Climate Signals in Earlywood, Latewood and Tree-Ring Width Chronologies of Sessile Oak (Quercus petraea (Matt.) Liebl.) from Majdanpek, North–Eastern Serbia

In this article, the dependence of the sessile oak (Quercus petraea (Matt.) Liebl.) radial growth (tree-ring, earlywood, and latewood widths) on climate (the mean monthly temperature and precipitation totals) was studied in the Majdanpek area, north-eastern Serbia. The growth response of the oak trees to the prevailing climate conditions was dendroecologically investigated, by applying the correlation and response function, as well as by pointer years analysis. The site chronology covered 159 years (1855-2013). We found that latewood and total tree-ring width contain the imprinted positive response to the amount of precipitation in summer months (June and July) of the current growing season. The earlywood width showed no direct dependence on climate data, but it was signifi cantly affected by the previous-year latewood width. Moreover, 40 % of the variation in the latewood width is explained by the earlywood variation in the same season. The temperature was not found to have any signifi cant effect on the growth of oak at the study site. The use of pointer years, determined by applying several calculation procedures, has highlighted previous results, indicating that the precipitation in summer months was the deciding climate factor leading to the occurrence of the years with exceptionally wide or narrow tree-rings and latewood. To enhance our understanding of the response of the sessile oak growth at south-oriented sites with a shallow soil profi le to precipitation and temperature variations, and expand the current database and knowledge, future studies should be undertaken.


UVOD
Although oaks are among the most investigated tree species in Europe in terms of dendrochronology and dendroclimatology, oak dendrochronology is a dynamic and continuously evolving discipline (Haneca et al., 2009). The majority of oak chronologies have been developed in Central and Western Europe (Čufar et al., 2014b). In the Balkan Peninsula, the highest number of oak dendrochronological and dendroclimatological studies have been performed in Slovenia (Čufar and Levanič, 1999 constructed a pedunculate oak radial growth chronology from Kobiljak and investigated the oak climate-growth relationships for this site near Zagreb. This chronology and an oak chronology from Serbia have been used within a network of 41 local oak tree-ring chronologies to detect common climatic signals in oak tree rings in SE Central Europe (Čufar et al., 2014b). The pedunculate oak sensitivity to climate and hydrological parameters in the fl oodplain forests in the lowland Croatia was studied by Mikac et al. (2018). In Serbia, similar research of the pedunculate oak growth and mortality in oak fl oodplain forests, depending on the change of water regime and climate, was performed by Stojanović et al. (2015). Some facts about the dendroclimatological behaviour of sessile oak in Serbia were presented by Stajić et al. (2015). The relationships between the radial increment and stable carbon isotope of Q. robur and Q. cerris and climatic variables have been most recently examined by Kostić et al. (2019), etc.
Despite numerous studies that have been conducted across Europe in the past decades, information about the oak growth-climate response is still lacking in South-Eastern Europe (Stajić et al., 2015). There-fore, given the substantial potential signifi cance of SE Europe for dendrochronological research (Ważny et al., 2014), as well as the insuffi cient number of such studies in Serbia, the fi rst goal of this research was to construct the sessile oak tree-ring chronology for the study site conditions near Majdanpek (North-Eastern Serbia). After defi ning the new oak chronologies, further research efforts will be focused on identifying the most important climatic elements that infl uence the radial growth of this species. To obtain the strongest climatic signal possible, we developed sessile oak chronologies separately for the total ring width (TR), earlywood width (EW), and latewood width (LW).

MATERIJALI I METODE
The research was carried out in a pure mature sessile oak forest (Quercetum montanum tilietosum tomentosae) in a regeneration phase, in the locality of Debeli Lug, Majdanpek, in north-eastern Serbia (Figure 1). The sampling (stem discs at breast height) was conducted at around 300 m a.s.l, on shallow soil of slightly steep, south-facing terrain. The climate data were obtained from the "Debeli Lug" meteorological station (290 m a.s.l.). The available climate data covered the period from 1949 to 2013, but the records were quite incomplete. To fulfi l the gaps and prepare the data for a dendroclimatological analysis correctly, we applied the Inverse distance interpolation with fi ve nearest nearby meteorological stations (Žagubica, Kučevo, Negotin, Donji Milanovac and Bor), and the data were tested for homogeneity. The average annual temperature (9.8 °C) is close to the average value in Serbia amounting to 10.1°C. The warmest and coldest months are July (20.3 °C) and January (-1.3 °C). The annual sum of precipitation is 734 mm/m 2 and it also refl ects the country's average (Popović et al., 2005).
To study oak radial growth-climate relationships, we sampled 19 dominant sessile oak trees with a diameter ranging from 42.6 to 72.3 cm. The measurements of EW, LW, and TR widths were performed in two directions using a Lintab device (38 empirical RAW series). The obtained series were cross-dated visually and .........Radaković, Stajić: Climate Signals in Earlywood, Latewood and Tree-Ring Width... statistically in the R environment (R Development Core Team, 2008). At the same time, the conventional statistical parameters were calculated. The stand w regularly managed and, therefore, the series were detrended by using a cubic smoothing spline having a 50 % cut-off at 67 % of the series length (Cook and Peters, 1981). The series of standardised indices were obtained after the measured widths were divided by the estimated values. The "prewhitened" series were then established as a residual of the autoregressive modelling, where the order for the individual series was determined by the Akaike Information Criterion. Both types of indices were subsequently averaged by applying a biweight robust estimation of the mean value (Cook et al., 1990). Thus, we developed a standard (STD) and a residual (RES) chronology for each part of the radial increment separately. The presence and the strength of the common signal in the series were evaluated by computing several widely-applied dendrochronological parameters: the mean total, the correlation within and between the trees (r tot , r wt , r bt ), Expressed Population Signal (EPS) and Signal-to-Noise Ratio (SNR).
The infl uence of climatic factors on the width of EW, LW, and TR was studied using (1) the correlation analysis between the developed chronologies and climate and (2) the response function analysis. We used the monthly sum of precipitation and the mean temperature data of the previous August to October of the current year (30 independent variables). Besides, TR, EW, and LW from the current and the previous year were also included in the analysis. The signifi cance of the obtained correlation coeffi cients was determined following the bootstrapping procedure (Zang and Bi-ondi, 2015). The temporal stability of the radial growth-climate relationships was evaluated using moving windows. Because of the higher number of predictors included in the regression, the coeffi cients were calculated for 50-year long periods, moved for one year across the common period. Furthermore, pointer years in the longest common period of RES were examined by using three different approaches, i.e., nine calculation variants.

REZULTATI I RASPRAVA
To establish a sound basis for a dendroecological study in a new region, it is necessary to understand the characteristics of the sampled trees of the locally avail- The results of the correlation statistics of the indexed series (standardised -STI and residual -RES) show that the average values of the calculated parameters (r tot , r wt , r bt , EPS and SNR) are almost the same for TR and LW and higher compared to EW (Table 1), indicating that EW contains the lowest common signal. Having averaged the individual series, the STD and RES site chronologies were obtained (Figure 2), and their basic statistical parameters are shown in Table  1B. As the RES chronologies are slightly more sensitive than the STD ones, we decided to use the RES chronologies of TR, EW, and LW for further analysis.
The results of the climate-growth correlations reveal that the EW chronology is signifi cantly positively correlated (r = 0.41, p <0.01) with the LW formed in the previous year (LW t-1 ) and not directly related to the monthly climate data (Figure 3). The fi ndings that EW is correlated with LW t-1 are also confi rmed by the results of the response function (Figure 3), indicating that 16 % (p <0.01) of the total variation are related to LW t-1 .
The insensitivity of the EW chronologies to the studied monthly climate parameters was also determined by Nechita and Popa (2011) for oak growing in Vaslui region, Romnia. In some cases, when the direct climate-growth relationship could not be confi rmed, the dependence of EW on the LW of the preceding year can be noticed (García González, Eckstein 2003; Sohar et al., 2013). Since LW refl ects the amount of summer precipitation, this can be seen as a kind of "indirect" climate infl uence on the EW width in the following year. Contrary to EW and LW, the TR chronologies  (1880 -2013), (B) the main statistical parameters of the STD and RES oak chronology (1855 -2013). Abbreviations: r tot -total correlation, r wt -within-trees correlation, r tb -between-trees correlation, EPS -Expressed Population Signal, SNR -Signal-to-Noice Ratio, SD -standard deviation, MS -mean sensitivity and AC1 -Autocorrelation order 1. . Kratice: r tot -ukupna korelacija, r wt -korelacija unutar stabala, r tbkorelacija između stabala, EPS -izraženi signal populacije, SNR -omjer signal : šum, SD -standardna devijacija, MS -srednja osjetljivost, AC1 -autokorelacijski red 1. were not signifi cantly related to any of the analysed growth independents from the previous growing season ( Figure 3). Further, the TR and LW chronologies were positively infl uenced by the precipitation in the current June (r TR = 0.42, r LW = 0.41) and July (r TR =0.32, r LW = 0.34), indicating that the high precipitation during the summer months of the current year is of the utmost importance for the formation of sessile oak TR and LW in the studied conditions. The LW chronology was also strongly correlated (r = 0.61) to the EW of the current year (EW t ). The results of the conducted response function additionally clarifi ed the obtained fi ndings, suggesting that the widest LW and TR are expected to be produced in the years with abundant June precipitation. Namely, these results showed that TR was positively infl uenced by the current June precipitation and this variable explained 17 % (p<0.001) of the total variation in the TR chronology.

Chronology type
Climate response and correlation models for LW found June precipitation and EW t to be statistically signifi cant. These variables (16 % -June precipitation and 40% -EW t , p <0.001) explained almost 50 % of the total variance in the LW chronology. It must be noted that neither the correlation nor the response analysis found any signifi cant infl uence of monthly temperatures. Such dependency of the sessile oak radial growth on water availability in early summer is not typical for the growth of sessile oak in the northwestern part of Serbia, on Fruška Gora Mountain (Stajić et al., 2015).
These authors found no statistically signifi cant relationships between the sessile oak growth and the current summer monthly precipitation data. The main reason for the observed divergence in the growth reactions of sessile oak is the fact that the soil of the stand on Fruška Gora Mountain is characterized as deep, highquality, and well-drained soil with a good water-air regime. On this soil, oak trees withstand a lack of precipitation and high temperatures in June and July without much diffi culty (Stajić et al., 2015). However, in case of further exposure to high temperatures in August, followed by small amounts of precipitation, it can be expected that typically low growth of oak in August could be even signifi cantly lower, as reported by Stajić et al. (2015) for oak in the north-western part of Serbia. On the other hand, oak trees at warmer, south-oriented sites with a shallow soil profi le are highly exposed to summer droughts, as is the case in the present study. In such circumstances, during dry summer months, trees suffer from water defi cit that acts as a growth-limiting factor.
Besides that, June is characterized by a prolonged photoperiod, which can be an additional reason for increased demand for water supply. Therefore, in case of wet and sunny conditions in June, the maximum wood growth can be expected in many sites in SE Central Europe (Čufar et al., 2014a). However, Čufar et al. (2014a) concluded that there were some similarities, but they also found some differences concerning the  did not fi nd a statistically signifi cant relationship between the oak growth and monthly temperature or precipitation data of the current year. Also, no signifi cant impact of precipitation on the oak growth in Serbia was found for the pedunculate and Turkey oak growing in the lowland area in the northern part of the country (Stojanović et al., 2015).
To evaluate the temporal stability of the dendroclimatic signal observed, we implemented the bootstrap moving response function. The months, which had been previously highlighted as important for the radial growth, were included in this procedure. The results indicated that the detected impact of monthly climate conditions established quite a stable correlation across the common period ( Figure 4). The July correlation is signifi cant in each investigated position of the window, while the June correlation showed occasional insignificance. A similar performance was ascertained for the response coeffi cients in the TR, while the most unstable correlation was revealed for the LW chronology.
The detected link between the sessile oak TR and LW growth and the precipitation in summer months was confi rmed by the results of the pointer year analysis ( Figure 5). The determination of the pointer years represents a common practice applied in dendroclimatological studies, but the existence of various calculation procedures has caused a sort of selection bias. Therefore, Stajic et al. (2017) and Jetschke et al. (2019) highlighted the need for standardisation, which would enable reliable comparisons of various results. To address this problem, we have proposed that only those years that are determined by two or more methods used here can be marked as pointer years. Following these recommendations, we determined positive (1997,1982,1969,1926,1920,1910,1903, and 1884) and negative (2006,1962,1956,1940,1902 and 1891) pointer years common to LW and TR. The positive years of 1994 and 1887 were identifi ed only for TR, while an exceptionally narrow LW portion was formed during 1998, 1984, and 1886. A strict routine in terms of selected thresholds and the use of more than one of the applied methods resulted in the identifi cation of noticeably fewer pointer years (16 for TR and 17 for LW) than in some other oak dendroclimatologial studies. The applied procedures for the pointer year calculation detected the absence of the years with an exceptionally wide or narrow EW portion of a tree ring. The detected pointer years occurred with precipitation anomalies. Both types of pointer years for TR and LW are related to the below and above long-term averages of precipitation in summer months. Three common positive pointer years for TR and LW were 1997, 1982, and 1969, which were characterized by exceptionally abundant precipitation summer months. Compared to the average reference period values , these years in August (1997), July (1982), and June (1969) had 172 %, 209 %, and 257 % more precipitation. Three common positive pointer years for TR and LW that occurred in the years with exceptionally abundant precipitation in July (1982), June (1969), and August (1997) were equal to or higher than the reference period values .
Within the group of negative pointer years, pronouncedly low values of TR and LW were determined by different dry periods of the current year. Namely, these common pointer years (2006, 1962, and 1956) were characterized by the following prevailing lowamounted precipitation drivers: Sep-Oct, May and August, respectively, when the amount of precipitation was 67-91 % below the 30-year average value. Extremely narrow LW widths are related to very pronounced anomalies in the precipitation in June (1998) and August (1984), with 53 % and 19 % of the 30-year average precipitation value, respectively.
The present climate-growth results are mostly in concordance with the common feature of the oak's response in the Balkans and surrounding countries. In Croatia, the pedunculate oak growth was positively

CONCLUSIONS 4. ZAKLJUČAK
Conversely to many other regions and countries, dendroclimatological studies have not been intensively carried out in Serbia. The results of this study showed that the main factor limiting the process of the TR and LW growth formation in the sessile oak from Majdanpek (Eastern Serbia) was summer precipitation, especially in June. The fact that a higher amount of summer precipitation increases the TR and LW width was confi rmed by the multi-approach dendroclimatological procedure. The absence of a signifi cant negative reaction of sessile oak to temperature is not such a common feature of the oak response in many European and Balkan countries. Additionally, the impact of the previous year's climate on the sessile oak growth was not detected. Finally, the main limitation of this research is the spatial representativeness of the sampled stands, since the obtained results represent the fi rst insight into the dendroclimatological behaviour of sessile oak in this part of Serbia. To achieve a better and deeper insight into the reaction of the sessile oak growth at south-oriented sites with a shallow soil profi le to precipitation and temperature variations, as well as to ex-  pand the existing database and knowledge, future studies should be undertaken.