; significant trends are plotted as solid lines. The left y-axis is for the warm spell days [days] (dots), and the right y-axis is for the inter-annual variability [days] (bars). See text (Section 2.2) for details.
222
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224 Table 5 Seasonal trends of growing degree days [° C/decade] in summer, covering different time periods. The numbers are given for station-based (GSOD) and ERA40 data (see for details in the text of Section 2). Statistically significant trends are marked with an asterisk. GSOD west
Summer 51 yrs (1958–2008) 44 yrs (1958–2001) 33 yrs (1969–2001) 22 yrs (1980–2001)
GSOD east
GSOD
ERA40
GSOD
ERA40
37.36⁎ 19.15 61.92⁎ 104.95⁎
19.06 63.30⁎ 108.52⁎
84.65⁎ 97.19⁎
74.40⁎ 100.86⁎
Again, the interpolated ERA40 magnitude and variability fit well with the station data.
3.5. Growing season length
Fig. 8. Seasonal patterns and trends of growing degree days for summer, based on ERA40 data, 1958–2001. The colors show the seasonal mean [° C] and the black isolines show the seasonal trend [° C/decade]; positive (negative) trends are depicted by solid (dashed) black lines. The isoline spacing is 1 °C/decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
with decreasing length of the analyzed period, indicating that the warming increased over the last two decades. Similar results were found by Førland et al. (2004) who analyzed station records over western Greenland, the North Sea and Scandinavia. Although they use a slightly different definition of growing degree days, they find that most analyzed stations show an increase of growing degree days over l976–2000, but a clearly larger increase is found for the period 1990– 2002. The recent data after 2001 for western Russia does not confirm the further increase in growing degree days (Fig. 9). The figure further displays that the inter-annual variability in the western part is larger than in the eastern part. Moreover, a pronounced decadal variability in the growing degree days is seen in west. In addition, an increase in the inter-annual variability by 30 °C is evident for the eastern stations if one compares the variability magnitude of 100 °C within the period 1995–2002 with that before 1995 (70 °C).
Fig. 10 shows the spatial distribution of the growing season length. The geographical pattern agrees with that of the growing degree days. Along the Arctic coast, the growing season is less than 10 days, but in the inland the higher summer temperatures allow a longer season. Highest values (up to 180 days) are found over western Russia, lowest values (20–40 days) occur over the Taimyr Peninsula and parts of the Ellesmere and Banks Islands. The presented ERA40 data agree well with satellite observations which determine the days of thaw and freeze and calculate the growing season length as the number of days in between (Smith et al., 2004; McDonald et al., 2004). Therefore the determined growing season length is systematically longer than our calculations, but the general spatial distributions are very similar. The conclusions from the ERA40 trends calculated for the growing season length agree with those for the growing degree days, indicating a slight cooling over the east Siberian mountains and parts of the Canadian Archipelago, and a warming mostly everywhere else. The calculated trends in the growing season length are significantly positive (up to 6 days/decade) in the southern West Siberian Plain, western European Russia, northern and southern Alaska, and parts of northwestern Canada. Significant negative trends occur sparsely over parts of eastern Siberia. The calculated trend patterns can be found similarly in Smith et al. (2004), though there they are based on a much shorter time period (1988–2002). The only region with an obvious difference in the trend of the growing season length is the Taimyr Peninsula, where Smith et al. (2004) calculated a decrease while here an increase is found. This is more consistent with Euskirchen et al. (2006), who calculate an increase from their terrestrial ecosystem model. Differences in the trend slopes are due to our usage of periods of warm/cold days for calculation. Those change more slowly than days of freeze/thaw as used by Smith et al. (2004) and therefore may cause our smaller trend slopes for positive as well as negative trends. The results calculated from ERA40 data are also in accordance with Groisman et al. (2003), who determine the growing season length from a station data set using 10 °C as a threshold and find
Fig. 9. Year-to-year variability and trends of growing degree days for summer, 1958–2008, based on station data and ERA40. The left panel shows GSOD west, the right panel GSOD east. The corresponding numbers of the trends and their significance are given in Table 5; significant trends are plotted as solid lines. The left y-axis is for the growing degree days [° C] (dots), and the right y-axis is for the inter-annual variability [° C] (bars). See text (Section 2.2) for details.
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Fig. 10. Seasonal patterns and trends of growing season length, based on ERA40 data, 1958–2001. The colors show the seasonal mean [° C] and the black isolines show the seasonal trend [days/decade]; positive (negative) trends are depicted by solid (dashed) black lines, the line of the zero trend is shown white. The isoline spacing is 2 days/ decade. The white (grey) dotted areas indicate significant (zero) trends. The land–sea mask is illustrated as thick grey line.
223
cold spell days shows an overall increase in lasting warm periods and decrease in lasting cold periods. This is associated with a shift in both daily maximum and minimum temperatures towards warmer temperatures as well as atmospheric circulation changes leading to more (less) stable conditions for warm (cold) spells. Additionally, a decrease in the variability of cold spells and a momentary increase in the variability of warm spells is detected, which is particularly pronounced in winter, and is expressed by the occurrence of fewer and shorter cold spell days and coevally longer warm spell events. The comparison between station data and interpolated ERA40 data shows good agreement for all calculated indices, though the model performs better on indices using relative thresholds (cold and warm spell days) than on indices using absolute thresholds (frost days). This is due to a better reproduction of the temperature variability in comparison to absolute temperature values. The growing degree days show positive trends over most parts of the Arctic and are significant in some areas like northern Alaska and northeastern Canada, based on ERA40. The regional analysis of Russia station data even shows an increase of trend magnitudes with time, though some calculated trends are not significant. Similar results are obtained from the analysis of the growing season length, which increases over most of the Arctic except eastern Siberia and parts of the Canadian Archipelago. Finally, we conclude that the warming in annual and seasonal mean temperature found e.g. by Serreze and Francis (2006) can also be partly confirmed in temperature-related climate indices. Additionally, it is indicated that these indices are characterized by a large interannual variability in the Arctic which in turn shows a pronounced decadal variability. Thus, we see high changeability even in our relatively short period of analysis, demonstrating that a combination of high natural variability in the Arctic and an increased radiative forcing due to higher greenhouse gas concentrations is an important factor when analyzing trends. Acknowledgements
the most prominent positive trends over Alaska (ca. 3 days/decade) and western Russia (ca. 2 days/decade). 4. Summary and conclusions In this study, an analysis of climate indices based on daily temperature data for 1958–2008 is presented for the pan Arctic domain with a special focus on the Russian Arctic. Data from the ERA40 reanalysis as well as station data have been used to gain information spatially and on a more regional scale. The focus is on the changes in temperaturerelated indices expressing temperature extremes, i.e. frost days, warm and cold spell days, growing degree days and growing season length. The spatial analysis of frost days clearly signs to a warming along the sea ice boundaries in both transition seasons. Also, few areas over land with increasing frost days were found. The high inter-annual variability in frost days is emphasized; it causes in fact only few significant trends. Distinct regional differences in the variability as well as in the amount of frost days are demonstrated in the comparison of eastern and western Russian stations and are attributed to the different regional climate conditions. The calculation of cold and warm spells shows that their spatial distribution is strongly dependent on the predominant atmospheric circulation patterns in the different years. A significant decrease of cold spells is calculated over marine areas (up to 6 days/decade), associated with the shifted sea ice boundary, in winter and transition seasons. A negative trend in cold spells is also found for the Russian stations, though it is not statistically significant. Examination of trends in warm spell days based on ERA40 data indicates a positive and significant trend in summer and the transition seasons. Regional station analysis provides positive, but non-significant trends. The investigation of both warm and
This research has been funded by the European Union project CARBO—North. We are grateful to I. Hebestadt for programing support. We thank two anonymous reviewers which helped to improve the manuscript. References ACIA, 2005. Arctic Climate Impact Assessment. Cambridge University Press. 1042p. Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank, A., Haylock, M., Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A., Ambenje, P., Kumar, K.Rupa, Revadekar, J., Griffiths, G., Vincent, L., Stephenson, David B., Burn, J., Aguliar, E., Brunet, M., Taylor, M., New, M., Zhai, P., Rusticucci, M., Vazquez-Aguirre, J.L., 2006. Global observed changes in daily climate extremes of temperature and precipitation. J. Geophys. Res. 111, D05109. Euskirchen, E.S., McGuire, A.D., Kicklighter, D.W., Zhuang, Q., Clein, J.S., Dargaville, R., Dye, D.G., Kimball, J.S., McDonald, K.C., Melillo, J.M., Romanovsky, V.E., Smith, N.V., 2006. Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Glob. Chang. Biol. 12, 731–750. Frich, P., Alexander, L.V., Della-Marta, P., Gleason, D., Haylock, M., Klein Tank, A.M.G., Peterson, T., 2002. Observed coherent changes in climatic extremes during the second half of the twentieth century. Clim. Res. 19, 193–212. Førland, E.J., Skaugen, T.E., Benestad, R.E., Hanssen-Bauer, I., Tveito, O.E., 2004. Variations in thermal growing, heating and freezing indices in the Nordic Arctic, 1900–2050. Arct. antarct. alp. res. 36 (3), 347–356. Groisman, P.Ya., Sun, B., Vose, R.S., Lawrimore, J.H., Whitfield, P.H., Førland, E.J., Hanssen-Bauer, I., Serreze, M.C., Razuvaev, V.N., Alekseev, G.V., 2003. Contemporary Climate Changes in High Latitudes of the Northern Hemisphere: Daily Time Resolution. International Symposium on Climate Change, Beijing, China, 31 March– 3 April, 2003, 1172. World Meteorological Organisation Publication, pp. 51–55. IPCC, 2007. IPCC, Climate Change 2007: The Physical Scientific Basis. In: Solomon, S., Qin, D., Manning, M. (Eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge. United Kingdom and New York. Kiktev, D., Sexton, D.M.H., Alexander, L., Folland, C.K., 2003. Comparison of modelled and observed trends in indices of daily climate extremes. J. climate 16, 3560–3571.
224
H. Matthes et al. / Global and Planetary Change 69 (2009) 214–224
Klein Tank, A.M.G., Können, G.P., 2003. Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99. J. Climate 16, 3665–3680. McDonald, K.C., Kimball, J.S., Njoku, E., Zimmermann, R., Zhao, M., 2004. Variability in springtime thaw in the terrestrial high latitudes: monitoring a major control on the biospheric assimilation of atmospheric CO2 with spaceborne microwave remote sensing. Earth interact. 8 (20), 1–23. Moberg, A., Sonechkin, D.M., Holmgren, K., Datsenko, N.M., Karlèn, W., 2005. Highly variable northern hemisphere temperatures from low- and high-resolution proxy data. Nature 433, 613–617. Peterson, T.C., Folland, C., Gruza, G., Hogg, W., Mokssit, A., Plummer, N., 2001. Report on the Activities of the Working Group on Climate Change Detection and Related Rapporteurs 1998–2001. WMO, Rep. WCDMP-47, WMO-TD 1071, Geneve. Rinke, A., Dethloff, K., 2008. Simulated circum-Arctic climate changes by the end of the 21st century. Glob. Planet. Change 62, 173–186.
Serreze, M.C., Francis, J.A., 2006. The arctic amplification debate. Clim. change 76, 241–264. Smith, N.V., Saatchi, S.S., Randerson, J.T., 2004. Trends in high northern latitude soil freeze and thaw cycles from 1988 to 2002. J. Geophys. Res. 109, D12101. Uppala, S.M., Kålberg, P.W., Simmons, A.J., Andrae, U., Bechtold, V., Da Costa, Fiorino, M., Gibson, J.K., Haseler, J., Hernandez, A., Kelly, G.A., Li, X., Onogi, K., Saarinen, S., Sokka, N., Allan, R.P., Andersson, E., Arpe, K., Balmaseda, M.A., Beljaars, A.C.M., Berg, Van De, L., Bidlot, J., Bormann, N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, M., Fuentes, M., Hagemann, S., Hólm, E., Hoskins, B.J., Isaksen, L., Janssen, P.A.E.M., Jenne, R., McNally, A.P., Mahfouf, J.-F., Morcrette, J.-J., Rayner, N.A., Saunders, R.W., Simon, P., Sterl, A., Trenberth, K.E., Untch, A., Vasiljevic, D., Viterbo, P., Woollen, J., 2005. The ERA-40 re-analysis. Quart. J. Roy. Meteor. Soc. 130, 2961–3012.