Examination of Long-Term Temperature Change in Türkiye: Comparative Evaluation of an Advanced Quartile-Based Approach and Traditional Trend Detection Methods

Abstract

The fact that 2023 and subsequently 2024 were the hottest years in history makes it even more important to monitor temperature changes over time. In this study, trends in the mean, maximum, and minimum temperature data of 81 provinces in Türkiye were examined using three traditional methods (Mann–Kendall, Linear Regression Analysis and Sen’s slope), one innovative method (ITA), and the QuarTrend (QT) method proposed in this study, which uses quartiles of the data series. The objectives of this research are (1) to determine and evaluate the long-term temperature trends in Türkiye (1960–2022) and (2) to comparatively evaluate the trend results of the proposed QT method, traditional trend detection methods, and ITA. In the study, a statistically significant (p < 0.05) increasing trend was found in the mean (0.027 °C/year), maximum (0.031 °C/year), and minimum (0.038 °C/year) annual temperatures of Türkiye. While traditional trend tests detected similar trends with ITA and QT for mean temperatures; ITA and QT detected more trends than traditional methods for maximum and minimum temperatures. The results have direct implications for the impacts of climate change in the study region. The results have the potential to support the development of climate-resilient and adaptive policies for effective water resource planning and management to sustain the environment and agricultural productivity in Türkiye.

1. Introduction

During the first quarter of the 21st century, population growth, unplanned urbanization, land use, agricultural activities, technological advancements, and the energy expended for all of these activities have led to the absorption of all waste and residue into the atmosphere in gaseous form. As a result, the chemical structure of the troposphere began to change, and short-wave extraterrestrial radiation became unable to escape back into the atmosphere. Consequently, the accumulation of various gases, particularly carbon dioxide, has led to warming in the troposphere. The IPCC’s fourth and fifth assessment reports [1,2] explain in detail how these emissions contribute to global warming. The terms global warming and climate change are used interchangeably for the average temperature increase in the Earth’s climate system. According to the IPCC’s fifth assessment report [2], scientists are more than 95% confident that global warming is the result of greenhouse gas emissions caused by various human-related activities. It is also stated that in the 21st century, even if emissions are kept at their lowest limits, the global surface temperature is expected to increase by an additional 0.3 to 1.7 °C. The maximum increase limits are predicted to be between 2.6 and 4.8 °C. A study of the Climatic Research Unit (CRU) at East Anglia University examining temperature anomalies since 1850 showed that 2024 is the hottest year on record (Figure 1). Various research teams have used different methods with multiple datasets to estimate global temperature change from millions of individual temperature records, but all come to an agreement that 2024 was the warmest year on record. The fact that different datasets and methodologies (HadCRUT5, Berkeley, NOAA, GISTEMP, Copernicus, JRA-3Q) yield similar results shows that the results are not sensitive to a particular method. Although the outputs of different teams (Figure 1) show slight differences (usually the GisTEMP curve is at the bottom and the Berkeley curve is at the top), the common result of all is a noticeable linear trend, specifically after 1980 [3,4].

NASA reported that 2024 was the hottest year on record, concluding that the warming trend in recent decades is due to heat-trapping carbon dioxide, methane, and other greenhouse gases. The world experienced record increases in carbon dioxide emissions from fossil fuels in 2022 and 2023. The concentration of carbon dioxide in the atmosphere has risen from pre-industrial levels in the 18th century (about 278 parts per million) to about 420 parts per million today. NASA scientists showed that the average surface temperature of the earth in 2024 was the highest on the historical records. Global temperatures in 2024 were 1.28 °C above NASA’s 20th century baseline, beating the previous record set in 2023. A new record came after 15 successive months (from June 2023 to August 2024) marking an “unprecedented temperature streak.” NASA Administrator Bill Nelson indicated that 2024 was the warmest year since record keeping started in 1880. The scientists of NASA estimate Earth in 2024 was about 1.47 °C warmer than the mid-19th century average (1850–1900). In a large time frame of 2024, the average temperatures were more than 1.5 °C above the reference point. The annual average temperature may have exceeded this level for the first time. This number (1.5 °C) is important since 195 nations agreed to limit global warming below this number in the 2015 Paris Agreement [5].

The most recent assessment report of the IPCC (AR6) has underscored with high confidence that human influence has warmed the climate at a rate unprecedented in at least the last 2000 years, with observed increases in the frequency and intensity of hot extremes across most land regions. The Mediterranean region, which includes Türkiye, has been identified as a climate change hotspot, projected to experience warming at a rate exceeding the global average [6].

The conditions and findings mentioned above increase the importance of examining temperature changes. Scientists have conducted studies on temperature trends for various regions of the world, especially in the last two decades. For example, trend analyses including temperature data were carried out by [7] for Canada using monthly and annual temperature data with the Mann–Kendall (MK) and the discrete wavelet transform (DWT) method; by [8] for Serbia using seven meteorological datasets (temperature, relative humidity, vapor pressure, etc.) at seasonal and annual scales with MK and Sen’s (SSE) methods; by [9] for Ethiopia using seasonal and annual timescale temperature and rainfall data with linear regression analysis (LRA); by [10] for South China using annual temperature and rainfall data with the MK method; by [11] for the Syr Darya Basin in Central Asia using annual timescale temperature and rainfall with the MK method; by [12] for East India using seasonal and annual temperature and rainfall data with MK, SSE, LRA, and Spearman’s rank (SRA) tests; by [13] for Iraq using temperature data at seasonal and annual timescales with the MK method; by [14] for the eastern Hindu Kush in northern Pakistan using annual temperature data with the MK and SSE; by [15] for 18 selected megacities from six continents using seasonal and annual temperature and rainfall with the MK and LRA methods; by [16] for Iran using annual and seasonal temperature data with the MK method; by [17] for Western India using annual and seasonal temperature and rainfall data with the MK, SRA, and innovative trend analysis (ITA) methods; by [18] for the southwestern United States using seasonal and annual timescale temperature and rainfall data with MK and SSE methods; by [19] for Nigeria using seasonal and annual temperature and rainfall data with MK and SSE methods; by [20] for Türkiye using seasonal and annual timescale temperature data with MK and LRA methods; by [21] for the Black Sea Coast of Romania using monthly, seasonal, and annual timescale temperature data with MK, LRA, and SSE methods; by [22] for the Soan River Basin in Western Pakistan using monthly temperature data with ITA; by [23] for Thailand using annual temperature, relative humidity, and rainfall data with MK and ITA methods; by [24] for Morocco using monthly and seasonal temperature data with MK, SSE, and ITA methods.

In almost all of the above studies, which include mean, maximum, or minimum temperature records at different timescales, the presence of a positive trend is significant.

In trend analysis, traditional methods (Mann–Kendall, Linear Regression analysis and Sen’s slope) and the relatively new ITA (innovative trend analysis) have been frequently used in the literature. In order to use traditional methods in trend analysis, the independence and the normality of the time series and sufficient data length must be provided. Unfortunately, in practice, assumptions regarding independence, normality and data length of the series are often ignored. Although in some studies trends are detected after eliminating serial dependency by pre-whitening [25], some researchers [26,27,28] do not recommend this because pre-whitening has the risk of removing some portion of the trend. Innovative ITA has the advantage of examining non-monotonic trends in “low”, “medium”, and “high” data regions of a data series compared to traditional methods. As for ITA, although it is often claimed in the literature that ITA does not rely on any assumptions, the numerical trend test of ITA is based on the assumption that the variances of the two half-series are the same. On the other hand, refs. [29,30] showed that the inflated variance of the ITA slope estimator can cause a Type I error (detecting a nonexistent trend) due to the persistent structure of the dependent series. Therefore, in almost all previous studies, ITA detected trends in more series than other trend tests [31,32,33,34,35,36,37,38,39,40,41,42,43,44,45].

Our motivation in this study is to examine the temperature changes in Türkiye as well as to propose a new trend detection approach which avoids the normality and independence assumptions in traditional trend analysis methods and the Type I error resulting from the inflated variance of the ITA slope estimator. By proposing and validating the QuarTrend method, this research aims to enhance the toolkit for detecting climate trends, particularly in series with non-normal distributions or extreme values, which are characteristic of a warming climate. The proposed method makes it almost impossible to detect a false trend because it compares all quartiles corresponding to different probabilities (25%, 50%, 75%) within the sub-series. Furthermore, quartiles are more robust to outliers and can reduce the interference of highly skewed data.

In this context, a comparative analysis of trend methodologies was conducted for the mean, maximum, and minimum annual temperature records of 81 provinces in Türkiye over the past six decades. Temperature trends were examined by traditional methods, ITA, and QT methods. The results were discussed, compared with temperature trends in the region and the world, and the contribution and applicability of the proposed quarter-based trend approach was evaluated.

2. Materials and Methods

2.1. Study Area and Data

Türkiye is situated in Northern Hemisphere between the temperate and subtropical zones, and between 36° and 42° N and 26° and 45° E (Figure 2). It is a transcontinental country, straddling both Europe and Asia. The country is at the crossroads of the East and West, bordered by eight countries and surrounded by three different seas: the Aegean Sea to the west, the Mediterranean Sea to the south, and the Black Sea to the north. Türkiye is largely characterized by high, mountainous terrains and vast plateaus. The Anatolian Plateau lies in the central part of the country, surrounded by mountain ranges like the Taurus Mountains in the south and the Northern Anatolian Mountains in the north. The Eastern Anatolia region is particularly mountainous. The fact that Türkiye is surrounded by seas on three sides, the extension of the mountains, and the diversity of landforms have led to the emergence of different climate types. The country experiences several different climate zones, with significant variations in temperature, rainfall, and humidity across regions. The Mediterranean climate is most common along the southern and southwestern coasts. Summers are hot and dry, with summer temperatures frequently exceeding 30 °C; winters are mild and rainy. The Oceanic climate is most common along the northern coasts. Summers are cool and relatively rainy. Winters are mild and rainy and average temperatures rarely drop below freezing. This climate has more consistent precipitation throughout the year, especially in winter. The Continental climate is dominant in the interior, especially on the Central Anatolian Plateau. Summers are hot and dry, with summer temperatures often exceeding 30 °C. Winters are very cold with snowfall. The temperature difference between summer and winter is extreme, with hot days in summer and freezing temperatures in winter. The Mountain climate is common in high altitude areas, such as the Eastern Anatolian Plateaus. Winters have heavy snowfall and summers are cool. These areas receive significant amounts of precipitation, and temperatures can vary greatly depending on altitude.

The study focused on the trends of temperatures in Türkiye. Trend analyses were applied to mean, maximum, and minimum annual temperature data measured in 81 city centers of the country. Temperature data was provided by the Turkish State Meteorological Service (MGM), which subjects its records to a rigorous, multi-stage quality control (QC) procedure as per their official protocols. This process involves six automated tests applied to the data at the end of each month: a validity test against station-specific climatological limits, a step test for plausible changes between consecutive measurements, a persistence test to identify faulty sensors, internal logic checks, a comparison with similar instruments, and a field test against neighboring stations. Data are flagged based on these checks, and only values ultimately classified as ‘Good’ after expert verification are incorporated into the official archives. Furthermore, to ensure the reliability of long-term trend analysis by minimizing non-climatic biases from factors like station relocations or instrument changes, we utilized MGM’s homogenized annual temperature series. These series are processed using standard relative homogenization techniques, thereby providing a robust foundation for detecting statistically reliable climatic trends [46].

The spatial representativeness of the 81 stations is high, as they are distributed across all seven geographical regions of Türkiye, capturing the diverse climatic zones from coastal to continental and high-altitude environments (Figure 2). Stations located in major urban centers (e.g., Istanbul, Ankara, Izmir) were included as they are essential for understanding climate impacts on population centers; however, the pervasive nature of warming trends across both urban and rural stations (as shown in our results) indicates that the observed signals are dominated by regional climate change rather than localized urban heat island effects alone.

The geographic information, statistical parameters and observation periods of the meteorological data are given in Appendix A. T_mean, T_max and T_min are the mean, maximum and minimum annual temperatures, respectively. $\overline{x}$ and C_s are the average and skewness coefficients of the series. The averages of T_mean series vary between 4 °C and 19 °C, the averages of T_max series vary between 31 °C and 43 °C, and the averages of T_min series vary between 2 °C and −30 °C for 81 meteorological stations of Türkiye. The skewness (C_s) of the T_max series are mostly positive (skewed to the right), while the skewness of the T_min series are mostly negative (skewed to the left), as expected. Most stations have at least 60 years of data, although a few stations have shorter observations.

2.2. Methodology

The trends were investigated for the mean, maximum and minimum temperature data using the MK trend test, LRA, Sen’s slope, ITA and an advanced quartile-based trend detection method.

2.2.1. Mann–Kendall Test

The Mann–Kendall test is one of the most frequently applied methods for identifying trends in climatic and hydro-meteorological time series (e.g., [38,39,42,43,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64]). It is a nonparametric test for determining whether monotonic trends exist in time series. The null hypothesis (H₀) assumes there is no trend and that the data are independent and identically distributed. The alternative hypothesis (H₁) proposes the presence of a trend—either increasing or decreasing. The Mann–Kendall statistic (S) is computed to reveal the trend direction.

$$S={\sum }_{i=1}^{n-1}{\sum }_{j=i+1}^{n}sgn(x_j-x_i),$$

(1)

$$sgn\left(x_j-x_i\right)=\left\{\begin{array}{c}+1, if \left(x_j-x_i\right)>0\\ 0, if \left(x_j-x_i\right)=0\\ -1, if \left(x_i-x_j\right)>0\end{array}\right.,$$

(2)

In the equations, n represents the total number of data, and x_i, x_j are the observed data at times i and j, respectively (j > i). The sign (sgn) function is computed with Equation (2). A positive value of S suggests an increasing (positive) trend, while a negative value of S indicates a decreasing (negative) trend.

When ties occur among data values and the dataset contains more than 10 elements, the variance is estimated using the following formula, assuming a normal distribution:

$$Var\left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-\sum _{i=1}^P{t}_i(t_i-1)({2t}_i+5)}{18}},$$

(3)

where P denotes the number of tied groups and t_i the number of data points in the i-th tied group. The standardized Z value is then calculated as

$$Z=\left\{\begin{array}{c}{\displaystyle \frac{S-1}{\sqrt{Var\left(S\right)}}} ; If S>0\\ 0 ; If S=0\\ {\displaystyle \frac{S-1}{\sqrt{Var\left(S\right)}}} ; If S<0\end{array}\right..$$

(4)

2.2.2. Linear Regression Analysis (LRA)

LRA, a parametric method, has also been extensively employed in trend studies of hydro-meteorological and climatic data [31,33,38,41,44,48,52,65,66,67,68,69,70,71,72]. LRA is based on the assumption that the data in each group (Y, the data and X, the time) follow a normal distribution. The following formula is used to determine the regression coefficient:

$$R={\displaystyle \frac{\sum (X-\overline{X})(Y-\overline{Y})}{\sqrt{\sum {\left(X-\overline{X}\right)}^2{\left(Y-\overline{Y}\right)}^2}}}.$$

(5)

With n − 2 degrees of freedom, t statistics were used to determine the regression’s significance as follows:

$$t={\displaystyle \frac{R(n-2)}{\sqrt{1-R^2}}},$$

(6)

Here n represents the size of the dataset. If the calculated t exceeds the critical value (t_α) at a chosen significance level α, the trend is considered statistically significant. A positive value of t indicates an increasing (positive) trend, whereas a negative value of t indicates a decreasing (negative) trend.

2.2.3. Sen’s Slope Test

Sen’s slope Test is a non-parametric method for estimating the magnitude of a trend in a time series. It does not assume any probability distribution of the series and it is robust to the outliers [27,73,74]. The slope (Q) between all data pairs is calculated as:

$$Q=median\left({\displaystyle \frac{x_j-x_i}{(j-i)}}\right) for all i<j$$

(7)

where x_i and x_j are the data at times i and j, respectively. The median of these slopes represents the overall trend magnitude. A positive Q shows an increasing (positive) trend, while a negative Q shows a decreasing (negative) trend.

2.2.4. Innovative Trend Analysis (ITA)

Innovative Trend Analysis (ITA) is a graphical and statistical method used to identify trends in a time series by dividing the series into two equal halves and plotting the first half against the second half. The data pairs are plotted on a 1:1 line, and deviations from this line indicate the presence and direction of a trend [3,36,75]. The trend slope is calculated as:

$$Trend Slope={\displaystyle \frac{2({\overline{y}}_2-{\overline{y}}_1)}{n}}$$

(8)

where $\overline{y}$₁ and $\overline{y}$₂ are the means of the first and second halves of the dataset, respectively, and n is the total number of observations. ITA is particularly useful for identifying non-monotonic trends and trends in datasets with non-normal distributions.

2.2.5. Quartile Based Trend (QuarTrend) Analysis

Recent studies have shown that although ITA provides some advantages over classical trend tests, it can also detect a trend that does not actually exist (Type I error) in some cases. For this reason, as stated in the Introduction section, in almost all comparative trend analysis studies in the literature, ITA has quantitatively detected more trends than classical trend tests. The fact that classical trend tests are based on assumptions such as normality and independence, and the drawbacks of ITA explained above, have led us to a new trend analysis approach. In the presented quartile-based trend (QuarTrend) analysis methodology, the data will be divided into two chronological subgroups (the first half, FH and second half, SH) and the trend will be identified by comparing the quartiles of these two groups. The decision-making mechanism of quartile-based trend analysis is as follows:

• Increasing Trend: If Q₁(SH) > Q₁(FH), Q₂(SH) > Q₂(FH), and Q₃(SH) > Q₃(FH);
• Decreasing Trend: If Q₁(SH) < Q₁(FH), Q₂(SH) < Q₂(FH), and Q₃(SH) < Q₃(FH);
• No Trend: If all three quartiles in one of the two half series are not greater than those in the other,

where Q₁, Q₂, and Q₃ represent the 1st, 2nd, and 3rd quartiles, respectively.

3. Results

3.1. Trend Analysis of Temperature Series in Türkiye

Trends of mean, maximum, and minimum annual temperature data at 81 meteorological stations in Türkiye were analyzed using classical methods (i.e., the MK, LRA, and Sen’s slope tests), the ITA, and the QuarTrend (QT). The results are summarized in

Table 1. The trends of mean annual temperature in Türkiye.

Nr	Station	Slope (Sen’s)	Slope (Linear)	LRA	Sen’s Slope	MK	ITA	QT
1	Adana	0.017	0.017	↑	↑	↑	↑	↑
2	Adıyaman	0.032	0.032	↑	↑	↑	↑	↑
3	Afyon	0.031	0.029	↑	↑	↑	↑	↑
4	Ağrı	0.050	0.064	↑	↑	↑	↑	↑
5	Aksaray	0.045	0.041	↑	↑	↑	↑	↑
6	Amasya	0.021	0.020	↑	↑	↑	↑	↑
7	Ankara	0.035	0.034	↑	↑	↑	↑	↑
8	Antalya	0.016	0.016	↑	↑	↑	↑	↑
9	Ardahan	0.040	0.037	↑	↑	↑	↑	↑
10	Artvin	0.022	0.019	↑	↑	↑	↑	↑
11	Aydın	0.028	0.026	↑	↑	↑	↑	↑
12	Balıkesir	0.046	0.046	↑	↑	↑	↑	↑
13	Bartın	0.024	0.022	↑	↑	↑	↑	↑
14	Batman	0.025	0.024	↑	↑	↑	↑	↑
15	Bayburt	0.038	0.035	↑	↑	↑	↑	↑
16	Bilecik	0.030	0.028	↑	↑	↑	↑	↑
17	Bingöl	0.023	0.020	↑	↑	↑	↑	↑
18	Bitlis	0.027	0.019	↑	↑	↑	↑	↑
19	Bolu	0.024	0.023	↑	↑	↑	↑	↑
20	Burdur	0.023	0.020	↑	↑	↑	↑	↑
21	Bursa	0.027	0.026	↑	↑	↑	↑	↑
22	Çanakkale	0.031	0.031	↑	↑	↑	↑	↑
23	Çankırı	0.023	0.022	↑	↑	↑	↑	↑
24	Çorum	0.023	0.023	↑	↑	↑	↑	↑
25	Denizli	0.044	0.041	↑	↑	↑	↑	↑
26	Diyarbakır	0.010	0.010	–	–	–	↑	↑
27	Düzce	0.031	0.031	↑	↑	↑	↑	↑
28	Edirne	0.033	0.032	↑	↑	↑	↑	↑
29	Elazığ	0.031	0.027	↑	↑	↑	↑	↑
30	Erzincan	0.041	0.037	↑	↑	↑	↑	↑
31	Erzurum	0.001	0.000	–	–	–	↓	↓
32	Eskişehir	0.032	0.030	↑	↑	↑	↑	↑
33	Gaziantep	0.043	0.040	↑	↑	↑	↑	↑
34	Giresun	0.027	0.027	↑	↑	↑	↑	↑
35	Gümüşhane	0.022	0.023	↑	↑	↑	↑	↑
36	Hakkari	0.042	0.046	↑	↑	↑	↑	↑
37	Hatay	0.017	0.015	↑	↑	↑	↑	↑
38	Iğdır	0.043	0.042	↑	↑	↑	↑	↑
39	Isparta	0.024	0.022	↑	↑	↑	↑	↑
40	İçel	0.054	0.052	↑	↑	↑	↑	↑
41	İstanbul	0.029	0.028	↑	↑	↑	↑	↑
42	İzmir	0.027	0.026	↑	↑	↑	↑	↑
43	Karabük	0.016	0.019	–	–	–	↑	↑
44	Karaman	0.034	0.034	↑	↑	↑	↑	↑
45	Kars	0.040	0.038	↑	↑	↑	↑	↑
46	Kastamonu	0.019	0.017	↑	↑	↑	↑	↑
47	Kayseri	0.031	0.027	↑	↑	↑	↑	↑
48	Kilis	0.030	0.028	↑	↑	↑	↑	↑
49	Kırıkkale	0.035	0.034	↑	↑	↑	↑	↑
50	Kırklareli	0.033	0.033	↑	↑	↑	↑	↑
51	Kırşehir	0.026	0.026	↑	↑	↑	↑	↑
52	Kocaeli	0.030	0.029	↑	↑	↑	↑	↑
53	Konya	0.020	0.020	↑	↑	↑	↑	↑
54	Kütahya	0.030	0.027	↑	↑	↑	↑	↑
55	Malatya	0.038	0.034	↑	↑	↑	↑	↑
56	Manisa	0.020	0.019	↑	↑	↑	↑	↑
57	Maraş	0.036	0.034	↑	↑	↑	↑	↑
58	Mardin	0.037	0.037	↑	↑	↑	↑	↑
59	Muğla	0.018	0.017	↑	↑	↑	↑	↑
60	Muş	0.052	0.051	↑	↑	↑	↑	↑
61	Nevşehir	0.029	0.027	↑	↑	↑	↑	↑
62	Niğde	0.038	0.034	↑	↑	↑	↑	↑
63	Ordu	0.039	0.037	↑	↑	↑	↑	↑
64	Osmaniye	0.037	0.029	↑	↑	↑	↑	↑
65	Rize	0.032	0.030	↑	↑	↑	↑	↑
66	Sakarya	0.037	0.039	↑	↑	↑	↑	↑
67	Samsun	0.025	0.023	↑	↑	↑	↑	↑
68	Siirt	0.029	0.030	↑	↑	↑	↑	↑
69	Sinop	0.026	0.026	↑	↑	↑	↑	↑
70	Şırnak	0.038	0.044	↑	↑	↑	↑	↑
71	Sivas	0.033	0.031	↑	↑	↑	↑	↑
72	Tekirdağ	0.027	0.027	↑	↑	↑	↑	↑
73	Tokat	0.031	0.027	↑	↑	↑	↑	↑
74	Trabzon	0.024	0.023	↑	↑	↑	↑	↑
75	Tunceli	0.032	0.030	↑	↑	↑	↑	↑
76	Şanlıurfa	0.034	0.032	↑	↑	↑	↑	↑
77	Uşak	0.025	0.025	↑	↑	↑	↑	↑
78	Van	0.048	0.043	↑	↑	↑	↑	↑
79	Yalova	0.030	0.030	↑	↑	↑	↑	↑
80	Yozgat	0.030	0.026	↑	↑	↑	↑	↑
81	Zonguldak	0.016	0.017	↑	↑	↑	↑	↑
Increasing Trend (%) ⟶				96	96	96	100	100

,

Table 2. The trends of maximum annual temperature in Türkiye.

Nr	Station	Slope (Sen’s)	Slope (Linear)	LRA	Sen’s Slope	MK	ITA	QT
1	Adana	0.020	0.023	–	–	–	↑	↑
2	Adıyaman	0.034	0.034	↑	↑	↑	↑	↑
3	Afyon	0.014	0.017	–	–	–	↑	↑
4	Ağrı	0.048	0.046	↑	↑	↑	↑	↑
5	Aksaray	0.027	0.026	↑	↑	↑	↑	↑
6	Amasya	0.040	0.040	↑	↑	↑	↑	↑
7	Ankara	0.032	0.034	↑	↑	↑	↑	↑
8	Antalya	0.017	0.023	↑	↑	↑	↑	↑
9	Ardahan	0.011	0.018	–	–	–	↑	–
10	Artvin	0.027	0.018	–	–	–	↑	↑
11	Aydın	0.038	0.041	↑	↑	↑	↑	↑
12	Balıkesir	0.002	0.005	–	–	–	↑	–
13	Bartın	0.028	0.022	–	–	–	↑	–
14	Batman	0.006	0.005	–	–	–	–	–
15	Bayburt	0.040	0.041	↑	↑	↑	↑	↑
16	Bilecik	0.044	0.040	↑	↑	↑	↑	↑
17	Bingöl	0.003	0.001	–	–	–	↑	–
18	Bitlis	0.031	0.032	↑	↑	↑	↑	↑
19	Bolu	0.015	0.015	–	–	–	↑	↑
20	Burdur	0.047	0.047	↑	↑	↑	↑	↑
21	Bursa	0.005	0.005	–	–	–	↑	–
22	Çanakkale	0.031	0.035	↑	↑	↑	↑	↑
23	Çankırı	0.052	0.052	↑	↑	↑	↑	↑
24	Çorum	0.047	0.047	↑	↑	↑	↑	↑
25	Denizli	0.050	0.054	↑	↑	↑	↑	↑
26	Diyarbakır	0.006	0.005	–	–	–	–	–
27	Düzce	0.007	0.005	–	–	–	↑	↑
28	Edirne	0.045	0.046	↑	↑	↑	↑	↑
29	Elazığ	0.029	0.026	↑	↑	↑	↑	↑
30	Erzincan	0.038	0.034	↑	↑	↑	↑	↑
31	Erzurum	0.043	0.043	↑	↑	↑	↑	↑
32	Eskişehir	0.025	0.026	↑	↑	↑	↑	↑
33	Gaziantep	0.030	0.030	↑	↑	↑	↑	↑
34	Giresun	0.018	0.013	–	–	–	↑	–
35	Gümüşhane	0.022	0.022	–	–	–	↑	↑
36	Hakkari	0.015	0.016	–	–	–	↑	↑
37	Hatay	0.037	0.035	↑	–	–	↑	↑
38	Iğdır	0.041	0.040	↑	↑	↑	↑	↑
39	Isparta	0.032	0.033	↑	↑	↑	↑	↑
40	İçel	0.040	0.038	↑	↑	↑	↑	↑
41	İstanbul	0.028	0.028	↑	↑	↑	↑	↑
42	İzmir	0.014	0.016	–	–	–	↑	↑
43	Karabük	0.042	0.043	–	–	–	↑	↑
44	Karaman	0.011	0.013	–	–	–	↑	↑
45	Kars	0.038	0.039	↑	↑	↑	↑	↑
46	Kastamonu	0.047	0.049	↑	↑	↑	↑	↑
47	Kayseri	0.026	0.024	↑	↑	↑	↑	↑
48	Kilis	0.014	0.016	–	–	–	↑	↑
49	Kırıkkale	0.056	0.054	↑	↑	↑	↑	↑
50	Kırklareli	0.027	0.028	–	–	–	↑	↑
51	Kırşehir	0.055	0.053	↑	↑	↑	↑	↑
52	Kocaeli	0.012	0.012	–	–	–	↑	–
53	Konya	0.022	0.023	↑	↑	↑	↑	↑
54	Kütahya	0.038	0.039	↑	↑	↑	↑	↑
55	Malatya	0.048	0.045	↑	↑	↑	↑	↑
56	Manisa	0.022	0.023	–	–	–	↑	↑
57	Maraş	0.043	0.046	↑	↑	↑	↑	↑
58	Mardin	0.030	0.032	↑	↑	↑	↑	↑
59	Muğla	0.040	0.040	↑	↑	↑	↑	↑
60	Muş	0.034	0.036	↑	↑	↑	↑	↑
61	Nevşehir	0.035	0.035	↑	↑	↑	↑	↑
62	Niğde	0.040	0.037	↑	↑	↑	↑	↑
63	Ordu	0.038	0.044	↑	↑	↑	↑	↑
64	Osmaniye	0.018	0.022	–	–	–	↑	–
65	Rize	0.043	0.042	↑	↑	↑	↑	↑
66	Sakarya	0.024	0.023	–	–	–	↑	↑
67	Samsun	0.000	−0.004	–	–	–	↑	–
68	Siirt	0.011	0.009	–	–	–	–	–
69	Sinop	0.039	0.046	↑	↑	↑	↑	↑
70	Şırnak	0.029	0.026	↑	↑	↑	↑	↑
71	Sivas	0.057	0.053	↑	↑	↑	↑	↑
72	Tekirdağ	0.040	0.047	↑	↑	↑	↑	↑
73	Tokat	0.050	0.045	↑	↑	↑	↑	↑
74	Trabzon	0.033	0.033	↑	↑	↑	↑	↑
75	Tunceli	0.020	0.019	–	↑	↑	↑	↑
76	Şanlıurfa	0.029	0.027	↑	↑	↑	↑	↑
77	Uşak	0.025	0.027	↑	↑	↑	↑	↑
78	Van	0.025	0.020	–	↑	↑	–	–
79	Yalova	0.035	0.037	↑	↑	↑	↑	↑
80	Yozgat	0.053	0.052	↑	↑	↑	↑	↑
81	Zonguldak	0.009	0.012	–	–	–	↑	↑
Increasing Trend (%) ⟶				64	63	63	96	84

and

Table 3. The trends of minimum annual temperature in Türkiye.

Nr	Station	Slope (Sen’s)	Slope (Linear)	LRA	Sen’s Slope	MK	ITA	QT
1	Adana	0.018	0.020	–	–	–	↑	–
2	Adıyaman	0.010	0.014	–	–	–	↑	↑
3	Afyon	0.056	0.063	↑	↑	↑	↑	↑
4	Ağrı	0.126	0.124	↑	↑	↑	↑	↑
5	Aksaray	0.032	0.039	–	–	–	↑	↑
6	Amasya	−0.016	−0.017	–	–	–	–	–
7	Ankara	0.084	0.083	↑	↑	↑	↑	↑
8	Antalya	0.000	−0.003	–	–	–	↑	–
9	Ardahan	0.087	0.092	↑	↑	↑	↑	↑
10	Artvin	0.029	0.030	–	–	–	↑	–
11	Aydın	0.035	0.034	↑	↑	↑	↑	↑
12	Balıkesir	0.080	0.086	↑	↑	↑	↑	↑
13	Bartın	0.060	0.057	↑	↑	↑	↑	↑
14	Batman	−0.014	−0.017	–	–	–	↓	–
15	Bayburt	0.080	0.062	↑	↑	↑	↑	↑
16	Bilecik	0.044	0.047	↑	↑	↑	↑	↑
17	Bingöl	0.073	0.061	↑	↑	↑	↑	–
18	Bitlis	0.043	0.049	↑	↑	↑	↑	↑
19	Bolu	0.126	0.127	↑	↑	↑	↑	↑
20	Burdur	−0.010	−0.008	–	–	–	↑	–
21	Bursa	0.077	0.090	↑	↑	↑	↑	↑
22	Çanakkale	0.049	0.044	↑	↑	↑	↑	↑
23	Çankırı	0.042	0.041	–	–	–	↑	↑
24	Çorum	0.033	0.032	–	–	–	↑	–
25	Düzce	0.100	0.093	↑	↑	↑	↑	↑
26	Denizli	0.053	0.050	↑	↑	↑	↑	↑
27	Diyarbakır	−0.015	−0.014	–	–	–	↓	↓
28	Edirne	0.042	0.051	↑	↑	↑	↑	↑
29	Elazığ	0.040	0.039	–	–	–	↑	–
30	Erzincan	0.112	0.092	↑	↑	↑	↑	↑
31	Erzurum	−0.151	−0.146	↓	↓	↓	↓	↓
32	Eskişehir	0.081	0.076	↑	↑	↑	↑	↑
33	Gaziantep	0.080	0.086	↑	↑	↑	↑	↑
34	Giresun	0.034	0.041	↑	↑	↑	↑	↑
35	Gümüşhane	0.032	0.020	–	–	–	↑	↑
36	Hakkari	0.043	0.042	–	–	–	↑	↑
37	Hatay	0.008	0.006	–	–	–	↑	↑
38	Iğdır	0.072	0.078	–	↑	–	↑	↑
39	Isparta	0.006	0.015	–	–	–	↑	–
40	İçel	0.093	0.094	↑	↑	↑	↑	↑
41	İstanbul	0.033	0.042	↑	↑	↑	↑	↑
42	İzmir	0.040	0.037	↑	↑	↑	↑	↑
43	Karabük	0.011	0.013	–	–	–	↑	–
44	Karaman	0.036	0.033	–	–	–	↑	↑
45	Kars	0.069	0.070	↑	↑	↑	↑	↑
46	Kastamonu	0.050	0.053	↑	↑	↑	↑	↑
47	Kayseri	0.139	0.127	↑	↑	↑	↑	↑
48	Kilis	0.057	0.066	↑	↑	↑	↑	↑
49	Kırıkkale	0.020	0.046	–	–	–	↑	↑
50	Kırklareli	0.043	0.040	–	–	–	↑	–
51	Kırşehir	0.006	0.011	–	–	–	↑	–
52	Kocaeli	0.047	0.042	↑	↑	↑	↑	↑
53	Konya	−0.022	−0.013	–	–	–	↓	↓
54	Kütahya	0.046	0.045	–	↑	–	↑	↑
55	Malatya	0.028	0.027	–	–	–	↑	–
56	Manisa	0.031	0.033	↑	↑	↑	↑	↑
57	Maraş	0.027	0.024	–	–	–	↑	↑
58	Mardin	0.023	0.024	–	–	–	↑	–
59	Muğla	0.021	0.023	–	–	–	↑	↑
60	Muş	0.080	0.080	–	–	–	↑	↑
61	Nevşehir	0.034	0.035	–	–	–	↑	↑
62	Niğde	0.033	0.029	–	–	–	↑	–
63	Ordu	0.036	0.033	↑	↑	↑	↑	↑
64	Osmaniye	0.033	0.033	↑	↑	↑	↑	↑
65	Rize	0.020	0.021	–	–	–	↑	–
66	Sakarya	0.076	0.091	↑	↑	↑	↑	↑
67	Samsun	0.012	0.017	–	–	–	↑	–
68	Siirt	0.031	0.033	–	–	–	↑	↑
69	Sinop	0.033	0.035	↑	↑	↑	↑	↑
70	Şırnak	0.007	0.005	–	–	–	↑	↑
71	Sivas	0.076	0.063	↑	↑	↑	↑	↑
72	Tekirdağ	0.056	0.050	↑	↑	↑	↑	↑
73	Tokat	0.000	−0.004	–	–	–	↓	↓
74	Trabzon	0.015	0.015	–	–	–	↑	–
75	Tunceli	0.097	0.097	↑	↑	↑	↑	↑
76	Şanlıurfa	0.027	0.027	–	–	–	↑	↑
77	Uşak	0.021	0.021	–	–	–	↑	↑
78	Van	0.109	0.114	↑	↑	↑	↑	↑
79	Yalova	0.054	0.051	↑	↑	↑	↑	↑
80	Yozgat	0.049	0.049	↑	↑	↑	↑	↑
81	Zonguldak	0.037	0.034	↑	↑	↑	↑	↑
Increasing Trend (%) ⟶				51	53	51	93	73

and Figure 3, Figure 4 and Figure 5. The symbols ↑ show an increasing trend, the symbol ↓ show a decreasing trend, and the symbol—indicates a non-significant trend at the significance level of α = 5%. The results were compared to assess the consistency and robustness of the detected trends.

Table 1 shows that an increasing trend in mean temperatures was detected in almost all of the stations. In the series where no trend was detected for average mean (Diyarbakır, Erzurum, Karabük stations), three traditional methods gave the same results. On the other hand, two innovative methods showed a negative trend for Diyarbakır and Erzurum, and a positive trend for Karabük station.

For maximum annual temperature series, three traditional trend tests detect an increasing trend in approximately two-thirds of the stations. On the other hand, ITA detects the highest rate of increasing trend (96%), while the increasing trend detected by QT remains at 84%.

Considering the minimum annual temperature series, the three traditional trend tests detect an increasing trend in about half of the stations. On the other hand, ITA detects the highest increasing trend rate (93%), while QT detects an increasing trend in 73% of the stations.

3.2. Trend Slopes

The temperature averages of all stations were used and linear trends were determined for the mean, maximum, and minimum temperature series to evaluate the trends for the whole country. The trends and the increasing slopes of the temperatures in Türkiye are given in Figure 3, Figure 4 and Figure 5. The average of the increasing trends for Türkiye was calculated as 0.027 °C/year, 0.031 °C/year, 0.038 °C/year in mean, maximum, and minimum annual temperatures, respectively.

4. Discussion

4.1. Comparison of the Results with Previous Studies Conducted in Türkiye, in the Region, and Around the World

In this study, where more than 60 years of data from 81 meteorological stations were used, Türkiye’s annual mean, maximum, and minimum temperatures were evaluated using three traditional and two innovative trend analysis methodologies. The results of the study noticeably indicated that the mean annual temperatures showed a significant increasing trend in almost all stations according to all trend analysis methods. For annual maximum and minimum temperatures, ITA detected an increasing trend in almost all stations. Traditional trend analysis methods detected increasing trends in two-thirds of annual maximums and half of annual minimums. The QuarTrend (QT) method proposed in this study quantitatively determines a trend between traditional methods and ITA. An increasing trend is detected in 84% of maximum temperature series and in 73% of minimum temperature series with the QT method.

The significant increasing trends in mean, maximum, and minimum temperatures found across Türkiye are consistent with the global warming effects confirmed by the IPCC AR6. Our observed warming rate of 0.027 °C/year for mean annual temperature aligns with the enhanced warming signal identified for the Mediterranean region [76].

There have been studies in the literature on the changes in temperatures in Türkiye. These include trend analyses and estimations based on climate scenarios. The data taken into consideration are station observations in some studies and reanalysis datasets in others. To give examples of these studies, in a comprehensive study conducted in the early 2000s based on temperature observations covering the period 1929–1999 [77], mean, maximum, and minimum temperatures were examined and temperature increases were detected in 13% of the mean, 30% of the maximum, and 57% of the minimum annual temperature series. The Turkish State Meteorological Service [78] developed climate predictions with three different global models for 2016–2099 in order to estimate the future effects of climate change on the country. In the study, three global model datasets were used and dynamic downscaling method was used with the regional RegCM4.3.4 model and projection results were obtained for Türkiye with 20 km resolution for the future periods 2016–2040, 2041–2070, and 2071–2099 according to RCP4.5 and RCP8.5 scenarios with 1971–2000 reference period. According to the RCP4.5 scenarios, annual average temperatures are expected to increase by 1.4 °C on average in the first half of the century and by 2.2 °C on average in the second half of the century. According to the RCP8.5 scenarios, annual average temperatures are projected to increase by 1.7 °C on average in the first half of the century and by 3.8 °C on average in the second half of the century. In a more recent study, Ref. [79] worked with reanalyzed datasets of mean annual temperatures covering the period 1901–2014 and detected an increasing trend in almost all of the 81 stations (except Bingöl, Diyarbakır, and Erzurum). Our station-based observations of a ~1.6 °C increase since 1960 (0.027 °C/year × 60 years) are consistent with reanalysis datasets for the region [79] and fall within the range of early 21st century projections made by (MGM) for the RCP4.5 scenario, providing observational validation for these models. Ref. [79] used the Climate Research Unit Time Series (CRU TS V3.23) dataset (1901–2014) and temperature trend was analyzed with the Mann–Kendall test. In 79 of 81 stations (98%) of Türkiye, mean annual temperatures showed a significant increasing trend. In the future projections made by ref. [78] according to different climate scenarios, an increase in average temperatures is predicted for all of Türkiye’s provinces (100%), even in the optimistic scenario (RCP4.5). Another recent study [20] investigated annual mean temperature trends over the 50-year period (1969–2018). According to the study results based on geographical regions, an increasing trend was detected at 96% of the stations. The temperature increase slope varies from one region to another and remains between 0.030 °C and 0.035 °C annually. Ref. [80] investigated the effects of climate change on temperature in the Eastern Black Sea region of Türkiye. For the RCP8.5 scenarios an increase of 0.035 °C/year in the interior and 0.03 °C/year in the coastal parts of the region are estimated. For the RCP4.5 scenarios the increase is estimated to be 0.025 °C/year and 0.02 °C/year, respectively. Most recently, ref. [81] examined the monthly and annual average temperatures in the Kızılırmak basin and in the Seyhan Basin in Türkiye between 1957 and 2022 and stated that the test results showed a significant increasing trend. Ref. [63] analyzed trends of meteorological parameters in the Marmara basin, Türkiye, and stated that the Mann–Kendall, Spearman Rho, and Triangle ITA methods detected an increase in average temperatures. Ref. [64] examined the temperature trend of 13 meteorological stations in the Antalya basin, Türkiye, and found that average temperatures increased by 1.74 °C in a 54-year period (0.032 °C/year).

When temperature trend studies from countries around Türkiye are examined, the findings regarding the temperature changes in the region are remarkable. Ref. [13] conducted a study to evaluate the trends in temperature series of Iraq from 1965 to 2015. MK trend test identified significant increases in maximum and minimum annual temperature data at all stations. According to the study, increases in minimum annual temperatures were determined between 0.048 and 0.117 °C/year and in maximum annual temperatures between 0.025 and 0.101 °C/year. Ref. [16] studied the long-term temperature trends at 34 meteorological stations in Iran over a period of 50 years (1961–2010). The trends were analyzed on seasonal and annual temperature data using the Mann–Kendall trend test, and the results showed increasing trends in temperature over the country in both annual and seasonal data. Significant increasing trends were detected in many regions of the country. An increase in mean annual temperature was determined in 62% of the stations. Ref. [82] used 5 models with synoptic data of 95 stations for projecting minimum and maximum temperatures in Iran and predicted an increasing trend in minimum and maximum temperatures, although the minimum temperatures were more significant. Ref. [21] analyzed the trends of the mean, maximum, and minimum temperatures using linear regression, Mann–Kendall, and Sen’s slope tests in the Romanian Black Sea coastal area. Results showed an increasing trend in the mean, maximum, and minimum annual temperatures. The average of the positive linear trends was calculated as 0.04 °C/year, 0.062 °C/year, and 0.064 °C/year in mean, maximum, and minimum annual temperatures, respectively.

An increasing trend has been witnessed in almost all temperature trend studies conducted in various parts of the world recently. Ref. [83] recently presented a comparative trend analysis of various climate indicators in Europe and the world, examining monthly and annual data. The study reported annual mean temperature increases of 0.035 °C/year for Europe and 0.021 °C/year globally. In another study conducted for Europe, ref. [84] examined the changes in daily minimum and maximum temperatures for the Iberian Peninsula using a high resolution downscaling simulation. They used the Weather Research and Forecast (WRF) model and found an annual increase of 0.019 °C/year in maximum temperatures and 0.013 °C/year in minimum temperatures with the ERA-driven simulations. Ref. [85] investigated the change in extreme climate events in Spain between 1951 and 2021. According to the study, annual maximum temperatures have increased by 0.014 °C/year to 0.026 °C/year.

Considering the case studies for the Asian continent, ref. [86] analyzed the spatial and temporal variation in climate extremes using daily minimum and maximum temperature data at 303 meteorological stations in China during 1961–2003. They found that both minimum and maximum temperatures show significant warming trends. Minimum temperatures have a greater tendency to increase than maximums. In a later study, ref. [87] investigated the mean maximum and minimum temperature data in China from model simulations during the period 2006–2100. The analysis focused on the periods 2011–2040 and 2061–2090 under two different RCP scenarios. Under the high emission scenario (RCP8.5), T_mean, T_max, and T_min show a significant warming rate of 0.043 °C/year, 0.042 °C/year, and 0.045 °C/year during the years 2011–2040 and 0.072 °C/year, 0.070 °C/year, and 0.076 °C/year during the years 2061–2090. In the low emission scenario (RCP4.5), T_mean, T_max and T_min show consistent increases with annual trends of 0.029 °C/year, 0.029 °C/year, and 0.030 °C/year, respectively, between 2011 and 2040, and the increases calm down up to 0.014 °C/year for T_mean, T_max, and T_min between 2061 and 2090. Ref. [88] investigated temporal trends on the meteorological data for Lanzhou of Gansu province in northwestern China from 1951 to 2016. They indicate that the mean, maximum and minimum temperature have a significant upward trend. Ref. [17] investigated trends in 35 years (1979–2013) of minimum and maximum temperatures (T_min and T_max) in an arid region of India. The results indicate that increasing trends are more significant in T_min (0.046 °C/year) than T_max (0.021 °C/year). Ref. [89] studied the trends in temperature in India for 107 years (1901–2007). Temperatures (T_mean, T_max, and T_min) were showing a significant increasing trend, particularly in the post-monsoon time and in the winter. An extensive variation was noticed all over India for minimum temperatures. Mean annual temperature showed an increasing trend for all regions, especially in the Western Himalayas. Also, an increase in maximum temperature trend is observed in all the regions. Highest increase in maximum temperature data in the Western Himalayas is signifying warming of the cold region. Minimum annual temperature showed an increasing trend everywhere except for Northwest India. Ref. [90] evaluated the spatial distribution of annual and seasonal temperatures and trends in temperature extremes over Pakistan. Daily temperature data with 1°  ×  1° spatial resolutions were used to evaluate the trends during the period 1960–2013. The results show that there is an increase in the annual mean of daily minimum and maximum temperature series over 92% and 99% areas of Pakistan at 95% confidence level. The minimum temperature (0.017–0.037 °C/year) is increasing faster than the maximum temperature (0.017–0.029 °C/year). Ref. [91] studied the variation in temperature in the Upper Narmada Basin of central India. The mean annual temperatures indicated significant increasing trends with the highest and lowest rates of 0.047 °C/year and 0.029 °C/year, respectively. Ref. [92] investigated the spatial, temporal, and climate variability of a data-sparse Himalayan watershed. The results of the trend analysis showed that temperature is increasing on an annual scale (0.011 °C/year). In another study conducted in the Asian continent, ref. [23] analyzed the annual changes in temperatures in six regions of Thailand between 2001 and 2020 and showed that mean annual temperatures had an increasing trend for all six regions.

As examples of temperature trend studies conducted on the African continent, ref. [24] examined seasonal, monthly and annual maximum and minimum temperatures between 1979 and 2019 for Morocco. Despite seasonal changes, they found a general increasing trend in both annual maximum and annual minimum temperatures throughout the period. Another study conducted for Morocco [93] projected an increase in mean annual temperature towards 2050, and simulations show an increase under RCP scenarios 2.6, 4.5, and 8.5. Ref. [9] analyzed the trends in mean, maximum, and minimum temperature data in the Upper Blue Nile River Basin, Ethiopia. In the study covering the period from 1981 to 2010, at annual scale, maximum and minimum temperatures increased significantly at a rate of 0.01 and 0.015 °C per year, respectively, over more than 33% of the study area. Minimum temperatures increased at a higher rate than maximum temperatures. Ref. [19] studied the maximum and minimum temperature trends in Nigeria for the period from 1970 to 2010, finding a significant increasing trend in most of the locations across the country. Ref. [94] conducted a study on extreme temperature trend and return period mapping in the Upper Tekeze River Basin of Northern Ethiopia. The results of the study showed that future (50-year) extreme temperature will increase by 0.6–2.2 °C (1.2–4.4 °C/year). Ref. [95] analyzed the temperature changes and trends for the Niger Central Hydrological Area of Nigeria, from 1911 to 2015. The change over the study area for T_mean, T_max, and T_min was found to be 0.0043 °C/year, 0.0007 °C/year, and 0.009 °C/year, respectively.

The examples of temperature trend studies conducted in the American continent also indicated increasing trends. Ref. [18] examined annual maximum and minimum temperature data from stations in the southwestern United States for the period between 1902 and 2017. A total of 75% of the stations showed an increasing trend in annual maximum temperature at rates ranging from 0.006 °C/year to 0.03 °C/year. Annual minimum temperature increased at rates ranging from 0.01 °C/year to 0.08 °C/year at 69% of the stations during the last century. Two relatively older studies have considered the Canadian temperature trend, the first of which ref. [96] used temperature data for the period 1970–2000 and found an average increase of 0.02 °C/year, 0.03 °C/year, and 0.04 °C/year in mean, maximum, and minimum annual temperatures, respectively. Another study [97] examining Canadian temperature trends for the period 1948–2009 found a temperature increase in mean annual temperatures up to 0.9 °C over the 60-year period (0.015 °C/year). Ref. [98] showed significant increasing trends in temperature in both study regions using three methodologies for northwest Mexico. Maximum temperature records have increased by approximately 3 °C in 50 years (0.06 °C/year).

A comprehensive temperature trend study covering each of the six continents (three mega cities with a population of more than 5 million people from each continent) was conducted for urban and peri-urban areas of the cities [15]. In this study conducted for 18 mega cities, it was found that temperatures increased in 17 cities (Tokyo, Beijing, New York, Los Angeles, Mexico City, Lagos, Johannesburg, Cairo, Sao Paulo, Buenos Aires, Santiago, Moscow, Berlin, Madrid, Sydney, Alice Spring, Perth) and showed a decreasing trend in only 1 city (Delhi). Annual and seasonal temperature increasing trends in urban areas (0.01 °C/year) are relatively higher compared to the increasing trends (0.008 °C/year) in peri-urban areas. However, it is crucial to distinguish the scales of analysis. The aforementioned study focused on urban and peri-urban areas, where trends can be significantly amplified by the Urban Heat Island (UHI) effect. In contrast, the present analysis for Türkiye integrates data from 81 stations representing a mix of urban and non-urban environments, a standard methodology for capturing the large-scale, regional climate trends [76,99]. While stations located within growing cities like Istanbul or Ankara are undoubtedly influenced by UHI [100], the consistent and statistically significant warming signal detected across the entire network—including rural and high-altitude stations—strongly indicates that the results are dominated by regional climate change. This pattern of background warming amplified locally by UHI is consistent with findings in other Mediterranean regions [101]. Thus, the national-scale trend identified here provides the essential background warming upon which local UHI effects are involved.

A summary of global studies conducted on an annual scale using observed data is provided in

Table 4. Summary of studies conducted on an annual scale with observed data.

Study	Temperature Data	Site/Region—Country	Period	Trend Test 1	Trend	Trend Slope (S) (°C/Year)
[8]	Annual max. and min.	Zhujiang River Basin—China	1980–2010	MK, SSE	increasing	0.018 < STmax < 0.023 0.009 < STmin < 0.013
[7]	Annual mean	Ontario and Quebec, Canada	1967–2006	MK, DWT	increasing
[9]	Annual max. and min.	Upper Blue Nile River Basin, Ethiopia	1981–2010	LRA	increasing	STmax = 0.016 STmin = 0.021
[11]	Annual mean	Syr Darya Basin, Central Asia	1881–2011	MK	increasing	STmean = 0.014
[13]	Annual max. and min.	Iraq	1965–2015	MK, mMK	increasing	0.025 < STmax < 0.101 0.048 < STmin < 0.117
[15]	Annual mean	18 mega cities in 6 continents	1901–2008	MK, LRA	increasing	STmean = 0.009
[16]	Annual mean	Iran	1961–2010	MK, mMK	increasing
[18]	Annual max. and min.	Southwest of USA	1902–2017	MK, SSE	increasing	0.006 < STmax < 0.031 0.01 < STmin < 0.08
[20]	Annual mean	Türkiye	1969–2018	MK	increasing	0.029 < STmean < 0.035
[21]	Annual mean, max. and min.	Black Sea Coast, Romania	1990–2020	LRA, MK, SSE	increasing	STmean = 0.04 STmax = 0.062 STmin = 0.064
[23]	Annual mean	Thailand	2001–2020	MK, ITA	increasing
[90]	Annual max. and min.	Pakistan	1960–2013	mMK	increasing	0.017 < STmax < 0.029 0.017 < STmin < 0.037
[63]	Annual mean	Marmara Basin, Türkiye	1976–2021	MK, SRA, ITA	increasing
[86]	Annual mean	China	1961–2003	MK, SSE	increasing	STmean = 0.027
[88]	Annual mean, max. and min.	Lanzhou city, China	1951–2016	MK, SSE, SRA	increasing	STmean = 0.074 STmax = 0.04 STmin = 0.104
[91]	Annual mean	Upper Narmada Basin, India	1901–2002	MK, SSE, ITA	increasing	0.029 < STmean < 0.047
[92]	Annual mean	Northwest Himalayas, India	1981–2018		increasing	STmean = 0.011
[96]	Annual mean, max. and min.	Greater Toronto Area, Canada	1970–2000	MK	increasing	STmean = 0.02 STmax = 0.03 STmin = 0.04

¹ DWT, the discrete wavelet transform; MK, the Mann–Kendall test; mMK, modified Mann-Kendall; SSE, Sen’s slope estimator; LRA, linear regression analysis; SRA, Spearman rank correlation.

. This table provides a comparison with similar studies in the literature regarding the type of observed temperature data, study region, data period, trend analysis method, and trend result.

It is remarkable that in almost all of the temperature trend studies summarized above, which have been conducted in the recent past in Türkiye, in the region and around the world, the mean, maximum, and minimum temperatures have shown an increasing trend, which is consistent with the results of this study. However, increasing trends show changes depending on the geography. In our study, the average of increasing trends for Türkiye was calculated as 0.027 °C/year, 0.031 °C/year, 0.038 °C/year for mean, maximum, and minimum annual temperatures, respectively, while these increase rates in the region and around the world vary from 0.01 °C/year to 0.07 °C/year for mean temperatures, from 0.01 °C/year to 0.10 °C/year for maximum temperatures, and from 0.01 °C/year to 0.11 °C/year for minimum temperatures.

Another important finding in our study is that the minimum temperature is increasing faster than the maximum temperature, and the maximum temperature is increasing faster than the mean temperature, which is consistent with the literature summarized above. Future minimum temperatures rising faster than maximum temperatures indicate an increasing risk to crop yields due to increased moisture loss from soil and plants. The finding that minimum temperatures are rising faster than maximum temperatures (0.038 °C/year vs. 0.031 °C/year) is a pattern observed in many warming climates. This can be linked to climate feedback mechanisms, such as increased nighttime heat retention due to higher humidity or changes in cloud cover. Furthermore, the rapid warming of minimum temperatures has significant implications for agriculture and water resources, potentially leading to increased evapotranspiration and moisture stress [102]. While our analysis at the national scale captures the overarching regional trend, it is important to acknowledge that urban stations may exhibit amplified warming signals due to the urban heat island effect. Future studies could focus on disentangling these contributions, but the consistency of the trend across diverse stations underscores that the primary driver is large-scale climate change.

4.2. Comparison of the Results of the Trend Analysis Methodologies

In the study, temperature trends were examined with traditional methods such as Mann–Kendall, Linear Regression Analysis, and Sen’s slope, which are frequently used in the literature, as well as Innovative Trend Analysis and also the QuarTrend (QT) method, which is proposed here for the first time and is based on quartiles. In the examination of mean annual temperatures, the traditional trend detection methods identified an increasing trend in almost all stations (78 out of 81 stations). However, in three stations where traditional methods did not detect a trend, ITA and QT identified a negative trend in two stations and a positive trend in one station. For the maximum annual temperature series, the three traditional trend tests detected increasing trends in approximately two-thirds of the stations. On the other hand, ITA detected the highest increasing trend rate (96%), while the increasing trend detected by QT remained at 84%. When the minimum annual temperature series are considered, the three traditional trend tests detected an increasing trend in approximately half of the stations. On the other hand, ITA detected the highest increasing trend rate (93%), while QT detected increasing trends in 73% of the stations.

In summary, ITA detected the highest number of trends in mean, maximum, and minimum temperature series (100%, 96%, 93%) compared to the traditional trend analysis methods. This situation is consistent with other trend analysis studies in the literature [31,32,33,34,37,41,43,44,67,103,104,105]. However, Refs. [29,30] it is noted that the inflated variance of the ITA slope estimator may lead to detecting a trend that does not actually exist. This is thought to be one reason why ITA detects trends in a larger number of series than traditional trend tests. Considering all these facts and the fact that the QT method is somewhere between ITA and traditional methods in detecting trends quantitatively, it is suggested that the QT method proposed in the study is worth considering in trend analysis.

According to linear regression analysis, as a result of examining trend slopes for mean, maximum, and minimum temperature series:

(a)

In mean annual temperatures, the average slope of the series without trend was calculated as 0.01, while the average slope of the series with trend was calculated as 0.03;

(b)

In maximum annual temperatures, the average slope of the series without trend was calculated as 0.015, while the average slope of the series with trend was calculated as 0.039;

(c)

In minimum annual temperatures, the average slope of the series without trend was calculated as 0.02, while the average slope of the series with trend was calculated as 0.068.

The average of the slopes of the trendless series is considered to constitute a lower threshold for the slopes of the series in the same population (i.e., mean temperatures). Thus, this lower threshold of the slope for the mean, maximum, and minimum annual temperature series is accepted as 0.01, 0.015, and 0.02, respectively.

The comparative evaluation of the trend detection results of traditional trend analysis methods, ITA, and QT method showed that:

(a)

For the mean annual temperature series

-

Almost all methods showed the same (increasing) trend for 78 of the 81 stations.

-

Only 1 station (Erzurum) had a trend slope below the trend slope threshold (0.01).

(b)

For the maximum annual temperatures series

-

All the 5 trend methods showed the same result at 54 stations (51 increasing trend and 3 no trend).

-

Three traditional trend tests fail to detect trends in 17 of 81 stations (21%) with trend slopes above the lower slope threshold (0.015).

-

In 9 stations only ITA detected an increasing trend. Less than half of these trends (33%) were above the trend slope threshold (0.015).

-

In 14 stations, only ITA and QT detected an increasing trend, where most of the linear trend slopes of these stations (79%) were above the trend slope threshold (0.015).

(c)

For the minimum annual temperatures series

-

All the 5 trend methods showed the same result in 41 stations (39 increasing trend, 1 decreasing trend and 1 no trend).

-

Three traditional trend tests fail to detect trends in 20 of 81 stations (25%) with trend slopes above the lower slope threshold (0.02).

-

In 16 stations only ITA detected an increasing trend. Less than half of these trends (44%) were above the trend slope threshold (0.02).

-

For the maximum annual temperature series in 14 stations only ITA and QT detected an increasing trend, where most of the linear trend slopes of these stations (79%) were above the trend threshold (0.02).

As a result of the above comparisons, we can summarize the effect of trend analysis methods on the analysis results as follows:

-

Although all trend analysis methods showed almost the same results for the mean temperature series (T_mean), significant differences between the methods occur in the extreme value series (T_max, T_min).

-

It has been determined that three traditional trend tests are not always sufficient to detect a trend when evaluated in terms of trend slope.

-

The fact that “ITA sometimes detects trends that do not really exist”, which has been stated in some previous studies, is confirmed here as well.

-

The QT method proposed in this study is positioned somewhere between traditional trend methods and ITA in terms of trend detection and makes more rational trend detections when the trend slope is taken into account.

-

The reason why QT method gives rational results, especially for extreme value series, is that both positive and negative skewness are taken into account by comparing the first and third quartiles (Q₁ and Q₃) as well as the medians (Q₂) of the half series.

5. Conclusions

The fact that 2023 and subsequently 2024 are the hottest years on record in history reveals the importance of examining temperature changes. In this perspective, monitoring significant changes in the temperature series becomes more important over time. Ref [106] stated that scientists are more than 95% sure that global warming is a result of greenhouse gas emissions caused by different human (anthropogenic) activities.

This study provides a comprehensive, nation-wide analysis of temperature trends in Türkiye from 1960 to 2022, firmly placing the observed changes within the context of anthropogenic global warming as assessed by [76]. The results demonstrate a significant warming signal across the country, with minimum temperatures increasing at the fastest rate, followed by maximum and mean temperatures. This pattern is consistent with global and regional observations and underscores the pervasive impact of climate change on annual temperature values. Temperature fluctuations from year to year are highest in minimum temperatures, then in maximum temperatures, and lowest in average temperatures.

When trend methods are compared, traditional trend analysis methods such as Mann–Kendall, LRA, and Sen’s slope give similar results to ITA and QT methods for mean temperatures, while they detect trends for fewer stations in maximum and minimum temperatures. As mentioned in the literature and confirmed in this study, ITA is the trend analysis method that detects trends in the most series. When trend slopes are taken into account, traditional methods miss some trending series, while ITA detects trends in non-trending series. In order to overcome this problem, an innovative QT trend analysis method is proposed, which uses the quartiles of the half-series (Q₁, Q₂, Q₃). This method, which is situated between the traditional methods and ITA in terms of trend detection, gives more rational results according to the trend slopes. The QT method should be expected to be more successful than ITA especially for highly skewed extreme value series, as it considers both positive and negative skewness by comparing the first and third quartiles (Q₁ and Q₃) of the chronologically divided half-series as well as the median (Q₂).

The innovations brought by the study and its contribution to the literature are summarized as follows:

• As to our knowledge, there is no current temperature trend study covering the whole country for Türkiye. The studies conducted are either regional or do not take into account all the mean, maximum and minimum temperatures.
• The study did not only focus on the results of trend analyses at a certain level of significance but also examined the relationship between the detected trends and the trend slope.
• The QT method proposed in this study is a robust trend analysis method based on quartiles and it is suitable for detecting trend analysis of positively or negatively skewed series.
• The QT method, which uses order statistics, is less sensitive to outliers than trend analysis methods based on means, medians, or linear regression.

Although the proposed QT method robustly determines the existence of a trend by checking each of Q₁, Q₂, and Q₃, it does not provide detailed information about the fluctuation or density of the trend. To address this deficiency, the method can be evaluated in conjunction with the trend slope and the graphical representation of the trend.

To generalize and validate the proposed QT method for hydro-meteorological trend analysis, it is recommended that follow-up studies should be conducted by considering various regions of the world and using different types of data (precipitation, flow, evaporation, humidity, etc.) and even using synthetic data.