Seasonal trends in the Early Twentieth Century Warming (ETCW) in a centennial instrumental temperature record from Central Europe 3Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.DOI: 10.15201/hungeobull.73.1.1 Hungarian Geographical Bulletin 73 2024 (1) 3–16. Introduction Changes in weather and climate patterns due to anthropogenic forced global climate change have direct and indirect effects on human life and socio-economic systems. The final two decades of the 20th century, along with the first two decades of the 21st, have been successively warmer than any decade preceding 1850. What is more, global surface temperatures between 2001 and 2020 were 1.1 °C higher than 1850–1900 (IPCC, 2021). This increase in temperatures can be traced back to the beginning of the past century, occurring in two distinct periods, from the mid-1920s to mid-1940s (Przybylak, R. et al. 2021) and from 1978 to the present day (Delworth, T.L. and Knutson, T.R. 2000). However, between these two periods of accelerated warming, a stagna- tion was observed from the 1940s to the 1970s, and a less steep trend from 2000 to approxi- mately 2013 (Hegerl, G.C. et al. 2018). 1 Department of Methodology for Business Analysis, Faculty of Commerce, Hospitality and Tourism, Budapest Business University. Alkotmány utca 9-11., H-1054 Budapest, Hungary. Corresponding author’s e-mail: magyar.norbert@uni-bge.hu 2 Department of Meteorology, Faculty of Science, ELTE Eötvös Loránd University. Pázmány Péter sétány 1/A., H-1117 Budapest, Hungary. 3 Institute for Geological and Geochemical Research, HUN-REN Research Centre for Astronomy and Earth Sciences. Budaörsi út 45., H-1112 Budapest, Hungary. 4 Department of Meteorology and Water Management, Georgikon Campus, Hungarian University of Agronomy and Life Sciences. Festetics utca 7., H-8360 Keszthely, Hungary. Seasonal trends in the Early Twentieth Century Warming (ETCW) in a centennial instrumental temperature record from Central Europe Tímea KOCSIS1, Rita PONGRÁCZ 2, István Gábor HATVANI 3, Norbert MAGYAR 1, Angéla ANDA 4 and Ilona KOVÁCS-SZÉKELY 1 Abstract The goal of the present paper is to investigate whether any objectively defined and statistically significant changes can be discovered in one of the longest homogenized instrumental temperature records in East-Central Europe. Thus, it is hoped that the present analysis will add to earlier attempts and elucidate the persistence of the warming period observed in the early 20th century. Similar to the global tendency, the Early Twentieth Century Warming (hereinafter, ETCW) period can be identified between 1931 and 1951 in the annual mean temperature time series of Keszthely, a small town in Hungary. The Mann-Kendall trend test was used to determine whether a monotonic trend was present, as it is not possible to regard the residuals of the linear trend as normally distributed. A signifi- cant rising trend can be observed in the warming period in spring of the years between 1925 and 1951. In case of summer and autumn, this period cannot be characterized as having any significant identifiable trend. A rise in the mean can, however, be recognized. Overall, the specific regional manifestation of the global ETCW may clearly be illustrated in this study via detailed statistical analysis of the temperature records for Keszthely, a location with one of the longest temperature records in Hungary. However, other regions surrounding Hungary show similar climatic trends, emphasizing the fact that the behaviour presented here is not unique to Central and Eastern Europe. Keywords: change-point detection, Early Twentieth Century Warming (ETCW), temperature records, time series analysis Received April 2023, accepted February 2024. mailto:magyar.norbert@uni-bge.hu Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.4 Out of these warming periods the one in the first decades of the 20th century was the more prominent, called the “Early Twentieth Century Warming” (ETCW) (Hegerl, G.C. et al. 2018). It was found that, besides oth- er factors, natural variability made a large contribution to the ETCW, and particular- ly to regional anomalies in the 1920s and 1930s. Nonetheless, the ETCW represents a phase of a robust surface warming globally (Semenov, V.A. and Latif, M. 2012), which we might add, was concentrated in higher latitudes (Yamanouchi, T. 2011). On the re- gional scale, the warming observed the early 20th century displays a noticeable maximum at higher latitudes in the Northern Hemisphere (Delworth, T.L. and Knutson, T.R. 2000). Exploring regional temperature features over the course of the ETCW is much more enlightening than looking at the global mean (Brönnimann, S. 2009). The ETCW was most evident in the Arctic, and in winter (Bokuchava, D.D. and Semenov, V.A. 2021), with the Arctic surface temperature warming more than twice as fast as global temperatures (Svendsen, L. et al. 2018). Atmospheric circu- lation fluctuations are important for the early 20th century Arctic warming (Tokinaga, H. et al. 2017). A considerable part of regional manifestations of the ETCW appeared out- side of the Arctic, including across the USA, Western Europe, and the North and South Atlantic (Hegerl, G.C. et al. 2018; Bokuchava, D.D. and Semenov, V.A. 2021). The end of the ETCW in Central Europe was marked by summer heatwaves, droughts, and cold winters (Hegerl, G.C. et al. 2018). The ETCW was probably caused by the combined ef- fects of long-term natural climate variations in the North Atlantic and North Pacific and the influence of natural radiative forcing, as well as the growing concentration of green- house gases in the atmosphere (Bokuchava, D.D. and Semenov, V.A. 2021). Utilizing a Bayesian change-point detection method, a change-point can be observed in the data of Greenland ice cores at the beginning of the 20th century (centred on 1933) (Hatvani, I.G. et al. 2022). Research of climate features during the ETCW is mainly based on conventional weather observations from the continental surface, measurements taken at sea by ships, and some climate proxy data; even if data- saving initiatives aim at supplementing the global climate record in this period (Allan, R. et al. 2011 – in: Hegerl, G.C. et al. 2018), the main sources of information about climate change remain ordinary measurements. The existing historical observations recorded by national meteorological and hydrological services, although considered to be incom- plete, may nonetheless augment our knowl- edge of key climate processes and climate change (Dee, D. et al. 2021). Meteorological instruments have been widely used since 1850 in Europe, the Eastern USA and some other regions, in- cluding, sporadically, the oceans (Stott, P. et al. 2018). Instrumental meteorologi- cal measurements were conducted from the 18th century in Italy, Belgium, Sweden, Spain, and Russia (Moberg, A. et al. 2000; Camuffo, D. and Jones, P.D. 2002). National weather yearbooks began to be published around 1860 in many countries (Jones, P.D. 2001). Instrumental temperature time series are available from 1760 for the Alpine zone (Böhm, R. et al. 2001). The longest instru- mental temperature time series in the Czech Republic is for Prague-Klementinum, start- ing in 1775 (Brázdil, R. et al. 2012). In Poland, the mean annual temperature records for the periods 1779–1998 (Warsaw) and 1792–1995 (Cracow) are analysed by Przybylak, R. (2010). In Budapest, regular meteorological observations started in 1780 (Camuffo, D. 2018). Official instrumental meteorological observations began in 1851 in Austria, and in 1870 in Hungary (this was the epoch of the Austro-Hungarian Monarchy). Keszthely station was one of the earliest working sta- tions in the Hungarian network, a network that was organized according to the Western European standards of the time. It is also in- teresting because there is a special microcli- mate at that location in consequence of the presence of the nearby large lake, Balaton. 5Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. Despite the long time series, in the literature there is a marked lack of research dealing with the ETWC in Central Europe. The aim of the present study is to explore the presence of objectively defined and statistical- ly significant changes occurring in one of the longest East-Central European homogenized instrumental temperature records. Hence, the present study complements previous efforts and elucidates the persistence of the observed warming period in the early 20th century. Materials and methods Keszthely centennial homogenized temperature time series The Keszthely (E 17o14’, N 46o44’ – Figure 1) monthly temperature time series spans 1871 to 2018 and was measured at the long-estab- lished Georgikon Academy of Agriculture in Keszthely, and later, at a meteorological sta- tion of the Hungarian Meteorological Service. Keszthely is located on the northern shore of Lake Balaton, a 17,000–19,000-year-old water body and the largest shallow lake in Central Europe (surface area: about 600 km2, average depth: 3.25 m). As the last overall publica- tion dealing with the climate of Balaton was published 50 years ago (Béll, B. and Takács, B. 1974), and due to the absence of a perma- nent measuring network around the lake, the meteorological dataset of weather conditions over and around the lake originates from two permanent lakeshore meteorological stations (Keszthely and Siófok) that have made ob- servations over the long term (Anda, A. et al. 2023). Using the categories of the Koppen-Gei- ger classification, the climate of Keszthely is temperate continental (Cfb), with an annual long-term mean air temperature of 10.5 °C, with a monthly minimum of -1.03 °C during January, and a maximum of 21.14 °C in July. The average annual precipitation sum is 673.3 ± 137.9 mm, though with a large monthly variation of between 32.7 and 76.1 mm in the driest January, and wettest July, respectively. In winter, the lake may be covered with ice, sometimes thicker than 0.5 m, factor impact- ing the air temperature during early spring. The time series explored here were quality controlled and homogenized by the Hungarian Meteorological Service. Homogenization was conducted using the MASH tool (Szentimrey, T. 1999, 2008), which eliminates errors and inhomogenei- ties while filling the gaps in the data (Izsák, B. and Szentimrey, T. 2020). The MASH pro- cedure was originally developed for the ho- mogenization of monthly series. It is a rela- tive method and depending on the distribu- tion of the meteorological elements under ex- amination, additive (e.g., for temperature) or multiplicative (e.g., for precipitation) models can be employed (Szentimrey, T. 2006). Fig. 1. Location of Keszthely, Hungary (46°44′N, 17°14′E, elevation 114.2 m above Baltic Sea level). Source: Redrawn from Kocsis, T. et al. 2020. Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.6 Methodology After giving an overall description of the Keszthely mean temperature time series (KT) regarding long-term trends (see the first part of section ‘Results and discussion’), the data was aggregated into annual and seasonal (winter: DJF, spring: MAM, summer: JJA, autumn: SON), and abrupt changes in the time series were sought using change-point analysis. There are numerous change-point detection techniques applied in the explo- ration of climatic (proxy) variables (e.g., Ducré-Robitaille, J.-F. et al. 2003; Reeves, J. et al. 2007; Ruggieri, E. 2013; Topál, D. et al. 2016; Hatvani, I.G. et al. 2020), and of these, Bayesian change-point analysis (BCPa) (see the second part of section ‘Results and dis- cussion’) was selected for the present study, due to its ability to provide uncertainty es- timates of the number (even multiples) and timing of change-points, a key advantage over a frequentist approach (Ruggieri, E. 2013). In the segments, thus, determined, numerous aspects of the mean temperature time series are described and assessed for the presence of prevailing significant trends. Preprocessing and change-point analysis To achieve common variability and increase the signal-to-noise ratio relative to the mag- nitude of the supposed changes a 10-year low-pass filter was applied to all the time series. This aided the detectability of valid change-points (Ruggieri, E. and Antonellis, M. 2016; Ruggieri, E. 2018). The time series were then centred and subjected to the lin- ear model of the Bayesian change-point algo- rithm of Ruggieri, E. (2013) (BCPa for short). The BCPa involves three key steps: 1. The calculation of the probability of the data given by the model between any two time points. This encompasses all potential climate regimes (in terms of timing) within the dataset. 2. The utilization of the probabilistic calcula- tions from step 1 as building blocks to assem- ble these segments recursively, incrementally adding one possible change-point at a time. 3. The application of Bayes’ rule to cal- culate the posterior distribution of the pa- rameters of interest – here the number and location of change-points – as well as the parameters of the regression model in each interval (Hatvani, I.G. et al. 2020). In a nutshell, BCPa can generate the pos- terior distribution on both the number and location of change-points in a data set. For an additional overview on change-point analy- sis (see Ruggieri, E. 2013, 2018). In terms of parameterization, the model has several parameters which need to be specified, including those of the prior dis- tribution of regression coefficients and the residual variance, which are known as hy- per-parameters (Quiang, R. and Ruggieri, E. 2023). Here, the parameterization followed that the method of Hatvani, I.G. et al. (2022) on climate proxy time series, specifically: – prior on the regression coefficients k0 = (0.001, 0.01); – prior on the variance of the residuals, v0 = 1, σ0 = variance of each time series; – the minimum distance between successive change-points, dmin = 10 steps (years in the present case); – the maximum number of change-points kmax = 50; – the number of sampled solutions, num. samp = 10,000. The prior parameters were deliberately selected to represent a minimum of prior in- formation, allowing the inference to be pri- marily driven by the actual data. In general, in any case where the model is sensitive to the k0, v0, and σ0 parameters, the number of change-points is influenced while their loca- tion or distribution is left unaffected. In sim- pler terms, adjusting these parameters can modify the detection threshold of the model, but will not result in the shift of a change- point from one section of the data to another. For a detailed explanation of the parameters, the reader is referred to Ruggieri, E. (2013) and the supplemental online material of Hatvani, I.G. et al. (2022). 7Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. Because BCPa determines the probability of a change-point at a given location – thus, providing uncertainty estimates – and the change-point time series obtained were passed through a 5-year centralized rolling summary function in order to cumulate the probabili- ty dispersed between consecutive years. This cumulative probability is the probability of the change-point(s) shown in the study. Only “practically certain” change-points with a 5-year sum probability > 80 percent were con- sidered in the evaluation. In addition, a 5-year window was considered when evaluating the coincidence of the differently aggregated T time series, as was done in Hatvani, I.G. et al. (2022), but even more rigorously. Trend analysis and descriptive stats The aim was to determine the general tenden- cies of the time series. As the temperature data underwent homogenization using MASH, there was no need to check the data for homogeneity. The common linear regression (trend) based on ordinary least squares method was used, along with a control to conform to the demand of normal distribution of the residuals using the Shapiro-Wilk normality test (Shapiro, S.S. and Wilk, M.B. 1965). The decision over significance was made at the 5 percent significance level. Beside the ten- dencies, absolute deviation from the mean was calculated for every year on annual, sea- sonal, monthly levels to obtain information about the tendencies of the variability. After searching for change-points in the time se- ries, they were segmented according to the change-points. Any tendencies in the seg- mented time series were determined using linear trend. In cases where the distribution of the residuals is significantly different from normal distribution, the Mann-Kendall (MK) non-parametric trend test was applied to de- termine the tendency of the segments at α = 5 percent. The simple MK trend test is based upon the work of Mann, H.B. (1945) and Kendall, M. (1975), and it is closely related to Kendall’s rank correlation coefficient. A detailed description of the methodol- ogy of the simple MK trend test is given by Gilbert, R.O. (1987), and Hipel, K.W. and McLeod, A.I. (1994). The test statistic is based on S, which represents the number of signs between the elements of the times series. A positive value for S means that there is a ris- ing trend, a negative value for S, the contrary. With the help of descriptive statistics (mean, standard deviation) and trend analysis, the reasons for the presence of a change-point were explored. The simple MK trend test does not consider the possible autocorrela- tion of the data. The presence of positive au- tocorrelation in the data increases the chance of detecting trends when actually none exist, and vice versa (Hamed, K.H. and Rao, A.R. 1998). This effect of the existence of autocor- relation in data is often ignored: Hamed, K.H. and Rao, A.R. (1998) supposed a modified non-parametric trend test suited to auto-cor- related and gave a detailed description of it. Modified MK trend test for auto-correlated. In the case of the detection of a non-parametric trend, Sen’s slope estimator (Sen, P.K. 1968) was applied. This is a non-parametric method that can calculate the change per unit time (direction and volume). Sen’s method uses a linear model to estimate the slope of the trend (Da Silva, R.M. et al. 2015). In the case of a proven autocorrelation in the dataset, the mod- ified version of the MK trend test was applied (Figure 2). AR1 values were higher than the critical value for the annual mean temperature data, for summer seasonal mean time series, and July (Figure 3). The autocorrelation of the data and the normal distribution of the residuals were taken into account in the anal- ysis of the variability as well. The trend estimations and the change- point analysis were performed in the R statis- tical environment (R Core Team, 2019). The modified MK stats and trend packages were used for trend analysis, band-pass filtering was performed with the band-pass function of the astrochron package (Meyers, S.R. 2014) and change-point detection was performed using the Bayesian change-point algorithm (Ruggieri, E. 2013). Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.8 Fig. 2. Decision making flow-chart for choosing method. Source: Authors’ own elaboration. Fig. 3. Autocorrelation (AR1) of the time series (significant autocorrelations are in red). Source: Authors’ own elaboration. Results and discussion Overall tendencies in the centennial Keszthely mean temperature time series (KT) The autocorrelation was taken into account using the Hamed and Rao method (Hamed, K.H. and Rao, A.R. 1998) for the analysis of the annual mean temperature data as the presence of autocorrelation was proved and an overall significant monotonic increasing tendency (Figure 4, A – Sen’s slope: 0.004, S = 1905, p = 2.506 10-02) was found for the an- nual mean, corresponding to a rise on aver- 9Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. age of 0.4 °C per 100 years. Significant trends were seen for the winter and spring (Figure 4, E – winter, Sen’s slope: 0.0084, S = 1305, p = 2.896 10-02), with the simple MK test and Figure 4, B – spring, slope: 0.0048, t(146) = 2.234, p = 2.703 10-02, with a linear trend), with an increase of 0.84 °C and 0.48 °C per 100 years, respectively. Regarding the summer and autumn, no overall significant (p > 0.05) trend was observed. Among the months, linear increasing trends were detected in January (slope: 0.0119, t(146) = 2.259, p = 2.534 10-02) and in November (slope: 0.0086, t(146) = 2.197, p = 2.956 10-02), respectively. Variability was assessed by examining the absolute deviation from the mean (annual, seasonal, monthly, respectively). The distri- butions of the residuals cannot be accepted to be normal, therefore the MK test was used. Significant autocorrelation can be detected in the analysed data in September, so an auto- correlated MK test was applied. It is interesting to see the tendencies of the variability, but the only significant monotonic trend can be detect- ed in December. The absolute deviation from the mean temperature in December displays a significant linear decreasing trend (Sen’s slope: -0.0042, S = -1235, p = 4.070 10-02). This means that in December, fewer extremities occur. Abrupt changes and decadal trends in the Keszthely mean temperature time series (KT) A change-point can be found in the annual mean temperature data in 1895, with a prob- ability of 99 percent, and another in 1931 with 90 percent probability. The change-point (CP) in 1931 is in good coincidence with the beginning of the ETCW. In 1952, a negative shift can be seen with 80% probability. This may be considered as indicating the end of the ETCW period (see Figure 4, A). In the sea- sonal data, change-points can be detected in 1925 with 99 percent probability in spring, in 1927 in the summer mean temperature and in 1924 in the average temperature of winter. In the time series of autumn, a CP can be seen in 1924 with 80 percent probability. In autumn, a change-point can be seen in 1969 with 99 percent probability. A secondary CP can be identified in 1952 and 1953 in the spring and summer time series, respectively. These change-points are also in good agreement with dates for the beginning (Delworth, T.L. and Knutson, T.R. 2000) and end (Przybylak, R. et al. 2021) of the ETCW (see Figure 4, B-D). The period of the ETCW can be detected in the annual data and in spring and summer, but in autumn and winter only the beginning can be identified, while the end is missing. The time series were divided at the CPs that have a probability of 80 percent at least. The year in which the CP was detected be- comes the first year of each separate period. Linear trend, or MK trend, was fitted, and the slope coefficient was examined at the 10 and 5 percent significance levels (Table 1). Among the separated parts of the annual mean temperature, significant periods of increase can be observed between 1931 and 1951, and 1985 and 2018. The timing of the first is contemporary with the ETCW. The second is the significant temperature rise be- ing experienced nowadays. The previous division was also appli- cable to the times series of the seasons, as well. In spring all of the separate periods show significant rising tendencies at the 10 percent significance level. A significant in- creasing trend can be found in the period of the ETCW. Surprisingly, in the summer the ETCW period does not exert a significant in- fluence, but the previous and the next period displayed significant change. Between 1871 and 1926 a significant decreasing trend may be observed in the summer mean tempera- ture (see Table 1). The average temperature, however, is higher than the mean of both the previous and subsequent periods. Between 1953 and 2018 a significant temperature rise can be detected (Figure 5). This result is some- what at odds with the findings of Hegerl, G.C. et al. (2018), namely, that in Central Europe droughts and heatwaves marked the end of the ETCW in the late 1940s and early 1950s. Studies analysing the time series of the Palmer Drought Severity Index (PDSI) Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.10 Fig. 4. Keszthely annual (A), spring (B), summer (C), autumn (D), and winter (E) raw (grey), 10-year low-passed (orange) temperature time series spanning 1871–2018. Change-points of posterior probability > 80%, and < 90% are indicated in orange, and those > 90% with a red vertical column. The blue line in part (A) is the linear trend of the annual time series. Source: Authors’ own elaboration. 8 10 12 14 16 Ai rT 18 20 22 24 Ai rT (° C) 8 10 12 14 Ai rT -6 -4 -2 0 2 4 1870 1890 1910 1930 1950 1970 1990 2010 Ai rT Year 8 10 12 14 1870 1890 1910 1930 1950 1970 1990 2010 Ai rT Year A B C D E yt = 0.004t + 10.171 11Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. for the 20th century in Hungary state that sig- nificant wet sub-periods are detectable in the first half of the 20th century, while dry breaks occurred in the second half of the century (Horváth, Sz. et al. 2005). Szinell, C.S. et al. (1998) also confirmed that droughts generally occurred more frequently in the second half of the century in Hungary. Szinell, C.S. et al. (1998) analysed the PDSI series between 1881 and 1995, and found no change at Keszthely stations in terms of drought. Makra, L. et al. (2005) detected significant wet periods in the first part of the 20th century (between 1901 and 1940) for East Hungary. Hegerl, G.C. (2018) also mentions that the ETCW period was also a period of cold winters. The Hungarian Meteorological Service publishes the time series of extreme cold and hot indices for Keszthely, among other sta- tions in the country, on its homepage (www. met.hu). The graphs published on the web- site suggest that in the ETCW period days of extreme cold (e.g., number of cold days, number of frosty days) and the number of hot days were more frequent. Examining only the data for autumn, the ETCW cannot actually be detected. The only significant period is between 1969 and 2018, with an increasing trend in the seasonal mean tem- perature data. In the time series for winter, no significant trend can be detected in any segment. The beginning of the last segment (1924) is contemporary with the beginning of the ETCW, but the end cannot be detected. With help of the descriptive statistics (mean, standard deviation) of the segments shown in Table 2, and trend analysis, the rea- sons for the presence of a change-point were explored. The change-point in the annual data in 1895 was probably caused by a rise of the mean, and a decline in SD. Between 1931 and 1951, a significant increasing trend is accompanied by a rise in the mean and in SD. From 1952, the mean and SD are lower than in the preceding segment. Between 1985 and 2018, a significant increasing trend can be identified as the reason for the change- point. In spring, all three segments display a significant increasing trend, and between 1925 and 1951, a higher mean and SD are observable. A cooling period can be found in summers between 1871 and 1926, while from 1927 the mean and SD are higher than previously. This is probably the reason for the change-point in 1927. The last period of the examined time series displays a sig- nificant rising trend. The change-point in autumn, 1924, may be the consequence of a rise of the mean and decline in the SD, imply- ing a lower degree of autumn temperature variability compared with warmer years. The period 1969–2018 shows a significant increas- ing trend according to the MK trend test. In the three segments of the winter, an upward shift in the mean followed by a downward shift can be seen. In case of SD, the opposite pattern can be detected. For the ETCW period, significant rising trends can be detected in the annual mean temperature and in spring. The mean tempera- ture is higher than the previous or subsequent periods’ average in the case of the annual data, Table 1. Segments of the time series in which trends can be identified Annual 1871–1894 1895–1930 1931–1951** (Sen’s slope = 0.059) 1952–1984 1985–2018** (slope = 0.042) Spring 1871–1924** (slope = 0.018) 1925–1951* (slope = 0.062) 1952–2018** (slope = 0.018) Summer 1871–1926** (slope = 0.014) 1927–1952 1953–2018** (slope = 0.018) Autumn 1871–1923 1924–1968 1969–2018** (Sen’s slope = 0.024) Winter 1872–1895 1896–1923 1924–2018 *Significant at 10%, **at 5% significance level. http://www.met.hu http://www.met.hu Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.12 and also individually in spring, summer and autumn, too. In comparison, the last part of the 20th century (i.e., well within the current warming period) displays significant warming tendencies for spring, summer and autumn, while the mean temperatures of these periods are slightly lower than those in the ETCW pe- riod. These facts suggest that in the ETCW pe- riod a shift in the mean was observable, while in the current warming period, a slow but significant increasing tendency is occurring. Keszthely has local importance as a tour- ist destination and as a biodiversity reser- vation in Hungary. The high dependence of Keszthely weather on lakeshore conditions determines its characteristics, and this is likely to continue. The significant declin- ing precipitation trend, together with rising temperature trends, may cause declining lake water levels, disturbing tourist activi- ties. The specialty of its meteorological data- set lies in its unique uninterrupted length of 150 years (Kocsis, T. and Anda, A. 2006), and its particular importance in the assess- ment of the water quality of the largest shal- low freshwater lake in Central Europe, Lake Balaton (Hatvani, I.G. et al. 2022), including its semi-constructed water protection sys- Fig. 5. Trends within segments in the summer season. Source: Authors’ own elaboration. Table 2. Mean (M) and standard deviation (SD) of the segments defined in Table 1 Seasons M ºC /SD ºC/ (Period) Annual 10.023 /0.707/ (1871–1894) 10.513 /0.509/ (1895–1930) 10.852 /0.893/ (1931–1951) 10.360 /0.642/ (1952–1984) 10.740 /0.803/ (1985–2018) Spring 10.644 /1.011/ (1871–1924) 11.036 /1.464/ (1925–1951) 10.860 /1.071/ (1952–2018) Summer 20.053 /0.818/ (1871–1926) 20.829 /0.943/ (1927–1952) 20.009 /0.897/ (1953–2018) Autumn 10.469 /1.215/ (1871–1923) 11.344 /1.009/ (1924–1968) 10.560 /1.194/ (1969–2018) Winter -0.975 /2.094/ (1872–1895) 0.715 /1.583/ (1896–1923) 0.367 /1.856/ (1924–2018) 13Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. tem. Worldwide, wetlands cover 5–8 per- cent of total land surface (Mitsch, W.J. et al. 2013), including the Kis-Balaton wetland in the neighbourhood of Keszthely, and these wetlands constitute 20–30 percent of the to- tal global carbon pool. They therefore play a crucial role in climate change mitigation (Dang, A.T.N. et al. 2021). The release of a wetland’s CO2 may also augment the local impacts of global warming. Similar to global trends, the ETCW period can be detected between 1931 and 1951 in the annual mean temperature time series of Keszthely: between 1925 and 1951 a signifi- cant increasing trend is observed for spring mean temperature, and although in the case of summer and autumn, this period cannot be described as displaying any significant trends, an upward shift in the mean can be observed. Similar to the regional results presented in this paper, evidence of significant increments in temperature between 1920 and 1950, then starting from 1985 until nowadays, are also reported in Italy (Herrera-Grimaldi, P. et al. 2018). Brunetti, M. et al. (2000) analysed the time series of annual mean temperature for the period of 1866–1995 and reported that the Italian climate has become warmer and drier, especially in the South, since about 1930. In the continental mid-latitudes of the Northern Hemisphere warm anomalies are often ac- companied by dry conditions. So, besides warming trends during the ETCW, southeast of the present study area, in Bulgaria, sev- eral dry periods were detected in the 1940s by Alexandrov, V. et al. (2004). Moreover, in the 1930s the East European Plain was affected by intense long-lasting droughts (Popova, V. et al. 2022), and these are also considered to be connected to the ETCW. While the previously mentioned studies re- port droughts linked to the ETCW period, in Hungary Makra, L. et al. (2005) detected significant wet periods in the first part of the 20th century (between 1901 and 1940), while dry periods were found in the second half of the century, that is, including under the current warming period. Unfortunately, no literature could be found on investigations of the ETCW and its effects on the Central European region, nor on the Carpathian- Basin, nor on Hungary itself. Conclusions This paper wanted to fill the gap between the wide number of studies about the effects of ETCW in the Artic and limited investi- gations concerning its impact on Europe. Keszthely, with its uninterrupted tempera- ture records since 1871, serves as a focal point for analysing long-term temperature trends. The research uncovers evidence of ETCW in Keszthely’s annual, spring, and summer temperatures, but its indicators remain elu- sive during fall and winter. Furthermore, the study notes that the characteristics of ETCW in Central Europe slightly differ from those re- ported for the broader region. Unfortunate- ly, due to a lack of studies on ETCW in the Carpathian Basin, comparisons could not be made, highlighting a critical area for future research. Acknowledgements: The dataset was provided by the Department of Meteorology and Water Management, Georgikon Campus, Hungarian University of Agriculture and Life Sciences (Keszthely). The authors thank Eric Ruggieri for his constructive com- ments on the methodology description. REFERENCES Alexandrov, V., Schneider, M., Koleva, E. and Moisselin, J.-M. 2004. Climate variability and change in Bulgaria during the 20th century. Theoretical and Applied Climatology 79. (3): 133–149. https://doi.org/10.1007/s00704-004-0073-4 Allan, R., Brohan, P., Compo, G.P., Stone, R., Luterbacher, J. and Brönnimann, S. 2011. The inter- national Atmospheric Circulation Reconstructions over the Earth (ACRE) initiative. Bulletin of the American Meteorological Society 92. (11): 1421–1425. https://doi.org/10.1175/2011BAMS3218.1 Anda, A., Simon, Sz. and Simon-Gáspár, B. 2023. Impacts of wintertime meteorological variables on decomposition of Phragmites australis and Solidago canadensis in the Balaton System. Theoretical and Applied Climatology 151. 1963–1979. https://doi.org/10.1007/s00704-023-04370-y https://doi.org/10.1007/s00704-004-0073-4 https://doi.org/10.1175/2011BAMS3218.1 https://doi.org/10.1007/s00704-023-04370-y Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.14 Béll, B. and Takács, B. (eds.) 1974. A Balaton éghajlata (Climate of the Lake Balaton). Budapest, OMSZ. (in Hungarian) Bokuchava, D.D. and Semenov, V.A. 2021. Mechanisms of the early 20th century warming in the Arctic. Earth-Science Reviews 222. 103820. https:// doi.org/10.1016/j.earscirev.2021.103820 Böhm, R., Auer, I., Brunetti, M., Maugeri, M., Nanni, T. and Schöner, W. 2001. Regional temperature variability in the European Alps: 1760–1998 from homogenized instrumental time series. International Journal of Climatology 21. (14): 1779–1801. https://doi. org/10.1002/joc.689 Brázdil, R., Zahradníček, P., Pišoft, P., Štěpánek, P., Bělínová, M. and Dobrovolný, P. 2012. Temperature and precipitation fluctuations in the Czech Republic during the period of instrumental measurements. Theoretical and Applied Climatology 110. (1): 17–34. https://doi.org/10.1007/s00704-012-0604-3 Brönnimann, S. 2009. Early twentieth-century warm- ing. Nature Geoscience 2. (11): 735–736. https://doi. org/10.1038/ngeo670 Brunetti, M., Maugeri, M. and Nanni, T. 2000. Variations of temperature and precipitation in Italy from 1866 to 1995. Theoretical and Applied Climatology 65. (3): 165–174. https://doi.org/10.1007/ s007040070041 Camuffo, D. and Jones, P.D. 2002. Improved under- standing of past climatic variability from early daily European instrumental sources. In Improved Understanding of Past Climatic Variability from Early Daily European Instrumental Sources. Eds.: Camuffo, D. and Jones, P., Dordrecht, Springer Netherlands, 1–4. https://doi.org/10.1007/978-94-010-0371-1_1 Camuffo, D. 2018. Evidence from the archives of socie- ties: Early instrumental observations In The Palgrave Handbook of Climate History. Eds.: White, S., Pfister, C. and Mauelshagen, F., London UK, Palgrave Macmillan, 83–92. https://doi.org/10.1057/978-1- 137-43020-5_7 Dang, A.T.N., Kumar, L., Reid, M. and Anh, L.N.T. 2021. Modelling the susceptibility of wetland plant species under climate change in the Mekong Delta, Vietnam. Ecological Informatics 64. 1–15. https://doi. org/10.1016/j.ecoinf.2021.101358 Da Silva, R.M., Santos, C.A.G., Moreira, M., Corte- Real, J., Silva, V.C. and Medeiros, I.C. 2015. Rainfall and river flow trends using Mann-Kendall and Sen’s slope estimator statistical tests in the Cobres River basin. Natural Hazards 77. 1205–1221. https://doi.org/10.1007/s11069-015-1644-7 Dee, D., Thorne, P., Noone, S., Baddour, O. and Tassone, C. 2021. The critical role of observations in informing climate science, assessment and policy. Boletín-Organización Meteorológica Mundial 70. (2): 27–33. Delworth, T.L. and Knutson, T.R. 2000. Simulation of early 20th century global warming. Science 287. (5461): 2246–2250. https://doi.org/10.1126/sci- ence.287.5461.2246 Ducré-Robitaille, J.-F., Vincent, L.A. and Boulet, G. 2003. Comparison of techniques for detection of discontinuities in temperature series. International Journal of Climatology 23. (9): 1087–1101. https://doi. org/10.1002/joc.924 Gilbert, R.O. 1987. Statistical Methods for Environmental Pollution Monitoring. London, John Wiley & Sons. Hamed, K.H. and Rao, A.R. 1998. A modified Mann- Kendall trend test for autocorrelated data. Journal of Hydrology 204. 182–196. https://doi.org/10.1016/ S0022-1694(97)00125-X Hatvani, I.G., de Barros, V.D., Tanos, P., Kovács, J., Kovács, I.S. and Clement, A. 2020. Spatiotemporal changes and drivers of trophic status over three de- cades in the largest shallow lake in Central Europe, Lake Balaton. Ecological Engineering 151. 105861. https://doi.org/10.1016/j.ecoleng.2020.105861 Hatvani, I.G., Topál, D., Ruggieri, E. and Kern, Z. 2022. Concurrent changepoints in Greenland ice core & delta: 180 records and the North Atlantic Oscillation over the past millennium. Atmosphere 13. (1): 93. https://doi.org/10.3390/atmos13010093 Hegerl, G.C., Brönnimann, S., Schurer, A. and Cowan, T. 2018. The early 20th century warming: Anomalies, causes, and consequences. WIREs Climate Change 9. (4): e522. https://doi.org/10.1002/ wcc.522 Herrera-Grimaldi, P., García-Marín, A., Ayuso- Muñoz, J.L., Flamini, A., Morbidelli, R. and Ayuso-Ruíz, J.L. 2018. Detection of trends and break points in temperature: The case of Umbria (Italy) and Guadalquivir Valley (Spain). Acta Geophysica 66. (3): 329–343. https://doi.org/10.1007/s11600- 018-0118-1 Hipel, K.W. and McLeod, A.I. 1994. Time Series Modelling of Water Resources and Environmental Systems. Amsterdam, Elsevier. Horváth, Sz., Makra, L. and Mika, J. 2005. 20th cen- tury variations of the soil moisture in East Hungary. Acta Climatologica et Chorologica, Universitatis Szegediensis Tom. 38–39. 85–95. Hungarian Meteorological Service – on its home- page www.met.hu: https://www.met.hu/eghajlat/ eghajlatvaltozas/megfigyelt_hazai_valtozasok/ hosegindexek/, and https://www.met.hu/eghajlat/ eghajlatvaltozas/megfigyelt_hazai_valtozasok/ hidegindexek/(in Hungarian) IPCC 2021. Summary for policymakers. In Climate Change 2021: The Physical Science Basis Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Eds.: Masson-Delmotte, V. et al., New York–Cambridge UK, Cambridge University Press, 3−32. https://doi. org/10.1017/9781009157896.001 Izsák, B. and Szentimrey, T. 2020. To what extent does the detection of climate change in Hungary https://doi.org/10.1016/j.earscirev.2021.103820 https://doi.org/10.1016/j.earscirev.2021.103820 https://doi.org/10.1002/joc.689 https://doi.org/10.1002/joc.689 https://doi.org/10.1007/s00704-012-0604-3 https://doi.org/10.1007/s00704-012-0604-3 https://doi.org/10.1038/ngeo670 https://doi.org/10.1038/ngeo670 https://doi.org/10.1038/ngeo670 https://doi.org/10.1007/s007040070041 https://doi.org/10.1007/s007040070041 https://doi.org/10.1007/978-94-010-0371-1_1 https://doi.org/10.1057/978-1-137-43020-5_7 https://doi.org/10.1057/978-1-137-43020-5_7 https://doi.org/10.1016/j.ecoinf.2021.101358 https://doi.org/10.1016/j.ecoinf.2021.101358 https://doi.org/10.1007/s11069-015-1644-7 https://doi.org/10.1126/science.287.5461.2246 https://doi.org/10.1126/science.287.5461.2246 https://doi.org/10.1002/joc.924 https://doi.org/10.1002/joc.924 https://doi.org/10.1016/S0022-1694(97)00125-X https://doi.org/10.1016/S0022-1694(97)00125-X https://doi.org/10.1016/j.ecoleng.2020.105861 https://doi.org/10.3390/atmos13010093 https://doi.org/10.1002/wcc.522 https://doi.org/10.1002/wcc.522 https://doi.org/10.1007/s11600-018-0118-1 https://doi.org/10.1007/s11600-018-0118-1 http://www.met.hu https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hosegindexek/ https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hosegindexek/ https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hosegindexek/ https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hidegindexek/ https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hidegindexek/ https://www.met.hu/eghajlat/eghajlatvaltozas/megfigyelt_hazai_valtozasok/hidegindexek/ https://doi.org/10.1017/9781009157896.001 https://doi.org/10.1017/9781009157896.001 15Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16. depend on the choice of statistical methods? GEM – International Journal on Geomathematics 11. (1): 17. https://doi.org/10.1007/s13137-020-00154-y Jones, P.D. 2001. Early European instrumental records. In History and Climate. Eds.: Jones, P.D., Ogilvie, A.E.J., Davies, T.D. and Briffa, K.R., New York, Springer Science, 55–77. https://doi. org/10.1007/978-1-4757-3365-5_4 Kendall, M. 1975. Rank Correlation Methods. London, Charles Griffin. Kocsis, T. and Anda, A. 2006. A keszthelyi meteoroló- giai megfigyelések története (The history of the mete- orological measurements at Keszthely). Keszthely, Hungary, PE-GMK Nyomda. (in Hungarian) Kocsis, T., Kovács-Székely, I. and Anda, A. 2020. Homogeneity tests and non-parametric analy- ses of tendencies in precipitation time series in Keszthely, Western Hungary. Theoretical and Applied Climatology 139. (3): 849–859. https://doi. org/10.1007/s00704-019-03014-4 Makra, L., Mika, J. and Horváth, Sz. 2005. 20th century variations of the soil moisture content in East Hungary in connection with global warming. Physics and Chemistry of the Earth 30. 181–186. https:// doi.org/10.1016/j.pce.2004.08.035 Mann, H.B. 1945. Nonparametric tests against trend. Econometrica 13. (3): 245–259. https://doi. org/10.2307/1907187 Meyers, S.R. 2014. Astrochron: An R package for astrochronology. https://cran.r-project.org/ package=astrochron Mitsch, W.J., Bernal, B., Nahlik, A.M., Mander, Ü., Zhang, L., Anderson, C.J., Jørgensen, J.E. and Brix, H. 2013. Wetlands, carbon, and climate change. Landscape Ecology 28. 583–597. https://doi. org/10.1007/s10980-012-9758-8 Moberg, A., Jones, P.D., Barriendos, M., Bergström, H., Camuffo, D., Cocheo, C., Davies, T.D., Demarée, G., Martin-Vide, J., Maugeri, M., Rodriguez, R. and Verhoeve, T. 2000. Day-to-day temperature variability trends in 160- to 275-year-long European instrumental records. Journal of Geophysical Research: Atmospheres 105. (D18): 22849–22868. https://doi. org/10.1029/2000JD900300 Popova, V., Matveeva, T. and Bokuchava, D. 2022. The early 20th century warming in the East European plain climate: Extreme drought in 1920– 1940, atmospheric circulation anomalies and links with the sea ice variability. Environmental Sciences Proceedings 19. (1): 57. https://doi.org/10.3390/ ecas2022-12864 Przybylak, R. 2010. Instrumental observations. In The Polish Climate in the European Context: An Historical Overview. Ed.: Przybylak, R., Dordrecht, Springer Netherlands, 129–166. https://doi.org/10.1007/978- 90-481-3167-9_5 Przybylak, R., Svyashchennikov, P.N., Uscka- Kowalkowska, J. and Wyszyński, P. 2021. Solar radiation in the Arctic during the early twentieth- century warming (1921–50): Presenting a compila- tion of newly available data. Journal of Climate 34. (1): 21–37. https://doi.org/10.1175/JCLI-D-20-0257.1 Quiang, R. and Ruggieri, E. 2023. Autocorrelation and parameter estimation in a Bayesian Change Point Model. Mathematics 11. 1082. https://doi. org/10.3390/math11051082 R Core Team 2019. R: A Language and Environment for Statistical Computing. Vienna, Austria, R Foundation for Statistical Computing. https://www.R-project.org/ Reeves, J., Chen, J., Wang, X., Lund, R.B. and Lu, Q. 2007. A review and comparison of changepoint detection techniques for climate data. Journal of Applied Meteorology and Climatology 46. 900–915. https://doi.org/10.1175/JAM2493.1 Ruggieri, E. 2013. A Bayesian approach to detecting change points in climatic records. International Journal of Climatology 33. (2): 520–528. https://doi. org/10.1002/joc.3447 Ruggieri, E. 2018. A pruned recursive solution to the multiple change point problem. Computational Statistics 33. (2): 1017–1045. https://doi.org/10.1007/ s00180-017-0756-9 Ruggieri, E. and Antonellis, M. 2016. An exact approach to Bayesian sequential change point detection. Computational Statistics & Data Analysis 97. 71–86. https://doi.org/10.1016/j.csda.2015.11.010 Semenov, V.A. and Latif, M. 2012. The early twentieth century warming and winter Arctic sea ice. The Cryosphere 6. (6): 1231–1237. https://doi.org/10.5194/ tc-6-1231-2012 Sen, P.K. 1968. Estimates of the regression coefficient based on Kendall’s tau. Journal of American Statistical Association 63. 1379–1389. https://doi.org/10.1080/0 1621459.1968.10480934 Shapiro, S.S. and Wilk, M.B. 1965. An analysis of variance test for normality (Complete sam- ples). Biometrika 52. (3–4): 591–611. https://doi. org/10.2307/2333709 Stott, P., Strangeways, I. and the Climate Science Communications Group, 2018. Measuring meteoro- logical variables for studying the climate. Weather 73. (10): 332–332. https://doi.org/10.1002/wea.3369 Svendsen, L., Keenlyside, N., Bethke, I., Gao, Y. and Omrani, N.-E. 2018. Pacific contribution to the early twentieth-century warming in the Arctic. Nature Climate Change 8. (9): 793–797. https://doi. org/10.1038/s41558-018-0247-1 Szentimrey, T. 1999. Multiple analysis of series for ho- mogenization (MASH). In Proceedings of the Second Seminar for Homogenisation of Surface Climatological Data. Budapest, Hungary, WMO, 27–46. Szentimrey, T. 2006. Development of MASH homogenization procedure for daily data. In Proceedings of the Fifth Seminar for Homogenization and Quality Control in Climatological Databases, Vol. 29. Budapest, Hungary, 123–130. https://doi.org/10.1007/s13137-020-00154-y https://doi.org/10.1007/s00704-019-03014-4 https://doi.org/10.1007/s00704-019-03014-4 https://doi.org/10.1016/j.pce.2004.08.035 https://doi.org/10.1016/j.pce.2004.08.035 https://doi.org/10.2307/1907187 https://doi.org/10.2307/1907187 https://cran.r-project.org/package=astrochron https://cran.r-project.org/package=astrochron https://doi.org/10.1007/s10980-012-9758-8 https://doi.org/10.1007/s10980-012-9758-8 https://doi.org/10.1029/2000JD900300 https://doi.org/10.1029/2000JD900300 https://doi.org/10.3390/ecas2022-12864 https://doi.org/10.3390/ecas2022-12864 https://doi.org/10.1007/978-90-481-3167-9_5 https://doi.org/10.1007/978-90-481-3167-9_5 https://doi.org/10.1175/JCLI-D-20-0257.1 https://doi.org/10.3390/math11051082 https://doi.org/10.3390/math11051082 https://www.R-project.org/ https://doi.org/10.1175/JAM2493.1 https://doi.org/10.1002/joc.3447 https://doi.org/10.1002/joc.3447 https://doi.org/10.1007/s00180-017-0756-9 https://doi.org/10.1007/s00180-017-0756-9 https://doi.org/10.1016/j.csda.2015.11.010 https://doi.org/10.5194/tc-6-1231-2012 https://doi.org/10.5194/tc-6-1231-2012 https://doi.org/10.1080/01621459.1968.10480934 https://doi.org/10.1080/01621459.1968.10480934 https://doi.org/10.2307/2333709 https://doi.org/10.2307/2333709 https://doi.org/10.1038/s41558-018-0247-1 https://doi.org/10.1038/s41558-018-0247-1 Kocsis, T. et al. Hungarian Geographical Bulletin 73 (2024) (1) 3–16.16 Szentimrey, T. 2008. An overview on the main methodological questions of homogenization. In Proceedings of the Fifth Seminar for Homogenization and Quality Control in Climatological Databases. Climate Data and Monitoring WCDMP. No. 71. 1–6. Szinell, C.S., Bussay, A. and Szentimrey, T. 1998. Drought tendencies in Hungary. International Journal of Climatology 18. 1479–1491. https://doi.org/10.1002/ (SICI)1097-0088(19981115)18:13<1479::AID- JOC325>3.0.CO;2-P Tokinaga, H., Xie, S.-P. and Mukougawa, H. 2017. Early 20th-century Arctic warming intensified by Pacific and Atlantic multidecadal variabil- ity. Proceedings of the National Academy of Sciences 114. (24): 6227–6232. https://doi.org/10.1073/ pnas.1615880114 Topál, D., Matyasovszkyt, I., Kern, Z. and Hatvani, I.G. 2016. Detecting breakpoints in artificially modified- and real-life time series using three state- of-the-art methods. Open Geosciences 8. (1): 78–98. https://doi.org/10.1515/geo-2016-0009 Yamanouchi, T. 2011. Early 20th-century warming in the Arctic: A review. Polar Science 5. (1): 53–71. https://doi.org/10.1016/j.polar.2010.10.002 https://doi.org/10.1002/(SICI)1097-0088(19981115)18:13%3C1479::AID-JOC325%3E3.0.CO;2-P https://doi.org/10.1002/(SICI)1097-0088(19981115)18:13%3C1479::AID-JOC325%3E3.0.CO;2-P https://doi.org/10.1002/(SICI)1097-0088(19981115)18:13%3C1479::AID-JOC325%3E3.0.CO;2-P https://doi.org/10.1073/pnas.1615880114 https://doi.org/10.1073/pnas.1615880114 https://doi.org/10.1515/geo-2016-0009 https://doi.org/10.1016/j.polar.2010.10.002 Hungarian Geographical Bulletin Vol 73 Issue 1 (2024) Tímea KOCSIS - Rita PONGRÁCZ - István Gábor HATVANI - Norbert MAGYAR - Angéla ANDA - Ilona KOVÁCS-SZÉKELY: Seasonal trends in the Early Twentieth Century Warming (ETCW) in a centennial instrumental temperature record from Central Europe