Documentary evidence of historical floods and extreme rainfall events in Sweden 1400 – 1800

Introduction Conclusions References


Introduction
The purpose of this article is to give an overview of major historical flood events in Sweden in the pre-instrumental period (1400-1800) based on documentary sources.A few data concern Finland.Focus will be on river floods driven by rainfall (summer and autumn) and snowmelt (spring).First, a general presentation of the basic orographical and hydrological features of Sweden will be given, followed by a presentation and critical evaluation of available sources in terms of reliability and validity.An indexation on magnitude will be given and an attempt to identify flood-rich and flood-poor sub-periods will be made.A catalogue of floods and extreme rainfall events 1400-1800 is found in Table A1, and a catalogue of possible flood-related harvest failures 1200-1600 is found in Table B1.The study intends to align with prevalent recommendations in methodology and observation periods in order to enhance the possibilities of synoptic reconstruction, calibration and general conclusions on flood regimes in Europe in the pre-instrumental period.

Basic orographical and hydrological characteristics
The Scandinavian mountain range (with a maximum altitude of 2469 m above sea level) runs in a north-south direction on the western side of the Scandinavian Peninsula.The continental divide largely coincides with the border between Sweden and Norway.Most rivers in Sweden flow down on the eastern slopes of the mountain range in a southeasterly direction through the largely flat lands into the Bothnian Sea and the Gulf of Bothnia.In south-central Sweden a number of large lakes are found -Vänern, Vättern, Mälaren and Hjälmaren -which catch waters to constitute the main basins of large catchment areas (see Fig. 1).In the southernmost part of the country the modestly elevated Småland highlands, with a maximum altitude of 377 m, is the source of a number of smaller rivers that run both into the Baltic Sea to the east and the south and into the Kattegatt-Skagerrak of the North Sea in the west.
The most important catchment areas are Dalälven, Norrström, Göta älv and Motala ström (see Fig. 1).Dalälven is Sweden's longest river with a total extension of 520 km.The total catchment area is 28 954 km 2 .Lake Mälaren constitutes a basin collecting water from a wide range of smaller rivers, totalling a catchment area of 22 650 km 2 , all flowing into the Baltic Sea at Stockholm.The main outlet is Norrström, north of the Old Town of Stockholm, which has given the name of the entire catchment area.Sweden's largest lake, Vänern, catches waters running down from the higher altitudes in the province of Värmland as well as in Norway and lets its waters continue to the North Sea by the Göta älv River.At its mouth, the second largest city of modern Sweden, Gothenburg, is located, though only founded as late as 1628.The total area of the catchment is 50 229 km 2 .The Motala ström catchment area with 15 481 km 2 , is constituted by the waters running from Lake Vättern to the Baltic Sea at Norrköping.
The geographical distribution of hydrological data in documentary sources mirrors the economic geography of medieval and early modern Sweden.Population density was highest in southern Sweden where consequently agriculture, the most important economic activity and especially sensitive to variations in hydrological patterns, was concentrated.Mining, the second most important economic activity, was concentrated in the less populated areas of Bergslagen in the provinces of Västmanland and Dalarna in the areas north and west of Lake Mälaren.Already in the Bronze Age rich mineral resources were found here, giving rise to an early mining activity which used the streaming rivers as a source of power for the roasting and smelting of the raw ore.The mining area predictably became an important zone for the Swedish economy.The rivers are often subject to intense spring floods when the snow in the mountains to the northwest melts rapidly.The seasonality of floods in the mining area is therefore concentrated to the spring season and is explained by a combination of snow storage in the mountains and the rate of melting in the spring.Lake Mälaren was originally a bay of the Baltic Sea but was separated from it and transformed into a lake by the continuous postglacial rebound around 1000 BC.The outlet was for long confined to Södertälje and the two narrow canals Norrström and Söderström in Stockholm, founded around AD 1250 and later to become the capital of Sweden.The combination of these three factors -the importance of mining to the Swedish economy, the location of the mines near rivers subjected to spring floods and the location of Sweden's most prominent early city -produce a number of hydrological data in contemporary historical sources.
Although many catchment areas in Sweden are quite large (15-50 000 km 2 ), most rivers in south-central Sweden are, with a few exceptions, not suitable for navigation due to their small size and the presence of rapids.It led authorities at an early point to explore the possibilities of building canals and locks, but such projects were never carried out on any significant scale before the 19th century (Meyersson, 1943;Bring, 1911).Also dredging projects were few and limited before the 19th century.The only systematic dredging of Swedish rivers seems to have been a consequence of increased log driving, predominantly in the northern provinces and in the first half of the 18th century (Ahlbäck and Albertsson, 2006;Wik, 1950).Serious alterations of runoff through engineered modifications only occurred in the second half of the 20th century with the development of hydropower plants in the north.The hydrological events prior to the 19th century are therefore to a large extent the result of natural factors.The hydrologically most vulnerable point was the city of Stockholm.It is located at the outlet of Lake Mälaren, where a floodgate was constructed already before the 16th century to control the spring-flood water and where most works of this kind were carried out (see e.g.Almquist, 1903, 241 pp.;Handl. rör. Skand. hist. 19, 1834, 183 pp.;Almquist, 1913, p. 82).In the early 15th century some dredging works were carried out at the outlet of Södertälje and again in the late 17th century but had little impact on the hydrology of the lake (Bring, 1924).
Consequently, the human impact on river streambeds and floodplains has been limited during the period in concern here.The pressure of urbanization, population increase, deforestation, and other land use changes as well as surface alterations and irregularities in channel alignment can be considered to be negligible due to the sparse population of Sweden and the low-intensity utilization of rivers.Hypothetically, climate, i. e. precipitation and temperature, would be the main driver behind any observable flood regime change before 1800 (Glaser et al., 2010;Wetter et al., 2011).Exceptionally, other natural factors than climate explain floods.For example, according to locals changes in the water levels of Lake Vänern were due to winds over the large lake surface rather than floods in the tributary rivers or drought (Elvius, 1751(Elvius, -1752, p. 39, p. 39 The present study has been mainly based on printed letters, diaries, travel notes, annals and chronicles, as well as secondary sources such as regional topographical descriptions.Some data have been found in the Swedish National Archives (Riksarkivet) in Stockholm.There are also some compilations of general weather data from the 18th century (Ferner, 1756;Falkengren, 1781;Ekman, 1783).Further data could still be found e.g. for the 18th century in newspapers but it is argued here that the main trends would not change substantially.The survey covers the period up to 1800, approximately a century before the beginning of systematic instrumental hydrographic measurements (Lindström and Alexandersson, 2004).The period has been chosen in order to avoid complications in the analysis due to the increased interference of anthropogenic factors in the 19th century.
As for most of Europe, the amount of documentary sources in Sweden is meagre for the Middle Ages but increases dramatically from ca. 1520 due to successful centralization efforts of the central authorities by King Gustavus Vasa (1523-1560) as well as fortunate preservation circumstances (Retsö and Söderberg, 2015a).Thus, for the 12th and 13th centuries, most climatological and parameteorological proxy data are found in chronicles and annals, written long after the events were described and most often of Danish or north German origin.This type of source material is notoriously difficult to use for historical reconstruction, but with specified methodology it is not useless especially concerning spectacular and severe events like floods (Wetter et al., 2011;Retsö and Söderberg, 2015b).Geographical specificity is not very great -in earlier sources it is confined to general terms (Sweden, Norway, Denmark).The earliest mentioning of a hydrological extreme found in Scandinavian sources possibly relevant for Sweden is from a Danish annal written sometime after 1288, which states that the year 1195 was characterized by "extreme wetness" (yuerwaetis vaedher) (Jørgensen, 1930, p. 179).The only primary sources of a uniform kind from the Swedish Middle Ages are the diaries of the Birgittine monastery in Vadstena and the Franciscan order of Visby (Gejrot, 1996;Odelman and Melefors, 2008), but they contain very little of hydrological data.
The quantitative increase of documentary sources in general after 1520 also implies greater reliability since the number of independent data also increases and the basic requirements for documentary sources such as nearness in time and space and neutrality are better complied with, as well as the specific requirements on data for the study of long-term structures and parameteorological phenomena such as floods, namely, regularity, frequency, uniformity, high time resolution and geographical specificity (Bell and Ogilvie, 1978;Brázdil et al., 2005Brázdil et al., , 2010)).In addition, the degree of detail as to the causes and impact on society is greater.There are several uniform individual records produced by the same person (e.g.Brahe, 1920;Hausen, 1880;Lewenhaupt, 1903), whole individual letter suites (e.g.Sjöberg, 1911Sjöberg, , 1915;;Wijkmark, 1995), and a number of institutional records such as letters from bailiffs and civil servants throughout the country (Retsö, 2002;Almquist 1868Almquist , 1875Almquist , 1877Almquist , 1893Almquist , 1902Almquist , 1903Almquist , 1913;;Styffe, 1893;Edén, 1905;Ahnlund, 1930).
Hydrological data are limited to statements on extreme flood events or general characterizations of an entire year.The approach chosen here is the threshold approach (Hall et al., 2014), i.e.only floods and rainfall events that have been perceived by contemporaries to be beyond normality have been included.Concerning floods, the sources tell us about two cases: floods due to excessive precipitation and extreme spring floods.However, it is most often impossible to assess the magnitude of floods in quantitative terms.Some exceptions are the floods at Uppsala in 1622, at Söderköping in 1684 -for which the only known floodmark has been found (Broocman, 1760, p. 149) -at Holmen in 1646, at Ekby 1709, and at Stockholm and Uppsala in 1780.
The magnitude is normally described in vague qualitative terms, e.g. as the worst "in living memory" (mannaminne).It is argued here that such implicit comparisons with previous floods are indications of perceived absolute magnitude and not relative to real magnitude.The threshold approach inevitably involves an element of interpretation based on an analysis of terminology, the basic understanding of which may have varied somewhat over time and between persons but has nevertheless been mainly constant.For example, "severe spring flood" (svår vårflod) must have meant a spring flood above normal expectations, and the same is the case with "much wetness" (mycket väta).
The data used have thus been restricted to such data that can be confirmed to be reliable and valid and above the threshold of perceived normality.A commonly recommended 3-scale indexation of the magnitude is used here, based on the criteria of duration, spatial extension and material damage/human casualties (Sturm et al., 2001;Llasat et al., 2005;Glaser et al., 2010;Wetter et al., 2011): (1) floods on a regional scale with little material damage and/or short duration, (2) floods of significant regional or supra-regional magnitude with considerable material damage and/or average duration, and (3) floods of regional or supra-regional magnitude with disastrous material damage and/or long duration.Following Hall et al. (2014), the survey intends to identify flood-rich periods in order to facilitate cross-continental comparisons.Due to lack of reliable data at this stage no attempt will be made to assess discharge.All dates are adjusted according to the Gregorian calendar (New Style), introduced in Sweden in 1753.

Results
With all these prior observations of the source material, the result of the survey is as follows.A total of 157 floods or extreme rainfall events have been found for the period 1400- 1800, of which 107 can be unambiguously defined as floods (see Table A1).Catchments particularly hit by floods were Norrström, Göta älv and Dalälven (see Fig. 1).
There is no clear picture of flood frequency during this period.Yet, the data are clearly sufficient to make a preliminary identification of flood-rich and flood-poor periods (see Figs. 2 and 3).There is a clear tendency to more frequent floods in the 17th and 18th centuries in general.Especially two periods stand out as particularly flood-rich: 1591-1670 with two intermediate sub-periods with fewer floods in the 1610s and the 1650s, and the early 18th century.On a decadal timescale the highest number of floods is found in the 1640s (12 events), the 1700s and the 1720s (11 each), followed by the 1630s (10) and the 1620s (9).Years of significantly severe floods were 1649 (6 events), 1622 and 1780 (5 each), and 1596, 1640, 1661, 1677, 1707, 1709 and 1728 (4 each).Particularly serious was the flood in the province of Östergötland in August 1649 (the so-called Olsmässofloden).According to one assessment considerably more than 100 mm, perhaps as much as 200-300 mm of rain may have fallen over certain locations in the southern and central parts of Östergötland in a few days (Alexandersson and Vedin, 2001).The flood in May 1650 seems to have been equally serious; the situation caused the authorities to initiate works to widen the outlet at Stockholm and also to investigate the possibilities to widen the outlet through the Södertälje Canal.The same happened in the spring of 1661 and the authorities sped up the work at Södertälje (Bring, 1924, p. 16).As for magnitude, 32 % of all events were of the third category (floods of regional or supraregional magnitude with disastrous material damage and/or long duration), and 44 % of the second category (floods of significant regional or supraregional magnitude with considerable material damage and/or average duration) (Fig. 3).The impression of the period 1591-1670 as one of dramatic hydrology is substantiated by the fact that almost one-sixth, or 27, of all third category events occurred during that period.

Comparison with flood-rich and flood-poor periods in continental Europe
The result does not neatly coincide nor with tendencies in flood frequency or particular events observed in continental Europe (compare e.g.Wetter et al., 2011;Benito et al., 2003;Glaser et al., 2010;Glaser and Stangl, 2004;Elleder, 2013;De Kraker, 2006).For example, both Brázdil et al. (1999) and Schmocker-Fackel and Naef (2010) found few floods in northern Switzerland in the first half of the 16th century but there was a flood frequency peak in 1560-1590, whereas a first peak in Sweden is found only in the following 2 decades.As for the 17th century, there was a low in Switzer-land until the first decades of the 18th century (Pfister, 1999;Schmocker-Fackel and Naef, 2010).The latter half of the century was again a period of high frequency, which only partially coincides with the documentary data from Sweden.
There are no traces of similarities with Sweden on single extreme years in the same region (Wetter et al., 2011) or in central Europe at a larger scale (Glaser et al., 2010).There are only slight similarities with the flood chronology of Spain, in particular for the Llobregat and Tagus catchments in the extended period 1580-1620 (Llasat et al., 2005;Benito et al., 2003).

Relation to quantity of source data
The first question to address is whether this can be explained by a deficiency in the Swedish source material.It is held here that more documentary sources could doubtlessly improve the picture in its details but not substantially change the general pattern.For example, the increase in reliable flood data in the 17th and 18th centuries is not entirely a reflection of a total increase in documentary sources.Indeed, the total quantity of preserved documentary sources rises considerably already in the 1520s but the rising frequency of floods does not occur until the 1590s.It can thus be concluded that the data most probably reflect a real increase in flood events towards the late 16th century.Consequently, it can also be presumed that floods really were more rare in the source-poor late Middle Ages.As has been pointed out by Wetter et al. (2011), it is highly improbable that spectacular events like major floods would pass unnoticed by chroniclers.

Comparison with North Atlantic Oscillation (NAO) reconstructions
The meteorological/climatological causes behind these Swedish flood data require further research to be identified.In a number of studies the great variability in flood frequencies in Europe has been explained by large-scale atmospheric circulation patterns, particularly on a decadal timescale (Schmocker-Fackel and Naef, 2010; Casanueva et al., 2014).In particular, the NAO together with other, related atmospheric circulation patterns is normally seen as the main explanation for climatic variability in northern Europe (Lindholm et al., 2009), especially winter precipitation in the NAO positive phases (Hurrell and Van Loon, 1997;Barker et al., 2004;Casanueva et al., 2014) as well as for river discharges, snow accumulation and flooding (Prudhomme and Genevier, 2011).However, there is no clear connection between the existing NAO reconstructions and flood frequency in Sweden.In one reconstruction (Luterbacher et al., 1999) a few winter NAO indices coincide with flood peaks in Sweden (early 18th century, the 1770s and perhaps also the 1740s), while in others the picture is somewhat different (see e.g.Luterbacher et al., 2002;Glueck and Stockton, 2001).For example, while it would seem that a tendency to a 10-12 years cycle, a timescale which is close to one of the suggested NAOs (Hurrell and Van Loon, 1997;Cook et al., 1998), could be seen in the Swedish sources between 1620 and 1661 there is no sign of it after that.Similarly, one NAO winter index reconstruction (Cook et al., 2002) identifies a positive phase until about 1640 whereafter it went into a neutral or negative phase which would not be able to explain the flood peaks in Sweden in the following decades.If the chronology of Swedish flood events is compared with the NAO index found in Luterbacher et al. (2002), then no correlation at all can be seen.According to expectations, the NAO would have a marked influence on precipitation and streamflow, particularly in its positive mode when westerly winds bring moist and warm air over Scandinavia.NAO is also expected to have a stronger effect in the winter than in the summer, and a stronger effect in northern Sweden than in the south (Uvo, 2003).However, the documented floods are slightly more frequent in negative NAO phases (83 events) than in positive NAO phases (64 events).Furthermore, floods related to the winter season, i.e. spring floods, are about as many (70) as those related to the summer season (77).There is no clear tendency in either high altitude or high latitude catchments (defined as catchments no.1-5 and 7-8 in Fig. 1): the number of events (106) in the former catchments is indeed greater than the number of events (34) in the low latitude/altitude catchments but exhibits a perfectly equal distribution between positive and negative NAO phases (53 events each).
Previous attempts have failed to establish an unambiguous connection between NAO, winter precipitation and floods, in general, particularly for northern Europe (Bouwer et al., 2006;Uvo, 2003;Casty et al., 2005).One reason for that is undoubtedly that NAO operates on a great variety of timescales and is interfered with by other local conditions as well as other circulation patterns (Jacobeit et al., 2003;Lavers et al., 2012).Also changes in flood frequencies are obviously the result of the workings of several driving forces at the same time but to different degrees at different times and places, particularly in colder climates as in Sweden.For example, it seems that NAO-related precipitation patterns east of the Scandinavian mountains, i.e. in Sweden, are overplayed by other climatic factors (Uvo, 2003;Linderholm et al., 2003).However, it is conspicuous that the great majority of the worst flood events have been recorded in catchments that are particularly subjected to spring floods fuelled by melting snow from high altitudes or latitudes (Norrström, Göta älv, Dalälven, Torneälven, Piteälven, Ljungan, Indalsälven) and where lake evapotranspiration is lower and water storage capacity higher (see Fig. 1).Furthermore, if the average winter temperature of Stockholm (Leijonhufvud et al., 2010) is taken as a proxy for a general meteorological pattern in southern and central Sweden, the frequency of floods has a clear correlation with cold and snowy winters.Although the correlation of winter precipitation with NAO is generally weaker in Sweden than on the Atlantic coast of Norway (Uvo, 2003) and although temperature tends to increase under positive NAO, precipitation in the winter at northern latitudes would under all conditions and NAO modes come as snow and therefore generate a larger storage of water in the mountains.Thus, the melting of large amounts of snow in the spring would affect Sweden as well and thereby contribute to spring floods whose intensity would depend on the evolution of temperature in the spring.
Plausibly, a decline in evaporation due to decreasing mean temperatures, probably in connection with heavy winter precipitation in the form of snow and increased spring precipitation due to NAO, generated considerably higher levels of runoff, notably at higher altitudes (cf.Burt and Howden, 2013) where the wellsprings of most Swedish catchments are located.The combination of soil saturation, huge snow amounts and spring rain has been pointed out a an important trigger for spring floods (Wetter et al., 2011).This allows for the conclusion that the NAO can account for precipitation patterns, mainly in the winter, but not necessarily all flood peaks, whereas climate plays the main role for the frequency of floods.The peaks in Sweden's flood history would then be characterized as cases of "complex extremes" (Benestad and Haugen, 2007), involving both temperature and precipitation.In some cases, the correlation between snow-rich winters and spring floods is explicit in contemporary sources; for example, the winters of 1543, 1544, 1601 and 1780 and the following disastrous spring floods.This correlation between flood frequencies and the so-called Little Ice Age has also been noted for other areas of Europe (Brázdil et al., 1999;Pfister, 1999;Glaser, 2008).

Medieval floods and harvest failures without stated causes
There are no unambiguously reliable data on floods in Swedish medieval sources before the 15th century.In Danish and German chronicles reports are found of heavy raining and/or floods in 1287,1315,1336,1347,1357 (Jansson, 2003, p. 148).It should be noted that there are no indications in Swedish medieval sources, as in central Europe, for floods in the 1340s or in 1501 (cf.Rokoengen et al., 2001;Brázdil et al., 2005;Rohr, 2007;Kiss, 2009;Elleder et al., 2013).It is also uncertain whether the statements in Danish chronicles are relevant for Sweden.The same is the case with the report in Heinrich of Balsee's chronicle on a flood in northern Germany in December 1374 (Crull, 1878, 165 pp.).In many cases the magnitude of floods in the early modern period is related to the damage on crops (see e.g.Jämtl. räk., 1564Jämtl. räk., -1571, 38 pp.;, 38 pp.;Sommarström, 1935, p. 285;Ekström, 1949, p. 417;Lindblom, 1793, p. 121;Strömbeck, 1993, p. 170).Some medieval data tell about severe harvest failures and famine without stating the causes (see Table B1).At the present stage no details can be found to support that these extreme events were caused by floods and undoubtedly some of them are connected to drought.But it is also clear that several of them may have been caused by floods.Already Emmanuel Le Roy Ladurie warned about the difficulty to establish a strict causal connection between climate and crop failures unless the precise cause is stated or the data can be supported by other contemporary data (Le Roy Ladurie, 1971, 275-6 pp.).The purpose of presenting the Swedish data here is to furnish a point of departure for future research and comparative analyses that can shed more light on the matter.

Conclusions
Two periods stand out as particularly flood-rich in the preinstrumental period in Sweden according to documentary records: 1591-1670 and the early 18th century.In particular, there are clusters of floods in the 1640s, the 1700s and the 1720s.One-third of all events were floods of regional or supra-regional magnitude with disastrous material damage and/or long duration, and half of them occurred in the period 1591-1670.
The spatial scale of spring floods and their temporal concentration in clusters suggest causality on a large timescale, i.e. meteorological conditions connected with the Little Ice Age rather than atmospheric circulation patterns such as the North Atlantic Oscillation (NAO) and a reflection of regional response to climatic variability.NAO could very well explain winter precipitation patterns as well as the flood peaks between 1591 and 1650 but only in combination with a Little Ice Age cooling, which in turn is the more plausible explanation for the peaks in the early 18th century.Given the high degree of continuity in demographic and economic conditions in the 1400-1800 period, it therefore seems reasonable to conclude that among the potential drivers of flood regime change are the changes in precipitation and temperature, i.e. climatic change, that mainly account for the longterm variability of historical floods in this period.Although there is a natural time lag in relation to temperature, there is a clear correlation between the seasonality and the chronology of spring floods, on the one hand, and, on the other, rapid and late melting of larger snow storages in combination with spring precipitation from ca. 1600.This is further confirmed by the observable spatial coherence of major flood events.

D. Retsö :Figure 1 .
Figure 1.Catchment areas and major lakes in Sweden mentioned in the text.

Figure 2 .
Figure 2. Decadal frequency of floods and extreme rainfall events in Sweden, 1400-1800, according to documentary sources.

Figure 3 .
Figure 3. Documentary evidence of floods and extreme rainfall events in Sweden, 1400-1800, with levels of magnitude.

Table A1 .
Documentary evidence of floods and extreme rainfall events in Sweden 1400-1800 (bf -before).