1. Introduction
In the last few decades, there has been an accelerated transformation of the climate system. The general opinion is that the primary cause of this is the uncontrolled anthropogenic emission of gases with the greenhouse effect, primarily carbon dioxide (CO
2). It is generally believed that heat waves, rising sea levels, extreme rainfall and floods, melting ice, long–term droughts and forest fires, devastating tropical hurricanes and other extreme events are being registered more often and are most likely linked to global warming. The sixth report of the Intergovernmental Panel on Climate Change (IPCC) confirms, with a high degree of certainty, the supremacy of the anthropogenic factor on today’s global climate. The Paris Agreement from 2015 sets out a global framework by limiting global warming to below 2 °C, i.e., the global average temperature should be 1.5 °C higher by the end of this century compared to the pre–industrial level. If the global temperature rises by 2 or more degrees compared to the pre–industrial period, irreversible and possibly large negative changes in natural and human systems can be expected according to the models [
1]. However, in the aforementioned IPCC Report, it is said that the global temperature nowadays is higher by 1 °C and that an increase of 1.5 °C will be happening in the next 20 years (until 2040), because the progression of atmospheric CO
2 concentration has not stopped.
Therefore, the CO
2 of anthropogenic origin remains the greatest threat to climate change. In addition, a recent announcement by NOAA [
2] highlights the concern over the record increase in the concentration of methane (CH
4) in the atmosphere during 2020 and 2021. Methane is the second most powerful anthropogenic factor in global warming, right after CO
2. This gas stays in the atmosphere for a much shorter time than CO
2, approximately for nine years. However, CH
4 has a far stronger greenhouse effect, as it retains even 31 times more heat than CO
2, and has a strong short–term impact. It is very likely that the record increases in CH
4 concentrations are due to the expansion of oil and gas drilling, as methane tends to leak from wells and pipelines. This gas is also released from agriculture (cattle farming and landfills).
Even though they take place in different time distances, anthropogenic climate change and COVID–19 are currently the most pressing crisis [
1,
3,
4,
5,
6]. It is considered that the solution to the climate crisis is to stop the anthropogenic emission of CO
2 and other gases that pollute the atmosphere, respectively, the decarbonization of the global economy, which means stopping the burning of fossil fuels and switching to the use of green energy [
7,
8,
9]. Regarding modern climate change, the most up–to–date information for Europe and the European Union is based on data from the Copernicus Climate Change Service (C3S), which operates within the European Center for Medium–Range Weather Forecasting from Reading (ECMWF). The latest data released by C3S shows that globally, 2020 was on par with 2016, which has been rated as the warmest on record since 1851. For Europe, 2020 was the warmest year on record (since 1851), with a margin of 1.6 °C compared to the average of the reference period 1981‒2010, which is 0.4 °C above the previous warmest year in 2019. The fact that the 5 warmest years were registered in the last decade, 2011–2020 [
10] indicates significantly of the warming of Europe.
Regional differences regarding changes in precipitation are more pronounced than those regarding temperature. Certain regions and countries are becoming more arid, some are becoming more humid, while there are also those areas where dry and extremely rainy periods alternate with floods. Research indicates that for almost 120 years (1901–2019), central Europe and the Mediterranean region became drier, while northern Europe became wetter [
11]. Changes in temperature and precipitation, as well as more frequent and intense weather extremes, are also registered in the Mediterranean region, a region that is becoming warmer, but also drier [
12,
13,
14,
15,
16,
17].
In the area of the Western Balkans, a sub region to which Montenegro belongs, there has been a rise in temperature and more frequent extreme weather events, such as extreme temperatures, heat waves, droughts, and extreme precipitation, yet the changes in annual precipitation sums remain insignificant [
18]. In the area of neighboring Serbia, during the second half of the 20th and the beginning of the 21st century, an increase in temperature, a negative trend of precipitation and more frequent individual extreme events were recorded [
19,
20,
21,
22,
23,
24,
25,
26,
27]. The warming trend is also present in neighboring Bosnia and Herzegovina [
28,
29,
30]. Contemporary climate change and extreme weather events have not bypassed Montenegro, a small Mediterranean country. Over the past 20 years, records of temperature, precipitation and wind have been registered in the territory of Montenegro [
31,
32,
33].
The main purpose of the article is to identify changes in temperature and precipitation in the territory of Montenegro during 1961–2020. A large database was used, as we focused on daily extreme temperatures (maximum and minimum) as well as daily precipitation from 18 meteorological stations (MS). When the analysis is conducted on a daily basis, numerous climate indices for temperature and precipitation are used in the research of climate change. Generally, indices with fixed and variable thresholds are used, followed by those that demonstrate frequency and intensity. For the purposes of this article, we used indices that are calculated using the percentile distribution. Indices calculated using fixed thresholds, such as tropical days, summer days, frosty days, icy days, rainy days and others, do not give an objective picture of changes in these two most important climatic elements. For instance, summers in both the coastal region of Montenegro and in the area of the capital (Podgorica), have tropical days and tropical nights as a frequent occurrence (Tmax ≥ 30 °C, Tmin ≥ 20 °C). Conversely, these days are almost non–existent in the higher mountainous areas in the north of the country. Winters with frosty days (Tmin < 0 °C) are common in the north, while they are almost non–existent in the south of the country. Consequently, we focused on percentiles, for they have a significant advantage over indices with fixed thresholds. They are calculated according to the same empirical distribution; thus, variable thresholds are obtained in this way. For work purposes, percentages that indicate the frequency of extreme events (temperature and precipitation) were used. Trend calculations were made for eight percentile indices on a seasonal and annual level. Percentiles indicating potentially extreme events were selected—for maximum and minimum temperature of the 10th and 90th (Tx10p, Tn10p, Tx90p and Tn90p), and for precipitation of the 95th, 75th, 25th and 5th percentile. Therefore, the percentile method demonstrates the most realistic changes in temperature and precipitation, and this is the main contribution of the results presented in this article. In the context of contemporary climate change, no research was conducted in Montenegro; thus, the results presented in this paper will help decision makers to better understand this current issue.
2. Study Area
The area of research is Montenegro, a country located in Southeast Europe, i.e., on the Balkan Peninsula. The area of the country is 13,812 km
2, with a population of about 620,000 inhabitants. It extends to the Adriatic Sea for a length of about 100 km (air distance). In terms of relief, most of the country has a mountainous character. The coast of Montenegro is a narrow region along the Adriatic Sea, from a few meters wide to a maximum of 10 km. In the immediate vicinity of the Montenegrin coast, the Dinaric Mountains, with peaks up to 1894 m.a.s.l., rise. In the Central region, there is the Skadar lake basin (water level at about 6 m), the Zetsko–Bjelopavlićka plain (10–60 m), karst fields whose bottoms are at about 650 m (Cetinjsko, Nikšićko) and mountainous areas. In this region is located the two largest cities in Montenegro—Podgorica and Nikšić. The mountainous region occupies the northern half of the country, with peaks over 2000 m.a.s.l. Here, is the highest peak of Montenegro—the peak of Zla Kolat on Prokletije, with an altitude of 2534 m.a.s.l. (
Figure 1). Apart from relief fragmentation, the primary factors that influence the climate of Montenegro are the mathematical–geographical position, proximity to the sea and synoptic conditions, i.e., air masses within cyclones and anticyclones [
34].
Certainly, the great dissection of the terrain (0–2534 m.a.s.l.) and the unevenness of the relief conditions a true climatic mosaic. In general, based on the Köppen classification, two climates (C and D), three types (Cs, Cf and Df) and five subtypes of climate are distinguished in Montenegro: Csa, Csb, Cfb, Dfb and Dfc. For the period of 1961–1990, the average annual temperature along the coast and in the area of the Zetsko–Bjelopavlić lowland and Podgorica was from 14.8 °C to 15.8 °C. The average annual temperature in places up to the 600–700 m altitude was around 10 °C, and at around 1000 m altitude it was around 6–7 °C, while in the north of the country, at around 1500 m altitude, it was around 4.5 °C. Most of the country has a Mediterranean pluviometric regime, which means that the summers were dry, while the greatest amount of precipitation occurs during autumn and winter. Average annual precipitation varies widely. The town of Crkvice, located in the southwest of the country, at an altitude of 1000 m on the slopes of the coastal mountain Orjen, collects an average of 4600 mm of precipitation per year [
35]. On the other hand, Pljevlja, in the far north of the country, collects the least amount of precipitation in Montenegro; the annual average for the period of 1961–1990 was 802 mm [
36].
3. Database and Methods
For the purposes of the study, daily data on the maximum and minimum temperature and precipitation from 18 meteorological stations for the period 1961–2020 were used. The data were taken from the state Institute of Hydrometeorology and Seismology of Montenegro. The time series were complete, and only few data from 0.01% to 3% of the total number of data cases were missing. The only exceptions were three meteorological stations (MS) that started working later: Plav (start of work in 1966), Rožaje (1968) and Golubovci (1978). All data sets were tested for relative homogeneity using the MASH (Multiple Analysis of Series for Homogenization) method. The MASHv3.02 version was used, because it enables the quality control of daily data, as well as the estimation of missing data in the series [
37,
38,
39]. Those parts of the time series that were determined to be inhomogeneous were excluded from the series, which is only 0.3–1.0% of the data. Those data that were excluded (0.3–1.0%), as well as those that were missing (0.01–3%) and missing for the three mentioned MS, were estimated based on the data of the surrounding stations using MASHv3.02 methods. In this way, all data sets for the period of 1961–2020 were completed.
In order to more accurately detect temperature and precipitation extremes, the percentile method is used as one of the most relevant indicators of today’s climate change. To be precise, climate indices defined using percentile methods represent specific days, not with fixed but with variable thresholds. Percentile values range from 0 to 100. Percentiles indicate the position of data in a given increasing sequence, that is, how many percentage members of the observed sequence were above or below the observed value. For example, the 10th percentile indicates that 10% of the data in the observed time series has a value less than its absolute value. Compared to standardized deviations and other methods, the advantage of percentiles is that they more accurately identify rare (extreme) events. In particular, during certain seasons, it is impossible to analyze climate indices that were defined by means of fixed thresholds—e.g., there were no summer and tropical days and tropical nights (TR) in Montenegro during the winter, and there were no frosty and icy days in summer (in the north of the country, in the higher mountain areas, there is no TR even in summer).
Taking into account the above mentioned, for the purposes of this study, 8 percentile climate indices were analyzed—4 for temperature and 4 for precipitation (
Table 1). Temperature indices (Tx10p, Tx90p, Tn10p and Tn90p) were defined using the 10th and 90th percentiles of the maximum (Tx) and minimum (Tn) daily temperature. Climatic indices for precipitation were determined using the 95th, 75th, 25th and 5th percentiles of daily precipitation sums (Rd) of 1 mm or more. The percentile thresholds for the mentioned temperature and precipitation indices were calculated according to the rules proposed by the Expert Team on Climate Change Detection and Indices (ETCCDI), which is supported by the World Meteorological Organization (WMO) Commission on Climatology, the Joint Commission on Oceanography and Marine Meteorology (JCOMM) and the Research Program on Climate Variability and Predictability (CLIVAR) [
40,
41,
42].
The procedure for calculating thresholds for temperature indices is more complex than for precipitation percentile indices. The percentile thresholds for temperature were determined for each calendar day of the year for the standard climate period. For example, the threshold for Tn90p for January 10 was calculated based on 150 data, i.e., daily minimum temperatures of a 5–day window (8–12 January) for the standard climate period. When the threshold was calculated, days with a minimum temperature higher than the obtained value (90th percentile) in the period of 1961–2020 were counted. The same procedure was applied for each calendar day from 1 January to 31 December, for all stations included in the analysis. Thresholds for Tn10p, Tx10p and Tx90p were determined in the same way. This means that 365 (366) thresholds for one temperature percentile index were calculated for each MS.
The methodology for obtaining percentile thresholds for precipitation was simpler, because they were calculated from a sample of all days with a precipitation amount of 1 mm or more (Rd) for a given time unit (month, season, year). When the threshold was calculated for, e.g., the 75th percentile for the winter season (December, January and February), the counted days with precipitation were above that threshold in the mentioned season for the period of 1961–2020. The same procedure was applied for other seasons. At the annual level, the percentile threshold for precipitation was calculated based on the daily precipitation of 1 mm or more for all 12 months of the year for the standard climate period, and then the days with precipitation above the obtained value were counted in the observed 60–year period (1961–2020).
Climate indices defined using percentiles were calculated according to the same empirical distribution, which means that the obtained results can be compared from different parts of the world. It is considered to be the main advantage of the analysis of weather extremes that are determined using percentile thresholds. Secondly, in the analysis of extreme weather events, the importance of indices determined using percentile thresholds is also in the fact that they consider more moderate extremes and therefore instead of one with an absolute value, more potentially dangerous phenomena can be identified in the observed time unit (season, year).
For the purposes of this study, 5 time series were formed for one index (for seasons and on an annual level), that is, for all 8 percentile indices used, a total of 40 time series were formed for one MS. Next, a trend was calculated for each time series and its significance was examined. The Sen Method was used to calculate the trend, and the significance of the tendency was examined using the Mann–Kendall (MK) test at the risk level of 0.001, 0.01, 0.05 and 0.1, i.e., the degree of accuracy of the hypothesis of 99.9, 99, 95 and 90% [
43,
44,
45].
The main advantage of Sen’s slope estimates and MK test is that they are non–parametric methods, i.e., they are less demanding than parametric ones, since they take into account a smaller number of assumptions required for their implementation. Accordingly, these methods are often used to detect trends in meteorological and hydrological variables. The MK test is based on the calculation of variance (S).
The statistic
S is obtained by Equation [
45]:
The signum function (sgn) is calculated according to the formula:
where
n is the length of the sample,
xk and
xj are from
k = 1, 2, ...,
n–1 and
j =
k + 1, ...,
n. If
n is bigger than 8, statistic
S approximates to normal distribution. The mean of
S is 0 and the variance of
S can be acquired as follows:
The
Z statistic is calculated using the formula:
Sen’s method assumes that the trend is linear:
where:
Q—slope,
B—constant and
t—time.
The estimate of the slope of the trend (
Q) is calculated according to the formula:
The indices j and k denote a time instance (e.g., years) and j > k.
The significance of the trend is evaluated using the Z value obtained by the formula and the trial given in the two–sided test table for the level of: α = 0.10, 0.05, 0.01 and 0.001. A positive (negative) value of Z indicates an upward (downward) trend.
The trend calculations were tabulated. For the purpose of visualization, the spatial distribution of the trend of individual temperature and precipitation percentile indices was also shown. Additionally, the SURFER computer program was used for these purposes. This program contains a number of interpolation techniques. The most reliable method for dissected terrains has been the Kriging method. Using Kriging interpolation algorithms based on the data of all nearby stations, the daily value of the given percentile index was estimated in 100 × 100 m grids.
It should be mentioned that the World Meteorological Organization (WMO) recommended new climate standards to be used at the end of 2020, i.e., from the period of 1991–2020 [
46]. Respecting the recommendation of the WMO regarding the use of a new reference climate period (the previous base climate period was 1961–1990), for the purposes of this study, 1991–2020 was used as the standard or base period.
5. Discussion
The results of the calculation of four temperature percentile indices clearly indicated significant warming trend in Montenegro. The percentile temperature indices demonstrated the smallest changes in winter, which was not expected. Namely, under the conditions of the anthropogenic effect of the greenhouse, we should expect more intense warming at night and in winter than during the day and in summer. The opposite has been happening in Montenegro. Firstly, the maximum temperature rose faster than the minimum temperature (cold/warm days (Tx10 and Tx90) decrease/increase more intensively than cold/warm nights (Tn10 and Tn90)), in general. Secondly, changes in the frequency of the considered percentile indices, for both the maximum and minimum temperature, were more intense in summer than in winter. As for precipitation, there is a slight aridization, but the frequency of extremes has not changed significantly. The general view is that modern climate changes are occurring as a consequence of the anthropogenic greenhouse effect. On the other hand, some authors point out that changes in temperature, precipitation, cloudiness and other climatic elements can be partially or completely attributed to natural factors, such as variations in atmospheric and oceanic oscillations (teleconnections). It is recognized that the phenomenon of El Niño–Southern Oscillation (ENSO) has an impact on the global climate. Furthermore, it affects certain telecommunications, with the strongest intensity in the tropics of the Pacific and Indian Oceans. [
47,
48,
49,
50]. Some studies demonstrate that ENSO does affect European climate, e.g., [
51], and it was determined that there is a relationship between ENSO and the decadal variability of precipitation in Serbia [
52]. During 1951–2010, it was determined that there was a connection between the temperature change in Montenegro and several global and regional teleconnections, such as [
53]: East Atlantic Oscillation (EA), Arctic Oscillation (AO), Atlantic Multidecadal Oscillation (AMO), East Atlantic–West Russian Oscillation (EAWR), Scandinavian Pattern (SCAND) and North Atlantic Oscillation (NAO), specifically with the Mediterranean Oscillation (MO) and Western Mediterranean Oscillation (WeMO). The mentioned teleconnections, but also some others, such as the Summer North Atlantic Oscillation (SNAO), Polar–Eurasian Oscillation (POLEUR), North Sea–Caspian Pattern (NCP) and South Oscillation (SOI), affect the variability of cloudiness over Montenegro [
54].
At any rate, Montenegro is not released from extreme weather events. In the last two decades, the frequency and intensity of extreme weather and climate events related to temperature have increased. High temperatures and heat waves occur more often. Regarding precipitation, we mentioned that the frequency of extremes has not changed significantly. Though, it would be wrong to conclude that there are no changes in precipitation or that they are minor. In this study, not intensity indices but extreme frequency indices were analyzed. According to previous research, precipitation extremes have become more intense (extreme). For example, during 2010, Montenegro faced heavy rain and floods three times—in January, November and especially at the beginning of December. Likewise, we can mention the examples of extreme weather on the Montenegrin coast on 9 June 2018, then in November 2019 [
55]. The results for the two neighboring countries, Serbia and Croatia, based on the analysis of percentile and other climate indices, also demonstrate that in the period of 1961–2010, no significant changes were registered in the frequency of precipitation extremes [
56,
57].
Therefore, this study indicated that the frequency of extremely hot/cold weather has increased/decreased in Montenegro. There have been extreme daily precipitation-related events before (heavy rains, dry intervals, etc.), but their frequency has not significantly changed in recent times. However, in the last two decades, such phenomena have become more intense, i.e., more extreme (not more frequent, but more extreme). On the basis of the aforementioned, it could be assumed that there are different percentile indices (some show frequency, others intensity) and one should be careful when interpreting the obtained results [
58].
The results presented in this paper can assist decision makers to take the climate change issue seriously in Montenegro and develop adaptation and mitigation strategies. Strategies and plans for mitigating and adapting to contemporary climate change should be expertly planned and future research should be focused in that direction.
6. Conclusions
For the purposes of this work, the percentile method was applied to detect the tendency frequency of extreme temperatures and precipitation that occurred as a consequence of modern climate change in the territory of Montenegro. The research included data from 18 meteorological stations for the period of 1961–2020, and all calculations were made on a seasonal and annual level. A total of eight percentile indices were considered, four each for temperature and precipitation.
The obtained results indicated that the number of cold days and cold nights (Tx10p and Tn10p) was decreasing in the territory of Montenegro, and the number of warm days and warm nights (Tx90p and Tn90p) was increasing. In most cases, the trend of changes was statistically significant, both at the level of seasons and during the year. The decrease in the cold and the increase in the warm percentile indices was a clear indicator of the warming tendency of Montenegro. For the observed 60–year period (1961–2020), the most intense warming was registered in the warmer part of the year, especially during the summer season, when the trend of all considered temperature percentile indices was significant at the highest level of hypothesis acceptance (p < 0.001) in all MS. In summer, the tendency of the Tx10p and Tn10p indices was up to −6.9 and up to −7.1 days/decade, and Tx90p and Tn90p up to 1.7 and up to 2.0 days/decade.
For the observed 60–year period, the trend of changes of the frequency in all four precipitation indices (R95p, R75p, R25p and R5p) was minor and in most cases statistically insignificant. During the last two decades, the absolute daily, seasonal and annual rainfall records were registered in most of Montenegro for the instrumental period, but the frequency of extremes did not demonstrate significant changes in one direction or the other. In contrast to the temperature, where there was uniformity in terms of the sign of the trend in the entire area of Montenegro, the obtained results for precipitation indices did not demonstrate this. Namely, the same precipitation index had a positive trend in some regions of the country, and a negative trend in others. For example, the trend of the R75p index in winter ranged from −0.6 to 0.5 days/decade. As for the different signs of the trend, sometimes even in close places, it is plausible that they are a consequence of relief diversity (altitude and relief unevenness).