Patterns in the Multiannual Course of Growing Season in Central Europe Since the End of the 19th Century

Next to atmospheric precipitation, air temperature is one of the primary factors determining vegetation development, including growing season start (GSS), end (GSE) and length (GSL). Research on the changes and variability of the growing season (GS) correspond with issues related to climate changes and adaptation to such changes (Chen et al. 2000). GS start and end are considered to be among the best indicators of the vegetation dynamics and long-term biological effects of climate change (Peng et al. 2017, Duarte et al. 2018, Cui, Shi 2021).

In the majority of publications, based on three types of data (phenological, satellite, and climatological), the authors evidenced a considerable increase in GSL in different regions around the globe over the recent decades, caused by its earlier start date and later end date (Chmielewski, Rötzer 2002, Walther, Linderholm 2006, Christidis et al. 2007, Barichivich et al. 2013, Liu et al. 2016, Park et al. 2016, Cui et al. 2017, Koźmiński et al. 2021) A review of results of other studies on GSL until the end of the 20th or first years of the 21st century was performed and published by Linderholm (2006).

According to Xia et al. (2013), it is difficult to compare results globally regarding GSL variability in the last century due to limited access to multiannual daily measurement series. Therefore, in their paper (Xia et al. 2013), the authors applied a linear correlation between the GS start and end date and mean monthly temperature in April and October in the Northern Hemisphere. They determined that in the period 1901–2009, GSL averaged for the study area increased, primarily due to its earlier start. Trends of GSL changes, however, were lower than those resulting from analyses based on data from the second half of the 20th century (Xia et al. 2013). Also, according to Dong et al. (2013), GSS and GSE fluctuations are closely correlated with the mean monthly temperature (Tmean) of April and October, respectively, and GSL was stronger correlated with minimum temperature (Tmin) in spring (March–April) and autumn (September–October). In the GS in Poland in the period 1966–2015, greater changes occurred in thermal than precipitation conditions (Tomczyk, Szyga-Pluta 2019).

The GS and its multiannual changes are usually associated with a potential change in the terms of planting, development, and maturing of plants and harvest yields. GSL, however, does not only affect the vegetation, including plant yield but also the hydrological cycle: snow melt, surface runoff, evapotranspiration, and soil moisture, and translates into an increase (in the case of longer GSL) or decrease (in the case of shorter GSL) in water use (Peterson et al. 2002, Groisman et al. 2004, Backlund et al. 2008, Christiansen et al. 2011). Earlier snow melt negatively affects the shift of maximum flow rate in some rivers (Barnett et al. 2005), and tourism in skiing areas (Beniston 2003, Nicholls 2005).

According to research by Ryan and Archer (2008) and Fagre et al. (2009) in the USA, the observed changes in GSL will also contribute to changes in plant species in forests. Prolongation of GSL also affects life cycles, particularly those of migratory birds, and insect invasions (Cotton 2003, Logan et al. 2003, Janetos et al. 2008). For example, GS prolongation in Germany (4.2 days per decade) and Slovenia (1.0 days per decade) contributed to the prolongation of phenological seasons of oak, birch, and beech, but simultaneously to the reduction of the season of chestnut (−12 days per decade) caused by attacks of pests (Menzel et al. 2008). Many plant and animal species expanded their ranges towards the poles over the last century (Parmesan, Yohe 2003, Haggerty, Mazer 2008). It should be remembered, however, that GS prolongation can cause competitive relations between plants in different ecosystems (Kolářová et al. 2014). Unfavourable effects of GS prolongation are also observed in the reduction of planning of certain crops, among others rice (Peng et al. 2004).

The analysis of the GS is currently based on three types of data: meteorological (air temperature values), phenological observations, and satellite data (Normalized Difference Vegetation Index – NDVI) (Cui, Shi 2021). Results of comparative research from these three types of data conducted by Cui and Shi (2021) in northern China show that the dates of GSS and GSE determined based on air temperature are earlier than those from phenological observations, and those determined from the NDVI index proved the latest. Unfortunately, each of these methods has its advantages and drawbacks. Multiannual (>100 years) GS variability can be determined only based on the former type of data, namely air temperature measurements or alternatively, phenological observations, but these have been and are conducted only locally due to their time and labour-consuming character (Studer et al. 2007). Satellite data are limited only to the period of the last several decades. According to Walther and Linderholm (2006), the spatial and temporal assessment of changes in the GS can only employ measurements of air temperature, because it is considered the primary factor limiting vegetation growth. Moreover, access to air temperature measurements is much simpler, and even measurement series dating back to the 19th or 18th century can be used.

There is no commonly adopted method to thermally define the GS. Therefore, it is difficult to compare results published by different authors. In research on the GS where authors have used meteorological data, i.e. about air temperature, it was usually defined employing constant thermal thresholds. The difficulty in its calculation lies in the determination of temperature thresholds that should be universally applicable at the GSS and GSE (Kexin et al. 2021). For average and higher latitudes, the threshold value commonly adopted as the start of the GS is 5°C, because it is considered the lowest temperature for the growth of plants (Frich et al. 2002, Barichivich et al. 2013, Cui et al. 2018).

The threshold of 5°C was applied among others in research by Jones and Briff (1995), Carter (1998), and Frich et al. (2002), whereas the research considered mean daily or minimum air temperature (Menzel et al. 2003). For example, Frich et al. (2002) defined GSL as a period at the start of which mean daily temperature was above 5°C for >5 subsequent days, and at the end of which it is maintained below 5°C for >5 days. Linderholm et al. (2008) modified this definition so that GSS occurs on the last day of the first 6-day period with a mean daily temperature above 5°C after the last frost, and GSE is the first day of a 10-day period with a temperature below 5°C. Bootsma (1994) defined the analysed season as a period between a 5-day weighted average temperature maintained above 5.5°C and falling below 5.5°C, respectively, and Jones et al. (2002) – start/end as the first/last 5-day period with a temperature above 5°C occurring after/before the last/first frost of the winter season. Unfortunately, differences in the way of determination of specific dates of temperature transition through a given temperature threshold still exist (Walther, Linderholm 2006, Qian et al. 2009). Nonetheless, Xia et al. (2015) suggest that a relevant combination of mean monthly air temperature values can be used as an approximation of the season start.

Few studies analyse the GS based on larger than 100-year air temperature measurement series. Meteorological stations with such long measurement series are located in large cities. Therefore, they do not represent areas under agricultural use or natural vegetation cover. Research by Jones et al. (2002) used a 200–250-year-long series of air temperature measurements in Europe from four locations: Central England, Stockholm, Uppsala, and St. Petersburg. According to the authors, the temperature in the GS increased in Fennoscandia before 1860, and in Central Europe and central England, GSs at the end of the 20th century were approximate to those occurred before 1860. More than 100-year-long data series was also applied by Walther and Linderholm (2006). The authors analysed the Greater Baltic Area and determined that the greatest fluctuations of trends of changes in the GS occurred in Denmark (the differences exceeded half a month), and in Stockholm (Sweden) only small differences were observed. Carter (1998) used measurements during the period 1890–1995 from 10 stations in the Nordic region (Finland, Sweden, Norway, and Iceland). He determined that GS in all stations increased, but the rate of changes, particularly after 1960, was somewhat different, especially in Iceland. The obtained results presented by different authors largely depend on the applied index characterising the GS and a given area, although according to Walther and Linderholm (2006), the criterion of GS determination defined by Jones et al. (2002) (start) and Carter (1998) (end) proves effective in northern Europe.

The objective of this paper is to determine patterns in the multiannual course of dates of the start (GSS), end (GSE), and length (GSL) of the thermal GS in selected cities of Central Europe since the end of the 19th century. Particular attention was paid to the tendency of changes in selected parameters of the analysed season.

The study was based on values of mean monthly air temperature in six stations in Central Europe: Kraków, Potsdam, Poznań, Prague, Toruń, and Vienna (Fig. 1). The analysis covered the longest available period, that is, 1893–2020, and even for all stations. Data from Potsdam and Prague were obtained from national meteorological services, namely Germany's National Meteorological Service and Czech Hydrometeorological Institute, respectively. In the case of the stations in Poznań and Toruń, it was based on archive data provided in publications by Kolendowicz et al. (2019) and Pospieszyńska and Przybylak (2019). Data from recent years were obtained from the archive of the Institute of Meteorology and Water Management – National Research Institute. Data from Kraków were obtained from the research station of the Department of Climatology of the Institute of Geography and Spatial Management of Jagiellonian University. Data from Vienna were obtained from publicly available bases of the historical instrumental climatological surface time series of the Greater Alpine Region (HISTALP). The used data were verified in terms of quality and uniformity, as were relevantly explained in the source materials of such data.

Fig. 1

The aforementioned data provided the basis to first determine the dates of the start and end of the GS, defined as a period with mean daily air temperature of >5°C. The above definition has been previously adopted in numerous studies (Carter 1998, Skaugen and Tveito 2004, Linderholm et al. 2008). The determination of GSS and GSE dates employed mathematical formulas proposed by Gumiński (1948). The method was based on the following assumptions: mean monthly temperature occurs on the 15th day of the month, each month has 30 days, and month-to-month changes in temperature occur evenly. The following formulas were applied: x=30[(tp-t1)/(t2-t1)] for GSS,x=30[(t1-tp)/(t1-t2)] for GSS, \matrix{{{\rm{x}} = 30\left[ {\left( {{\rm{tp}} {\text -} {{\rm{t}}_1}} \right)/\left( {{{\rm{t}}_2} {\text -} {{\rm{t}}_1}} \right)} \right]\;{\rm{for}}\;{\rm{GSS,}}} \hfill \cr {{\rm{x}} = 30\left[ {\left( {{{\rm{t}}_{\rm{1}}} {\text -} {\rm{tp}}} \right)/\left( {{{\rm{t}}_1} {\text -} {{\rm{t}}_2}} \right)} \right]\;{\rm{for}}\;{\rm{GSS,}}} \hfill \cr } where: –

tp – threshold temperature (5°C),

The number of days calculated based on the above formulas is added to the 15th day of the month preceding the threshold temperature. If the resulting number is >15, the addition should consider the actual number of days in a given month. The obtained data is the GS start or end. The described method is commonly adopted in the determination of the GS as well as the other thermal seasons of the year (e.g. Skowera, Kopeć 2008, Szyga-Pluta 2011, Kępińska-Kasprzak, Mager 2015, Tomczyk, Szyga-Pluta 2019). Bartoszek and Siłuch (2015) evidenced considerable conformity of average GSS terms determined using the Gumiński method (1948) and satellite teledetection in the decade 2001–2010.

The next stage involved the calculation of GSL in particular years. Moreover, the five earliest and latest GSS and GSE dates were compared, as well as the shortest and longest GS for each of the discussed stations.

Then, changes in the start and end, as well as the length of the GS in the analysed multiannual period were determined. The direction and rate of changes were assessed using linear regression, and trend significance was verified using a t-Student test at a level of 0.05. The study involved verification of the trend change points. The estimation of optimal breaks in the trend of long-term series, i.e. years pointing to a change in direction or trend intensity in shorter periods, employed the strucchange R package (Zeileis et al. 2002), where the procedure proposed by Bai and Perron (1998) was applied. Then, for the designated subperiods, the direction and rate of changes in the analysed parameters were determined.

In the years 1893–2020, GSS in the analysed stations was characterised by high year-to-year variability, particularly in those located more southwards, i.e. in Prague and Vienna, where the standard deviation was 16.3 days and 14.6 days, respectively. The majority of the latest start dates occurred in the first half of the 20th century. Increasingly earlier GSS was observed from the 1990s in all the analysed stations except for Kraków, where the acceleration of changes occurred already at the beginning of the 1970s. The rate of changes in the analysed period ranged from 1.05 days/10 years in Potsdam to 2.09 days/10 years in Prague. It was determined that the intensification of the rate of the occurring changes was recorded in all stations at the end of the 20th and in the 21st century.

A smaller variability occurred in GSE dates in the years 1893–2020 in the analysed stations, as evidenced by lower values of the standard deviation, ranging from 10.2 days (Toruń) to 13.2 days (Prague). A delay in GSE occurred in all cities by an average from 1.0 day/10 years in Vienna to 1.28 days/10 years in Kraków. More intensive changes in GS end dates occurred at the turn of the 20th century and in the 21st century.

In the years 1893–2020 in selected stations in Central Europe, GSL lasted from 219 days in Toruń to 250 days in Prague and was characterised by approximate variability in all stations (standard deviation from 17 days to 19 days). The length of the analysed season increased in the studied cities from 2.14 days/10 years in Toruń to 3.88 days/10 years in Prague. An increase in season duration was also determined in different regions of Poland, among others by Skowera and Kopeć (2008), Tylkowski (2015), Graczyk and Kundzewicz (2016), Tomczyk and Szyga-Pluta (2019), Koźmiński et al. (2021). The greatest increase in GSL in Poland in the years 1966–2015 was determined at the coast (Szyga-Pluta, Tomczyk 2019), and in the years 1971–2020 in the west of the country (Koźmiński et al. 2021). The increasing tendency of GSL in the regions of coastal forest assemblages at the eastern Baltic coast decreases eastwards (Tylkowski 2015), and is higher than in north-eastern Europe, where it averages 1.5 days/10 years (Linderholm et al. 2008). A similar direction of changes was also observed in other regions of Europe (Carter 1998, Menzel et al. 2003, Jaagus 2006, Linderholm et al. 2008, Irannezhad, Kløve 2015, Potopova et al. 2015) and in China (Dong et al. 2013, Cui et al. 2017). An increase in GSL in Europe by 10.8 days from the 1960s based on 30-year-long observations was also determined by Menzel and Fabian (1999), as confirmed by later phenological observations (Menzel 2000, Stenseth et al. 2002), and by Walther and Linderholm (2006) by 20 days in the Baltic region. The length of the analysed season considerably increased in Eurasia (12.6 days in the years 1950–2011). In North America the change was smaller (6.2 days) (Barichivich et al. 2013). In the boreal and Arctic zone, GSL increased by 2.6 days per decade (in the years 1982–2014), whereas a greater rate of changes was observed in Eurasia and in boreal regions than in North America and Arctic regions (Park et al. 2016).

The rate of changes increased in Kraków, Potsdam, Poznań, and Toruń at the turn of the 20th century and in the 21st century, and in Prague and Vienna approximately a decade earlier. Earlier study results point to a similar tendency (Nieróbca et al. 2013, Tomczyk, Szyga-Pluta 2019). Nieróbca et al. (2013) calculated GSL using approximation of the value of multiannual daily temperature and determined that in the period 2001–2009, the season was longer by 8 days in comparison with that in the years 1971–2000. This is also confirmed in phenological studies because over the last three decades, the thermal sensitivity of spring phenophases has also increased (Wang et al. 2014, Jabłońska et al. 2015). According to Liu et al. (2016), in the period 2000–2013 in comparison to the years 1986–1999, the tendency for prolongation of GS in China weakened. Results by Xia et al. (2013) show that in the years 1901–2009, an increase in GSL was weaker than estimated at the end of the second half of the 20th century due to the multidecadal variability of climate (MVC).

The causes of longer GS are not straightforward. In Toruń and Potsdam, its increase was caused by a somewhat greater shift of the end date, and in the remaining stations, it was determined by its earlier start date. Previous research of the authors in Poland points to a stronger effect of GSE delay on its longer duration (Tomczyk, Szyga-Pluta 2019). Such variability also occurred in Finland: in the north, the end was delayed, the beginning was early in the centre, and in the southwest coast, the shift of both GS start and end was the same (Irannezhad, Kløve 2015). Song et al. (2010) determined the prolongation of the GS primarily as a result of its start earlier by 1.7 days/10 years in northern China, and in the southern part, it was caused by an even shift of the start and end dates. In northern and central Europe, more considerable changes occurring in spring were observed based on phenological changes (Menzel 2000). It is therefore difficult to unequivocally predict the future state of the agroclimate and the resulting level of agricultural production, because irrespective of global tendencies, the role of local conditions will be substantial (Szwejkowski et al. 2017).

Phenological research conducted in the years 1951–1998 showed that in West and Central Europe the acceleration of spring phenophases of plants reached 4 weeks and their delay in East Europe reached 2 weeks (Ahas et al. 2002). An increase in GSL in Europe by 10.8 days from the 1960s based on 30-year-long observations was also determined by Menzel and Fabian (1999), as confirmed by later phenological observations (Menzel 2000, Stenseth et al. 2002). In the Czech Republic, as a result of earlier GS start and later end, the GS was prolonged by an average of 23.8 days over 35 years (Kolářová et al. 2014). Christidis et al. (2007) evidence that an increase in GSL is currently caused by its earlier start, although in the future the shift of both the start and end date will contribute to an increase in its length. Based on multiannual measurement series, the authors of this study determined that these changes occur increasingly faster, as already observed earlier by Song et al. (2010). According to Skaugen and Tveito (2004), the average GSL along the southern coast of Norway will be the same in the years 2021–2050 as in the southern part of Great Britain, Holland, and northern Germany in the years 1961–1998, i.e. approximately 200–210 days.

Earlier studies evidenced linear dependencies between air temperature and vegetation indices in different regions of the world (Xia et al. 2015). Regression results in different places around the globe show that an increase in mean air temperature by 1°C in a relevant combination of monthly values corresponds with a shift of GS in time (particularly its start) (Chmielewski et al. 2004, Piao et al. 2006, Dai et al. 2014, Ge et al. 2014, Wang et al. 2014). According to results by Karlsen et al. (2007), an increase in spring temperature by 1°C generally corresponds with a shift of GS start to 5–6 days earlier, but an evident regional tendency is observed depending on the degree of the oceanic character of the climate. An increase in temperature by 1°C in the oceanic climate corresponds to a shift of GS start to approximately 7–9 days earlier in comparison to <5 days earlier in the continental parts. Research conducted by Parmesan and Yohe (2003) revealed considerable shifts in the range of many plant and animal species by an average of 6.1 km per decade towards the poles (or metres per decade upwards) and considerable average acceleration of spring activity by 2.3 days per decade. The trend analysis by Czernecki and Jabłońska (2016) unequivocally points to the acceleration of the term of flowering of common lilac and European horse-chestnut by an average of 1.7 days per decade and is primarily caused by a rapid increase in temperature observed since the 1990s. In earlier decades, late spring phenophases showed no unequivocal change trends.

Changes in dates of occurrence of phenological changes are determined by temperature fluctuations dependent on changes in the circulation regime (Degirmendžić et al. 2000, Chmielewski, Rötzer 2002, Aasa et al. 2004). According to Bartoszek and Węgrzyn (2011), important predictors of GS start in East Poland include the North Atlantic Oscillation (NAO) index accounting for 13%–29% of its variability, and the year-to-year variability of GS start dates is related to the character of zonal circulation.

A considerable increase in temperature occurred among others in the warm season of the year. It translates into the prolongation of the intensive GS and a substantial increase of heat resources in the period of active plant growth (Żmudzka 2012). Bartoszek and Banasiewicz (2007) also determined that the tendency for the occurrence of GS warmer than the norm that was evident in Central Europe in the decade of 1991–2000 is still maintained. In GS in Poland in the period 1966–2015, greater changes in thermal conditions occurred (an increase in mean air temperature, an increase in air temperature total) than precipitation conditions (Tomczyk, Szyga-Pluta 2019).

Future agroclimatic scenarios developed based on global models show that the observed trends can still increase (Qian, Gameda 2010). Research by Xia et al. (2015) reveals that in the majority of terrestrial areas of the Northern Hemisphere, GS start will be from 4.7 to 10.1 earlier in the period 2040–2059 than in the years 1985–2004, and in the years 2080–2099 it will be from 4.3 days to 21.6 days earlier.

In the period 1893–2020 in Central Europe, GS was prolonged, although these changes in particular stations selected for the analysis occurred unevenly, and were simultaneously the result of different factors. Higher variability of the start of the analysed season occurred in stations in the south of the analysed area (Prague, Vienna). Smaller variability is observed for its end date. The prolongation of GS in Vienna and Prague was determined to the highest degree by its start date, and in the remaining stations – its later end. It is worth emphasising that in the years 1893–2020, two subperiods can be designated differing in the intensity of the rate of changes in the start, end dates, and length of GS. It intensified in all stations at the end of the 20th and in the 21st century. The rate of changes increased in Kraków, Potsdam, Poznań, and Toruń at the turn of the 20th century and the 21st century, and in Prague and Vienna approximately a decade earlier.

The knowledge of trends of changes in the GS is important for the development of plans of adaptation and mitigation of their potential effects, not only in the agricultural sector but also in urbanised areas, including cities analysed in this paper. The variability of the GS also affects the structure and function of ecosystems also occurring in cities.

The performed calculations can be useful for municipal services dealing with the strategy of adaptation to climate change, including those responsible for urban green areas and the choice of appropriate plant species in parks and other green areas.

Over the upcoming years, the observed changes in the GS in urbanised areas will not only have an increasingly strong impact on the phenology of green areas, but also have economic, epidemiological, and bioclimatic consequences to which they will need to adapt. Urban planners, not only of green areas, will have to undertake increasingly frequent strategic activities to mitigate those effects on the environment.