Reconstructions of droughts in Germany since 1500 – combining hermeneutic information and instrumental records in historical and modern perspectives

The present article deals with the reconstruction of drought time series in Germany since 1500. The reconstructions are based on historical records from the virtual research environment Tambora (tambora.org, 2018) and official instrumental records. The historical records and recent data were related to each other through modern index calculations, drought categories and their historical equivalents. Historical and modern written documents are also taken into account to analyze the climatic effects and consequences on the environment and society. These pathways of effects are derived and combined with different drought categories. The derived historical precipitation index (HPI) is correlated with the standardized precipitation index (SPI). Finally, a historical drought index (HDI) and a historical wet index (HWI) are derived from the basic monthly precipitation index (PI) from 1500 onward. Both are combined for the historical humidity index (HHI). On this basis, the long-term development of dryness and drought in Germany since 1500, as well as medium-term deviations of drier and wetter periods and individual extreme events, is presented and discussed.

In addition to many climatological, ecological and social specifications, long-term reconstructions of 31 droughts are helpful for a better, more holistic understanding. They contribute significantly to 32 answering questions about long-term trends, accumulations, recurrence times and the variability of 33 extreme events. There is also evidence about societal contextualisations, especially the impacts and 34 responses on environment and societies, which have changed fundamentally through time (Erfurt et  35 al. 2019). 36 Droughts are generally referred to as periods of extremely dry weather that persist long enough to 37 cause a severe deficit in the water balance, which in turn causes environmental and social impacts and 38 damages (Wilhite & Glantz 1985, Glaser & Erfurt 2019. From a statistical point of view, a drought is an 39 exceptional event with a rare recurrence probability (Benestad 2003). According to a widely used 40 scheme, droughts are subdivided into four types that reflect their chronological development: A 41 meteorological drought describes a period of considerable precipitation deficit, usually in comparison 42 with a reference period. High air temperatures and wind speeds, intensive solar radiation and cloudless 43 skies can aggravate the precipitation deficit (Wilhite & Glantz 1985). With continued duration, the 44 amount of plant-available soil water is reduced, with negative effects on plant growth and harvest 45 yields. In this case, we speak of an agricultural drought (Bernhofer et al., 2015). If the drought 46 continues to progress, reduced surface runoff, sinking water levels and ultimately sinking groundwater 47 levels occur -a so-called hydrological drought. Lastly, extremely low groundwater levels or baseline 48 flows are referred to as groundwater droughts. Additionally, the term socio-economic drought is used 49 when it comes to impacts on people and the environment. The degree of severity depends on the 50 assets, adjustment options and resilience of the affected society (Wilhite & Glantz 1985, McKee et al 51 1993. 52 Droughts can be defined numerically according to very different criteria, which result in a large number 53 of indices. They differ in the type of input data, temporal and spatial coverage and the consequences 54 for different sectors. While the input data used for meteorological droughts are temperature and 55 precipitation, assessments of hydrological droughts are based on gauging data, groundwater levels 56 and runoff. The timeframes range from days over weeks and months to years (Bernhofer et al., 2015). 57 Similarly, the size of the study area varies according to the question. 58 Common drought indices include the Standardized Precipitation Index (SPI), the Standardized 59 Precipitation Evapotranspiration Index (SPEI) and the Palmer Drought Severity Index (PDSI). In the US, 60 the PDSI is the most common drought index, which is also used as the basis for the US Drought Monitor 61 (Palmer 1965 inter alia soil water information and additional temperature data. The present study is based on the 68 SPI because of its wide distribution, simple calculation and its ability to integrate historical events. 69 Numerous studies on droughts with an explicitly historical perspective have been presented in recent 70 years, for example by Gil-Guirado et al (2019) precipitation, dryness and drought records (red), and impacts and consequences (green).] 113 The data relevant for the analysis are consistently available from 1500 onwards as classified monthly 114 hygric indices using a seven-level scale (PI) (Fig. 3). 115 The monthly PI reveals a differentiated picture of drier and wetter periods since 1500. 116 In general, a particularly large number and higher differentiation of sources document outstanding 117 drought events. The second, modern data set used for the analysis consists of the official precipitation data for 126 Germany from 1881 onward. These values, recorded, averaged and provided by the DWD (Deutscher 127 Wetter Dienst, the official German Weather Service) from the national official network stations, 128 respresent the area of modern-day Germany. This study also draws upon the official drought 129 categories D0-D4 and their definitions by the DWD (2019). The hygric indices (PI) were derived from the written evidence of the tambora sources via semantic 141 profiles, a method well established in historical climatology (Glaser 1991, 1996, 2013 Riemann 2009, Pfister 1999, Brazdil et al., 2005. Therefore, direct hygric indications as well as the 143 descriptions of impacts and consequences are hierarchically ordered according to their intensity and 144 assigned to the appropriate index value. A seven-scale index scheme, ranging from -3 to +3 with 145 index 0 representing the average situation, has proven to be appropriate for the classification of 146 historical records (Glaser 1991, 1996, Glaser & Riemann 2009, Glaser 2013. 147 Direct descriptions mostly refer to the absence of rain or the corresponding lack of clouds. In many 148 cases, the duration is also indicated. The documents related to specific consequences describe 149 effects on harvest results, the phaenological and ecological situation, but also hydrological 150 consequences and impacts on economy, society, and their reactions. This correlates very well with 151 the definitions of meteorological, agrarian, hydrological, groundwater and socio-economic drought in 152 modern classifications (NDMC 2018). 153 Generally, extreme events are represented by a larger number of sources in historical documents. 154 Such information also spans wider areas, especially if the records refer to droughts. Additionally, the 155 information is more detailed and quite often severe events are compared to previous ones. Such 156 long-term memories persist across generations. 157

158
The hierarchical class assignment and its typical indicators for the precipitation indices (PI) -1 to -3 159 are presented as follows: 160 Index -1 is indicated by descriptions of a beginning rainfall deficit. There are often indications of 161 higher damages relating to the harvest of rain-sensitive products such as hay, vegetables and other 162 garden products. 163 Index -2 relates to a longer duration of lack of precipitation, prolonged heat and dryness. Average 164 crop losses for main crops are reported as well as low water levels in smaller bodies of water and 165 reduced spring fills. Heat stress on plants, premature leaf discoloration and the death of plant parts 166 are observed, also dry cracks in soils, occasional forest fires and the impairment of infrastructure, for 167 example related to shipping and water mills. 168 Index -3 represents extreme dryness revealing a chain of effects: After a prolonged period of dryness 169 and heat, the agrarian consequences include severe crop losses and even harvest failures as well as 170 emergency slaughteries due to fodder shortages. If the dryness lasts for weeks, several months or 171 even seasons, there are integrating effects that correpond to reports like low water levels in greater 172 lakes, ponds and larger river systems as well as the drying up of springs and wells. In addition, reports 173 of excessive water shortage and the appearance of "hunger stones" are common. Ecological impacts 174 include a generally visible heat stress of the vegetation, premature leaf discoloration and the 175 withering of plants. In addition, dry cracks in soils, dust veils, dust storms and effects of wind erosion 176 are indicated. There are diverse descriptions of a shift of the phaenological phases, e.g. early 177 flowering, ripening and harvest, but also expression like "wine of the century" indicate dry years. The 178 same is true for reports of forest fires and fish kills. The impairment of infrastructure, especially the 179 termination of shipping and the failure of mills are frequently mentioned socio-economic impacts. 180 The direct consequences for human health and well-being are also documented, e. g. through 181 indications of heat stress and death, increased death rates, the outbreak of epidemics, diseases and 182 hunger crisis due to a lack of food. In addition, the reports include price increases and speculations. 183 Documented authorities´ reactions range from restrictions and regulations on water access or 184 rationing to the declaration of a state of emergency. Societal reactions like supplications, 185 processions, pilgrimages, increasing irrational explanations and interpretations are quite common. 186 The sources also report begging, moving around in order to seek for food, riots and protests, theft, 187 looting, robbery and social excesses. These integrating effects allow conclusions to the preceding 188 months, and in many cases, the exact dates of meteorological droughts are indicated by the name 189 day of saints. 190 The indexing process is similar to modern classifications and definitions of drought categories. Such 191 modern drought catgories also take into account the descriptions of impacts and societal 192 consequences and reactions (McKee 1993, NDMC 2018). 193 Weather diaries with daily records exist for 60 years from the period 1500 to 1800 also on 194 precipitation days (Glaser 1996, Glaser 2013. These records are compared with modern precipitation 195 data on a monthly scale, enabling a comparison of numerical rainfall data with the classified written 196 evidences, which serves as an additional verification of the index levels. 197 198 For the numerical records since 1800, we used a classification scheme based on normal distribution, 199 index "0" reflecting the monthly average with a plus/ minus 0.75 -fold standard deviation. Index "-1" 200 ranges betwen -0.75 and -1.5 standard deviation, Index "-2" between -1.5 and -2.25 fold standard 201 deviation and "-3" below -2.25 standard deviation. The positive indices refer to the appropriate 202 positive ranges. The period 1951-1980 was chosen as reference period. These numerically derived 203 indices were combined with the hermeneutically derived ones. 204 The positive hygric index correponds to the humid and wet situations and is derived in the same 205 manner. The summary of the monthly PI for Germany from 1500 onward is given in Fig including an increased mortality following the outbreak of epidemics, often due to poor water quality. 232 Harvest losses lead to price increases and subsequently to hunger (Nees & Kehrer 2002). Religious rites 233 such as prayer services for rain or processions, but also official measures such as water rationing are 234 taken. If the drought persists, the conditions described become more acute. In the historical context, corresponding time periods of one to twelve months as SPI1 to SPI12. 250 The drought categories D0-D4 and the characterization of droughts as well as the duration and the 251 description of the consequences were also taken over from the scheme of the DWD (DWD 2018), see 252 Tab.1, first to forth column. 253

Derivation of the Historical Precipitation Index (HPI) 254
The derivation of the Historical Precipitation Index (HPI) is based on the monthly Precipitation Indices 255 (PI). We calculated the HPI as the sum of the PIs of the corresponding number of the relevant months. 256 This was done for time windows from one to twelve months in order to map the accumulative effects 257 of dryness and lack of rainfall, analogous to the SPI. For example, HPI3 of June results from the sum of 258 the PIs April to June. We included also positive values for humid and wet conditions. 259

Correlation of the Historical Precipitation Index (HPI) with the modern SPI 260
To compare the HPI with the modern SPI, a correlation analysis for the overlapping period 1881-1996 261 was applied. The results show a very high correlation of 0.65 to 0.74 between SPI and HPI. The 262 correlation reveals a high connectivity between the two parameters (HPI vs SPI). The strength of the 263 statistical relationship and its shape are shown in Fig. 4. There is also an obvious connectivity between 264 the specific inclination and the duration. Therefore, we introduced a duration dependent 265 compensation factor. 266 HSPI3, HSPI6 and HSPI12 (Fig. 6). 281 [ Fig. 6: HSPI3, HSPI6 and HSPI12 for Germany since 1500 for the duration of three, six and twelve 282 months] 283 284

Derivation of the Historical Drought Index (HDI) and comparison with the Modern Drought 285
Index (MDI) 286 For the determination of the Historical Drought Index (HDI), we stepwise interpolated the classes 287 linearly defined in Tab. 1, using the negative SPIs with different durations (Fig. 7). We also used this approach to calculate the last 500 years. 293

Derivation of the Historical Wet Index (HWI) and Historical Humidity Index (HHI) 313
To include not only dryness and drought aspects, humidity has also been considered in the analysis by 314 including the positive hygric indices as a Historical Wet Index (HWI). Its derivation was analogous to 315 the class boundaries of the drought categories. The dominating effects of the HDI and the HWI are 316 combined into the Historical Humidity Index (HHI). The monthly results are shown in Fig. 8, along with 317 the frequency-filtered signals for 1 and 5 years (Fig.8). 318 In addition to the calculations, the chains of effects extracted from the written hermeneutic evidence 332 were used to confirm extreme events. These are characterized by detailed evidence, emphasizing 333 strong impacts on agriculture, forestry, water circle and water supply as well as socio-economic 334 effects like rising prices and hunger. Additionally, ecological effects were considered, e.g. wildfires, 335 algae bloom, fish kills and soil erosion. There is also evidence about the societal contextualisation, 336 especially about the societies´ coping and adaptation strategies, which reveal their vulnerability and 337 resilience capacity. 338 These aspects have changed through the ages: In the agrarian-feudal age, societies coped with 339 drought very differently than during the industrialisation period when, for instance, migration 340 became an option. There was also a great shift in the past 100 years: To visualize the outstanding single drought years, we also derived a "yearcloud"of classified years 349 since 1500. It supports the comparison of the related extreme droughts to minor ones through time, 350 especially between the modern and historical period. At first, a seven-scale index scheme is used to deduce a monthly Precipitation Index (PI) since 1500. 367 This method is well established in historical climatology (Glaser 1991, 1996, 2013 2009, Pfister 1999, Brazdil et al., 2005. Hygric indications as well as the descriptions of impacts and 369 consequences, here referred to as chain of effects are hierarchically ordered according to their 370 intensity and assigned to the appropriate index value. The process is similar to modern classifications 371 and definitions of droughts. In difference to modern drought-indices like SPI, SPEI or PDSI derived from 372 modern instrumental records, these classifications directly take into account the descriptions of 373 impacts and societal consequences and reactions (see also Erfurt et al 2019). 374 The derivation of the historical precipitation index (HPI) from the monthly precipitation indices (PI) -375 including the positive deviations -is the sum of the corresponding number of months of the period 376 1881-1996, analogous to the SPI. We calculated the HPI for time windows from one to twelve months 377 in order to map the accumulative effects of dryness and lack of rainfall as well as for humid and wet 378 conditions. The derived Historical Precipitation Index (HPI) is correlated with the Standardized 379 Precipitation Index (SPI). A calibration factor had been calculated and applied to derive the Historical 380 Standardized Precipitation Index (HSPI). In this sense, the HSPI reflects the longer rainfall deficits and 381 dryness in a more comparable way to modern statistical approaches. Finally, a Historical Drought Index 382 (HDI) and a Historical Wet Index (HWI) are derived from the hygric indices seperately. Both are 383 combined for the Historical Humidity Index (HHI). 384 The reconstructed monthly HHI for Germany since 1500 clearly reveals the annual structures of 385 dryness and wetness. It allows to easily identify dry months and longer periods of dryness and 386 droughts. In total, the synopsis reveals the generally high variability of dryness and wetness through 387 time. Additionally, the time series clearly show that not only summer, but also winter precipitation 388 deficits occur. In this context, it is noticeable that in comparison to the analysis of summer droughts 389 comparatively few studies on spring, autumn or winter droughts are available. The same can be stated 390 for the humidity aspect. An exception is, for example, Pfister et al. Even if the derived HHI shows some remarkable changes and a high variability on the 5-year scale, 403 there is an extensive stationarity of annual humidity as expressed by the HSPI in the long-term 500-404 year perspective. In opposition to that there are remarkable shifts and seasonal trends on the longer 405 scale, for instance winter humidity has increased while summer precipitation has decreased slightly 406 during the last 150 years. There are also sections with comparable seasonal shifts and trends like an 407 increase in winter humidity between 1590 and 1725 and in summer humidity between 1540 and 1690, 408 as well a decrease in winter humidity between 1725 and 1800. 409 In the historical context, for example, water-driven mills played a key role in food security. Due to a 426 lack of water, horse mills or hand mills were set into operation. In today's context, on the other hand, 427 the lower energy production of hydropower plants, or the shutdown of nuclear power plants due to 428 low water or high water temperatures plays a major role (DWD 1947a,b, Loßnitzer 1947, BNN 1947a The comprehensive data collections and derived time series also enable to identify outstanding and 449 correspondingly well-documented extreme drought events, as given in the one-year filtered time 450 series. The written evidence often allows differentiated statements about the underlying climatic 451 causes and in many cases information is available about the temporal structure, particularly the onset, 452 the end and the course of droughts. They are thus similar in content to other Central European 453 historical sources (Pfister 1999, Brazdil et al. 2019). The written evidence covers not only the climatic 454 development but also the impacts and consequences as well as the responses and adaptation 455 strategies of societies. This allowed the application of the concept of pathway analysis, especially for 456 extreme drought events, which in turn served as baseline for the derivation of the evaluation scheme 457 (Tab.1). 458 The most outstanding drought events occurred in the 16 th century in 1590 and 1540, but also in 1536. 459 In