Reliability of flood marks and practical relevance for flood hazard assessment in south-west Germany

. Flood marks are rarely utilized in hazard assessment, mainly because of a lack in data availability, accessibility, and mistrust in their reliability. Challenging these common assumptions, we present an approach for evaluation and practical utilization of flood marks by the example of the Kinzig river, a Rhine tributary from the Black Forest with a history of severe 10 floods. We combined written documents on flood marks with field mapping at three study sites and collected 89 marks – about 50 % of them still preserved – which refer to  15 large floods between 1824 and 1991. The inclusion of a detailed historical mark survey allowed to identify and assess changes over time: they extend from small (+/- 15 cm) imprecisions in mark heights to considerable uncertainties in position, height, and displayed date for some modified marks. Plausibility checks with further data nevertheless demonstrated an overall good consistency. We then juxtaposed these marks with the current, modeled flood 15 hazard maps. A wide agreement is apparent, in that the large majority of the marks are situated at probable heights and within the modeled flooding area associated with extreme floods. For the few exceptions, we see plausible and historically sound reasons in changed local hydraulic conditions by flood protection walls, exceptional processes during a massive ice jam, and possibly also a local underestimation of hazard along Kinzig river tributaries. Overall, this study highlights (1) the broad availability of flood mark data, both on a larger spatial scale and with regard to already vanished marks, and (2) the significance 20 of the marks, verified by further data; and (3) it also demonstrates the possibility of a straightforward inclusion in hazard assessment. We thus encourage the systematic collection, maintenance, and integration of flood marks in a responsible risk management, not least regarding their value in the wider context of risk awareness and memory

based on early-instrumental measurements of water level maxima during floods  and monthly maxima  at the gauge Schwaibach by Bösmeier (2020). The updated Kinzig flood record comprises 308 flood events between 1500 and 2016. Discharge data could be assigned to 35 % of these events.

Flood mark collection
The research of flood marks at the study sites included both archives research and mapping of flood marks in the field. In 195 advance of the field work, the documentation of the historical survey by the CMH (CMH, 1908(CMH, , 1911 was consulted. Currently existing flood marks in the study sites were then mapped in 2018 and 2021. Mapping was carried out utilizing a handheld GNSS (Global Navigation Satellite System) receiver (a Trimble GeoExplorer® 6000 Series), a folding pocket rule for measuring the heights above the ground, and photo documentation to record flood mark locations, heights, and appearance.
For mapping, the following procedure was used: the historically documented flood mark locations were checked initially. The 200 search was then extended to the further vicinity, so that the main parts of the town centers close to the river were covered. In Wolfach, the marks within a radius of 750 m around the confluence between Kinzig and Wolf were recorded. In the smaller town Schiltach, the confluence between Kinzig and Schiltach as well as the area extending approximately 500 m upstream along both rivers was examined. In Haslach, the historical town center is not directly located along the Kinzig. Thus, the area north of the center including the industrial canal and the adjacent 500 m Kinzig river section were examined. Mapping 205 generally was undertaken without any claim of completeness.
Since satellite reception was limited in some narrow streets, the recorded coordinates were subsequently verified by reviewing their positions in Google Earth Pro (2020). In order to georeference the historical map sections in CMH (1908), the mapped marks were then compared with the historical maps which include the flood mark positions but do not display any coordinates.
This allowed to restore the coordinates of 'lost' historical marks which had been destroyed, removed by construction works, 210 or just had disappeared over time due to weathering of the rock. The mapped and photo-documented marks were finally collated with the detailed descriptions and illustrations in CMH (1908).

Assessment of flood mark plausibility and preservation
The gathered multi-temporal data set contains the results of the recent field survey on flood marks, the historical survey, and further qualitative and quantitative information on floods. These data not only present a unique basis to investigate which 215 marks were preserved or vanished over the years. It also presents a framework that can be used to rate the plausibility of a flood mark. Moreover, the data allow to check which of the older marks had been altered or moved since the beginning of the 20 th century and, if so, reveal to what extent these changes affected the marks regarding their location, height or inscriptions.
Such information is necessary for assessing the value of a mark as a substantial indication on the dimensions of historical floods and the potential extent of uncertainties. 220 In a first step, all collected flood marks including both preserved and historically documented marks were cross-checked with measurements and additional written descriptions. Therefore, the marks initially were grouped by associated flood events with the flood year as reference. It was checked, (1) whether the events documented by the flood marks are plausible with regard to the flood discharge record, and (2) whether they are backed with written descriptions, such as reports or chronicle entries.
(3) Finally, a qualitative comparison between the written evidence and the marks regarding the mark locations, heights, or 225 inscriptions followed. Noticeable compliance as well as clear discrepancies with the written evidence were noted. All marks were tagged as doubtful in regard to the further analysis if they could not be matched with a measured or described flood or if they showed discrepancies with written documents.
In a second step, for assessing flood mark preservation, a special focus was put on the historical flood mark survey at the beginning of the 20 th century. Flood marks that had been recorded during that survey documented by the CMH (1908CMH ( , 1911 were compared with the marks that were found to be still preserved at the sites. This allowed for (1) evaluating the overall https://doi.org/10.5194/egusphere-2022-223 Preprint. Discussion started: 17 May 2022 c Author(s) 2022. CC BY 4.0 License. mark preservation over the course of a century and drawing conclusions about potential reasons for the disappearance of marks.
(2) A consistency check between the historical survey and the preserved marks followed regarding the mark location, height, inscription, and appearance. This allowed to identify marks that very likely had been reinstalled or moved and to assess in how far the reinstallation had affected the mark. 235

Comparison with flood hazard maps
According to the European directive "On the assessment and management of flood risks" (2007/60/EC), member states had to carry out a preliminary risk assessment, produce maps for flood hazard and risk, and develop flood risk management plans until 2015. For the area of Baden-Wuerttemberg, FHMs were created using regionalized hydrological data, catchment models, rainfall-runoff models and hydrological studies (Reich et al., 2012). They display the extent of flooding which is likely to 240 occur in the case of HQ10, HQ50, and HQ100, which designate floods with the return periods of 10, 50, and 100 years, respectively. The maps also show an area which may be flooded during a very rare, extreme scenario ("HQextreme") which takes failure of protective measures and log jam at bridges or narrow passages into account (Ministerium für Umwelt, Klima und Energiewirtschaft Baden-Württemberg, 2016).
For this study, the current FHMs along the Kinzig between Schiltach and Haslach were provided by the LUBW (LUBW, 245 2018). In order to juxtapose the collected flood marks and the FHMs, a geospatial analysis was conducted. Firstly, it was assessed whether the mark sites were located within or outside of the flooding areas in the FHMs. This enabled a local validation of the extent of flooded area in the FHMs particularly regarding 'blind spots'. Then, the flooding depths of the FHMs at the flood mark sites were extracted. This allowed for a quantitative analysis of the flood mark heights with respect to the current flood hazard estimates. On the one hand, an event-specific relative comparison between mark heights and 250 flooding depths was carried out to test for consistency. On the other hand, exceptionally high marks were selected and individually examined with regard to their potential use in the flood hazard assessment. The analysis was carried out utilizing R (R Core Team, 2015) and the specific packages "raster" (Hijmans, 2020) and "rgdal" (Bivand et al., 2019).

265
The labels, which refer to the mark location by the site initials and to the utilized sequential numbering, e.g. H1, are displayed for particular marks. Basemap: distributed under the Open Data Commons Open Database License (ODbL) v1.0.
The majority of the flood marks were attached to houses along the river and to the river embankment, and some can be found at bridges or bridge abutments. Hence, they exist or existed in close proximity to the river. Yet, a few flood marks are also located at a distance of more than 100 m to the river (Fig. 2). While many single marks can be found, more than one mark has 270 been installed at most locations. A gateway in Haslach once even displayed a series of eight different marks referring to some of the largest flood events of the past centuries (Fig. 3a). A large majority of the currently preserved marks is well visible from public places. Only a small number is strongly weathered or somewhat hidden so that a close look at the right location is necessary to discover the mark.
With regard to type and appearance, the mapped markings are mostly made of notches engraved in bricks or house walls, 275 indicating the maximum flood level. A number of marks are located on particular stone slabs and a few on metal slabs. Some are rather painted than engraved, and often, weathered engravings have been repainted, obviously in order to preserve the signs ( Fig. 3d). Typically, the year of the flood event is inscribed next to the notch, frequently in combination with the labels "H.W." or "Hochwasser", which is German for "flood". Some marks even show the exact date of the flood, occasionally with the prefix "d.d." which is an abbreviation for the Latin de dato ("from the date", see

Plausibility of the flood marks
A large consistency was found between the flood marks in Haslach, Wolfach, and Schiltach and discharge measurements as 285 well as written documents. To begin with, the flood record in combination with documentary evidence confirmed that the majority of events represented by flood marks were severe on a local to regional scale. A long flood record at the main gauge in Schwaibach was combined by Bösmeier (2020) from early instrumental and systematic measurements (Fig. 4b). The gauge is located in the lower catchment, thus the return periods of measured events cannot be contrasted directly with single flood marks. The gauge measurements are nevertheless an approximate reference. In a synopsis with the flood mark record, they 290 underline the credibility of flood marks as evidence on historical floods (Fig. 4)  A further systematic plausibility check demonstrated that the large majority of the marks are also well supported by written sources, such as the chronicles of Schiltach (Trautwein, 1898) or Wolfach (Disch, 1920). These sources describein parts very detailedthe temporal evolution, extent or severity of the floods. Firstly, all flood marks, except one, could be associated 305 with flood events that were documented by written information. Reports were only missing for the already mentioned 1891 mark in Haslach, which as a consequence was classified as doubtful. Secondly, the qualitative comparison between the marks and the content of the associated written documents did not reveal explicit and considerable discrepancies. By contrast, some statements literally verify particularly high marks by describing the extent or maximum water level in reference to specific landmarks, buildings or bridges. One example represents the extreme flood in 1896, when the Kinzig tributary Wolf tore parts 310 of the graveyard in Wolfach away (Disch 1920). Another example is the disastrous ice jam in 1830: according to the chronicle of Wolfach, the flood followed a very cold winter. Since a large amount of ice, unrooted trees, and other material blocked the river bed in Wolfach, the Kinzig forged its way through the suburb (Disch, 1920). The consequence was an outstandingly high flood water level which even reached the first floor, as reports stated. This is documented by an extraordinary high mark in Wolfach (W28, Fig. 5c). 315

Flood mark preservation
The preservation and alteration of flood marks over the course of a century was assessed by comparing the historical flood mark survey at the beginning of the 20 th century (CMH, 1908(CMH, , 1911 with the status as of 2021 (Fig. 4a). Flood mark preservation turned out to be rather limited and appeared to be location dependent. Overall nearly a third of the 19 th century markings were still preserved, among them only 2 marks in Haslach (13 %), 6 marks in Wolfach (23 %), and 9 in Haslach 320 (50 % of the 18 marks documented for the 19 th century). In addition, one mark (W3) at a private site could not be accessed hence it was not added to the preserved marks, and two marks (W16, S11) from the 19 th century were not mentioned in CMH (1908) hence possibly installed post-hoc.
It appeared that flood marks were less likely to be 'lost' when installed at houses in the town (Schiltach and Wolfach) than positioned directly along the river (Haslach, compare Fig. 2). Generally, flood marks that had been attached to bridges before 325 the 20 th century were not found any more as the bridges across the Kinzig and its tributaries were frequently destroyed and rebuilt. Similarly, flood mark stones or pillars positioned along the river embankment or on roads close to the river were mostly not preserved. Likely reasons may be construction activities along the river such as the renewal or extension of dams, walls, https://doi.org/10.5194/egusphere-2022-223 Preprint. Discussion started: 17 May 2022 c Author(s) 2022. CC BY 4.0 License. roads or the river embankment. Besides that, flood marks within the settlement area might have received a higher level of attention and thus may have been preserved more easily. But even marks on house walls often did not outlast the 20 th century 330 due to rebuild or new construction of houses as well as weathering. Nevertheless, the positions of lost marks on buildings could be reconstructed fairly accurate utilizing the documented historical map sections and reference points such as other buildings in the street. Lacking such close points of reference, the positions of vanished flood mark stones along the river in Haslach could only be approximated. This also reduced their value for the comparison with FHMs.
The consistency check between the preserved marks which had been documented during the survey carried out by the CMH 335 in the early 20 th century also allowed to examine alterations over time. Several issues appeared and were regarded an indication for relocation or reinstallation during the past 115 years. (1) A significantly modified mark height or position relative to building characteristics, (2) modified mark components such as the inscription, its position relative to the notch, a correction of outdated orthography or other specifics, and (3) obvious signs of the renewal of a mark inscription, such as new paint or a new panel. Since building activity over the years often results in a change in ground elevation, it was decided to define a range 340 of tolerance: a mark height difference in comparison to the early 20 th century survey was designated as significant when exceeding 15 cm. Additionally, an imprecision in the range of a few centimeters easily may have resulted from the historical source because the mark heights above the ground or their relative distance were digitized from sketches in CMH (1908). In that source, some mark heights occasionally were not explicitly specified, thus they needed to be measured directly from the relatively small sketches which implied inaccuracies. 345 The results of this consistency check showed that 10 out of 17 preserved flood marks obviously had been moved and reinstalled ( Fig. 4a, in orange). Yet, a considerable (> 15 cm) height difference between the historical and the current position was only found for seven marks. One potential reason may be the integration of cornerstones with engraved flood marks into new buildings, regardless of their original position, inducing significant alterations. This might have caused the 0.5 m height difference between the documented and preserved marks H1 and H3 in Haslach (Fig. 3a,c). Then, 10 marks showed an 350 insignificant height difference to their historical position, among them seven marks without and three with clear indications of reinstallation. Hence, some flood marks have been maintained and reinstalled close to the original after a new building construction. Nevertheless, even some marks in a very good conditionapparently maintained and repaintedwere found to be significantly (about 30 cm) below their documented historical position. Thus, clear signs of flood mark conservation are no proof for their trustworthiness. As a consequence, the historical information in CMH (1908) was generally assigned higher 355 credibility if preserved marks showed considerable alterations. A precondition for this, however, was a successful plausibility check between the historical flood mark description and further written sources, as described above.
Nevertheless, the comparison between preserved marks and their historical documentation occasionally pointed towards some inaccuracies in the historical documents of the CMH. Again, the preserved marks in Haslach (Fig. 3a,c) serve as example: even though the marks are weathered, the deviation between the documented inscription ("1882") for H3 and the label recorded 360 during field work ("27 Dez ´82") is well visible. The preserved marks are most likely the remains of the historically recorded marks, and there was no apparent reasons for a later addition of the date. Hence, it appears as if the date was simply omitted during the survey of the CMH. This might be an isolated case. However, the repeated omission of a precise dating on flood marks would mean a significant loss of information, particularly for years with more than one severe flood, such as 1882.

Comparison of flood marks and flood hazard maps 365
The results of a geospatial analysis combining the flood mark positions and heights with the FHMs at the study sites demonstrated a wide agreement. They also reflected a range of uncertainty in the mark heights, particularly regarding the older marks. In addition, the analysis revealed a few inconsistencies, which required a closer examination and appeared to be of potential use in the flood hazard assessment. Note that the historical mark heights could not be reconstructed for all documented https://doi.org/10.5194/egusphere-2022-223 Preprint. Discussion started: 17 May 2022 c Author(s) 2022. CC BY 4.0 License. but since then vanished marks: in 7 (out of 60) cases, the documentation in CMH (1908) was imprecise or missing a reference 370 to the ground level.
To begin with, the collected flood marks could not indicate a blind spot in the FHMs: the current FHMs at Haslach, Wolfach, and Schiltach show considerable flooding depths at all but one of 89 collected mark locations (Fig. 2). Thereby, the majority of the marks locations in Haslach and Wolfach match with flooding areas assigned to a 50-year or 100-year flood. In Schiltach, many marks exist(ed) in places that in the meantime have been protected against floods up to a 100-year return period. 375 However, these locations are still likely to be flooded during more severe events. The only exception, where a flood mark coincided with no flooding depths in the FHMs, is the mark W3 in Wolfach. This exception is however plausible because the mark was located close to a former side channel leading to a saw mill (CMH, 1887). The area of the former side channel has been protected, and local hydraulic conditions have changed since then.
Then, an event-specific relative comparison between marks and FHMs demonstrated an overall good agreement, which can 380 be considered a mutual verification between marks and modeled hazard maps. In February 1990, for instance, the Kinzig catchment was struck by a severe flood with a recurrence interval between 20 and 50 years at Schwaibach in the lower catchment. Several marks remind of this incident (Fig. 5a). Since long gauge records were not available directly at the study sites, it was not possible to reconstruct and assign local return periods to the marks (Bösmeier, 2020). Therefore, the absolute FHM flooding depths could not be verified directly and precisely utilizing the flood marks. However, a relative comparison 385 was possible: all four marks in Wolfach are placed at heights approximately corresponding to HQ50. It is remarkable how clearly the mark heights mimic the FHM flooding depths, which points to a systematic and reliable installation. In Schiltach, one mark (S22) also points to a flood between HQ20 and HQ50. However, three other marks are situated within an area meanwhile protected against floods up to HQ100 by walls at the waterside of the Kinzig and its tributary river Schiltach.
Hence, local hydraulic conditions cannot be compared with the situation during the flood in 1990. This may also explain why 390 the recurrence intervals in Fig. 5a do not match with the other marks. Similar results, yet with larger deviations, were found for the extreme flood in December 1919, which was an event between HQ50 and HQ100 according to the measurements at Schwaibach (Fig. 5b).  Finally, a number of marks with extreme heights in comparison to the FHMs were found. Seven marks exceeded the modeled flooding depths of an HQextreme event, most of them quite considerably (Fig. 5c). Searching for factors that could explain these discrepancies, the individual cases were examined. Two of the seven marks are located along the Kinzig: W3, which appeared plausible as aforementioned, and W28. The latter is a reminder of the massive ice jam in Wolfach in February 1830 400 and its plausibility could be verified (Sect. 4.1.2). Strikingly, five of the seven marks above the HQextreme were situated along tributary rivers. In Wolfach, W11 and W14 existed along the Wolf roughly 500 m upstream of the confluence between Wolf and Kinzig. In Schiltach, S11, S15, and S16 were or still are situated along the Schiltach river. There, protective walls have been installed (Fig. 2), which considerably improved the flood situation and changed hydraulic conditions. Nevertheless, the height of the marks is exceptionalnote however, that S11 might have been hydrologic and hydrodynamic models, but critical scrutiny has to be applied to assumptions and potential limitations of models 495 with regard to processes beyond observations.
Then why is historical flood information not already systematically included in risk assessment? Kjeldsen et al. (2014) consider the main restraint in the difficulty to apply scientific findings in practice. This means, for instance, a limited access to historical data, or the complexity of assessing the reliability of qualitative data in contrast to standardized official measurements. As delineated above, the estimation of uncertainty in flood mark information is indeed time consuming and complex. Even if a 500 large amount of information is available, involved uncertainties may not be fully elucidated. However, significant uncertainty of systematic discharge measurements also should be taken into consideration. It can amount up to 30 % during extreme events (Kuczera, 1996). In this context, it appears reasonable to utilize flood mark information despite potential large uncertainty.
When there is no certainty that a mark is flawed, why not used it as a benchmark for a worst-case scenario? We therefore suggest to generally include the information on lost or preserved flood marks in two cases: firstly, if marks lie outside of areas 505 associated with the HQextreme in the FHMs, and secondly, if they are positioned considerably above the modeled local flooding depth. Subsequent to this initial screening, which may be more feasible than the extensive work performed in this study, the plausibility of the discovered 'extreme' marks can be tested in detail including further information. Finally, a review of one or more aspects in the process of hazard estimation may be appropriate depending on the number and reliability of these marks and in consideration of substantial changes, as described in the following. 510 A major reservation regarding flood marks and other historical evidence originates from the question of a realistic comparability between the historical and the current situation. In the Kinzig catchment, the assumption of stationary hydrological and hydraulic conditions can hardly be retained for the past 200 years in light of several alterations. The possibly most radical modification in the Kinzig catchment was the river rectification during the 19 th century. It affected large parts of the lower and middle reaches, and though the effects are hardly quantifiable, it is assumed to have aggravated historical floods 515 locally (Bösmeier, 2020). Therefore, the study focused on the upper catchment. Due to its geographical setting, the flood situation in Haslach may have been influenced by the rectification works, Wolfach and Schiltach, however, are located clearly above the cut off meanders. In fact, a few floods were represented by marks in Haslach (compare Fig. 3) but they were not installed at the other two sites. This may indicate a considerable difference of the historical flood hazard between middle and upper catchment. Nowadays, the situation is opposite in what concerns the effect of flood retention reservoirs: situated in the 520 lower and middle catchment, they currently do not protect Wolfach or Schiltach. Withal, continuous improvement of dams, modifications of bridges or of the river bed, and local protection walls have continuously affected flood hazard along the Kinzig. Flood protection structures along the river are displayed in the FHMs and can help to understand discrepancies towards flood marks, e.g. in Schiltach (Sect. 4.2). Furthermore, timber rafting, pursued on the Kinzig until the 1890s, occasionally may have contributed to an increase in the local flood level. By then, logs stored close to the river provided a particularly large 525 amount of material for river obstructions, yet driftwood and other material can also accumulate during a flood today. Whereas these various issues on the whole rather imply a decrease in flood hazard over the centuries, future flood hazard of the studied area is likely to increase due to current global warmingeven though massive ice jams, such as 1830, are increasingly unlikely.
Projections of future climate conditions point to a significant increase in winter precipitation and an increase in days with heavy precipitation even in a moderate scenario (RCP4.5) for the area of the Kinzig (Riach et al., 2019). Altogether, past 530 changes along the Kinzig are complex and not easily reproducible (Bösmeier, 2020), thus they could only be outlined but not quantified within this study. However, it appears justified to assign a higher relevance to the marks collected in Wolfach and Schiltach than to those in Haslach, based on our findings.
To conclude, we argue against a generalized rejection of historical evidence on the basis of untenable comparability between past and present. Certainly, the benefit from historical information has to be weighed against the influence of changes and the 535 efforts needed for the analysis. However, even if long-term changes and subsequent effects on flood hazard are hard to quantify, https://doi.org/10.5194/egusphere-2022-223 Preprint. Discussion started: 17 May 2022 c Author(s) 2022. CC BY 4.0 License. it might be a good idea to examine the local situation and fathom, whether the extent of historical floods might still be realistic as worst-case scenarioproving the opposite may be just as difficult.
Moreover, we belief that a large-scale, comprehensive collection and digitization of historical, lost, and still preserved flood marks is a desirable goal. The national collaborative platform of France (https://www.reperesdecrues.developpement-540 durable.gouv.fr), initiated in 2016, may serve as example. For Southwestern Germany, the historical documentation by the CMH, utilized for this study, represents a formidable foundation. For this area, information is particularly rich, but many flood marks are also documented for other German states, e.g. Thuringia, Saxony-Anhalt (Deutsch, 1997) and Bavaria (Brázdil et al., 1999), and for other European countries, for instance Czech Republic (Brázdil et al., 1999), France (e.g. Martin et al., 2018), Poland (Gorączko, 2021), Slovakia (Pekárová et al., 2013), Spain (e.g. Benito et al., 2021), Switzerland (e.g. Pfister, 545 1985Wetter, 2017) or the UK (e.g. Macdonald, 2007). A comprehensive data base on flood marks could not only strengthen flood risk management by enabling easy access to the information. A systematic collection and documentation of marks would also help to heritage-protect and maintain them, as Deutsch (1997) already incited. Finally, along the Kinzig, many marks are still preserved and highly visible at some places. This may not be the case for a large number of other communities along rivers, where all marks ever installed have been lost already. Then, it may even be advisable to reinstall lost marks at selected 550 locations, just like it is recently practiced in France. After all, flood marks are materialized memories of floods, as illustrated by McEwen et al. (2017) in their concept of sustainable flood memories: their visibility helps to build lay knowledge through generations. Such community knowledge of flood risk adds value to the expert knowledge and thereby supports a distributed flood risk management. In case of an emergency, this lay knowledge may be a crucial factor for taking the right decision, as Thieken et al. (2021) underline. 555

Conclusion
In light of relatively short systematic records, there is general agreement on the usefulness of historical flood information.
Nevertheless, these data are not systematically and routinely included in the hazard assessment. This can be explained by a limited access to data, a lack of confidence in historical information or an uncertainty about the comparability between the historical and the present flood situation. Flood marks in particular appear to be an underrated source, even though they can 560 provide more precise information than most descriptive documents. However, systematic studies on uncertainties and temporal continuance based on a large number of flood marks are rare.
In our study at three flood-prone sites along the Kinzig river, south-west Germany, we therefore collected 89 flood marks, which refer to at least 15 floods of the past 200 years. A qualitative plausibility check with historical documentation and early to recent gauge measurements showed that the marks almost exclusively refer to severe floods. Written documents support the 565 flood markssometimes they even literally verify particularly high marksand thereby underline the credibility of the marks as evidence on historical events. A historical survey of flood marks from the beginning of the 20 th century moreover helped not only in retrieving many already vanished marks (43 out of 89). Together with still preserved marks from the 19 th century, the survey also provided a base for assessing the extent of changes of the flood marks over time. The findings show that the effects range from small imprecisions in mark heights to considerable uncertainties in position, height, and displayed date due 570 to mark relocation or alteration. Not least considering the unresolved question of an initially adequate installation, the collected flood marks thus must be classified as uncertain pieces of information. However, in combination with further information that accounts for their credibility and supported by their number, they can be considered a useful additional information for hazard assessment. On this basis, we compared the collected marks with the modeled FHMs at the study sites, which had been created without the inclusion of historical flood information. A high agreement is apparent regarding the flood mark positions and 575 heights and the hazard maps, indicating a realistic local hazard assessment that also covers the extent of large historical floods.
For the few exceptions, plausible and historically sound reasons comprise changed local hydraulic conditions, for instance by https://doi.org/10.5194/egusphere-2022-223 Preprint. Discussion started: 17 May 2022 c Author(s) 2022. CC BY 4.0 License. flood protection walls, and exceptional processes during the massive ice jam of 1830. Additionally though, a few marks may also indicate a local underestimation of the flood hazard along Kinzig river tributaries. Since past changes in the Kinzig catchment are considerable, they have to be taken into account when comparing the historical with the current situation along 580 the river. The effects of these complex changes are hardly quantifiable. Therefore, the study deliberately focused rather on the upper catchment, where historical river rectification and the influence of flood retention reservoirs are not decisive.
Based on these findings, it appears reasonable to examine the local situation and fathom, whether historical floods could still be realistic as worst-case scenario nowadays, instead of generally rejecting flood mark information. This also applies for a larger scale and in a wider context of risk management. In Europe, numerous flood marks are documented or still preserved. 585 This provides the possibility of a systematic collection of these marks, as practiced for instance in France. After a plausibility check by documentary sources and possibly gauge measurements to identify potentially doubtful and definitely flawed marks, the data should be made easily accessible. Then, the information can contribute to heritage protection, maintenance, and even reinstallation of marks, where appropriate, which can help to build a risk culture and more flood-resilient communities.
Furthermore, the data can add substantial value to an integrated flood risk assessment in an effective and transparent way. 590 Specifically, we recommend to check the current hazard maps for systematic discrepancies towards plausible flood marks and, if apparent, attempt to understand their origin. Here, the rich historical information on flood marks for south-west Germany could be the basis of a follow-up study specifically regarding headwater catchments without flood protection reservoirs.
Finally, flood marks should be practically utilized in the flood hazard assessment at least as a routine benchmark for worstcase scenarios. This can help to prevent an underestimation of extreme floods and may also strengthen adequate disaster 595 management.
Data availability. Data on the collected flood marks and details on the plausibility checks with written evidence are available from the corresponding author upon request: annette.boesmeier@geographie.uni-freiburg.de Author contributions. IH and ASB developed the study concept and collected the data. ASB and SS analyzed the data and 600 discussed the results. ASB prepared the manuscript with contributions of both co-authors.