Observed changes of temperature extremes in Serbia over the period 1961 − 2010

    Observed changes of temperature extremes in Serbia over the period 1961 2010 Mirjana Ruml, Enike Gregori´c, Mirjam Vujadinovi´c, Slavica Radovanovi´c, Gordana Matovi´c, Ana Vukovi´c, Vesna Poˇcuˇca, Djurdja Stojiˇci´c PII: DOI: Reference:

S0169-8095(16)30254-X doi: 10.1016/j.atmosres.2016.08.013 ATMOS 3773

To appear in:

Atmospheric Research

Received date: Revised date: Accepted date:

26 February 2016 8 August 2016 17 August 2016

Please cite this article as: Ruml, Mirjana, Gregori´c, Enike, Vujadinovi´c, Mirjam, Radovanovi´c, Slavica, Matovi´c, Gordana, Vukovi´c, Ana, Poˇcuˇca, Vesna, Stojiˇci´c, Djurdja, Observed changes of temperature extremes in Serbia over the period 1961 - 2010, Atmospheric Research (2016), doi: 10.1016/j.atmosres.2016.08.013

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Observed changes of temperature extremes in Serbia over the period 1961−2010

PT

Mirjana Rumla,*, Enike Gregorića, Mirjam Vujadinovića, Slavica Radovanovićb, Gordana Matovića, Ana Vukovića, Vesna Počučaa and Djurdja Stojičićc a

RI

Faculty of Agriculture, University of Belgrade, Nemanjina 6, 11080 Belgrade, Serbia Republic Hydrometeorological Service of Serbia, Kneza Višeslava 66, 11000 Belgrade, Serbia c Faculty of Forestry, University of Belgrade, Kneza Višeslava 1, 11000 Belgrade, Serbia

SC

b

NU

ABSTRACT

AC CE P

TE

D

MA

The analysis of spatiotemporal changes of temperature extremes in Serbia, based on 18 ETCCDI indices, was performed using daily minimum and maximum temperature observations from 26 meteorological stations over the period 1961−2010. The observation period was divided into two sub-periods (1961−1980 and 1981−2010) according to the results of the sequential Mann–Kendall test. Temporal trends were evaluated by a leastsquares linear regression method. The average annual minimum temperature displayed a mixed pattern of increasing, decreasing, and no trends over 1961−1980 and a significant increasing trend over 1981−2010 across the whole country, with a regionally averaged rate of 0.48°C per decade. The average annual maximum temperature showed a decreasing trend during 1961−1980 and a significant increasing trend at all stations during 1981−2010, with a regionally averaged rate of 0.56°C per decade. Hot indices exhibited a general cooling tendency until 1980 and a warming tendency afterwards, with the most pronounced trends in the number of summer and tropical days during the first period and in the frequency of warm days and nights in the second. Cold indices displayed a mostly warming tendency over the entire period, with the most remarkable increase in the lowest annual maximum temperature and the number of ice days during the first period and in the frequency of cool nights during the second. At most stations, the diurnal temperature range showed a decrease until 1980 and no change or a slight increase afterwards. The lengthening of the growing season was much more pronounced in the later period. The computed correlation coefficient between the annual temperature indices and large-scale circulation features revealed that the East Atlantic pattern displayed much stronger association with examined indices than the North Atlantic Oscillation and East Atlantic/West Russia pattern. Keywords: temperature extremes, ETCCDI indices, Serbia, climate change, RClimDex *

Corresponding author at: Faculty of Agriculture, University of Belgrade, Nemanjina 6, 11080 Belgrade, Serbia. Tel. : +38163247523; fax number: +381112193659. Email address: [email protected] (M. Ruml).

ACCEPTED MANUSCRIPT

1. Introduction

AC CE P

TE

D

MA

NU

SC

RI

PT

Climate change is manifested not only by changes in average conditions, but also by changes in the occurrence of climate extremes (IPCC, 2013). The connection between the incidence of extremes and global warming may be nonlinear (Coumou and Rahmstorf, 2012) and relatively small changes in the mean temperature could produce substantial changes in the temperature extremes (Mearns et al., 1984; Hansen et al., 1988). The results of the recent study of Fischer and Knutti (2015) suggest that about 75% of the moderate daily heat extremes occurring over land worldwide are attributable to the observed warming, which is primarily human-induced. According to the IPCC’s Fifth Assessment Report, the globally averaged surface air temperature increased by 0.85°C over the 1880– 2012 period (IPCC, 2013). The rate of warming was especially high from 1951 to 2012 (0.12°C decade−1). Changes in the frequency, intensity, duration, and timing of extreme climate events are of particular importance due to the risk they pose to human and natural systems as highlighted in the IPCC Special Report on Extreme Events (SREX) of the Intergovernmental Panel on Climate Change (IPCC, 2012). Extreme weather events are among the most destructive natural disasters that may severely impact many sectors of society, such as agriculture, forestry, the energy sector, water resource management, urban planning, human health, tourism, etc. Therefore, besides climatologists, many other researchers and end-users have been increasingly interested in information about the historical and future changes in the frequency, intensity, and duration of extreme weather and climate events, as well as about the driving mechanisms underlying these trends. A large number of indices have been defined and applied in studying climate extremes, but the difference in methodologies used limits the scope of direct comparison between studies. The World Meteorological Organization’s Expert Team on Climate Change Detection and Indices (ETCCDI) recommended a suite of 27 core indices, which provides a common framework for assessing the frequency, duration, or severity of extreme temperature and precipitation worldwide (Zhang et al., 2011). User-friendly software was developed for standardised calculation of these indices and made available for public use. The suggested extreme indices describe so-called “moderate” or “soft” extreme events, with the annual number of occurrences sufficiently large to allow meaningful trend analysis within the typical length of daily meteorological data of 50 years. The ETCCDI indices have been widely used to investigate changes in temperature and precipitation extremes across the world on different spatial scales, both in historical data and climate projections (e.g. Orlowsky and Seneviratne, 2012; Sillmann et al., 2013; Thibeault and Seth, 2014). The ETCCDI regional workshops held worldwide resulted in many papers describing changes in climate extremes in various parts of the world, such as in Central and South Asia (Klein Tank et al., 2006), the western Indian Ocean (Vincent et al., 2011), the Indo-Pacific region (Caesar et al., 2011), parts of Africa (Aguilar et al., 2009), the Middle East (Zhang et al., 2005), the Arab region (Donat et al., 2013b), South America (Skansi et al., 2013), and the Caribbean region (Stephenson et al., 2014). At a national level, numerous studies have been conducted, such as in Canada (Zhang et al., 2000), Georgia (Keggenhoff et al., 2014), Iran (Rahimzadeh et al., 2009), New Zealand (Salinger and Griffiths, 2001), and Saudi Arabia (Athar, 2014). Observed changes in

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

climate extremes in Europe at the continental level were studied by Klein Tank and Können (2003) and Moberg et al. (2006). National trends were examined for Austria (Nemec et al., 2013), Greece (Kioutsioukis et al., 2010), Italy (Fioravanti et al., 2016), the extraCarpathians regions of Romania (Croitoru and Piticar, 2013), Spain (Brunet et al., 2007), and some other countries or their parts. The analyses of the indices from different parts of the world showed the widespread significant changes in temperature extremes in the recent past, especially in the last few decades. According to the majority of studies, minimum temperature extremes were warming faster than maximum temperature extremes on the global scale (Alexander et al., 2006; Donat et al., 2013a) or on a regional or national scale (e.g. Hundecha and Brádossy, 2005 – western Germany; Jones, 2005 – western United States; Moberg and Jones, 2005 – central and western Europe; El Kenawy et al., 2011 – northeast Spain; Stephenson et al., 2014 – the Caribbean region; Ghasemi, 2015 – Iran). However, Klein Tank and Können (2003) found that the pronounced warming between 1976 and 1999 in Europe was rather associated with an increasing trend in warm extremes than with a decreasing trend in cold extremes. Additionally, in several studies for the Mediterranean region (Kostopoulou and Jones, 2005; Efthymiadis et al., 2011; Burić et al., 2014; Fioravanti et al., 2016), stronger warming trends for hot rather than for cold extremes was detected. Most of the previous studies on observed temperature changes in Serbia were focused either on the mean temperature (Bajat et al., 2015) or on a few extreme temperature indices and a relatively small number of meteorological stations (Unkašević and Tošić, 2009a; Unkašević and Tošić, 2009b; Unkašević and Tošić, 2013; Knežević et al., 2014; Malinovic-Milicevic et al., 2015). Since a comprehensive analysis of changes in temperature extremes for Serbia is still lacking, the objective of this study was to investigate the trend characteristics and the variability of temperature extremes using the complete set of the core ETCCDI temperature indices, calculated for 26 meteorological stations with high-quality daily data over the period 1961–2010. In order to identify driving mechanisms contributing to the observed changes in temperature extremes, the relationship between the indices and large-scale circulation patterns (North Atlantic Oscillation (NAO), East Atlantic Pattern (EA), and East Atlantic/West Russia Pattern (EA/WR)) was examined. 2. Data and methods 2.1. Study area and data Serbia is located on the Balkan Peninsula (Fig. 1), in the southern part of the temperate zone, between latitudes 41°50′ and 46°10′ N. The northern part of the country is situated in the Pannonian plane, while the rest of the territory has a complex topography comprised of hills, low and medium-high mountains, and valleys. The temperature data, consisting of daily observations of maximum (TX) and minimum (TN) air temperature, were provided by the Republic Hydrometeorological Service of Serbia (RHMSS). A total of 26 meteorological stations selected for analysis are fairly uniformly distributed over the entire Serbian territory. Recent data were not available

ACCEPTED MANUSCRIPT

PT

for the Autonomous Province of Kosovo and Metohija, located in southwest of Serbia (Fig. 1), because this area has not been covered by the RHMSS since the late 1990s. Complete records were available for 19 stations, while 7 stations had 2% of values missing at most. The acronym, name, geographical coordinates, altitude, and percentage of missing data are presented for each station in Table 1.

RI

2.2 Temperature indices

AC CE P

TE

D

MA

NU

SC

To ensure straightforward comparison with results of other studies, standardised software packages developed and maintained by ETCCDI were used. A total of 18 ETCDDI indices (Table 2), 16 from a core set and two user-defined, were calculated on an annual basis using the R-based software package RClimDex, which is freely available from the ETCDDI website (http://etccdi.pacificclimate.org/). The indices, which describe different characteristics of warm and hot extremes, such as intensity, frequency, and duration, are defined as absolute or threshold indices (for definition see Table 2). Absolute indices include the hottest and coldest day, the hottest and coldest night, and diurnal temperature range. Threshold or ‘day-count’ indices are based on fixed or percentile-based thresholds. Fixed-threshold indices (summer, tropical, and hot days, frost and ice days, tropical nights, and growing season length) are less suitable for spatial comparisons than percentile-based (warm and cool days, warm and cool nights, warm and cold spell duration indicators), because they sample different parts of the temperature distribution at different sites. On the other hand, fixed-threshold indices are more relevant for impact assessments, as well as absolute indices, which are also often related to observed impacts. Two additional user-defined indices, based on fixed thresholds of 30 and 35°C for TX, were included in the study, because these temperatures are not rare and have biological significance in the studied region. The baseline period of 1971–2000 was used for the estimation of threshold values for the percentile-based indices. The percentile indices were modified from the original ETCCDI definition (Zhang et al., 2011) and expressed as the number of days instead of as a percentage. 2.3 Quality control and homogeneity test Prior to indices calculation, a quality control and homogeneity test of the data were done for each station. Even though the RHMSS performed technical and critical controls of the measurements, quality control was applied as the first step in the analysis procedures of the station data in the RClimDex. Temporal homogeneity of the monthly series of TN and TX was assessed by the RHtestV4 software package, based on the penalised maximal t-test (Wang et al., 2007) and the penalised maximal F-test (Wang, 2008a) set in a recursive algorithm (Wang, 2008b). Software and documentation are available at http://etccdi.pacificclimate.org/. The goal of climatic data homogenisation is to adjust observations, if necessary, so that the temporal changes in the adjusted data are caused solely by variation in climate, not by some non-climatic factors. Non-climatic factors may include changes in station location, environment, instrumentation, or observing methodologies, and may mask or strengthen real trends in data.

ACCEPTED MANUSCRIPT

NU

2.4. Large-scale atmospheric circulation patterns

SC

RI

PT

The RClimDex-based quality control did not reveal physically implausible values in the temperature time series. The results of the homogeneity test are given in Table 1. Altogether, eight step changes were detected, six in the mean monthly TN time series and two in the mean monthly TX time series. All detected steps were marked with ? in the RHtestV4 outputs, meaning that they may or may not be significant according to the applied homogeneity test (Wang and Feng, 2013). Since no potential causes for these shifts were found in the metadata and the steps occurred at the same or close date at least at two stations, we concluded that the detected step changes were due rather to climate variability than to measurement errors or other non-climatic reasons. Hence no adjustment of the data for the identified step changes was undertaken.

AC CE P

TE

D

MA

The variations and trends in circulation patterns may influence regional trends in temperature and precipitation and mask or intensify human-induced climate change (e.g. Abatzoglou and Redmond, 2007). Dominant atmospheric teleconnection patterns in the part of Europe where Serbia is situated are the NAO, EA, and EA/WR patterns. The NAO, with its centres of action located near the Icelandic low and the Azores high, affects the westerly airflow across the North Atlantic (Barnston and Livezey, 1987). The positive NAO phase is connected with above-average temperatures and precipitation across northern Europe and Scandinavia, and often with low temperatures and dry conditions over southern Europe and the Mediterranean Basin (Hurrell, 1995). This connection is most pronounced in winter and early spring, and substantially weaker in summer (Rogers, 1990). The EA pattern is considered to be the second dominant teleconnection pattern across the North Atlantic. The action centres of the EA are placed south-eastward of the NAO nodal lines. The positive phase of the EA pattern is associated with above-average temperatures in Europe throughout the year. The EA/WR pattern (referred to as the Eurasia-2 pattern by Barnston and Livezey (1987)) has different positions of action centres during different seasons. The EA/WR positive phase is usually linked with northerly and north-westerly airflow across the Baltic Sea region and the East European plain, while the opposite airflow is linked with its negative phase. The pattern index values were downloaded from the NOAA Climate Prediction Center website (ftp://ftp.cpc.ncep.noaa.gov/wd52dg/data/indices/tele_index.nh). 2.5 Statistical methodology Temporal trends of indices for each station were determined using the least-squares linear regression method. The statistical significance of the change was calculated using a ttest. In addition, linear trends in temperature extremes for the entire Serbian territory were estimated using the average series for each index. Despite its simplicity, the least-squares linear regression model has proven to be a useful tool for the description of time series behaviour and offers the possibility of comparison with already reported climatic trends from different regions of the world. However, if the underlying assumptions of the linear regression are not satisfied, outcomes from a regression model may be seriously biased and misleading. Since temperature changes over the period 1961−2010 turned out to be non-

ACCEPTED MANUSCRIPT

SC

RI

PT

linear, the linear trends were determined for two sub-periods that did satisfy the assumptions of linear regression. To assess the dividing year, the sequential Mann–Kendall (SQ-MK) test for detecting abrupt changes in data series were utilised. The SQ-MK test is a sequential version of the Mann–Kendall rank statistic (Kendall, 1975) proposed by Sneyers (1990). It enables the detection of the approximate year at which a trend begins and whether a trend is statistically significant within the sample x1,. . ., xn. The number of cases xi > xj (where i = 1, . . ., n and j = 1 ,. . ., i-1) is counted and denoted by ni at each comparison. The test statistic calculated as (1)

MA

NU

is normally distributed with mean E(t) and variance Var(t) given by: (2) (3)

(4)

TE

D

The sequential statistic u(t) is calculated as:

AC CE P

Similarly, the sequential backward statistic u’(t) is computed starting from the end of the series. When the u(t) and u’(t) curves are plotted, the point where they cross each other indicates the approximate year at which the trend begins. If they diverge after the intersection and attain specific threshold values, then there is a statistically significant trend. The SQ-MK test was applied on time series of all indices for each station and for regional averages. A typical significance level of 0.05 was used. The Pearson correlation coefficient was computed for all paired combinations of temperature indices and between the indices and large-scale atmospheric circulation patterns in order to investigate their inter-relationships. 3. Results 3.1 Change points in the time series The time series of annual minimum (TNmean) and maximum (TXmean) temperatures averaged across 26 stations for the period 1961−2010 are shown in Figs. 2a and 3a, respectively. A clear shift towards warming can be seen from the early 1980s onwards, but also it is obvious that the first and principal underlying assumptions of linear regression, which is a linearity of the relationship between dependent and independent variables, is not satisfied. That was confirmed by the residuals plots (Figs. 2b and 3b), where it can be seen that the assumption of randomness and unpredictability of regression

ACCEPTED MANUSCRIPT

NU

SC

RI

PT

residuals were not satisfied. If one regression model wants to be used for the entire data set, a squared term has to be included to model the curvature in the middle of the data series. The polynomial model fits the data well, but interpreting it in a climate change context is not as intuitive as for the linear model. The results of the SQ-MK test (Fig. 4) revealed that an abrupt significant change in both the TNmean and TXmean time series occurred around 1980. The same sub-periods were used in trend analysis for all indices, since: (i) they were derived from TN and TX and most of them were significantly correlated with both TNmean or TXmean, while a few of them were significantly correlated either with TNmean or TXmean (Table 3); (ii) the abrupt changes in the time series of most indices according to the SQ-MK test occurred around 1980 at the majority of stations (not shown due to the large number of indices and stations); (iii) the periods 1961−1980 and 1981−2010 satisfied the assumptions of linear regression for all indices; (iv) having different period for each index and each stations would be very confusing and cause difficulties in drawing conclusions.

MA

3.2 Inter-correlation of extreme temperature indices

AC CE P

TE

D

Knowledge of inter-correlation between indices can be useful in impact assessment studies when not all of the needed indices are known. The correlation matrix presented in Table 3 revealed the presence of a significant correlation among most of the indices. The annual means of TN and TX were significantly correlated with most indices examined, except TNmean with DTR and TXmean with TXn, TNn, and CSDI. All hot extremes were significantly inter-correlated, with all correlation coefficients higher than 0.50. The cold indices also exhibited significant inter-correlation, except TXn, which was not significantly correlated with TX10p and FD0. The diurnal temperature range was not significantly correlated with TNmean and indices based on TN, which indicates that changes in DTR was dictated by changes in TX over the examined observational period. Growing season length showed a significant correlation with all indices, except with TXn. The strongest correlation was displayed with FD0, which indicates a close relationship between frost occurrence and GSL, and supports the idea of defining the growing season as a frost-free period. The strongest correlation exhibited for each index is shown in the last column of Table 3. 3.3 Changes in annual minimum and maximum temperatures In the period 1961−1980, TNmean showed a mixed pattern of upward, downward, and no trends (Fig. 5a), while TXmean showed a consistent pattern of a downward trend (Fig. 5b), which was insignificant at most stations (Table 4). A decreasing trend in TNmean was mostly presented in south-eastern Serbia, while the rest of the country experienced a increasing trend in TNmean. Changes in TNmean ranged from −0.47°C per decade at Vranje to 0.48°C per decade at Pozega, while changes in TXmean were between −0.06°C per decade at Loznica and −0.65°C per decade at Zrenjanin. There was no relationship between change rates of TNmean and TXmean (R = −0.14). For example, Pozega, the station with the largest increase in TNmean, was among the stations with a greater decrease

ACCEPTED MANUSCRIPT

NU

SC

RI

PT

in TXmean, while Vranje, the station with the largest decrease in TNmean, was one of two stations with the smallest decrease in TXmean. In the period 1981−2010, both TNmean and TXmean showed a marked, coherent, and significant increase throughout the country. On average for all stations, TNmean rose slightly less (0.48°C decade−1) than TXmean (0.56°C decade−1) during this period (Table 4), but trends of TNmean and TXmean differ more for individual stations (Figs. 5a and 5b). TXmean rose more than TNmean at 15 stations, while the opposite was observed at 4 stations. Nearly equal changes of TNmean and TXmean were recorded at 7 stations. Changes in TNmean ranged from 0.33°C per decade at Veliko Gradiste to 0.67°C per decade at Negotin, while changes in TXmean were between 0.43°C per decade at Kraljevo and 0.73°C per decade at Veliko Gradiste. As in the earlier period, there was no apparent association between decadal changes in TNmean and TXmean (R = −0.22). Again, the same station (Veliko Gradiste) had the greatest change rate in one temperature variable (TXmean) and the smallest in another (TNmean).

MA

3.4 Trends in hot extremes

AC CE P

TE

D

Changes in hot extremes (Fig. 6) closely matched the trend pattern of TXmean in both periods. In the period 1961−1980, all hot extremes exhibited downward trends at most stations (Table 4). On average for all stations, the downward trend was significant for the number of summer (SU25), tropical (TD30) and hot (HD35) days, warm days (TX90p), the hottest day (TXx), and the hottest night (TNx). An insignificant downward trend was observed for warm nights (TN90p), tropical nights (TR20), and warm spell duration indicator (WSDI). The most apparent change was displayed in SU25 and TD30 and the least in TR20 and WSDI. The reduction in SU25 and TD30 was the greatest at Veliko Gradiste (−15.7 and −16.3 days decade−1, respectively), concordant with changes in TXmean at this station in this period. The smallest reduction in SU25 (4.5 days decade−1) was observed at Loznica, the station with the smallest decrease in TXmean. The decrease in TD30 was the least at Zaltibor and Sjenica (less than 1 day per decade), the two highest stations with the smallest annual mean TD30 (3 days against the regional average of 28 days). Changes in TXx ranged from −2.2°C per decade at Veliko Gradiste and Zrenjanin to −0.8°C per decade at Vranje. Cooling of TNx was rather less, with the change rates ranging from −1.7°C per decade at Nis to +0.2°C per decade at Pozega. In the period 1981−2010, all hot extremes exhibited significant upward regional trends, except HD35, which also showed an upward but insignificant trend at most stations (Table 4). The increasing trend was significant at all stations for TX90p and TN90p. and WSDI. The largest decadal increase in WSDI and TX90p was recorded at the highest stations: Sjenica (7.0 and 16.8 days decade−1, respectively) and Zlatibor (6.9 and 16.1 days decade−1, respectively). Changes in TN90p ranged from 19.1 days per decade at Loznica to 8.6 days per decade at Smederevska Palanka. An increase in TXx, TNx, SU25, and TD30 in this more recent period was of smaller magnitude than their decrease in the earlier period (Figs. 6e and 6f). Warming of TXx was the greatest at Belgrade (1.1°C decade-1) and Leskovac (1.0°C decade-1), while the increase of TNn was the largest at Vrsac (1.0°C decade-1). The greatest increase in SU25 was recorded at Sjenica (8.4 days decade-1) and

ACCEPTED MANUSCRIPT

SC

RI

PT

Zlatibor (8.9 days decade-1), while the stations with the greatest increase in TD30 were Leskovac (8.6 days decade-1) and Nis (8.7 days decade-1).The largest significant increase in HD35 was observed along the South Morava River, the region that experienced the smallest decrease in HD35 (Fig. 6g) in the earlier period. An increase in TR20 at Belgrade stations (6.8 days decade−1) was much greater than at other stations (the next highest value was 3.3 days decade−1). But it has to be noted that in the earlier period Belgrade was the station with the greatest decrease in TR20 (−3.0 days decade−1). However, in the earlier period only four stations had a decrease in TR20 greater than 1 day per decade due to the fact that only six stations had a mean TR20 greater than 1 in this period. 3.5 Trends in cold extremes

AC CE P

TE

D

MA

NU

In contrast to hot extremes, which matched the trend pattern of TXmean, cold extremes did not closely follow changes in TNmean (Fig. 7). In both periods, all cold extremes exhibited warming trend, except cool days (TX10p) in the period 1961−1980 and the coldest day (TXn) in the period 1981−2010 (Table 4). The record cold events, the coldest day (TXn) and night (TNn), unexpectedly exhibited a significant increase during the earlier period. During the later period, insignificant trends were observed for both events (upward trends for TNn, and a mixture of upward and downward trends for TXn). Additionally, a reduction in the number of ice days (ID0) was much greater before 1980, despite less warming in that period. On the contrary, the frequency of frost days (FD0) was reduced more in the second than in the first period. In the later period, the number of cool nights (TN10p) increased significantly at all stations, while during the earlier period an increase of TN10p was less pronounced. The number of cool days (TX10p) displayed a mostly increasing, but insignificant, change in the first period and a decreasing, predominantly insignificant, change in the second. In the period 1961−1980, warming of TXn was the greatest at Dimitrovgrad and Krusevac (4.2°C decade−1) and the smallest at Negotin and Zajecar (<1°C decade−1), while the increase of TNn was the largest at Pozega and Smederevska Palanka (4.8°C decade−1) and the smallest at Negotin and Zajecar (<1°C decade−1). The only significant change in TX10p was observed at Leskovac (8.7 days decade−1), while the change of TN10p ranged from −13.0 days per decade at Rimski Sancevi to 4.4 days per decade at Vrsac. The only significant reduction in FD0 was recorded at Belgrade (7.7 days decade−1), while the reduction in ID0 was significant at all stations, except three, with the greatest decrease at Sombor (13.3 days decade−1). A decrease of CSDI was observed at all stations, ranging from −8.2 days per decade at Rimski Sancevi to −1.1 days per decade at Negotin. In the period 1981−2010, changes in TXn and TNn were small and showed no significance. A decrease in TN10p was of greater magnitude and significance than decrease in TX10p. The change in TN10p ranged −13.0 days per decade at Loznica to −6.5 days per decade at Veliko Gradiste, while chnage in TX10p varied between −6.1 days per decade at Nis to −2.0 days per decade at Kraljevo. The greatest reduction in FD0 was recorded at Zlatibor (−9.4 days decade−1) and the smallest at Veliko Gradiste (−2.5 days decade−1). The significant reduction in ID0 was recorded only at two stations, Leskovac (−2.5 days decade−1) and Pozega (−2.8 days decade−1), though the insignificant changes of ID0 at the

ACCEPTED MANUSCRIPT

high stations of Sjenica (−3.2 days decade−1) and Zlatibor (−3.3 days decade−1) were slightly greater.

PT

3.6 Trends in other extreme temperature indices

AC CE P

TE

D

MA

NU

SC

RI

DTR exhibited a decrease at the majority of stations in the first period and no change or an increase, mostly insignificant, in the second (Table 4). In the period 1981−2010, the greatest decrease in DTR was recorded at Pozega (−0.93°C decade−1) and Smederevska Palanka (−0.91°C decade−1), while an increase in DTR was found only at Vranje (0.39°C decade−1), the only station with a significant decrease in TNmean in this period (Fig. 8a). The lengthening of the growing season was more pronounced in the second period than in the first (Fig. 8b). On average for all stations, the increase of GSL was 14 days per decade over the last 30 years of the observational period (Table 4). The extension of the growing season in this period was the least and insignificant in the western part of the country (Fig. 8b), including Belgrade and Kragujevac in the central part, the two stations with the greatest mean GSL of all stations over the entire observational period (297 and 289 days, respectively). The greatest increase of GSL after 1980 was observed at Negotin (22 days decade−1) and Kursumlija (21 days decade−1), and the smallest at Pozega (5 days). Zajecar was among three stations with the largest increase of GSL (20 days decade−1) in the later period, and one of only two stations showing a decrease of GSL in the earlier period (−16 days decade−1) with a two-fold greater change than the other station displaying shortening of growing season (Veliko Gradiste). 3.7 Correlation between the temperature indices and large-scale circulation patterns Results of correlation analysis (Table 5) showed that the EA was much more associated with temperature indices than the NAO and EA/WR teleconnection patterns. The NAO was significantly correlated only with TXn and ID0, two indices that were correlated neither with the EA nor the EA/WR. The EA/WR pattern displayed significant correlation with more indices, mostly derived from TN, including TNmean, warm indices TNx, TN90p, and TR20, cold index FD0, and WSDI, the only index related to TX. However, all significant correlations between the EA/WR and temperature indices were of lower magnitude than the correlations of the same indices with the EA pattern. EA was significantly correlated with all indices, except with cold extremes TXn, TNn, ID0, and CSDI, representing severe cold events among those taken into consideration. The negative phase of the EA was associated with a greater frequency of cold days, cold nights, and frost days, while the positive phase was related to larger DTR, the lengthening of growing season, and the warming of all the examined hot indices. The prevailing positive phase of the EA since the late 1990s can be observed in Fig. 9, which displays time series of the EA pattern and two indices showing the highest correlation with EA (TXp90 and TN90p). Ghasemi (2015) demonstrated that even though the mean annual temperatures in Iran was significantly correlated with large-scale circulation features over the period 1961– 2010, teleconnection patterns did not play a significant role in the observed significant increases in mean annual temperature. To test whether the significant trends of temperature

ACCEPTED MANUSCRIPT

SC

RI

PT

extremes found in our study were partly forced by EA, regression analysis was done for all indices that exhibited significant correlation with this teleconnection pattern. New time series were constructed using linear regressions between extreme indices and EA and then tested for trend significance. The results showed that all time series predicted by EA exhibited significant trends. Observed and predicted time series and associated trends are displayed for TX10p (Fig. 10a), the index for which the observed trend and trend predicted by EA were nearly the same, and for TR20 (Fig. 10b), the index that displayed the largest disparity between the observed and predicted trends, with weaker but still significant predicted trend. Therefore, it can be concluded that the significant trends in temperature indices correlated with EA were partly forced by this large-scale circulation pattern.

NU

4. Discussion

AC CE P

TE

D

MA

The results of the study revealed that changes in temperature extremes in Serbia were quite different in the periods 1961−1980 and 1981−2010. The period 1961−1980 was characterised by a cooling trend in TXmean and hot extremes and by a mixed pattern of cooling, warming, and no trends in TNmean, while cold extremes displayed a mostly warming trend in this period. During this period, the most remarkable trends were cooling trends in SU25 and TD30 among the hot indices and warming trends in TXn and ID0 among the cold indices. The period 1981−2010 was characterised by a warming of both TNmean and TXmean, as well as of the majority of extreme indices, particularly hot extremes that exhibited a warming trend at all stations across the country. The most pronounced changes were displayed in TN90p and TX90p among the hot indices and in TN10p among the cold indices. The upward trends of hot extremes were significant at most stations in the period 1981−2010, except for the user-defined index HD35. Cold indices also exhibited an upward trend, but of less magnitude and significance. The detected change point in the studied temperature series in Serbia is in line with the reported turning points in European temperature time series. Klein Tank and Können (2003) reported that in Europe there has been a warming period followed by a cooling period since late 1970s. Toreti and Desiato (2008) found 1981 to be a change point in the annual mean temperature anomaly series in Italy, dividing the first cooling period and the second warming period. Fioravanti et al. (2016) identified a change point in 1977 for the minimum and maximum temperature in Italian annual series, with non-significant cooling trends during the sub-period 1961–1977 and significant warming trends during the subperiod 1978–2011. Spinoni et al. (2015) in their study on climatologies and trends of 10 variables in the Carpathian Region for the period 1961−2011, which included 20 weather stations from northern Serbia, concluded that solar dimming overcame global warming during the 1960s and 1970s, resulting in negative temperature anomalies. This finding might explain the cooling of TXmean and the inconsistent changes in TNmean before 1980, since TN is measured during the night and TX during the day in the presence of solar heating. The following observed changes in indices derived from TX points to different pattern of changes in TX throughout the year in the two studied periods: firstly, a significant warming of cold extremes based on TX, with a greater increase of TXn and

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

greater reduction of ID0 before 1980 than afterwards, and, secondly, a greater cooling rate of hot indices derived from TX over the first period than a warming rate in the second period despite less pronounced general warming and even a decrease in TXmean during this period. The highest stations in the study, Sjenica and Zlatibor, experienced the largest increase in TX90p, SU25, and WSDI and the largest decrease in ID0 (though not statistically significant) during recent decades, accompanied with strong warming of TXmean and moderate warming of TNmean. In addition, these two stations were among the stations with a greater warming of TXx and a greater decrease in FD0 (Zlatibor with the greatest) over the recent period. The fact that there was no correlation between trends in TNmean and TXmean for individual stations indicates that changes in TN and TX were not occurring simultaneously. The results that some stations experienced counter-trends in TNmean and TXmean confirm that changes in extreme temperatures and events are not always directly related to changes in the mean temperature. For example, at Pozega an upward trend in TNmean and a downward trend in TXmean of the same magnitude over the period 1961−1980 caused no considerable change in the mean temperature, but resulted in significantly smaller DTR and changes in other temperature extremes. Our results of the correlation between the large-scale circulation patterns and temperature indices are in line with the findings of Barnston and Livezey (1987) that the EA better describes the westerly airflow in central and southern Europe than the NAO. The stronger correlation of the EA with hot indices indicates that the EA positive mode is associated with the transport of warm air over Serbia.

5. Conclusion

This detailed analysis of trends in maximum and minimum temperatures and extreme temperature indices over Serbia, besides providing baseline information for scientists, end-users, and policy-makers regarding climate change at a national level, makes a contribution to enhancing understanding of regional patterns of climate change. Performed in a similar way as research in many regions worldwide, it offers the possibility of comparison with already reported trends from different parts of the world. The main findings are: (i) a decreasing trend in TXmean and a mixed increasing, decreasing, and no trend in TNmean in the period 1961−1980; (ii) a clear, spatially uniform, significant warming since 1980, with significant increasing trends in both TNmean and TXmean; (iii) a cooling tendency of hot indices until 1980 and a warming tendency afterwards; (iv) a general warming tendency of cold indices over the entire observational period; (v) no significant changes of DTR and significant lengthening of the growing season since 1980; (vi) annual temperature indices were much more strongly correlated with EA than with NAO and EA/WR pattern; (vii) the observed significant trends of extreme temperature indices were partly forced by EA. Finding that changes in some extremes were more apparent before 1980 than afterwards, despite the greater change of TNmean and TXmean in the later period, suggests that further research dealing with monthly and seasonal changes of extreme temperatures

ACCEPTED MANUSCRIPT

PT

and relevant indices should be undertaken in order to provide a more comprehensive representation of the changes in temperature extremes in Serbia in the recent past. Seasonal and monthly analysis might reveal a stronger association with NAO and EA/WR, since in that way the considerable inter-seasonal variability of these circulation patterns will be taken into account.

SC

RI

Acknowledgement: This study was realized as a part of the projects III43007, TR37005 and TR31005, financed by the Ministry of Education and Science of the Republic of Serbia. References

AC CE P

TE

D

MA

NU

Abatzoglou, J.T., Redmond, K.T., 2007. The asymmetry of trends in spring and autumn temperature and circulation regimes over western North America. Geophys. Res. Lett. 34, L18808. Aguilar, E., Aziz Barry, A., Brunet, M., Ekang, L., Fernandes, A., Massoukina, M., Mbah, J., Mhanda, A., do Nascimento, D.J., Peterson, T.C., Thamba Umba, O., Tomou, M., Zhang, X., 2009. Changes in temperature and precipitation extremes in western central Africa, Guinea Conakry, and Zimbabwe, 1955–2006. J. Geophys. Res. 114, D02115. Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank, A.M.G., Haylock, M., Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A., Rupa Kumar, K., Revadekar, J., Griffiths, G., Vincent, L., Stephenson, D.B., Burn, J., Aguilar, 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. Alexandersson, H., Moberg, A., 1997. Homogenization of Swedish temperature data. Part I: homogeneity test for linear trends. Int. J. Climatol. 17, 25–34. Athar, H., 2014. Trends in observed extreme climate indices in Saudi Arabia during 1979–2008. Int. J. Climatol. 34, 1561–1574. Bajat, B., Blagojević, D., Kilibarda, M., Luković, J., Tošić, I., 2015. Spatial analysis of the temperature trends in Serbia during the period 1961–2010. Theor. Appl. Climatol. 121, 289– 301. Barnston, A.G., Livezey, R.E., 1987. Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev. 115, 1083–1126. Brunet, M., Sigró, J., Jones, P.D., Saladié, O., Aguilar, E., Moberg, A., Lister, D., Walther, A., 2007. Long-term changes in extreme temperatures and precipitation in Spain. Contrib. Sci. 3, 331–342. Burić, D., Luković, J., Ducić, V., Dragojlović, J., Doderović, M., 2014. Recent trends in daily temperature extremes over southern Montenegro (1951–2010). Nat. Hazards Earth Syst. Sci. 14, 67–72. Caesar, J., Alexander, L.V., Trewin, B., Tse-ring, K., Sorany, L., Vuniyayawa, V., Keosavang, N., Shimana, A., Htay, M.M., Karmacharya, J., Jayasinghearachchi, D.A., Sakkamart, J., Soares, E., Hung, L.T., Thuong, L.T., Hue, C.T., Dung, N.T.T., Hung, P.V., Cuong, H.D., Cuong, N.M., Sirabaha, S., 2011. Changes in temperature and precipitation extremes over the IndoPacific region from 1971 to 2005. Int. J. Climatol. 31, 791–801. Coumou, D., Rahmstorf, S., 2012. A decade of weather extremes. Nat. Clim. Change 2, 491–496, Croitoru, A.E., Piticar, A., 2013. Changes in daily extreme temperatures in the extra-Carpathians regions of Romania. Int. J. Climatol. 33, 1987–2001. Donat, M.G., Peterson, T.C., Brunet, M., King, A.D., Almazroui, M., Kolli, R.K., Boucherf, D., AlMulla, A.Y., Nour, A.Y., Aly, A.A., Nada, T.A.A., Semawi, M.M., Dashti, H.A.A., Salhab,

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

T.G., El Fadli, K.I., Muftah, M.K., Eida, S.D., Badi, W., Driouech, F., Rhaz, K.E., Abubaker, M.J.Y., Ghulam, A.S., Erayah, A.S., Mansour, M.B., Alabdouli, W.O., Dhanhani, J.S.A., Al Shekaili, M.N. 2013b. Changes in extreme temperature and precipitation in the Arab region: long-term trends and variability related to ENSO and NAO. Int. J. Climatol. 34, 581–592. Efthymiadis, D., Goodess, C.M., Jones, P.D., 2011. Trends in Mediterranean gridded temperature extremes and large-scale circulation influences. Nat. Hazards Earth Syst. Sci. 11, 2199–2214. El Kenawy, A.M., López-Moreno, J.I., Vicente-Serrano, S.M., 2011. Recent trends in daily temperature extremes over northeastern Spain (1960–2006). Nat. Hazards Earth Syst. Sci. 11, 2583–2603. Fioravanti, G., Piervitali, E., Desiato, F., 2016. Recent changes of temperature extremes over Italy: an index-based analysis. Theor. Appl. Climatol. 123, 473–486. Fischer, E.M., Knutti, R., 2015. Anthropogenic contribution to global occurrence of heavy-precipitation and high-temperature extremes. Nat. Clim. Change 5, 560–564. Ghasemi, A.R., 2015. Changes and trends in maximum, minimum and mean temperature series in Iran. Atmos. Sci. Let. 16, 366–372. Hansen, J.I., Fung, I., Lacis, A., Rind, D., Lebedeff, S., Ruedy, R., Russell, G., Stone, P., 1988. Global climate changes as forecast by Goddard Institute for space studies three-dimensional model. J. Geophys. Res. 93, 9341–9364. Hundecha, Y., Brádossy, A., 2005. Trends in daily precipitation and temperature extremes across western Germany in the second half of the 20th century. Int. J. Climatol. 25, 1189–1202. Hurrell, J.W., 1995. Decadal trends in the North Atlantic oscillation and relationships to regional temperature and precipitation. Science 269, 676–679. IPCC, 2012. Summary for Policymakers, in: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L., Mastrandrea, M.D., Mach, K.J., Plattner, G.K., Allen, S.K., Tignor, M., Midgley, P.M. (Eds.), Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press, Cambridge, United Kingdom, pp. 3– 24. IPCC, 2013. Summary for Policymakers, in: Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom, pp. 3–29. Jones, G.V., 2005. Climate change in the western United States grape growing regions. Acta Hort. 689, 41–59. Keggenhoff, I., Elizbarashvili, M., Amiri-Farahani, A., King, L., 2014. Trends in daily temperature and precipitation extremes over Georgia, 1971–2010. Wea. Clim. Extremes 4, 75–85. Kendall, M.G., 1975. Rank Correlation Methods. Charles Griffin, London. Kioutsioukis, I., Melas, D., Zerefos, C., 2010. Statistical assessment of changes in climate extremes over Greece (1955–2002). Int. J. Climatol. 30, 1723–1737. Klein Tank, A., Können, G., 2003. Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99. J. Clim. 16, 3665–3680. Klein Tank, A.M.G., Peterson, T.C., Quadir, D.A., Dorji, S., Zou, X., Tang, H., Santhosh, K., Joshi, U.R., Jaswal, A.K., Kolli, R.K., Sikder, A.B., Deshpande, N.R., Revadekar, J.V., Yeleuova, K., Vandasheva, S., Faleyeva, M., Gomboluudev, P., Budhathoki, K.P., Hussain, A., Afzaal, M., Chandrapala, L., Anvar, H., Amanmurad, D., Asanova, V.S., Jones, P.D., New, M.G., Spektorman, T., 2006. Changes in daily temperature and precipitation extremes in Central and South Asia. J. Geophys. Res. 111, D16105.

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

Knežević, S., Tošić, I., Unkašević, M., Pejanović, G., 2014. The influence of the East Atlantic Oscillation to climate indices based on the daily minimum temperatures in Serbia. Theor. Appl. Climatol. 116, 435-446. Kostopoulou, E., Jones, P.D., 2005. Assessment of climate extremes in the Eastern Mediterranean. Meteorol. Atmos. Phys. 89, 69–85. Liang, L.Q., Li, L.J., Liu, Q., 2010. Temporal variation of reference evapotranspiration during 1961–2005 in the Taoer River basin of Northeast China. Agric. Forest Meteorol. 150, 298–306. Maidment, D.R., 1993. Handbook of Hydrology. McGraw-Hill, New York. Malinovic-Milicevic, S., Radovanovic, M. M., Stanojevic G., Milovanovic, B., 2015. Recent Changes in Serbian Climate Extreme Indices from 1961 to 2010. Theor. Appl. Climatol. doi:10.1007/s00704-015-1491-1 Mearns, L.O., Katz, R.W., Schneider, S.H., 1984. Extreme high temperature events: changes in their probabilities with changes in mean temperature. J. Clim. Appl. Meteorol. 23, 1601–1608. Moberg, A., Jones, P.D., 2005. Trends in indices for extremes in daily temperature and precipitation in central and Western Europe, 1901–99. Int. J. Climatol. 25, 1149–1171. Moberg, A., Jones, P.D., Lister, D.,Walther, A., Brunet, M., Jacobeit, J., Alexander, L.V., DellaMarta, P.M., Luterbacher, J., Yiou, P., 2006. Indices for daily temperature and precipitation extremes in Europe analyzed for the period 1901–2000. J. Geophys. Res. 111, D22106. Nemec, J., Gruber, C., Chimani, B., Auer, I., 2013. Trends in extreme temperature indices in Austria based on a new homogenised dataset. Int. J. Climatol. 33, 1538–1550. Orlowsky, B., Seneviratne, S.I., 2012. Global changes in extreme events: regional and seasonal dimension. Clim. Change 110, 669–696. Rahimzadeh, F., Asgari, A., Fattahi, E., 2009. Variability of extreme temperature and precipitation in Iran during recent decades. Int. J. Clim. 29, 329–343. Rogers, J.C., 1990. Patterns of low-frequency monthly sea level pressure variability (1899–1986) and associated wave cyclone frequencies. J. Clim. 3, 1364–1379. Salinger, M., Griffiths, G., 2001. Trends in New Zealand daily temperature and rainfall extremes. Int. J. Climatol. 21, 1437–1452. Sillmann, J., Kharin, V.V., Zwiers, F.W., Zhang, X., Bronaugh, D., 2013. Climate extremes indices in the CMIP5 multimodel ensemble: Part 2. Future climate projections. J. Geophys. Res. Atmos. 118, 2473–2493. Skansi, M.M., Brunet, M., Sigro, J., Aguilar, E., Arevalo Groening, J.A., Bentancur, O.J., Castellon Geier, Y.R., Correa Amaya, R.I., Jacome, H., Malheiros Ramos, A., Oria Rojas, C., Pasten, A.M., Sallons Mitro, S., Villaroel Jimenez, C., Martinez, R., Alexander, L.V., Jones, P.D., 2013. Warming and wetting signals emerging from analysis of changes in climate extreme indices over South America. Glob. Planet. Change 100, 295–307. Sneyers, R., 1990. On the statistical analysis of series of observations. Technical note no. 143, WMO-No. 415. World Meteorological Organization, Geneva, Switzerland. Spinoni, J., Szalai, S., Szentimrey, T., Lakatos, M., Bihari, Z., Nagy, A., Nemeth, A., Kovacs, T., Mihic, D., Dacic, M., Petrovic, P., Kržič, A., Hiebl, J., Auer, I., Milkovic, J., Štepanek, P., Zahradnicek, P., Kilar, P., Limanowka, D., Pyrc, R., Cheval, S., Birsan, M.V., Dumitrescu, A., Deak, G., Matei, M., Antolovic, I., Nejedlik, P., Štastny, P., Kajaba, P., Bochnicek, O., Galo, D., Mikulova, K., Nabyvanets, Y., Skrynyk, O., Krakovska, S., Gnatiuk, N., Tolasz, R., Antofie, T., Vogt, J., 2015. Climate of the Carpathian Region in the period 1961–2010: climatologies and trends of 10 variables. Int. J. Climatol. 35, 1322–1341. Stephenson, T.S., Vincent, L.A., Allen, T., Van Meerbeeck, C.J., McLean, N., Peterson, T.C., Taylor, M.A., Aaron-Morrison, A.P., Auguste, T., Bernard, D., Boekhoudt, J.R.I., Blenman, R.C., Braithwaite, G.C., Brown, G., Butler, M., Cumberbatch, C.J.M., Etienne-Leblanc, S., Lake, D.E., Martin, D.E., McDonald, J.L., Ozoria Zaruela, M., Porter, A.O., Santana Ramirez,

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

M., Tamar, G.A., Roberts, B.A., Sallons Mitro, S., Shaw, A., Spence, J.M., Winter, A., Trotman, A.R., 2014. Changes in extreme temperature and precipitation in the Caribbean region, 1961–2010. Int. J. Climatol. 34, 2957–2971. Thibeault, J.M., Seth, A., 2014. Changing climate extremes in the Northeast United States: observations and projections from CMIP5. Clim. Change 127, 273-287. Toreti, A., Desiato, F., 2008. Temperature trend over Italy from 1961 to 2004. Theor. Appl. Climatol. 91, 51–58. Unkašević, M., Tošić, I., 2009a. Changes in the extreme daily winter and summer temperatures at Belgrade. Theor. Appl. Climatol. 95, 27–38. Unkašević, M., Tošić, I., 2009b. An analysis of heat waves in Serbia. Glob. Planet. Change 65, 17– 26. Unkašević, M., Tošić, I., 2013. Trends in temperature indices over Serbia: relationships to largescale circulation patterns. Int. J. Climatol. 33, 3152–3161. Vincent, L.A., Aguilar, E., Saindou, M., Hassane, A.F., Jumaux, G., Roy, D., Booneeady, P., Virasami, R., Randriamarolaza, L.Y.A., Faniriantsoa, F.R., Amelie, V., Seeward, H., Montfraix, B., 2011. Observed trends in indices of daily and extreme temperature and precipitation for the countries of the western Indian Ocean, 1961–2008. J. Geophys. Res. 116, D10108. Wang, X.L., 2008a. Penalized maximal F-test for detecting undocumented mean shifts without trend-change. J. Atmos. Oceanic Tech. 25, 368–384. Wang, X.L., 2008b. Accounting for autocorrelation in detecting mean-shifts in climate data series using the penalized maximal t or F test. J. Appl. Meteor. Climatol. 47, 2423–2444. Wang, X.L., Wen, Q.H., Wu, Y., 2007. Penalized maximal t test for detecting undocumented mean change in climate data series. J. Appl. Meteor. Climatol. 46, 916–931. Wang, X.L., Feng, Y., 2013. RHtestsV4 User Manual, Published online, Climate Research Division, Atmospheric Science and Technology Directorate Science and Technology Branch, Environment Canada, Toronto, Ontario, Canada. http://etccdi.pacificclimate.org/software.shtml (accessed 20.07.2013). Zhang, X., Vincent, L.A., Hogg, W., Niitsoo, A., 2000. Temperature and precipitation trends in Canada during the 20th century. J. Atmos. Ocean Technol. 38, 395–429. Zhang, X., Aguilar, E., Sensoy, S., Melknyan, H., Taghiyeva, U., Ahmed, N., Kutaladze, N., Rahimzadeh, F., Taghipour, A., Hantosh, T.H., Albert, P., Semawi, M., Ali, M.K., Halal, M., Al-Shabibi, A., Al-Oulan, A., Zatari, A., Al Dean Khalil, I., Sagir, R., Demircan, M., Eken, M., Adiguzel, M., Alexander, L., Peterson, T.C., Wallis, T., 2005. Trends in Middle East climate extremes indices during 1950–2003. J. Geophys. Res. 110, D22104. Zhang, X., Alexander, L., Hegerl, G.C., Jones, P., Tank, A.K., Peterson, T.C., Trewin, B., Zwiers, F.W., 2011. Indices for monitoring changes in extremes based on daily temperature and precipitation data. Wiley Interdis. Rev.: Clim. Change 2, 851–870.

ACCEPTED MANUSCRIPT

Missing data (%) TX TN 0.0 0.0

Beograd

44.80

20.47

132

CU

Cuprija

43.93

21.37

123

0.0

0.0

DM

Dimitrovgrad

43.02

22.75

450

0.0

0.0

KI

Kikinda

45.85

20.47

81

0.0

0.0

KG

Kragujevac

44.03

20.93

185

0.0

0.0

KV

Kraljevo

43.72

20.70

215

0.0

0.0

KS

Krusevac

43.57

21.35

166

0.2

0.2

KU

Kursumlija

43.13

21.27

383

0.0

0.0

LE

Leksovac

42.98

21.95

230

1.7

1.5

LO

Loznica

44.55

19.23

121

0.0

0.0

NE

Negotin

44.23

22.55

NI

Nis

43.33

PA

Palic

46.10

PZ

Pozega

RS

Rimski Sancevi

SJ

Sjenica

SP

2.0

2.0

204

0.0

0.0

19.77

TE

102

0.4

0.4

43.83

20.03

310

0.0

0.0

45.33

19.85

86

0.0

0.0

43.27

20.00

1038

0.3

0.3

Smederevska Palanka

44.37

20.95

121

0.0

0.0

SO

Sombor

45.77

19.15

87

0.0

0.0

SM

Sremska Mitrovica

45.02

19.55

82

0.0

0.0

VA

Valjevo

45.77

19.15

87

0.7

0.7

VG

Veliko Gradiste

44.75

21.52

80

0.0

0.0

VR

Vranje

42.55

21.92

432

0.0

0.0

VS

Vrsac

45.15

21.32

84

0.0

0.0

ZA

Zajecar

43.88

22.28

144

0.3

0.0

D

42

21.90

AC CE P

MA

SC

BG

NU

RI

Latitude Longitude Elevation (°N) (°E) (m)

Code Name

PT

Table 1 Geographical coordinates and elevation of selected meteorological stations, percentage of missing data of daily maximum (TX) and minimum (TN) temperatures in the period 1961– 2010, and the results of the homogeneity test. Date of the detected stepa TX

TN

19931000 ?

19931100 ? 19931100 ?

19680500 ?

19680600 ?

20060900 ? 19940100 ? 20061000 ?

ZL Zlatibor 43.73 19.72 1028 0.0 0.0 ZR Zrenjanin 45.40 20.38 80 0.0 0.0 a Dates are given in the YYYYMMDD format; the flag ? denotes that detected step change in data may or may not be significant (Wang and Feng, 2013).

ACCEPTED MANUSCRIPT

Table 2 Definition, acronym (ID) and descriptive name of the ETCCDI temperature indices based on daily maximum (TX) and minimum (TN) temperatures calculated on an annual basis. Units

Highest TX Highest TX Number of days when TX > 90th percentile Number of days when TN > 90th percentile Number of days with TX > 25 °C Number of days with TX > 30 °C Number of days with TX > 35 °C Number of days with TN > 20 °C Number of days with at least 6 consecutive days with TX > 90th percentile

intensity intensity frequency frequency frequency frequency frequency frequency

°C °C days days days days days days

duration

days

Lowest TX Lowest TN Number of days when TX < 10th percentile Number of days when TN < 10th percentile Number of days with TN < 0 °C Number of days with TX < 0 °C Number of days with at least 6 consecutive days with TN < 90th percentile

intensity intensity frequency frequency frequency frequency

°C °C days days days days

duration

days

PT

Category

SC

NU

MA

D

AC CE P

Hot extremes Hottest day TXx Hottest night TNx TX90p Warm days TN90p Warm nights Summer days SU25 TD30 Tropical days HD35 Hot days Tropical nights TR20 Warm spell duration WSDI indicator Cold extremes Coldest day TXn Coldest night TNn TX10p Cool days TN10p Cool nights Frost days FD0 Ice days ID0 Cold spell duration CSDI indicator

Definition

RI

Descriptive name

TE

ID

Other temperature indices DTR GSL

Diurnal temperature range Average difference between TX and TN intensity Number of days between first span of at least 6 days Growing season length with Tmean > 5 °C and the first occurrence after 1 duration July of at least 6 consecutive days with Tmean < 5 °C

°C days

ACCEPTED MANUSCRIPT Table 3 The correlations matrix for the regionally averaged extreme temperatures indices over the period 1961−2010. GSL

Highest ra with

-0.85

-0.55

-0.56

-0.39

-0.21

0.76

0.71

TX90p (0.92)

0.25

PT

DTR

0.16

1

0.78

0.60 0.55

0.91

0.80

0.77

0.72

0.67

0.61

0.69

1

0.35 0.49 1

TNx TX90p TN90p SU25

0.63

0.87

0.48

0.42

0.37

0.60

0.54

0.32

0.43

-0.72

-0.86

-0.75

-0.51

-0.43

0.18

0.73

TN90p (0.87)

0.80

0.69

0.54

0.64

0.79

0.85

0.72

0.59

-0.16

0.04

-0.45

-0.13

-0.16

-0.05

0.12

0.57

0.31

HD35 (0.85)

1

0.66

0.70

0.65

0.79

0.73

0.89

0.61

-0.17

-0.05

-0.35

-0.26

-0.12

0.03

0.17

0.35

0.30

TR20 (0.89)

1

0.80

0.77

0.82

0.75

0.70

0.76

-0.03

0.05

-0.65

-0.34

-0.38

-0.20

-0.05

0.77

0.56

TXmean (0.91)

1

0.63

0.66

0.54

0.81

0.68

0.00

0.12

-0.58

-0.60

-0.57

-0.21

-0.14

0.33

0.64

TNmean (0.87)

1

0.80

0.63

0.59

0.53

-0.12

-0.05

-0.59

-0.27

-0.17

0.00

0.07

0.70

0.31

TD30 (0.80)

1

0.81

0.77

0.67

-0.18

-0.05

-0.47

-0.16

-0.17

0.03

0.17

0.69

0.43

HD35 (0.81)

1

0.71

0.70

-0.04

0.03

-0.44

-0.15

-0.14

-0.07

0.01

0.67

0.36

TXx (0.85)

0.66

-0.09

0.01

-0.37

-0.34

-0.24

-0.03

0.08

0.32

0.46

TNx (0.89)

1

0.10

0.04

-0.48

-0.33

-0.39

-0.13

-0.05

0.51

0.43

TX90p (0.76)

1

0.81

-0.31

-0.38

-0.22

-0.78

-0.65

-0.09

0.24

TNn (0.81)

1

-0.38

-0.47

-0.36

-0.77

-0.63

-0.05

0.35

TXn (0.81)

1

0.64

0.49

0.57

0.37

-0.59

-0.63

TXmean (-0.85)

1

0.65

0.49

0.60

0.02

-0.59

TNmean (-0.86)

1

0.45

0.32

-0.10

-0.76

GSL (-0.76)

1

0.64

-0.09

-0.47

TXn (-0.78)

1

0.12

-0.25

TXn (-0.65)

1

0.36

TX90p (0.77)

TD30 HD35 TR20

1

PT ED

WSDI TXn TNn TX10p

CE

TN10p ID0 CSDI DTR

AC

FD0

GSL Bold values significant at the 0.05 level. a

CSDI

Pearson coefficient of correlation (value shown in brackets)

RI

TXx

ID0

TX10p

SC

TNmean

FD0

TNn

NU

TXmean

TN10p

TXmean TNmean TXx TNx TX90p TN90p SU25 TD30 HD35 TR20 WSDI TXn

MA

Index

1

FD0(-0.76)

ACCEPTED MANUSCRIPT

Table 4 Regional decadal changes of temperature extremes and the number of stations with statistically significant (S) and non-significant (NS) trends in temperature indices.

1

6

2

9

8

TXmean (°C)

-0.36

5

21

0

0

0

-1.52 -0.64 -7.9 -4.3 -11.8 -10.8 -1.0 -0.5 -1.7

11 8 14 4 23 25 13 1 2

15 16 12 18 3 1 10 16 21

0 0 0 0 0 0 0 0 0

0 2 0 1 0 0 1 4 2

MA

3.36 3.01 3.6 -5.0 -3.4 -9.5 -5.3

0 0 0 9 1 23 21

0 0 0 13 18 3 5

11 2

1 0

PT

S

NS

0.48

0

26

0

0

0.56

0

0

26

0

0

0 0 0 3 0 0 2 5 6

0.82 0.52 11.8 12.7 6.4 6.2 1.6 1.2 4.7

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 1 0

15 15 26 26 21 23 5 15 25

11 11 0 0 5 3 20 9 1

0 0 0 0 0 0 1 1 0

23 8 1 0 0 0 0

3 18 22 3 6 0 1

0 0 3 1 1 0 2

0.05 0.27 -4.1 -9.8 -5.9 -1.5 -2.1

0 0 11 26 18 2 10

10 2 15 0 8 24 16

0 0 0 0 0 0 0

11 20 0 0 0 0 0

5 4 0 0 0 0 0

0 20

1 4

0.09 14.4

0 0

4 0

4 20

11 6

7 0

TE

TXn (°C) TNn (°C) TX10p (days) TN10p (days) FD0 (days) ID0 (days) CSDI (days)

AC CE P

Cold extremes

NS

0

NU

Hot extremes TXx (°C) TNx (°C) TX90p (days) TN90p (days) SU25 (days) TD30 (days) HD35 (days) TR20 (days) WSDI (days)

RI

0.04

S

SC

TNmean (°C)

1981−2010 Number of stations with Regional change Decreasing Increasing -1 No trend a a (decade )

D

Index

1961−1980 Number of stations with Regional change Decreasing Increasing -1 No trend (decade ) Sa NS Sa NS

Other temperature indices DTR (°C) GSL (days)

-0.40 3.5

13 0

Bold values significant at the 0.05 level. a Statistically significant at the 0.05 level.

ACCEPTED MANUSCRIPT

EA

EA/WR

0.15 -0.01

0.57 0.57

-0.18 -0.38

-0.05 -0.20 0.03 -0.15 0.04 0.01 0.04 -0.23 -0.06

0.44 0.61 0.62 0.66 0.49 0.49 0.37 0.59 0.51

-0.18 -0.37 -0.21 -0.39 0.04 -0.16 -0.13 -0.41 -0.24

Cold extremes TXn 0.31 TNn 0.27 TX10p -0.18 TN10p 0.05 FD0 -0.03 ID0 -0.28 CSDI -0.18

-0.04 0.06 -0.50 -0.37 -0.37 -0.21 -0.08

-0.09 -0.06 0.13 0.26 0.30 0.14 0.04

TXmean TNmean

RI

NAO

PT

Table 5 Correlation coefficients between the regionally averaged annual temperature indices and the North Atlantic Oscillation (NAO), East Atlantic (EA), and Atlantic/West Russia (EA/WR) annual indices.

Other temperature indices DTR GSL

NU MA

D

TE

AC CE P

TXx TNx TX90p TN90p SU25 TD30 HD35 TR20 WSDI

SC

Hot extremes

0.25

0.30

0.12

-0.07

0.45

-0.26

Bold values significant at the 0.05 level.

ACCEPTED MANUSCRIPT

Figure Captions

PT

Fig.1. Map showing the study area and location of meteorological stations used in the study (see Table 1 for explanation of station acronyms).

SC

RI

Fig. 2. (a) Time series of annual minimum temperature (TNmean) based on data from 1961 to 2010 averaged over 26 stations. The solid, dashed, and dotted lines represent linear regressions for the periods 1961–2010, 1961–1980, and 1981–2010, respectively; (b) Linear regression residual plot.

NU

Fig. 3. Same as Fig. 2 but for anual maximum temperature (TXmean). Fig. 4. Sequential Mann–Kendall test results for (a) TNmean and (b) TXmean. The dotted horizontal lines represent critical values corresponding to the 95% confidence interval.

AC CE P

TE

D

MA

Fig. 5. Spatial distribution of decadal trends and inter-annual variation of (a) annual minimum temperature (TNmean) and (b) annual maximum temperature (TXmean) in Serbia. Left and middle panels: station linear trends over the periods 1961–1980 and 1981– 2010, respectively. Upward/downward pointing triangles show increasing/decreasing trends. The triangle size is proportional to the magnitude of trend. The filled triangles correspond to statistically significant trends (significant at the 0.05 level), while dots indicate the stations with no trend. Right panel: time series of region averaged anomalies relative to the 1971–2000 mean. The curves show five-year moving averages. Fig. 6. Same as Fig. 5 but for hot extremes: (a) hottest day (TXx), (b) hottest night (TNx), (c) warm days (TX90p), (d) warm nights (TN90p), (e) summer days (SU25), (f) tropical days (TD30), (g) hot days (HD35), (h) tropical nights (TR20) and (i) warm spell duration indicator (WSDI). Fig. 7. Same as Fig. 5 but for cold extremes: (a) coldest day (TXn), (b) coldest night (TNn), (c) cool days (TX10p), (d) cool nights (TN10p), (e) frost days (FD0), (f) ice days (ID0) and (g) cold spell duration indicator (CSDI). Fig. 8. Same as Fig. 5 but for (a) diurnal temperature range (DTR) and (b) growing season length (GSL). Fig. 9. Annual time series of warm days (TX90p) departures from the 1970–2010 mean, warm nights (TN90p) departures from the 1970–2010 mean, and the East Atlantic (EA) pattern. The left vertical axis refers to TX90p and TN90p, the right vertical axis to EA. Fig. 10. (a) Cold days (TX10p) and (b) tropical nights (TR20) annual time series and trend lines of observed temperature extreme (solid blue line) and predicted by East Atlantic (EA) pattern (dashed red line).

TE AC CE P

Figure 1

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

(a)

PT

7.0

RI SC

6.0 5.5 5.0

NU

TNmean (oC)

6.5

4.0 1960

1970

1980

MA

4.5

1990

2000

2010

TE

D

Year

(b)

Residual

1 0.5 0 -0.5

AC CE P

1.5

-1 -1.5 1960

1970

1980

1990

Year Figure 2

2000

2010

ACCEPTED MANUSCRIPT

(a)

PT

19

RI SC

17 16

NU

TXmean (oC)

18

14 1960

1970

1980

MA

15

1990

2000

2010

1.5 1.0

Residual

TE

2.0

0.5 0.0 -0.5 -1.0

AC CE P

(b)

D

Year

-1.5 -2.0 1960

1970

1980

1990

Year Figure 3

2000

2010

ACCEPTED MANUSCRIPT

SC

RI

PT

5 u(t) u'(t) 4 3 2 1 0 -1 -2 -3 -4 -5 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

MA

NU

Test statistic

(a) TNmean

Year

D

(b) TXmean

AC CE P

Test statistic

TE

4 u(t) u'(t) 3 2 1 0 -1 -2 -3 -4 -5 -6 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Year Figure 4

Figure 5

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

Figure 6

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 7

Figure 8

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

TX90p

TN90p

EA 1.5

PT

60

20 0

NU

-20

MA

-40 -60

0.5 0

-0.5 -1 -1.5

AC CE P

TE

D

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Figure 9

EA

RI

1

SC

Number of days

40

ACCEPTED MANUSCRIPT

(a)

obs TX10p

pre TX10p

PT

70 60

RI

SC

40

NU

30 20 10

y = -0.060 x + 11.3

MA

Number of days

y = -0.069 x + 11.5 50

0

TE

D

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

(b) 8 7

AC CE P

obs TR20

y = 0.083 x - 162.9

Number of days

6 5 4

pre TR20

y = 0.046 x - 89.1

3 2 1 0

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Figure 10

ACCEPTED MANUSCRIPT

Highlights

SC

RI

   

Trends in extreme temperature indices in Serbia during 1961−2010 were investigated. Hot indices displayed a cooling trend until 1980 and a warming trend afterwards. A general warming tendency of cold indices over the entire period. No significant changes of diurnal temperature range. Significant lengthening of the growing season since 1980.

PT



AC CE P

TE

D

MA

NU

Running page head: Temperature extremes in Serbia