1 Introduction

During medieval and historically modern times, agriculture has played a key role in the maintenance of social stability, even the survival of political regimes (e.g. Federico 2008; van Cruyningen and Thoen 2012). Bad crop yields for two or more years running lead, in association with general political, socio-economic and demographic factors, to extensive famine. This has occurred in many European countries in the historical past (e.g. Alfani 2010; Campbell 2010; Pfister 2010; Dybdahl 2012; Gerrard and Petley 2013; Collet and Schuh 2017). Similar consequences and famines are, of course, also familiar to the Czech Lands (recently the Czech Republic), particularly those that occurred in the early 1280s, the 1310s, the 1430s and the early 1770s (e.g. Brázdil et al. 2001, 2017; Pfister and Brázdil 2006).

Such crises were also strongly influenced by crop failures. In the pre-industrial period, subsistence-oriented and natural/autarchic agriculture was particularly dependent on human labour and draught-animal power. The equal distribution of labour over the course of the entire year, often summarised in terms of the ‘drudgery-averse peasant’, and the minimisation of the risk of crop failure (the ‘risk-averse peasant’) were therefore crucial (Halstead and O’Shea 1989; Ellis 2003; Beck 2004; Vanhaute et al. 2011). Various land-use and complementary agricultural cultures were combined, which did not cover the same vegetation cycles and were supplemented in time segments by land tillage, care and harvest, even if, for some, the natural conditions were not optimal for the given site. The success of the yields depended on whether all the agricultural activities were managed in timely fashion and optimally executed (Landsteiner and Langthaler 2010). In the prevailing milieu of incomplete supplies and often inadequate marketing, every failure meant an existential threat to the populace, with the risk of famines and social upheaval (e.g. Katajala 2004; Suter 2004; Daim et al. 2011; van Molle and Segers 2013; Firnhaber-Baker and Schoenars 2017).

It is obvious that the agricultural cycle, so crucial to crop yields and quality, is significantly influenced by meteorological and climatological factors all the way from sowing, through growth, to ripening and harvesting. Climate and weather conditions influence all the key plant physiological processes, including photosynthesis, water and nutrient uptake, and phenological development (e.g. Porter and Gawith 1999; Larcher 2003). The climate directly influences not only general agro-climatic patterns (e.g. Eitzinger et al. 2013) but also the likelihood of extreme events (e.g. Peltonen-Sainio et al. 2010; Trnka et al. 2011b). It is also thought to shape key soil properties over the long-term (Jenny 1941). All these relationships are likely to be further pronounced in conditions of changing climate (e.g. Trnka et al. 2013). Among other things, the soil water regime is likely to alter over the whole of central Europe, affecting the balance of not only water but also of nutrients and carbon cycling (e.g. Hlavinka et al. 2014). The relationships between climatic conditions and pest outbreaks have been pointed out (e.g. Trnka et al. 2007), while the indirect pathways of weather/climate in influencing agricultural crops through pest, disease and weed populations have been highlighted e.g. by Porter et al. (1991). Shifts in the likelihood and severity of extreme weather events (especially heat waves, droughts and heavy precipitation) have been shown to increase considerably the risk of crop failure and exacerbate yield variability across Europe (e.g. Peltonen-Sainio et al. 2010; Semenov and Shewry 2011), including the Czech Republic (Kolář et al. 2014; Trnka et al. 2016).

Despite general progress in agro-technology, including modern tillage, plant protection approaches, precision agriculture, weather forecasting systems and selective crop breeding, agriculture still remains one of the economic sectors that depend strongly on seasonal weather. Recent analysis at a European scale has shown that inter-annual variability in regional yields of major crops remains virtually unchanged since the early 1900s, although it has increased significantly in absolute terms (e.g. Trnka et al. 2016). Some studies (e.g. Trnka et al. 2012) have reported that the crop-weather relationship has been deteriorating over time due to the changing climate, in spite of the incontrovertible success technological advances. Shifts in the overall agro-climatic conditions are inevitable in the region of the Czech Republic (Trnka et al. 2011a) and beyond, with the likelihood of agro-climatic extremes also increasing (Trnka et al. 2014, 2015). Even in the recent times, extreme weather events (drought, late spring frosts, heat waves, hailstorms, etc.) may result in very low yields (e.g. Kolář et al. 2014) and/or lower quality of agricultural crops (Žalud et al. 2017). This may have significant economic and socio-political impacts, as was, for example, the case after the extreme European drought of 1947 (Brázdil et al. 2016b), leading to famine (Below et al. 2007) and a sharp increase in suicide rates (Carleton 2017) in various parts of the world. Recently, in a period marked by increasing population, devastation of the environment and global change, climatic factors have proved serious enough to stimulate urgent debate about securing sufficient food sources on regional as well as global scales (e.g. Lesk et al. 2016).

There can be no doubt that long-term climate variability has significantly influenced the agricultural cycle over the centuries. This facilitates the use of long-term agricultural series to reconstruct climatic characteristics. For example, Možný et al. (2012) used a series of winter wheat harvest dates for the Czech Lands to reconstruct March–June temperatures for the 1501–2008 period. Series of grape-harvest dates for the Bohemian wine- and hop-growing region, located north-west of Prague, have been employed to reconstruct April–August temperatures from AD 1499 (Možný et al. 2016a) and April–August Standardised Precipitation Evapotranspiration Index (SPEI) (Možný et al. 2016b). Moreover, rich documentary evidence and an extended tradition of meteorological observations in the Czech Lands enables the reconstruction of temperature, precipitation and extreme weather patterns since AD 1501 (Dobrovolný et al. 2010, 2015), which may be supplemented by instrumental meteorological records from Prague-Klementinum since 1775 (Pejml 1975) and by measurements of other meteorological stations from the early nineteenth century onwards (Brázdil et al. 2012). All this opens up the possibility of joining series describing past agricultural cycles to climate variability.

The basic issue investigated in this contribution is how the agricultural cycle changed under the influence of long-term climatic variability in the Czech Lands over the past five centuries. This is analysed in terms of the Louny and Elbe regions in Bohemia in five model periods, from the sixteenth to the twenty-first century (1517–1542, 1561–1622, 1770–1815, 1871–1910 and 1971–2010). Their selection expresses the availability of agricultural data on the one hand and different stages of socio-economic and agricultural development in the country on the other. Section 2 briefly describes the areas studied, the basic data for the five model periods, and the available phenological and climatological data. Critical evaluation of approaches to documentary data and methods of statistical analysis are presented in Section 3. Section 4 provides the results of the analysis for the five model periods and comparison of them over the past 500 years. Section 5 discusses data uncertainty, the relation of the agricultural cycle to climatic variability and agricultural cycles in the long-term and European contexts. It is followed by some concluding remarks.

2 Area and data

2.1 Areas studied

The study concentrates on two model regions, for which information about the agricultural cycle reaches back the furthest. The Louny region in the north-west Bohemia consists of lowland positions around the lower River Ohře (Eger), approximately from Žatec to the confluence with the River Elbe (the locations of areas and places mentioned appear in Fig. 1, in the context of the recent territory of the Czech Republic). The Elbe region in central Bohemia includes lowland positions around the central part of the River Elbe, approximately between Kolín and Mělník. Both regions are very fertile agricultural areas, with prevailing chernozems in the Louny region. The Elbe region offers more diverse soil patterns. In addition to chernozems, calcaric regosols and arenosols are more frequent there (Kozák et al. 2009). Both regions also exhibit generally favourable agro-climatic conditions, with mean annual temperatures achieving 8.6–8.9 °C (14.9–15.3 °C in April–August), with the same temperature patterns. However, important differences exist in precipitation totals. The Louny region is much drier, with only ~ 470 mm annual total and ~ 250 mm in April–August, while in the Elbe region, the corresponding figures are ~ 540 and ~ 280 mm (data for the 1961–2000 period; see Tolasz et al. 2007).

Fig. 1
figure 1

Location of the two areas studied (1—the Louny region, 2—the Elbe region) and places within the territory of the Czech Republic reported in this paper

Full size image

2.2 Agricultural data

2.2.1 The Louny region (1517–1622)

Rich documentary evidence related to agricultural activities from the Louny region consists of books of accounts for the town of Louny, in which financial outlays for various kinds of work, including agricultural assistance, appear in the Ditributa section. The payment was always made on Saturdays (i.e. on the last day of the Christian week, which started on Sunday) so that the wages paid express the work for the whole week. For work in the fields, the record books related to farmsteads in the villages of Dobroměřice and Raná, belonging to the town, are the most important (see Brázdil and Kotyza 2000 for archival sources and further details). The corresponding books for Dobroměřice are the Registra špitální (the ‘Spital [hostel] Registers’) for 1517–1543, then taken up by four volumes of Registra vydání (Registers of Expenditure) for 1560–1571, 1571–1586, 1586–1605 and 1605–1622. Two volumes of Registra dvoru ranskýho (Registers of the Raná Farmstead), covering the years 1598–1613 and 1613–1622, cover the second village. The style of records in the Louny books changed in 1623, when weekly records of wages became monthly; the latter data are no longer suitable for our purposes.

An example of records, related to a part of the agricultural cycle in 1524, expresses to whom wages were paid, and for what kind of work (archival source 3 – further AS3, fol. 76–78; dates given in the Gregorian calendar): ‘[5 July] To reapers at Obora; [12 July] To farmers of Nečichy and Dobroměřice at the raking of hay in Trávník, for white bread as is usual; […] To ten servants for carrying hay for two days; [19 July] To labourers for carrying hay for two days. To 53 harvesters at the barley. To binders at the barley; [26 July] To 20 harvesters at the rye; [2 August] To harvesters on Monday (Tuesday, Wednesday, Thursday). To 14 harvesters on Saturday; […] [13 September] Given to reapers on aftermath […]; [11 October] From the hop-picking’. These records can be interpreted to indicate that, at Dobroměřice in 1524, the hay harvest started in the week before 5 July, the barley harvest in the week before 19 July, the rye harvest in the week before 26 July, the aftermath in the week before 13 September, and finally hop-picking in the week before 11 October.

The records reported, covering the 1517–1622 period with a gap of agricultural data in 1543–1560, enable close description of some parts of the agricultural cycle, represented particularly by harvest times for hay, grain and aftermaths. These were further supplemented by series of grape harvest dates for the Louny region compiled from various documentary sources: city council records from Louny and Litoměřice, financial records of the aristocratic Lobkowicz and Nostitz families, financial records of the Archdiocese of Prague, and records kept by individual farmers.

2.2.2 The Elbe region (1770–1815)

The Elbe region is represented here by the records kept by František Jan Vavák (Fig. 2), a peasant and reeve/steward living in the village of Milčice (Skopec 1907, 1908, 1910, 1912, 1915, 1916, 1918, 1924, 1936, 1938; Jonášová 2009). His Memoirs, a systematic and extended record, cover 46 years, consisting of seven volumes and almost 3000 pages. František J. Vavák, born in 1741, began to write his records in the early spring of 1770 and continued to do so almost until he died in 1816. In relatively systematic fashion, he noted variations in the weather, the seasonal course of agricultural work, the growth and harvest of crops, the course of the grazing of cattle and also the prices of agricultural products at the local and regional markets (Klír 2008; Klír and Vodáková 2017). He built upon this basic, routine and more-or-less stable framework with other records of diverse character and extent (Kutnar 1941; Jonášová-Hájková 1978, 1979).

Fig. 2
figure 2

a František Jan Vavák (1741–1816)—hypothetical portrait (Skopec 1907). b An example of Vavák’s manuscript in Memoirs from 1782 (Skopec 1908)

Full size image

For example, Vavák recorded the following information related to harvest-time in 1785 (Skopec 1910, p. 85): ‘On 2 August, when Portiuncula was celebrated, we started the harvest here; in two previous years we also paid reapers on this day. To a mile around us [the distance at which the harvest will begin] still in a week yet’. Some lines further on, he continues with more harvest information and also describes the weather patterns (ibid.): ‘Wet and cold weather started around the Assumption of the Virgin Mary [15 August]. Wheat, barley and oats had ripened insufficiently, and those [crops] that did ripen were difficult to harvest. It rained nearly every night, also during the day, due to which grain grew both horizontally and upright, and when the weather was warmer, more damage could be seen; because of the cold, the grain could not grow quickly [enough]. In the whole of June only 3 days and 5 nights were warm, in the whole of July 2 days and 2 nights, and in all of August 4 days and 2 nights, which I could observe very well. On the day of the Assumption of Virgin Mary, the cold was very severe and on 16–17 August even worse’.

Although Vavák’s Memoirs contain information about the key agricultural activities, such as ploughing and sowing, harvest of winter and spring crops, hay-making and the pasturage of cattle and sheep, only the harvest dates of rye are systematic (with just 3 years missing). This series was also supplemented by a series of grape-harvest dates compiled for the Elbe region from various documentary sources: city council records from Mělník and Litoměřice, the financial records of the aristocratic Lobkowicz family, the financial records of the Archdiocese of Prague and of the Order of Cistercians, and of individual farmers.

2.2.3 The Louny and Elbe regions (1871–1910, 1971–2010)

Agricultural series for the further two periods studied (1871–1910 and 1971–2010) were derived from the PHENODATA database of the Czech Hydrometeorological Institute (Svitáková et al. 2005), which contains phenological and agricultural data for after 1845. The sources used for creation of harvest series in the first of these two periods were, in particular, societies or bodies that collected and reported such data: (i) K. k. patriotisch-ökonomische Gesellschaft im Königreiche Böhmen (the I. R. Patriotic-Economic Society of the Kingdom of Bohemia), (ii) Landesculturrathes für das Königreich Böhmen (the Agricultural Council for the Kingdom of Bohemia) and (iii) Fysiokratická společnost Praha (the ‘Physiocratic’ Society, Prague). Among further agricultural and economic reports, those of the aristocratic Lobkowicz family and various church bodies such as the Archdiocese of Prague and the Order of Cistercians are worthy of note. The two regions studied were covered by 254 records of the grain harvest and 186 records of the grape harvest in the 1871–1910 period. At least two records were available for each year; the harvest of that year was then obtained by averaging the data available.

For the most recent period of 1971–2010, the PHENODATA database includes observations of field crops for 74 stations provided by voluntary observers, methodically managed by the CHMI and based on unified methods of observation (Hájková et al. 2012). In the two regions studied, a total of ten stations with complete observations were available, making up a total of 386 grain-harvest observations and 208 grape-harvest observations. More observations in any given year were again averaged.

2.3 Climatological data

Several reconstructions of climatological data from AD 1501 onwards may be used to characterise long-term climatic variability over the territory of the Czech Lands. Dobrovolný et al. (2010) calculated series of monthly temperatures for central Europe, based on monthly temperature indices for Germany, Switzerland and the Czech Lands derived from documentary evidence for 1501–1854, and on homogenised monthly mean temperatures for 11 climatological stations for 1760–2007 (Prague-Klementinum included), using the standard paleo-climatological method of reconstruction (Dobrovolný et al. 2009). Final series of monthly, seasonal and annual temperatures consist of reconstructed values up to 1759 and of measured temperatures since 1760. They are fully representative for the Czech Lands.

A similar approach was applied by Dobrovolný et al. (2015) to the reconstruction of series of annual and seasonal precipitation totals. Monthly precipitation indices derived from Czech documentary evidence for 1501–1854 and mean precipitation series for the Czech Lands in 1804–2012 (Brázdil et al. 2012) were used for the final reconstruction, which is based on precipitation indices arising out of documentary data up to 1803 and on instrumental measurements from 1804 onwards.

Both the temperature and precipitation series were used by Brázdil et al. (2016a) for calculation of long-term series of four drought indices from AD 1501 onwards: Standardised Precipitation Index (SPI), SPEI, Palmer Drought Severity Index (PDSI) and Palmer’s Z-index. They were created in order to describe short-term (seasonal), medium-term (summer half-year) and long-term (annual) droughts.

Brázdil et al. (2012) used long homogeneous temperature (10 stations) and precipitation (14 stations) instrumental measurements starting in the nineteenth century to compile series of mean monthly temperatures and mean monthly precipitation over the entire territory of the Czech Lands/Republic for the 1800–2010 and 1804–2010 periods, respectively.

3 Methods

The work with agricultural and meteorological/climatological documentary data is based on critical evaluation of existing sources obtained by extraction of data from the books of accounts for the Louny region (Brázdil and Kotyza 2000) and from the Memoirs of František Jan Vavák (Skopec 1907, 1908, 1910, 1912, 1915, 1916, 1918, 1924, 1936, 1938; Jonášová 2009). These editions of original records are guarantees of source credibility. Records extracted from them were first established in place and time, then analysed with respect to content and interpreted with the aim of deriving relevant information prepared in the form of time series and used for further analyses. If necessary, these were supplemented with other documentary records from the Czech Lands (e.g. in descriptions of individual years) kept in the historical-climatological database of the Institute of Geography, Masaryk University, Brno (AS1).

The analysis of the relationship of the agricultural cycle to climatic variability is based on comparison of existing agricultural data with climatological data in five model periods: 1517–1542, 1561–1622, 1770–1815, 1871–1910 and 1971–2010. Agricultural data are presented as series expressing the dates of any related work in the fields (e.g. harvest), for which mean dates, variability expressed by standard deviation, earliest and latest dates, and Pearson’s correlation coefficients with air temperature, precipitation and SPEI were calculated. The statistical significance of correlation coefficients was evaluated by means of t test at a significance level of α = 0.05. In addition to harvest beginnings for individual crops (such as hay, oats and aftermath), data from the Louny region for the entire 1517–1542 and 1561–1622 periods permitted the creation of series for the beginnings and ends of grain harvests (barley, wheat, rye and oats are considered together; records did not consistently specify the type of grain harvested). From these dates, the duration of the grain harvest (number of days) was then calculated (Table 1 and Fig. 3).

Table 1 Basic statistical characteristics of agricultural series in the Louny region and their correlations with selected Czech climatological variables (T temperature, P precipitation) during the 1517–1622 period (GH grain harvest, STD standard deviation, A–S summer half-year). For grain harvest duration, the mean and the shortest and longest duration/year are expressed in days. Statistically significant correlation coefficients for α = 0.05 appear in italics
Full size table
Fig. 3
figure 3

Fluctuation in dates upon which wages were paid for harvesting work at Dobroměřice (hay harvest, beginning, end and duration of grain harvest, oat harvest, and aftermath) in comparison with MAMJ temperatures, MAM and JJA precipitation, MAM and JJA SPEI in the Czech Lands during the 1517–1542 and 1561–1622 periods. Smoothed by 5-year Gaussian filter

Full size image

Climatic variables were presented for the summer half-year (April–September) in the case of grape-harvest dates and separately for spring (MAM), summer (JJA) or other combinations of months showing the highest correlations to harvest dates (e.g. AMJ or MAMJ). Temporal climatic variability for the model periods was documented in the form of fluctuations in corresponding climatic variables, expressed—with regard to the series available (see Section 2.3)—as anomalies with respect to the 1961–1990 reference period in the first three periods (Figs. 3, 4, 5 and 6) and in real values for the last two periods (Figs. 7, 8, 9 and 10). Fluctuations in agricultural and climatic series were smoothed by 5-year Gaussian filter.

Fig. 4
figure 4

Fluctuations in the beginning of the grape harvest in the Louny region in comparison with summer half-year (A–S) temperatures, MAM and JJA precipitation and summer half-year (A–S) SPEI in the 1517–1622 period. Smoothed by 5-year Gaussian filter

Full size image
Fig. 5
figure 5

Fluctuations in the beginning of the rye harvest at Milčice in comparison with MAMJ temperatures, MAM precipitation and MAM SPEI in the Czech Lands in the 1770–1815 period. Smoothed by 5-year Gaussian filter

Full size image
Fig. 6
figure 6

Fluctuations in the beginning of the grape harvest in the Elbe region in comparison with the summer half-year (A–S) temperatures, MAM and JJA precipitation and summer half-year (A–S) SPEI in the 1770–1815 period. Smoothed by 5-year Gaussian filter

Full size image
Fig. 7
figure 7

Fluctuations in the beginning of the spring barley and winter wheat harvests in the Louny and Elbe regions in comparison with climatological characteristics in the 1871–1910 period: AMJ temperatures, MAM precipitation and MAM SPEI for spring barley, MAMJ temperatures, MAMJ precipitation and MAM SPEI for winter wheat. Smoothed by 5-year Gaussian filter

Full size image
Fig. 8
figure 8

Fluctuations in the beginning of the grape harvest in the Elbe region in comparison with summer half-year (A–S) temperatures, MAM and JJA precipitation and summer half-year (A–S) SPEI in the 1871–1910 period. Smoothed by 5-year Gaussian filter

Full size image
Fig. 9
figure 9

Fluctuations in the beginning of the spring barley and winter wheat harvests in the Louny and Elbe regions in comparison with climatological characteristics in the 1971–2010 period: AMJ temperatures, MAM precipitation and MAM SPEI for spring barley, MAMJ temperatures, MAMJ precipitation and MAM SPEI for winter wheat. Smoothed by 5-year Gaussian filter

Full size image
Fig. 10
figure 10

Fluctuations in the beginning of the grape harvest in the Elbe region in comparison with the summer half-year (A–S) temperatures, MAM and JJA precipitation and summer half-year (A–S) SPEI in the Czech Lands in the 1971–2010 period. Smoothed by 5-year Gaussian filter

Full size image

4 Results

4.1 Climate and the agricultural cycle in the Louny region for 1517–1622

Fluctuations in the dates for harvesting work in the Louny region, expressed as the dates upon which wages for the previous week were collected, are shown for the Dobroměřice farmstead in Fig. 3. Generally decreasing trends in the beginning of the hay harvest as well as in the beginning and the end of the grain harvest (i.e. earlier onset) between 1517 and 1542 appear as a remarkable feature of inter-annual fluctuations. Later starts for grain harvests, as well as for aftermath, appear in the second period, from the 1560s.

Table 1 provides basic information about the variability of agricultural series and their relationships to climatic variables. Comparing the 1517–1542 and 1561–1622 periods, earlier mean beginnings for hay and aftermath harvest and higher variability in these dates characterised the second period, while the grain harvest started and finished earlier in the first period and oats were harvested in both before 15 August. The remarkably warm and dry year of 1540 (Brázdil et al. 2013; Wetter and Pfister 2013; Wetter et al. 2014; Orth et al. 2016) was reflected in the earliest dates for harvests of hay, oats and grapes and in the end of the grain harvest. In contrast to 1540, the latest occurrence for the same crops was in 1529, except for the grape harvest, but also for the longest duration of grain harvest. Documentary sources reported floods in that year, while there were wet and cold patterns in the summer half-year (Brázdil et al. 2013). In the 1561–1622 period, in similar fashion to 1540, the year 1616 had a particularly dry and warm summer (Brázdil et al. 2013), while the latest beginnings for harvests were not attributed to any particular year. The highest correlation coefficients for the oat harvest in both periods were obtained for JJA SPEI (0.591 in the first period, and 0.546 in the second). While grain harvest beginnings correlated significantly with JJA precipitation and JJA SPEI, the second period was additionally marked by MAMJ temperature (as the highest with −0.544) and MAM SPEI. The end of the grain harvest correlated particularly strongly with JJA SPEI in both periods (0.727 and 0.688, respectively); in the first period, it was supported only by significant JJA precipitation, while in the second period, only MAM SPEI was insignificant. The duration of the grain harvest correlated significantly only with precipitation and SPEI in JJA. The hay harvest showed statistically significant correlations with MAMJ temperatures and JJA SPEI in both periods. On the other hand, aftermath correlated significantly only with JJA precipitation in 1517–1542 and with JJA SPEI in 1561–1622. The grape-harvest days in the entire 1517–1622 period (Fig. 4) correlated significantly with all climatic variables (summer half-year temperatures at a value of −0.819, summer half-year SPEI, MAM and JJA precipitation totals).

4.2 Climate and the agricultural cycle in the Elbe region for 1770–1815

The variability of ploughing and sowing times, as well as other factors influencing the period of the growth and maturation of grain, was reflected in fluctuations in dates for starting the harvests. The progressive maturation and continuous harvest of grain, of various types in different areas, were considered optimal. Vavák’s reports for Milčice show the various courses for the rye harvests, with the earliest beginning on 24 June 1790 (mean start 12 July) and the latest on 2 August 1785, and graded harvest conclusions from early August to very late October. Vavák reported particularly dry weather for the earliest harvest in 1790 (Skopec 1912, p. 144): ‘Clearly this year [1790] would have been very happy, productive and profitable for all crops had such an unusual drought not carried on for so long, and had rain in quantity fallen once in the entire spring. We usually start the harvest around Saint Margaret’s Day, 13 July, but this year more than half of grain had been harvested by then; by then, many [people already] had [their] wheat and oats home’. Drought in 1790 is clearly confirmed by a deep negative anomaly in MAM precipitation totals, as well as very low MAM SPEI (Fig. 5). In contrast, Vavák reported bad ripening of grain for 1785 with prevailing cold and rainy weather during the summer months and only a few days of warmth (see Vavák’s report quoted in Section 2.2.2). Moreover, this coincides with the deepest negative MAMJ temperature anomaly during the entire 1770–1815 period (Fig. 5).

As follows from Table 2, the starts for rye harvests in Milčice correlated particularly well with MAMJ temperatures (correlation coefficient −0.764), and also with MAM SPEI and MAM precipitation; all these correlation coefficients were statistically significant at α = 0.05. A similar relationship with climatic variables was also obtained for grape-harvest days in the Elbe region (Fig. 6), which were statistically significant for summer half-year temperatures (−0.714) and SPEI, and for MAM precipitation totals. Extremes in dates for grape harvests followed in a short sequence: 30 September 1811 and 26 October 1814. In 1811, particularly higher temperatures and low precipitation in May–June led to an intense dry episode from May to July. Although it rained around the beginning of August, drought then set in until the end of September as witnessed, for example, for the Litoměřice region. Documentary sources speak of smaller quantities of grapes but wine of an excellent quality (Brázdil and Trnka 2015). Czech documentary sources for 1814 report a delayed grain harvest (even unripe rye harvested in September) and rain spells with wet patterns after 24 August for Nové Město na Moravě (Trnka 1912). According to the records of Šimon Hausner, a priest at Buchlovice in south-east Moravia, frequent rain in September complicated ploughing and autumn sowing; 8 days of rain around 15 October led to a vintage ‘on mud’ (AS2). Adverse weather patterns then resulted in small yields of grapes and bitter wine.

Table 2 Basic statistical characteristics of agricultural series of rye harvest (Milčice) and grape harvest (Elbe region) and their correlations with selected Czech climatological variables (T temperature, P precipitation) during the 1770–1815 period (STD standard deviation, A–S summer half-year). Statistically significant correlation coefficients for α = 0.05 appear in italics
Full size table

4.3 Climate and the agricultural cycle in the Louny and Elbe regions for 1871–1910

Fluctuations in the beginning of the spring barley and winter wheat harvests in the Louny and Elbe regions during the 1871–1910 period show nearly identical variability in both regions (Fig. 7). Generally decreasing linear trends, i.e. earlier starts for harvests, are in agreement with rising MAMJ or AMJ temperatures. While both crops differ in the two regions by only up to 1–2 days in mean commencements and the latest conclusions of harvests, the earliest harvest of both crops started in the Louny region 4 days before that in the Elbe region (Table 3). The earliest harvests in 1885 tally with a number of documentary sources reporting a warm April then dry patterns in May and June (‘It did not get properly wet from spring [onwards]’, ‘[…] meadows were not mowed due to drought’) (AS1). Relatively scarce documentary sources for 1871, which had the latest harvest, indicate a cold spring with delayed phenophases (e.g. trees blossoming in mid-June) or possibly long spells of rain (e.g. a flood on 5 August after 3 weeks of rain) (AS1). The starts of harvest correlated highly with AMJ temperatures for spring barley (−0.811 for the Elbe region and −0.815 for the Louny region) and with MAMJ temperatures for winter wheat (−0.823 and −0.827, respectively). The other two climatic variables, precipitation totals and SPEI, for the same combination of months, had no significant influence on beginning the harvests.

Table 3 Basic statistical characteristics of agricultural series in the Louny and Elbe regions and their correlations with selected Czech climatological variables (T temperature, P precipitation) during the 1871–1910 period (STD standard deviation, A–S summer half-year). Statistically significant correlation coefficients for α = 0.05 appear in italics
Full size table

For the grape harvest in the Elbe region (Fig. 8), the relationship to summer half-year temperature was much weaker (−0.547), although it was statistically significant at α = 0.05 (Table 3). No statistically significant correlation was found with summer half-year SPEI or MAM/JJA precipitation totals. The difference between the earliest vintage (7 October 1904) and the latest (5 November 1884) was almost a month. The 1904 year was characterised by slightly warmer July and August and by intense summer drought with below-mean precipitation totals. Particularly high temperatures and lack of precipitation were evident between 5 July and 17 August (Brázdil and Trnka 2015). For example, records from Plotiště nad Labem described summer 1904 as hot and dry, with no adequate rain from 5 May to 27 September, but with a damaging hailstorm on 21 June. Grass and clover did not grow, brooks and streams dried up, and there was a lack of water to drive mills. So dry and hard was the soil that it proved impossible to plough. There were also frequent forest fires. People prayed, and the church organised public processions of religious entreaty, for rain. Cattle were sold for low prices for want of feed, and there was a considerable shortage of milk (Pišl 1938). The latest vintage in 1884 correlates with documentary data reporting a cold, wet autumn and bitter wine (AS1).

4.4 Climate and the agricultural cycle in the Louny and Elbe regions for 1971–2010

Nearly identical fluctuations in the starting times for the spring barley and winter wheat harvests in the Louny and Elbe regions were also recorded in the 1971–2010 period (Fig. 9). A general increase in temperatures in this period led to earlier starts for crop harvests, culminating in 2007, while the latest commencements of harvest were recorded in 1980. In 2007, temperatures in all the months from January to August were markedly above the 1961–1990 mean. Although only April was significantly drier, just slightly lower precipitation totals in the next four months in combination with high temperatures resulted in a significant drought episode (Brázdil and Trnka 2015). On the other hand, rainy weather in 1980 complicated the harvest in the extreme: grain was wet and in some places it even proved impossible to use farm machinery (AS1).

Relationships to temperatures were clearly reflected in the highest, and statistically significant, correlation coefficients for spring barley dates with AMJ temperatures (−0.814 for the Elbe region and −0.793 for the Louny region) and of winter wheat with MAMJ temperatures (−0.861 and −0.846, respectively) (Table 4). Also statistically significant were the correlations with MAM SPEI, although correlations with corresponding AMJ/MAMJ precipitation totals remained insignificant. The harvest for both crops began on average 3 days earlier in the Louny region. This region also exhibited slightly higher variability (expressed in terms of standard deviation) compared to the Elbe region.

Table 4 Basic statistical characteristics of agricultural series in the Louny and Elbe regions and their correlations with selected Czech climatological variables (T temperature, P precipitation) during the 1971–2010 period (STD standard deviation, A–S summer half-year). Statistically significant correlation coefficients for α = 0.05 appear in italics
Full size table

The year 1980 was also unfavourable for grapes in the Elbe region; the vintage started on 26 October, the latest day in the 1971–2010 period (Fig. 10, Table 4). The earliest vintage, 14 September 2000, indicates that the range between the extremes reached 42 days (only in the Louny region during the much longer 1517–1622 period did a larger range, 45 days, occur—see Section 4.1). Mean monthly temperatures over the 1961–1990 means from February to June and in August occurred in the year 2000, with a severely dry episode continuing from April to June. Losses in agriculture due to low yields in 2000, in terms of compensation paid to farmers by the government, achieved 5 billion Czech crowns (Brázdil and Trnka 2015). Grape harvest dates were best correlated with summer half-year temperatures (−0.813); the correlation with summer half-year SPEI was also statistically significant (0.622).

4.5 Climate of the model periods from the sixteenth to the twenty-first centuries

The climatic character of the five periods studied is described in terms of temperatures, precipitation and SPEI values. Monthly (seasonal) temperatures and precipitations are expressed as anomalies with respect to a 1961–1990 reference period (Fig. 11a). As expected, 1971–2010 was the warmest period, with temperatures well over the 1961–1990 means (except September–October). Monthly precipitation totals were more-or-less stable and fluctuated slightly around the reference means. Temperatures remained close to the reference period in 1517–1542, except for cooler months in January–March. Reconstructed seasonal precipitation totals were below 1961–1990 means, with the exception of SON. The three remaining periods were cooler compared to 1961–1990, particularly in the winter half-year. Positive temperature deviations from the reference period appeared in 1770–1815 only in May and August. Precipitation totals were at their most notable in 1561–1622, with drier MAM and largely wetter JJA. Generally drier patterns prevailed in 1770–1815, particularly in MAM. Only anomalies of between 5 and − 5 mm from the reference period occurred in monthly precipitation totals in 1871–1910 (March, April, July, September and October were relatively wetter). Mean seasonal and summer half-year SPEI results for the five periods disclosed that all DJF SPEI were negative except 1770–1815, while drier patterns indicated by negative MAM, JJA and summer half-year SPEI were visible in 1517–1542, 1770–1815 and 1971–2010. The two remaining periods, 1561–1622 and 1871–1910, exhibited rather wetter patterns (particularly JJA and MAM, respectively).

Fig. 11
figure 11

Climatic variables in the Czech Lands/Czech Republic in the 1517–1542, 1561–1622, 1770–1815, 1871–1910 and 1971–2010 periods. a Annual variations in anomalies of mean monthly temperatures and seasonal/monthly precipitation (1961–1990 reference period) and in seasonal and summer half-year (A–S) SPEI. b Annual variations in variability of mean monthly temperatures (standard deviation STD) and seasonal/monthly precipitation (variation coefficient VARC; for seasons expressed as dots attributed to January, April, July and October)

Full size image

Climate variability in the periods studied was most obvious for temperatures in April, expressed as standard deviations (Fig. 11b). Clear consistency in seasonal precipitation expressed as variation coefficients was evident in the first three periods. Monthly variation coefficients showed their highest differences in the 1871–1910 and 1971–2010 periods in November, while in the months of the summer half-year, this occurred in July.

5 Discussion

5.1 The agricultural cycle and data uncertainty

Vavák’s Memoirs for Milčice, a unique narrative source even by wider European standards (see e.g. Lorenzen-Schmidt and Poulsen 1992, 2002; Peters 2003), facilitates the creation of a model for the optimal contemporary agricultural cycle (Fig. 12). The distribution of agricultural work during the year varied, depending on the course of weather. The system incorporated all of the traditional features: a three-field fallow system; the cultivation of the time-honoured spectrum of winter and spring crops, legumes and flax; non-systematic fertilisation; ploughing powered by draught animals and dependent on favourable physical features of the soil; turning over seedbeds; and eventual harvest with sickles. Until 1777, the serfs were burdened with corvée obligations (compulsory unpaid labour performed by serfs for their landlords/the nobility); corvée was shifted in that year to monetary payments as part of the reform of the chamber (crown) estates in the Czech Lands (Klír 2008; Klír and Vodáková 2017).

Fig. 12
figure 12

The optimal agricultural cycle at Milčice based on Vavák’s records

Full size image

All of the above were reflected in the agricultural characteristics of the three earliest model periods—1517–1542, 1561–1622 (the Louny region)—and 1770–1815 (Milčice) in the current study. However, since these are based exclusively upon documentary data, a degree of uncertainty must be taken into account. For example, data may be missing (e.g. the lost Registra špitální for the Louny region). Further, the authors’ focus of interest could change in the course of the year, as well as in the longer term, leading to incomplete records. Moreover, the way in which the books of accounts were kept in the Louny region does not permit exact dating of agricultural work, which may have been done anything from 1 to 6 days before the payments recorded on Saturdays.

Naturally, there were also differences between the various types of production unit. Peasant family households with limited opportunities to hire seasonal labour (Milčice) contrasted with manorial farms with flexible work forces at their disposal (Dobroměřice). Thus, peasant households were not always able to undertake essential tasks within optimal time segments. They were also limited to weaker draught teams and tools, under corvée obligations, and unable to increase the intensity of work in order to manage all of the activities in the shortest time intervals (see Ellis 2003; Beck 2004).

Data uncertainty could also arise out of the harvesting methods in the sixteenth to nineteenth centuries. Winter crops (rye, wheat) were cut exclusively with the sickle and then tied into sheaves; spring crops (oats, barley) were reaped with a scythe, then raked and bound (e.g. Černý 1930; Beck 2004). With a sickle, six to eight reapers took a whole day to harvest one Morgen (0.6 ha), while a single reaper with a scythe managed the same area in 1 day, accompanied by a woman to collect the product. The harvest proceeded quickly if the crop was sparse, while weeds and high seedbeds had a slowing effect (Slicher van Bath 1963). Under these conditions, the staged ripening of the grain and balanced crop level were especially crucial for peasant households. The rapid maturation of all of the grain types at one time was considered to be unfavourable by peasants, and a late start to the harvest had an entirely catastrophic impact, a collision with the preparation of the fields for the autumn sowing of winter crops and consequent threat to the entire agricultural cycle to follow. On the other hand, earlier maturation of grain did not matter, if it was progressive and did not clash with the haymaking. For the interpretation of narrative data, it is important to bear in mind that the farther the ripening of the grain was separated from its usual time segments, the more the likelihood that the harvest would be started too early. The opposite was also possible, with a delay of the harvest, most often caused by unfavourable rainy weather (after Vavák’s Memoirs; see Klír 2008).

The fragmentation of arable land into individual plots and their spatial distribution into various natural conditions (open fields), together with the ridge-and-furrow ploughing that shaped the surface of the field into characteristic vaulting, were among the most important agricultural strategies. They assured the progressive ripening of the grain and also minimised risks to pollination/fertility (e.g. Born 1979; Beck 2004; Klír 2008; Hall 2014). In the Czech Lands, all this started to change in the latter half of the eighteenth century, with agricultural reforms and improvement of marketing strategies, continuing with the mechanisation of agriculture introduced in the second half of the nineteenth century. Agricultural work could be executed quickly and simultaneously, while the consequences of possible crop failures could be overcome by supply. Ridge-and-furrow farming was therefore rendered obsolete in the second half of the nineteenth century; the surface of the fields became mirror-flat, the grain optimal for a given place could be selected, plots could be unified, and so on (Beranová and Kubačák 2010; for Europe e.g. Olsson and Svennson 2011; van Bavel and Thoen 2013). However, such modern agricultural production was associated with its own, new risks.

The question of changing varieties arises in any comparison of agricultural data at intervals analysed together. This is best documented for grapes, in which the Pinot variety (predominantly Pinot noir) prevailed in all periods, but with a later increase in the share of other varieties (1517–1622: 85% Pinot, 10% Savagnin blanc, 5% Gouais blanc; 1770–1815: 70% Pinot, 30% Savagnin blanc; 1871–1910: 60% Pinot, 20% Savagnin blanc, 10% Riesling, 10% Silvaner; 1971–2010: 40% Pinot, 20% Riesling, 20% Müller-Thurgau, 10% Silvaner, 10% St. Laurent). Turning to uncertainties in the dates of grain harvests, it is difficult to specify changes in individual grain varieties. However, Možný et al. (2012), analysing long-term winter wheat harvest dates, demonstrated no significant changes related to crops and agricultural technology during the twentieth century (when the most important changes in agriculture took place).

5.2 The agricultural cycle and climatic factors

The ways in which climate affects the agricultural cycle and the societies they support (e.g. Below et al. 2007; Carleton 2017) have been demonstrated in a plethora of studies from over the world (see e.g. Larcher 2003; Trnka et al. 2012). The existing relationships between agricultural cycles and climate variables has been used for many long-term reconstructions: for example, of temperature (e.g. Meier et al. 2007; Etien et al. 2008; Maurer et al. 2009; Garnier et al. 2010; Kiss et al. 2011; Wetter and Pfister 2011; Možný et al. 2012, 2016a; Molitor et al. 2016) and precipitation/drought (e.g. Možný et al. 2016b). The current contribution shows some of the existing relationships for five time-spans in the context of five centuries well-covered by agricultural data in Bohemia. This long period of time permits only the use of climatic variables reconstructed for the entire territory of the Czech Lands/Republic (see Section 2.3), not only for central or north-west Bohemia. Moreover, temperature and precipitation series in the pre-instrumental period were reconstructed from corresponding series of temperature and precipitation indices derived from documentary data. While monthly resolution for temperatures also relied upon temperature indices series from Germany and Switzerland and ten long-term meteorological stations in those two countries and Austria (Dobrovolný et al. 2010), precipitation reconstruction was confined to seasonal and annual resolution with territorial focus on the Czech Lands (Dobrovolný et al. 2015). Uncertainties in the climatic data decrease in the mean Czech series derived from instrumental records since the beginning of the nineteenth century, despite the fact that the number of meteorological stations used for calculation rose only slowly during this century (Brázdil et al. 2012). All these things may partly decrease the strength of the relationships in question.

The relationship of grain and grape harvests to temperature, precipitation and SPEI has been addressed here by using correlation coefficients, with their statistical significance, in five different model periods. Air temperature was quite dominant in 1871–1910, with no significant response of harvests to precipitation or SPEI (see Table 3). A similar predominance of temperature, but followed by statistically significant effects of SPEI, was recorded in 1971–2010 (see Table 4). The same situation also appeared in 1770–1815, but also the effect of MAM precipitation totals proved significant (see Table 2).

The most variable relationships appeared in the Louny region during the 1517–1622 period (see Table 1). While the influence of all three climatic variables proved significant for the grape harvest, the results for other crops were more diverse, due to the higher variety of agricultural characteristics intrinsic to their cultivation. The most notable temperature effect was recorded for the start of the grain harvest in 1561–1622, while the effect of JJA SPEI was dominant in the years 1517–1542 (even temperature was non-significant). JJA SPEI was also most important to the oat harvests in both periods, coinciding with weaker but statistically significant effects of temperature and JJA precipitation. JJA precipitation and JJA SPEI were the most important to the ending of the grain harvest, as well as in its duration. For the hay harvest, temperatures and JJA SPEI were significant in both periods, but temperature was more important in 1517–1542 and JJA SPEI in 1561–1622. Only JJA precipitation in the first period and JJA SPEI in the second were statistically significant for aftermath. Generally lower correlation coefficients in these periods than in following centuries can be also attributed to dating of agricultural work (‘before date of payment’).

5.3 The broader context

In a longer-term context, some reflection upon the five model periods can be found in examination of series of winter wheat and grape harvests for the Czech Lands from AD 1501, as compiled by Možný et al. (2012, 2016a), which appear in Fig. 13. Strong agreement between mean harvest days of the two crops can be found between 1517–1542 and 1971–2010 on the one hand and the three remaining periods on the other. While the means for the winter wheat harvest dates in these two groups were 21 July and 20 July, respectively, in the second group these were 30 July for 1561–1622, 29 July for 1770–1815 and 31 July for 1871–1910. The same mean wine grape-harvest date, 9 October, was recorded in 1517–1542 and 1971–2010, while in the three remaining periods it took place on 13 October for 1561–1622 and 1770–1815, and 15 October for 1871–1910. Summarising results from Fig. 13, the earliest commencements of harvest for both crops were typical of the most recent period, 1971–2010, with the highest temperatures and quite dry patterns. In contrast, the lowest temperatures with somewhat wet patterns in 1871–1910 led to the latest harvest dates. The highest variability in terms of harvest dates is found in the shortest period, 1517–1542, for winter wheat, and in the recent 1971–2010 period, for grapes.

Fig. 13
figure 13

Box-plots of agricultural and climatic series in the Czech Lands in the 1517–1542, 1561–1622, 1770–1815, 1871–1910 and 1971–2010 periods. a Winter wheat harvest, MAMJ temperatures, MAM SPEI. b Grape harvest, April–September temperatures and SPEI. Means of data for the five periods analysed are given as figures. Data from Dobrovolný et al. (2010), Brázdil et al. (2012, 2016a) and Možný et al. (2012, 2016a)

Full size image

Czech data for the five model periods may be partly compared with analogous data from Switzerland (Table 5). There, filling in for the mean dates for winter rye harvest (Wetter and Pfister 2011) that are missing in the recent 1971–2010 period, an earliest date of 19 July was recorded for 1517–1542 and the latest of 24 July for 1561–1622. This corresponds closely to a worsening of climate, particularly JJA, in the latter half of the sixteenth century compared to its first part (Pfister and Brázdil 1999; Brázdil et al. 2013). It also tallies with the situation in the Czech Lands for the winter wheat harvest, but the latest harvest of all occurred in 1871–1910. Moreover, the difference between the five periods achieved 11 days, while in Switzerland this was only 5 days. Further, Swiss grape-harvest dates (Table 5) gave a slightly lower range of values between the five periods (only 4 days, with the latest vintage on 18 October in 1561–1622, then 15 October in 1971–2010 and the earliest on 14 October in the three remaining periods) compared to the Czech Lands with 6 days (between 9 October in the first and last periods and 15 October in 1871–1910). The above data can be partly supplemented by Austrian figures from Maurer et al. (2009): mean beginning of grape harvest on 2 October in 1517–1542 (but 12 years missing) and on 5 October in 1561–1622 (9 years missing) in the Vienna area, and on 10 October in 1770–1815 (2 years missing) in the Klosterneuburg area.

Table 5 Mean dates of winter rye harvest (WRH) and grape harvest (GH) in Switzerland (Meier et al. 2007—data extended and supplemented by O. Wetter; Wetter and Pfister 2011)
Full size table

Comparison of the 1517–1542 and 1971–2010 periods yields the most remarkable results. Despite clearly warmer patterns in the last period (see Figs. 11 and 13), grain and grape harvests are not markedly different from the most recent and warmest period. As well as some degree of uncertainty in reconstructed temperatures for the first half of the sixteenth century (Dobrovolný et al. 2010), there are some indirect indications of warmer and drier patterns: trees planted in vineyards as protection against direct sunshine (Pejml 1966); melons from the Kolín region used as gifts in the course of negotiations with the Prague administration (Pejml 1974); and Martin Rakovský’s description of Louny town reporting ‘large orchards of Damascus plums, peaches, quinces, chestnuts, almonds and most various fruits; saffron and melons grow here’ (Okál 1974). This warm period also coincides with a May–July temperature reconstruction based on Lower Austrian grape-harvest series in which the decades at the beginning of the series in the sixteenth century were as warm as the 1990s (Maurer et al. 2009).

6 Conclusions

This paper analyses the relationships between the agricultural cycle, expressed through the starting times for different crops and grape harvests in the Louny and Elbe regions in Bohemia (current Czech Republic), to selected climatic variables considered on the longer-term scales for five model periods: 1517–1542, 1561–1622, 1770–1815, 1871–1910 and 1971–2010. The main results may be summarised as follows:

  1. 1.

    The recent warmest period, 1971–2010, have tended towards earliest harvest. This is comparable only with the situation in the first half of the sixteenth century (1517–1542), where possible data uncertainty should be taken in account.

  2. 2.

    Three other periods (1561–1622, 1770–1815 and particularly 1871–1910) show fairly dissimilar climatic patterns but were characterised by later harvest dates of grain, partly also grapes. For winter wheat, this could be explained by notably harsher winters.

  3. 3.

    The analysis demonstrates the dominant role of air temperatures on the agricultural cycle, combined with drought (SPEI) in all periods analysed except 1871–1910 in which, in the course of quite wet patterns, temperature was highly dominant. JJA precipitation was of significant influence in the three first periods.

  4. 4.

    Intensification of climatic effect on the agricultural cycle is expressed by generally increasing correlations between climatic variables and the agricultural cycle indicators used in the study. This can be attributed not only to the increasing accuracy of agricultural and climatic data, but also to the greater sensitivity of current high-intensity agricultural systems to weather variability.

  5. 5.

    Seasonal shifts of 10 to 14 days expected over next 30 years according to some studies (e.g. Eitzinger et al. 2013) would be unprecedented in the light of known data on the agricultural cycle from the past five centuries presented above, which indicates the magnitude of climate change considered likely.