Damage Analysis of the Eifel Route Railroad Infrastructure After the Flash Flood Event in July 2021 in Western Germany

Abstract

Extreme rainfall events characterized by small catchments with high-velocity flows pose critical challenges to infrastructure resilience, particularly the rail infrastructure, due to its partial location near rivers and in mountainous regions, and the limited availability of alternative routes. This can lead to severe damages, often resulting in long-term route closures. To mitigate flash flood damage, detailed information about affected structures and damage processes is necessary. Therefore, this study presents a newly developed multi-criteria flash flood damage assessment framework for the rail infrastructure and a QGIS-based analysis of the most frequent damages. Applying the framework to Eifel route damages in Western Germany after the July 2021 flood disaster shows that nearly 45% of the damages affected the track superstructure, especially tracks and bedding. Additionally, power supply systems, sealing and drainage systems, as well as railway overpasses or bridges, were impacted. Approximately 30% of the railway section showed washout of ballast, gravel and soil. In addition, deposit of wood or stones occurred. Most damages were classified as minor (47%) or moderate (34%). Furthermore, damaged track sections were predominantly located within a 50 m distance to the Urft river, whereas undamaged track sections are often located at a greater distance to the Urft river. These findings indicate that the proposed framework is highly applicable to assess and classify damages. Critical elements and relations could be identified and can help to adapt standards and regulations, as well as to develop preventive measures in the next step.

1. Introduction

1.1. Heavy Rainfall and Flood Event July 2021

Recent heavy rainfall events have led to significant flooding and flash floods in several European countries, notably the event in July 2021 in Western Europe. This resulted in extensive damage to property, and affected the rail infrastructure severely [1,2]. The July 2021 heavy rainfall event occurred from 12 July 2021 to 15 July 2021 and covered an area of 41,177 km², especially affecting North Rhine-Westphalia (NRW) and Rhineland-Palatinate (RP) with 22,716 km², as well as parts of Belgium, Luxembourg, France and the Netherlands [3]. The meteorological conditions that led to this event were characterized by the development of the low-pressure system “Bernd” over Central Europe from 12 July 2021. In the course of the weather situation, warm and humid air from the northern Mediterranean region was transported via Slovenia and Austria towards the Czech Republic and Poland and finally to northern Germany [4,5]. In Germany, rain gauges measured up to 165 mm in 48 h in Köln-Stammheim and Wipperfürth-Gardeweg in NRW [6], a value far beyond the multi-year average precipitation for the month of July for the observation period from 1991 to 2020 in Köln-Stammheim (82.5 mm) and Wipperfürth-Gardeweg (103.4 mm) [5]. This rainfall event was classified with a maximum return period of >1000 years [3].

The combination of high precipitation and low soil moisture storage capacity in RP and a medium storage capacity in the south-west of NRW resulted in limited infiltration and a high surface runoff [7]. Rivers such as the Ahr, Erft, Kyll and Rur were flooded [7] and severe flash floods occurred, particularly in sharply incised valleys like the Ahr valley. High flow velocities, in addition to high water levels, resulted in washout of soil at buildings or bank erosion, as well as transport of debris, which partially contributed to damming effects, thereby expanding the flooded area significantly [8,9]. This damaging event resulted in 184 fatalities in NRW and RP and 820 injured persons [6,10]. The total economic loss from the July 2021 event is estimated at up to EUR 46 billion [11] with estimated losses in Germany rating from EUR 33 billion [11] to EUR 40 billion [12], making the event the costliest on record in Europe.

1.2. Rail Infrastructure and Flood Damages

The rail infrastructure (refer to Figure 1) includes the rail track, which is divided into superstructure and substructure, control and safety systems, power supply, telecommunication, stations and other rail facilities [13]. Bridges, tunnels, culverts and drainage systems are also important elements of the rail infrastructure. Among the different modes of transport, railway infrastructure is particularly vulnerable to climate impacts and natural hazards [1] due to its partial location near rivers or in mountainous regions and the limited availability of alternative routes as they exist for roads. Severe damages often result in long-term line closures because of the complexity of the structure of railway lines.

Rail infrastructure damage caused by a flood or flash flood event (refer to Figure 2) includes damaged embankments or bridges, the accumulation of debris, and the clogging of drainage, as well as derailments, disruptions and closures of railway tracks as a result of the initial damages [14,15,16,17,18,19,20,21,22]. The hydraulic processes correlated with these damages are flooding with high water depth, high flow velocity and resulting forces. This can lead to scouring, erosion, washout of ballast, overtopping of bridges and transport of debris. The extent of these damaging processes can differ for the upper, middle and lower reaches of a watercourse due to their different river bed inclinations and slope gradients [22].

Flood-induced damage to the rail infrastructure can occur directly after the heavy rainfall event due to water contact (direct damage), e.g., submergence of tracks [17,21], or at a later point in time without water contact (indirect damage), e.g., disruption of freight operations due to damages and closures [14,16,23]. Furthermore, damages that can be monetized are defined as tangible damage, while intangible damage cannot be monetized and describes sentimental values, such as photos or cultural heritage, e.g., a railway museum [21] with historic items, that were destroyed by the flood. Tangible and intangible damage can be divided into primary and secondary damage; while primary damage is a result of the event, secondary damage is no longer directly related to the event itself [23].

Damage can also appear as a cascade effect with one damage triggering a second one. Examples for three different topographic settings are described in Figure 3. A heavy rainfall event in a low mountain region characterized by steep slopes can lead to flash floods with high flow velocities and shear stresses. This can result in the erosion of an embankment and further in the collapse of a bridge and therefore the derailment and damage of trains. In mountainous regions, this effect is even more pronounced and can also lead to mud and debris flows [24]. Here, the risk of derailment is even higher due to the high velocity and spontaneous occurrence of a debris flow. In the lowlands, the topography is less pronounced than in mountain regions and is predominantly flat. Precipitation runs off more slowly and leads to more prolonged flooding lasting up to days [25]. Damage to the rail infrastructure in these areas is characterized by water contact over a long period of time. Flooded sections of tracks can lead to track closures. This can result in delayed freight transportation, longer travel times, monetary loss for late shipment fees and even loss of businesses. This damage is classified as indirect damage.

Figure 1. Relevant elements of the rail infrastructure adapted from [13,26].

Figure 1. Relevant elements of the rail infrastructure adapted from [13,26].

Figure 2. Direct and indirect damages to the rail infrastructure due to river flood or heavy-rainfall-induced flash flood [14,15,16,17,18,19,20,21,22,23,27].

Figure 2. Direct and indirect damages to the rail infrastructure due to river flood or heavy-rainfall-induced flash flood [14,15,16,17,18,19,20,21,22,23,27].

Figure 3. Categorization of exemplary damage cascade effects to the rail infrastructure for a long-lasting flood in lowlands, a flash flood in a low mountain range and a mudslide in a mountainous area after a heavy rainfall event.

Figure 3. Categorization of exemplary damage cascade effects to the rail infrastructure for a long-lasting flood in lowlands, a flash flood in a low mountain range and a mudslide in a mountainous area after a heavy rainfall event.

The rail infrastructure in Western Germany sustained considerable damage due to heavy rain, flash floods and river floods in July 2021. Around 600 km of track was damaged [28], mainly affecting the Ahr route, Eifel route, Erft valley route, Volme valley route and Voreifel route [29]. Damages included over 50 bridges, 180 level crossings, nearly 40 signal boxes and over 1000 catenary and signal masts [30]. Additionally, energy systems, elevators and lighting systems were damaged. The damage to the rail infrastructure following the event in July 2021 along the Eifel and Ahr is estimated at around EUR 1.3 billion [31]. For context, Bubeck et al. [14] estimated annual flood-related damages to rail networks in Europe at around EUR 581 million. The reconstruction of the heavily affected railway lines takes several years and is still in progress. For instance, the Eifel route was reopened in summer 2025, while it will take until December 2025 for the Ahr route to be completely repaired [31,32,33].

The extent of these damages and the long period of time required to reconstruct the rail infrastructure highlight the vulnerability of rail infrastructure to flash floods. Climate models indicate an increase in heavy rainfall events in frequency and intensity in Europe including Germany as a consequence of climate change [34,35]. Consequently, more damages to rail infrastructure due to heavy rainfalls and flash floods are expected in the future. Since rail is an important mode of transport in the context of the European transport network for goods and people, a rail infrastructure more resilient to heavy rainfall is needed. Enhanced understanding of the damage potential of flash floods and heavy rainfall to the rail infrastructure is a key issue to develop recommendations for a more resilient rail infrastructure. Therefore, this study aims to develop a new multi-criteria damage classification for flood and flash flood damages to the rail infrastructure and to conduct a detailed damage analysis on the section of the Eifel route from Kall to Nettersheim resulting from the heavy rainfall and flood event in July 2021 by applying the newly developed multi-criteria framework. For this purpose, damage data was collected, and object categories, alongside damage processes and classification, were developed for uniform analysis purposes. Subsequently, affected elements within the rail infrastructure and correlating damage processes will be analyzed.

2. Materials and Methods

2.1. Development of a Multi-Criteria Framework

To date, there has been no unified data collection of heavy-rainfall-induced damage for rail infrastructures in Germany. As damage data is available in a variety of data types, such as photos and reports and from various sources like newspapers, literature and railroad track tapes, each of these data types reflects damage at a different level of detail. However, a uniform damage data basis and damage data classification is required for reliable data processing and analysis.

Only a few flood damage classification frameworks already exist in the literature. For example, Kellermann et al. [36] created a classification scheme to estimate flood damage to the railroad infrastructure according to three damage classes visualized with reference images. Damage class 1 refers to no or only little damage, whereas damage class 2 describes a complete inundation of the rail infrastructure and damages to the substructure. Additional damage to the superstructure, catenary and signals leading to a complete reconstruction are classified as damage class 3.

Further damage classifications with a greater level of detail of different areas were developed by Maiwald and Schwarz [37] for flood damage to buildings, Burghardt et al. [38] for flood damages to bridges, and Ril 80480 [39] by DB InfraGO for unspecified damages to engineering structures of the rail infrastructure. These damage classifications include five [38,39] to six damage classes [37] ranging from no damage to damages of different detail to complete destruction. Maiwald and Schwarz [37] and Burghardt et al. [38] additionally proposed a detailed description of damage and a visualization which were assigned to a damage class.

However, existing damage classification systems are either too generalized for the purpose of a detailed damage analysis as needed for this study [36] or were developed for different areas than the rail infrastructure [37,38,39]. Therefore, concepts of the existing literature were used as a reference for our study to develop a new multi-criteria (flash) flood damage assessment framework. The development of the framework and its application contains:

• A uniform damage description with newly defined object categories and damage processes.
• A uniform and user-friendly damage classification to five damage classes with reference images and exemplary photos, as a combination of the object category and the damage process.
• A class-based analysis of the most frequent damaged objects of the rail infrastructure, flood damage processes and correlating damage classes (refer to Section 2.2.4).
• The mapping of the most frequent damages using the Geographic Information System software QGIS (versions 3.28 and 3.40) and its tools (refer to Section 2.2.3).
• A flood damage exposure analysis of the most frequent damages with at least one parameter (refer to Section 2.2.4).

Object classification was defined on the basis of the existing literature [13], the urban flash flood (URBAS) database and various sources from Deutsche Bahn (DB), Germany’s largest railroad company, including existing data from DB InfraGO AG incident reports, and the common trades of DB AG, with the participation of experts from DB InfraGO. The defined object groups need to be detailed enough to allow for a precise analysis and summarize the most important elements of the rail infrastructure. Therefore, the defined object groups (refer to Figure 4) are structured hierarchically by level of detail and subdivided into seven top categories: track superstructure, track substructure, geotechnical and retaining structures, technical infrastructure, buildings and other structures as well as surrounding areas and miscellaneous. Overall, a total of 81 objects of the rail infrastructure were defined and assigned to seven top categories and 24 sub-categories. The number of defined objects and categories is therefore large enough to allow for a detailed damage classification without restricting a practical application.

(Flash) flood damage processes (refer to Figure 5) were defined and structured hierarchically in six top categories: erosion, flooding and inundation, transport and deposit of debris, water contact of cables and facilities, miscellaneous and unknown. For example, scouring and mass movement are classified as erosion, whereas flooding and inundation include saturation or flooded buildings. In total, 19 damage processes assigned to 6 top categories and 13 sub-categories were described.

Furthermore, five damage classes (refer to

Table 1. Damage class description.

Damage Class (D)	Description
D0	No damage
D1	Minor damage
D2	Moderate damage
D3	Severe damage
D4	Complete destruction

) were defined to classify flash floods and river flood damages of the railway infrastructure. Damage class D0 is associated with no damage, damage classes D1 to D3 represent increasing damage from minor to moderate to severe, and D4 describes the complete destruction of an element. Furthermore, to consider different effects of damages to different elements of the rail infrastructure, 32 detailed descriptions of damages (refer to

Table 2. Damage classes for rail infrastructure due to flood (refer to Table 1) and exemplary visualization or photo (photos: Schüttrumpf, 14 July 2021, 10 August 2021 and 28 December 2023; Winandy 24 October 2022; Schriewer 12 September 2024, 8 May 2025 and 16 May 2025).

No.	Description	Damage Class (D)	Reference Image	Exemplary Photo
	Cracks in structures
S-1	Minor cracks in structure (<0.5 m)	D1
S-2	Major cracks in structure (>0.5 m)	D2
	Mass movement and subsidence
M-1	Subsidence, slope collapses (<1 m3)	D1
M-2	Subsidence, slope collapses (>1 m3)	D2
M-3	No damage to substructure (earth body)	D0
M-4	Underwashed or outwashed substructure (earth body) (<1 m3)	D1
M-5	Underwashed or outwashed substructure (earth body) (>1 m3)	D2
M-6	Change in position, deformation or lowering of superstructure	D2
M-7	Underwashed or outwashed platform or superstructure (<1 m3)	D2
M-8	Underwashed or outwashed platform or superstructure (>1 m3)	D3
M-9	Underwashed or outwashed foundations (scouring), Washing away of structural element	D3
M-10	Collapse or failure of structural element	D3
M-11	Collapse of entire building or structure, destruction of upper and lower structure, roadway, platform, building, bridge	D4
M-12	Washing away of entire building or structure	D4
M-13	Subsidence in the area of the foundations	D3
	Contamination
C-1	Minor contamination of buildings and structures by non-hazardous substances	D0
C-2	Contamination by hazardous substances	D1
C-3	Deposition of mud, wood or debris in track bed (<1 m3)	D1
C-4	Deposition of dirt, or clogging of drainage systems	D1
C-5	Deposition of various substances on surrounding areas, in water bodies	D1
C-6	Failure of switch mechanisms due to siltation	D2
C-7	Deposition of mud, wood or debris in track bed (>1 m3) or branches with a diameter > 25 cm	D2
	Dampness
W-1	Water contact, dampness track bed without damage	D0
W-2	Dampness substrate, water contact with object excluding light, control, and signaling technology	D1
W-3	Tunnel or station flooded	D2
W-4	Water contact, dampness building (<0.5 m)	D1
W-5	Water contact, dampness building (>0.5 m)	D2
W-6	Water contact, dampness light, control, and signaling technology	D2
	Other damage
O-1	Short circuit or defective cable	D1
O-2	Damage to utility lines (water/electricity/etc.)	D2
O-3	Station closed	D1
O-4	Infrastructure disruption	D1

) for the 5 damage types cracks in structures, mass movement and subsidence, contamination, and dampness, as well as other damages, were developed as a guideline to assign damages to the previously described damage classes. Variations in the damage extent are considered for cracks in structures, subsidence or slope collapses, and underwashed or outwashed substructure or superstructure. A crack greater than 0.5 m (refer to Table 2, S-2) is considered more severe and classified as D2 compared to minor cracks below 0.5 m (refer to Table 2, S-1). Damages due to mass movement including slope collapses and underwashed superstructure are dependent on the volume. A mass movement greater than 1 m³ (refer to Table 2, M-2, M-5 and M-8) is assigned to a higher damage class than a mass movement below 1 m³ (refer to Table 2, M-1, M-4 and M-7). Additionally, a differentiation for the affected element was considered. Mass movement at the superstructure directly affecting the rail infrastructure is assigned to a higher damage class (D2 to D3) than mass movement on the side of the substructure (D1 to D2). A washout near foundations, e.g., bridges or the failure of a structural element (refer to Table 2, M-9) is associated with damage class D3, whereas the collapse of an entire building or whole structure (refer to Table 2, M-11) is assigned to damage class D4, complete destruction. The contamination of the rail infrastructure including the deposition of debris and hazardous substances (refer to Table 2, C-2 and C-3) is classified as minor damage (D1), whereas affected technical infrastructure due to contaminations or water contact (refer to Table 2, C-6) is classified as D2. Other variances in the damage class are possible but should be limited to ensure a uniform classification. Exemplary visualizations are displayed in Table 2.

2.2. Application of the Framework: Eifel Route from Kall to Nettersheim, July 2021

2.2.1. Study Area and Heavy Rainfall Event

This study chose the heavily affected Eifel route as the case study region. The Eifel route (route number 2631 in the DB railway network) is a 163.8 km long railway line running from Hürth-Kalscheuren close to the city of Cologne in NRW through the Eifel mountains to Trier-Ehrang in RP. It is used for regional passenger transport only and was not electrified before July 2021. As part of the reconstruction work, tracks, culverts and bridges were being renewed, embankments that have been washed away were rebuilt and signaling and interlocking technology were modernized [31]. In addition to the reconstruction, the electrification of the line is expected to be completed by 2026 so that continuous electric train operation will be possible from 2028 onwards [33].

For our study, the section from Kall to Nettersheim (Figure 6) was analyzed in detail. This section is located in NRW in Germany and covers a length of almost 10 km with 13 smaller bridges or culverts. It includes the three train stations Kall, Urft and Nettersheim. The railroad line runs along the Urft river and is located in a valley with adjacent slopes of up to 500 m above sea level. The catchment area of the Urft river covers 131.11 km² [46].

Between 13 July 2021 and 15 July 2021 many areas also received high precipitation sums in the proximity of the Eifel route. At Kall station, the daily precipitation equaled 124 mm on 14 July 2021 (ref. to Figure 6). For comparison, the long-term monthly average of precipitation from 1961 to 1991 is around 80 mm in this area [47]. The immediate vicinity of the railroad line was therefore affected by high precipitation totals, which resulted in high water levels. These were recorded by the Urft river gauging station “Kall-Sportplatz” in the Rur catchment area, which is located close to the train station Kall. The water level of the Urft river at gauge “Kall-Sportplatz” (refer to

Table 3. Significant water levels and discharges at the gauging station Kall-Sportplatz in the vicinity of the Eifel route from Kall to Nettersheim [46,50,52].

	Gauge Kall-Sportplatz
Operator	LANUV
Associated water body	Urft
Catchment area	Rur
Size of catchment area	131.11 km2
Nearest railroad station	Kall
Highest water level July 2021	3.55 m, 14 July 2021, 22:15
Water level at HQfrequent	2.05 m
Water level at HQ100	2.59 m
Water level at HQextreme	3.15 m
Highest discharge July 2021	400.67 m3/s, 14 July 2021, 22:15
HQ100	-
Highest measured discharge before July 2021	76.51 m3/s, 28 September 2007

) resulted in an estimated maximum discharge of around 400 m³/s at the peak of the water level in the late evening of 14 July 2021. This corresponds to a discharge of around 930 times the lowest discharge before the event of around 0.43 m³/s on 10 July 2021 at 2:45. The discharge of the Urft river and the water level at the gauging station were the highest ever recorded. This underlines the extent of the heavy rainfall and flood event. Unfortunately, the authors did not have access to further hydrological data evaluate this region. However, the available water level data from the Urft river can be used to assess the extent of the event.

Figure 6. (left): Map of the study area from the Eifel route (route number 2361) with the section from Kall to Nettersheim in the State of North Rhine-Westphalia (NRW) in Germany based on [48,49,50]. (right): Precipitation sum [mm] in a 1 km × 1 km grid for 14 July 2021 based on [51].

Figure 6. (left): Map of the study area from the Eifel route (route number 2361) with the section from Kall to Nettersheim in the State of North Rhine-Westphalia (NRW) in Germany based on [48,49,50]. (right): Precipitation sum [mm] in a 1 km × 1 km grid for 14 July 2021 based on [51].

2.2.2. Data Collection and Data Availability

The damage data of the Eifel route from Kall to Nettersheim after the flood and heavy rainfall event in July 2021 was collected and provided by DB InfraGO AG and DB E&C, as only a limited amount of damage data is available publicly. The data sources provided include photos, fault reports, an explanatory report and a route tape for the Eifel route. The data sources provided are described briefly and examined with regard to the data availability of the selected route section in the following. Unfortunately, financial data could not be provided for this study due to data protection regulations.

Georeferenced photos by DB InfraGO and DB E&C were the main source for the detailed damage analysis (ref. to

Table 4. Sum of provided damage data and correlating damage descriptions per data type.

Provided Data by DB AG	Data Source	Damage Data Descriptions [-]	%
Georeferenced photos (n = 197)	[48,53]	435	77.26
Explanatory report (n = 1)	[44]	100	17.76
Route tape (n = 1)	[54]	22	3.91
Fault reports (n = 6)	[43]	6	1.07
Total		563

). Photos include damage documentation during route inspection and drone recordings in August and September 2021. A total of 501 georeferenced photos depict the Eifel route from Kall to Nettersheim. Damage was visible on around 84% of the photos (n = 422). For each damage location, a representative photo was selected. In total, 197 photos were analyzed in detail.

Provided reports by DB InfraGO AG include an explanatory report and a route tape of the Eifel route. The reports contain damage descriptions with varying degrees of detail. The explanatory report describes damages assigned to the track kilometer and objects like bridges, culverts, railroad crossings, track superstructure, track substructure and drainage systems. Additionally, one damaged building is described. Damages to cables, control and safety technology, telecommunications and electrical energy systems, and specifically level crossing network connections are not included, as they are assigned to a different project. Additionally, reconstruction measures are described briefly. The route tape also links described damages to the track kilometer. However, damage descriptions are less detailed and the included photos are assigned broadly to damaged track sections.

Furthermore, fault reports from DB InfraGO AG were used to collect damage data on the rail infrastructure. These are available throughout Germany from 2017 and show reported faults in tabular form. The available reports were filtered in detail using the short text and the cause text of a fault message in order to establish a link to heavy rainfall or precipitation events. Only relevant fault messages with the keywords “rain”, “precipitation”, “water”, “flooding”, “flooded” and “mud” were used. Incident reports with the keywords “sleet” and “tree” or “tree fall” were excluded from further evaluation. Of the original 3796 fault reports, six fault reports were classified as relevant for the analysis of the Eifel route from Kall to Nettersheim for the July 2021 event.

2.2.3. Data Preparation and Damage Mapping

The original data from DB AG was described and classified by the authors using the presented framework to create a standardized data basis. Existing classifications of the original data were not evaluated and original damage data [43,44,48,53,54] is not shown in this work due to data protection. The described damages as seen on images or from reports were assigned to the correlating track route, the flood event, an object (refer to Figure 4), a damage process (refer to Figure 5) and a damage class (refer to Table 1 and Table 2), and collected in a tabular database. If multiple damage processes or objects were visible on one photo or described in one report, the description was multiplied for each damage type. Therefore, the number of recorded damages (refer to “damage data descriptions” in Table 4) can be greater than the quantity of images or reports. For example, 435 different damages were identified on 197 georeferenced photos (refer to Table 4).

Using this visual analyzation of images, direct tangible damages can be captured effectively. However, indirect damages emerging later, intangible damages or direct damages detected at a later time, for example, while reconstructing the damaged train section, could not be captured with the approach of visual image classification. Potential indirect damages could include disruptions to local or national transport links due to damaged railway lines or stations, construction noise from repair activities, and other secondary effects. These damages were not analyzed in this study.

Unfortunately, the given track route layer from DB [55] differed up to 5.0 m from the location determined on the OpenStreetMap [56] map and the location in the aerial photo [57]. Therefore, an adjusted track axis was determined according to the aerial photo from 18 April 2022 [57] (refer to Figure 7).

Data preparation for the QGIS analysis includes the determination of the three most frequent damages, which equals a sufficiently large number of damage descriptions for a reliable analysis. Therefore, the visual damages, washout of ballast and soil as well as the deposit of wood and gravel within a distance of up to 10.0 m to the track axis, were considered for damage mapping (refer to Figure 7). This buffer was chosen as deposit of debris (e.g., wood and gravel) close to the rail track can cause blocking of the drainage system resulting in an increased water level. Furthermore, floods or flash floods can cause saturation and lateral erosion of the superstructure, substructure and the surrounding embankment (refer to M-1 to M-6 in Table 2), which can further result in damages to the rail infrastructure, e.g., landslides and deposit of mud on the track bed (refer to M-1, M-2, C-2 to C-7 in Table 2).

These damages were mapped by hand as colorful lines in QGIS (versions 3.28 and 3.40) along the adjusted track axis using reference points for localization from the environment, e.g., buildings, roads or signs along the track route. The used data basis for localization were primarily the geo-tagged photos from August and September 2021 [48] and the aerial photo from 18 April 2022 [56], as debris removal work, e.g., removal of rail and trees, had already been carried out in 2022. The mapped damages consist of individual sections with length ranging from a few meters to a couple hundred meters. No visual damage to the rail track is defined as the absence of the previously described damages and was also mapped as a line along the adjusted rail axis.

Figure 7. Exemplary cross-section based on [13] and photographic aerial view [57] of the Eifel route track section near the Urft river [58] with indication of the buffer of 5.00 m (width of track bed including track embankment of a single-track route) and 10.00 m (surrounding area with drainage system and embankment of a single-track route) along the track axis [55] used for further analysis.

Figure 7. Exemplary cross-section based on [13] and photographic aerial view [57] of the Eifel route track section near the Urft river [58] with indication of the buffer of 5.00 m (width of track bed including track embankment of a single-track route) and 10.00 m (surrounding area with drainage system and embankment of a single-track route) along the track axis [55] used for further analysis.

2.2.4. Class-Based Analysis and Flood Damage Exposure Analysis

The class-based analysis was carried out using the tabularly collected and uniformly described damage data consisting of the provided photos and reports (refer to Table 4) according to the defined object categories (refer to Figure 4), the damage process (refer to Figure 5) and the damage classes (refer to Table 1 and Table 2). To calculate percentages of damages per object categories, damage processes and damage classes, the total number of the classified damages was used as a reference. These results were displayed as tables or diagrams.

As highlighted by Lehmkuhl et al. [22], it is important to analyze geomorphological processes during flood events as they can influence damage patterns. Therefore, as part of the flood damage exposure analysis, the locations of the most frequent damages were compared to the proximity of the Urft river, the river slope characteristics and the water depth from the hazard maps of NRW. The hazard maps used for this analysis are exemplified by the flood hazard map of an extreme event (HQ > 250), which describes a flood event with a statistical return period of 250 years and the flash flood hazard map for an extreme event, which was created using a simulation with uniform rainfall of 90 mm for one hour.

To analyze the distance of the mapped damage sections to the Urft river axis, points along the linear damage sections were determined every meter (refer to Figure A1). Afterwards, the shortest distance of these points to the river axis were determined and classified in four groups according to their distances ranging from less than 20 m to 150 m. The comparison of the distance to the river and its slope characteristics was carried out for the points assigned to a river cliff or a slip-off slope. A river cliff refers to the impact zone of the river, where erosion processes can be present. A slip-off slope describes the river zone, where mostly deposition of transported sediment is present. At some sections, the railroad track lies next to a river section with a large number of changes in the flow direction. These sections were defined as a combination of a river cliff and a slip-off slope and were not considered in combination with the distance to the Urft river. The correlating slope characteristics of the Urft river, river cliff, slip-off slope and a combination of a river cliff and slip-off slope in the vicinity of these mapped track sections, were assigned by hand using a visual analysis of the QGIS river layer.

The mean water depths from the flood and flash flood hazard maps were determined within a 5.0 m buffer (h_{mean, 0–5.0 m}) to take into account the width track bed, including the track embankment (around 2.50 m from the track axis on each side) and an additional 5.00 m (h_{mean, 5.0–10.0 m}) on each side of the track bed of the studied single-track route to consider the width of the surrounding area with any drainage system and embankment (refer to Figure 8 and Figure 9). The mean water depths were determined using the QGIS (versions 3.28 and 3.40) tool “zone statistics” and classified in four categories ranging from <0.1 m water depth to 1.0–2.0 m. For different track routes, e.g., with multiple tracks, a greater buffer should be chosen according to their width.

3. Results

3.1. Class-Based Analysis of Railroad Flood Damage

The class-based analysis was carried out with 563 tabularly collected damage descriptions from photos and reports containing the uniform classification according to the previously defined objects (refer to Figure 4), damage process (refer to Figure 5) and damage classes (refer to Table 1 and Table 2). Further, the collected data was analyzed with regard to these categories. The majority of the available damage data (refer to

Table 5. Quantity of damages to elements of the rail infrastructure (top category and sub-category) of the Eifel route from Kall to Nettersheim due to the heavy rainfall and flood event in July 2021 and its percentage in relation to the total number of damage descriptions of the available data set (refer to Table 4) with data limitations regarding cables, control systems and the substructure (refer to Section 2.2.2).

Object Group	Sub-Category	n	%	%
Track superstructure	Bedding and sleeper plate	126	22.38	44.93
	Other or not specified	66	11.72
	Track	61	10.83
Track substructure	Sealing and drainage system	35	6.22	16.87
	Railway overpass and bridge	30	5.33
	Earthworks and embankment	24	4.26
	Pedestrian crossing	3	0.53
	Other or not specified	3	0.53
Buildings and other structures	Railroad crossing	15	2.66	4.62
	Other structures and buildings	7	1.24
	Train station	4	0.71
Geotechnical and retaining structures	Foundation	2	0.36	0.36
Technical infrastructure	Power supply system	48	8.53	14.39
	Control, signal and safety technology (CST)	24	4.26
	Telecommunications facility	7	1.24
	Other or not specified	2	0.36
Surrounding areas	Grassland	42	7.46	18.83
	Water body	34	6.04
	Other properties	19	3.37
	Traffic area	11	1.95
	Total number of damage descriptions	563

) was assigned to the following object categories: track superstructure (44.93%), surrounding areas (18.89%), track substructure (16.87%) and technical infrastructure (14.39%). The most significant damage to the superstructure corresponds to the sub-categories bedding and sleeper plate (22.38%) and track (10.83%). Object categories with the least damages are buildings and other structures (4.62%), as well as geotechnical and retaining structures (0.36%).

The most frequent damage processes identified are erosion and transport or deposit of debris with 40.50% and 43.69% (refer to Figure 10 and

Table A1. Quantity and percentage (in relation to the total number of damage descriptions) of damage processes to elements of the rail infrastructure of the Eifel route from Kall to Nettersheim due to the heavy rainfall and flood event in July 2021 of the available data set (refer to Table 4) with data limitations regarding cables, control systems and the substructure (refer to Section 2.2.2).

Damage Process	Sub-Category 1 - Sub-Category 2	n	%	%
Erosion	Scouring - Washout of ballast/gravel/soil (n = 130) - Underwashing (n = 71) - Track distortion (n = 20) - Other (n = 1)	222	39.43	40.50
	Mass Movement	6	1.07
Miscellaneous	Other	7	1.24	1.24
Transport/Deposit of Debris	Deposit - Deposit of wood/branch/tree (n = 112) - Deposit of gravel/stones (n = 111) - Deposit of waste (n = 13) - Other (n = 8)	244	43.34	43.69
	Obstruction/Impact	2	0.36
Flooding/Inundation	Other or not specified	53	9.41	12.08
	Sweeping Off	7	1.24
	Floodplain	6	1.07
	Saturation	1	0.18
	Flooded Buildings	1	0.18
Water Contact of Cables and Facilities	Other or not specified	7	1.24	1.42
	Water Contact Cable	1	0.18
Unknown		6	1.07	1.07
	Total number of damage descriptions	563

). In particular, the damage process deposit (43.69%), including the sub-categories deposit of wood, branch and trees and deposit of gravel and stones, as well as the damage process scouring (39.43%), a sub-category of erosion, including the washout of ballast, gravel and soil, are present. The least frequent damage processes identified are flooding and inundation (12.08%) and water contact of cables and facilities (1.42%).

The most frequent combinations of an object group and a damage process are displayed in Figure 10 and

Table A2. Quantity and percentage (in relation to the total number of damage descriptions) of the most frequent damage processes of top category (t) sub-category 1 (s1) and sub-category 2 (s2) per object group (n = 563) of the available data set (refer to Table 4) with data limitations regarding cables, control systems and the substructure (refer to Section 2.2.2).

Object Group	Damage Process	n	% of Object Group
Track superstructure	Washout of ballast/gravel/soil (s2)	114	45.06
	Deposit of wood/branch/tree (s2)	68	26.88
	Deposit of gravel/stones (s2)	25	9.88
	Track distortion (s2)	20	7.91
	Other (t)	16	6.32
	Flooding/inundation (t)	10	3.95
	Total	253
Track substructure	Deposit of wood/branch/tree (s2)	29	30.53
	Flooding/inundation (t)	31	32.63
	Other (t)	13	13.68
	Washout of ballast/gravel/soil (s2)	12	12.63
	Deposit of gravel/stones (s2)	5	5.26
	Mass movement (s1)	5	5.26
	Total	95
Buildings and Other Structures	Flooding/inundation (t)	10	38.46
	Erosion/scouring (t)	6	23.08
	Unauthorized railroad crossing (s1)	5	19.23
	Transport/deposit of debris (t)	5	19.23
	Total	26
Geotechnical and Retaining Structures	Erosion/scouring (t)	2	100.00
	Total	2
Technical Infrastructure	Erosion/scouring (t)	54	66.67
	Flooding/inundation (t)	10	12.35
	Other (t)	10	12.35
	Water contact of cables and facilities (t)	7	8.64
	Total	81
Surrounding Areas	Deposit of gravel/stones (s2)	79	74.53
	Deposit wood/branch/tree (s2)	12	11.32
	Deposit of waste (s2)	7	6.60
	Floodplain (s1)	6	5.66
	Other (t)	2	1.89
	Total	106
	Total number of damage descriptions	563

. Overall, the washout of ballast, gravel and soil at the track superstructure (n = 114), the deposit of gravel and stones on surrounding areas (n = 79), the deposit of wood, branch and trees on the track superstructure (n = 68) and erosion, as well as scouring at technical infrastructures (n = 54), are predominant. However, the most dominant damage process differs per object group. In total, 45.06% of the damages to the track superstructure are assigned to washout of ballast, gravel and soil, and 26.88% to the deposit of wood, branches and trees. The track superstructure is mainly affected by flooding and inundation (32.63%) and the deposit of wood, branch and trees (30.53%). Buildings and other structures, as well as technical infrastructure, are mainly affected by flooding and inundation, as well as erosion and scouring. The deposit of gravel and stones is the most frequent damage process for surrounding areas, with 74.53%. Assigned damages to geotechnical and retaining structures are induced by erosion and souring (100%); however, only two damages were assigned to this object category in total.

Overall, the analyzed damages were mainly classified as D1 (minor damage) with 47% or D2 (moderate damage) with 34% (refer to Figure 11). Only 4% of the damages were classified as D4 (complete destruction). Damages assigned to the object top categories, track superstructure and track substructure, as well as buildings and other structures, are mainly classified as D1, ranging around 35% to 50% each (refer to Figure 12). Additionally, damage classes D2 to D4 are present. Damage class D2 was mainly assigned to the top category, technical infrastructure, with over 70%, damages to surrounding areas are mainly classified as D1, and damages to geotechnical and retaining structures are assigned to damage class D3 (severe damage). The top categories of the damage processes are mainly classified as D2 and D3 for erosion, transport and deposition of debris as D1 and water contact of cables and facilities as D2. Damage process flooding and inundation is equally assigned to damage classes D1 to D4.

3.2. Flood Damage Exposure Analysis

Geotagged photos (n = 197) and the most frequent damage processes, washout of ballast, gravel and soil, the deposit of wood, branches and trees, and the deposit of gravel and stones, were mapped and analyzed using the geographic information system QGIS (versions 3.28 and 3.40). The three linear mapped damages sometimes appear in the same location, especially the deposit of wood or gravel and the washout of gravel, as shown in Figure 13. In total, 2.90 km of the Eifel route was affected by the washout of ballast, gravel and soil, equaling 29.52% of the track length from Kall to Nettersheim. Additionally, the deposit of wood, branch and trees equals 0.90 km (9.13%), and the deposit of gravel and stones amounts to 1.84 km (18.79%). Undetected damage in photos indicated as “no visible damage” was mapped for 69.36% of the track route (6.80 km). It shows that almost half of the Eifel route section is affected, especially near the Urft river watercourse, around bridges, in narrow valleys, and near a meander (refer to Figure 13). Furthermore, damages were identified in the vicinity of 9 of the 13 bridges and culverts. Three out of the four intact bridges or culverts are located in the immediate vicinity of Kall station. For further analysis, the locations of the most recent damage processes were compared to the distance to the Urft river, its river cliff or slip-off slope and the water depth of the flash flood and flood hazard map for an extreme event. The results are described in the following sections.

3.3. Proximity to the Urft River and Slope Characteristics

To compare the distance of the railroad track to the slope characteristic of the Urft, 2664 damage points and 6786 points with no visible damage distinctly assigned to a river cliff or a slip off-slope were considered. Over 93% of the mapped deposit of gravel and stone and of the washout of ballast, gravel and soil is located within a maximum distance of 50 m to the Urft river with more than 57% within 20 m. Additionally, around 80% of the deposit of wood, branch and trees is mapped within 50 m to the Urft. Furthermore, the washout of ballast, gravel and soil and the deposit of gravel and stone was mainly detected at river cliffs of the Urft river with a maximum distance of 20 m (refer to Figure 14), whereas the deposit of wood, branches and trees was mainly located on the side of the slip-off slope of the Urft river.

No visible damage is located in the vicinity of Kall station, Nettersheim station, over a length of approximately 2 km between Nettersheim and Urft stations and with a greater distance to the Urft river compared to the damaged track sections (refer to Figure 14). Overall, the mapped damages are located at a maximum distance of 20 m, 20–50 m and 50–100 m to the Urft river with similar percentages. Furthermore, no visible damage is nearly equally detected at river cliffs and slip-off slopes within 100 m distance of the Urft river.

3.4. Water Depth from Flood and Flash Flood Hazard Maps

Furthermore, the flood plain area of the flood hazard map for an extreme event (>HQ 250) was compared to the mapped lines of the most frequent damages. The mean water depth for a 5.0 m buffer (left) and a 5.0 m to 10.0 m buffer (right) are displayed in Figure 15. Within the 5.0 m buffer, water depth of up to 2.0 m is present for all damages. The washout of ballast, gravel and soil, and the deposit of gravel and stone are mostly present between 0.1 m to 0.5 m (around 60% each). The deposit of wood, branches and trees mostly occurs with similar percentages below 0.1 m, between 0.1 m and 0.5 m and between 1.0 m and 2.0 m with around 30% each. No damages occur mostly below 0.1 m within the 5.0 m buffer. Considering the 5.0 m to 10.0 m buffer, damages mainly occur at water depths between 0.5 m and 1.0 m (around 60%), whereas no visible damage mainly occurs at water depths below 0.5 m (around 70%).

In addition, the mean water depths of the flash flood hazard map (refer to Figure 16) for an extreme event (90 mm/h) were analyzed. Damages mostly occur below 0.1 m, ranging from around 60% to 90% within the 5.0 m buffer and 30% to 70% for the 5.0 m to 10.0 m buffer. The washout of ballast gravel and soil, as well as the deposit of gravel and stone, is present with around 60% to 70% for a mean water depth of 0.1 m to 0.5 m within the 5.0 m to 10.0 m buffer. No visible damage occurs below 0.1 m and between 0.1 m and 0.5 m, with nearly 50% each within the 5.0 m buffer and with around 30% and almost 70% within 5.0 m to 10.0 m. Water depths higher than 1.0 m are not present in the area of the most frequent damages located.

4. Discussion

4.1. Development and Application of Multi-Criteria Framework

Recent flood events caused significant damages to the rail infrastructure. To determine frequently affected elements of the rail infrastructure and underlying damage processes, a multi-criteria framework designed to classify damages to railroad infrastructure caused by flooding events was developed as part of this work. This framework contains 81 hierarchically structured object groups and 19 damage processes, as well as a total of 32 visualized exemplary damages for 5 damage classes ranging from D0 (no damage) to D4 (complete destruction) with greater detail compared to the existing literature, like the work of Kellermann et al. [36].

The predefined damage descriptions of the object groups and damage process, the exemplary visualizations and the exemplary photos of damage could successfully be used by us to describe and classify damages from documents or photos, as well as to assign them to a damage class. The development of a user-friendly multi-criteria damage assessment framework was aimed for and therefore successfully developed. We are confident that this multi-criteria framework is also easily applicable for other study areas and by other users.

Further benefits of this multi-criteria framework are the detailed and object specific damage classification and analysis to establish a detailed damage database for the rail infrastructure. The damage data assessment can be simplified when using this proposed structure compared to inconsistent damage descriptions from different individuals. On the basis of the classified damage data, comparisons and correlations could be conducted, highlighting the relationships between proximity to rivers and their slope characteristics, as well as the flooded areas and water depths from existing flood and flash flood hazard maps. In particular, the comparison of the rail infrastructure with existing hazard maps represents an easy-to-implement assessment of potential damage to the rail infrastructure that can be applied beyond the boundaries of the study area. Additionally, the proximity to rivers and streams can be easily implemented by means of a comparison. However, the identification of river-cliff and slip-off slopes of a river requires additional groundwork.

4.2. Damage Data Availability and Data Preparation

In general, damage data analysis is limited to the available data set. In this study, damage descriptions were derived from photos and reports, e.g., the explanatory report (refer to Table 4) provided by DB. Additional data, like photos, or damage descriptions from other sources, like newspapers, were not considered or evaluated with regard to data availability and data usability in this study.

The provided photos were available in a high quantity, with damages displayed in detail. They could therefore be used as a reliable source for damage classification of the shown rail infrastructure. The photos used in our study were mainly taken up to two months after the flood event, i.e., almost no restoration work or limited removal of deposited material had taken place. This is ideal for detecting the exact position of deposited material and analyzing the geomorphological and hydrological processes that caused the erosion, transport and deposit processes. However, damages to the substructure and technical devices may not be visible on photos, possibly leading to an underestimation of the damage process “water contact of cable and facilities” and the object group “technical infrastructure”.

Damage descriptions of the explanatory report excluded control and safety technology, telecommunications, cables and electrical energy systems. These elements of the rail infrastructure could only be taken into account to a limited extent from the visual evaluation of the available photos, as they are hardly visible. We assume that damages to these elements are underrepresented in this study. However, the remaining object categories could be analyzed in great detail as the damage data was highly available for damages to the track superstructure.

Furthermore, only direct visible damages could be identified. To identify indirect damages occurring at a later point in time without water contact, an additional damage collection is proposed after an initial cleaning or reconstruction of the damaged area. In this way, additional damage can be recorded that was not previously visible, e.g., due to the deposit of wood or debris near bridges. Furthermore, additional data resources, e.g., earth observation (EO) data, could be used to determine emerging damages like deformations of bridges or embankments [61]. Additionally, we propose that a damage or track section specialist should assess the damage after its collection and document it adequately. Damage should also be mapped, e.g., using QGIS, and compared with the parameter set in this paper. This will enable a more accurate assessment of the damage and its classification.

4.3. Identified Damages

The class-based analysis of the damage data using the proposed framework shows that the majority of the damage data was assigned to the track superstructure, surrounding areas, the track substructure, and technical infrastructure. Additionally, the most frequent damage processes are erosion and transport and deposit of debris, especially including deposit of wood, branches and trees and deposit of gravel and stones. These results can be compared to damage descriptions from the literature (refer to Figure 2), e.g., damaged tracks and embankments due to overtopping, washout or submergence [14,16,17,21]. Additionally, the instability of structures, e.g., embankments [16] and the accumulation of debris at culverts and bridges [16,21,62] were also determined in this study. The clogging of drainage, as stated by Varra et al. [16] is not specified this study. However, we assume that clogged drainages are also present and can be estimated with the deposit of wood and stones near the track on the surrounding areas, which could lead to the clogging of drainages. Runoff or permeability tests of drainages were not considered as part of this study.

However, in comparison to described damages in the literature, some object groups are underrepresented in this study, for example, foundations and bridges [1,21,38,62]. It can be assumed that visual photo analysis, the main data source of this study, is not an appropriate method to detect damages like scouring at bridge foundations. Moreover, the QGIS analysis revealed several damages in close proximity of bridges. It can therefore be assumed that damage to the rail infrastructure in close proximity of bridges can be expected due to river flooding and bypassing of water at blocked bridges [62]. A detailed analysis of damaged bridges in the Ahr river valley after the flood in July 2021 conducted by Burghardt et al. [38] revealed that bridge damage was predominantly classified as D1 (minor damage), with a damage potential in urban areas three times higher compared to rural areas. Therefore, a high damage potential of urban railroad bridges is expected. To specify bridge damages in the study area and for other railroad tracks, a detailed damage analysis of the rail infrastructure near bridges should be conducted in further research.

In addition, the number of cable damages after contact with water is very low. It can be assumed that damage to a cable or electrical system occurred more frequently than shown since a short circuit may not be visible on photos. It is very likely that cables and electrical systems close to the ground were at least in contact with water; however, damage can only be clarified by a technical inspection. We also advise conducting further research on flood damages of cables and technical infrastructure.

Furthermore, surrounding structures and buildings were only documented to a limited extend. Therefore, we assume that flooded buildings are underrepresented in this study. It is expected that flooded buildings result in damage to the structure, e.g., due to dampness and contamination, and that, e.g., technical and mechanical, equipment inside of a building can be damaged by corrosion or siltation. A separate detailed analysis of flooded buildings is therefore recommended to analyze their vulnerability to floods.

The analysis of the assigned damage classes ranging from D0 (no damage) to D4 (complete destruction) showed that damage to the Eifel route was predominantly assigned to classes D1 and D2 (81% in total). In total, 15% of the damages were assigned to D3 and only 4% to D4. The quantity of no damage, D0, was not included in the quantitative evaluation of the Eifel route, as no representative data can be guaranteed. The results indicate mostly minor or moderate damages to the rail infrastructure. However, a combination of different damages in the same location can be considered more severe, as extensive reconstruction is needed. The cumulative appearance of damages in the same location and its impact on the damage class was not evaluated in this study. For further analysis, we propose to scale the damage class up if multiple minor or moderate damages are located within a small distance, e.g., within 20 m. Furthermore, the damage classification of the detected damages in this study area was conducted using our proposed damage classification in Table 2. For different damaging events and different track routes, adaptations or extensions of the described damages and their correlating damage class may be needed. Furthermore, the proposed damage classes only refer to the extent of the damage and do not imply the technical and financial aspects of clean-up and reconstruction. Therefore, the complexity of the damage repair cannot be derived from these results.

Overall, the most significant damages to the rail infrastructure as stated in the literature were determined in this study. Underrepresented damages could be detected with an additional damage analysis. Damage classification and evaluation can be conducted using the developed multi-criteria framework. Furthermore, the developed framework could be advanced if needed, e.g., to further specify damages to cables and technical infrastructure.

Large parts of the railroad network in Germany are very old and were built many years ago. These old railroads often follow river valleys to avoid topographically difficult terrain and steep sections or extensive tunnel sequences. Newer lines, especially those for high-speed long-distance transport, are built differently, ignoring the topography by constructing tunnels and bridges. However, these lines are not susceptible to flash floods, the main process our work deals with, as the bridges are located high above small river catchments. Since the construction method for these high-speed lines is different, our case study and the proposed classification scheme cannot be directly applied to this type of railroad. However, the high-speed lines are only a part of the country’s railroad networks, and so many lines, especially in European countries, are built in a similar way to the Eifel route. The Eifel route is therefore a representative case study and the results can be applied to similar routes in comparable topographical settings and with a similar distance to a river. An analysis of these boundary conditions of the railway infrastructure should be carried out before transferring the results of this study to other railroad routes. For different topographic settings, e.g., in lowland areas, different damages can be expected. The extended flood duration with lower flow velocities in lowland areas compared to mountainous regions with steep slopes and high flow velocities can lead to different damages, e.g., caused by long-term dampness and saturation of materials and structures. Therefore, further damage analysis in different topographic settings, in rural and in urban areas, is suggested. If needed, modifications to the multi-criteria framework and the damage analysis should be examined in future research.

4.4. Multi-Criteria Flood Damage Exposure Analysis Using QGIS

For the flood exposure analysis using QGIS (versions 3.28 and 3.40), the location of the three most frequent damage processes washout of gravel, deposit of gravel and deposit of wood were mapped using the provided photos and an aerial photo. Additionally, track sections with no visible damage in photos were classified as D0 (no damage) and mapped. It should be noted that data assigned to category D0 could still contain damages, as complete flooding of the track in the study area can be assumed. For example, invisible contamination of the track ballast due to the deposition of organic substances or fine sediments could have occurred, causing clogged drainages, affecting stability or even leading to hazardous contamination. However, to establish correlations between severely damaged track sections and other parameters, track sections with no visible damage (referring to undetected damage in the analyzed photos) were used for comparison.

The comparison of flood water levels of the Urft river and the maximum precipitation shows that the extent of the July 2021 event was greater than the simulated events of the flash flood hazard map and the flood hazard map. For example, water levels at the Kall-Sportplatz gauge in the study area show a maximum water depth of 3.55 m for the July 2021 event (refer to Table 3), whereas the water level for an extreme discharge (HQ extreme) equals 3.15 m. Furthermore, the daily precipitation sum equaled up to 142 mm in the immediate vicinity of the Eifel section on 14 July 2021 and the underlying flash flood hazard map for a maximum event was created for one-hour precipitation of 90 mm. Therefore, the flooded areas, water levels and flow velocities of the hazard maps can be transferred to the area under investigation and even expected to be higher than estimated in the hazard maps. Hence, track sections of the three most frequent damage processes and the absence of visible damage (indicated as no visible damage) were compared with hazard maps for river floods and heavy-rainfall-induced flash floods.

The determined mean water depths of the flood hazard map reach up to 2.0 m in the proximity of the damages and are higher compared to flash flood water depths of around 0.1 m. Therefore, the analysis shows that the damages in the study area are most likely linked to river flooding. Moreover, within a 5.0 m to 10.0 m distance to the damaged track section, the flood hazard map showed higher water levels (mostly 0.5–1.0 m) compared to a distance up to 5.0 m (mostly 0.1–0.5 m). In comparison to damaged track sections, the water depths of track sections without visible damage are nearly similar within 5.0 m. However, within a distance of 5.0 to 10.0 m, water depth near undamaged track sections is significantly lower compared to damaged track sections. Assuming that flooding leads to damage and that no damage occurs without flooding, flooding areas from the hazard maps are to be expected in the immediate vicinity of the railroad line. Water depth greater than 0.1 m is present both in the vicinity of damaged sections of track and in the vicinity of sections of track without visible damage. This suggests that the entire section was flooded by at least 0.1 m. The weak correlation of water depths and damages within 5.0 m suggests that further parameters must be taken into account when determining the damage potential of the rail infrastructure. At a distance of 5.0–10.0 m, the correlation of damage with higher water depths (mostly 0.5–1.0 m) is more pronounced compared to sections without visible damage (mostly < 0.5 m). It can therefore be assumed that it is possible to identify areas at risk for a 5.0–10.0 m buffer without taking other parameters into account. Furthermore, for a heavy rainfall event with a greater magnitude than the July 2021 event, higher water levels and an increase in the flooded area are to be expected. This may lead to a higher number of damages or even more severe damages.

In comparison with the flash flood hazard map, minor differences were determined for damaged and undamaged track sections for h_{mean, 0–5.0 m} and h_{mean, 5.0–10.0 m}. Correlating water depth are mainly below 0.5 m with a smaller percentage for 0–5.0 m compared to 5.0–10.0 m. Therefore, the water depth of the flash flood hazard map is not significant in determining the damage potential of the track route in this study. However, for different events and topographic settings, the flash flood hazard map may be more significant. We strongly advise comparing damaged track sections to flash flood hazard maps in further studies.

Flow velocities of the flood hazard map were not considered in this study because raster data, the preferred data source in this study, is not publicly available. Flow velocities of the flash flood hazard map were determined but did not show any significant relation to damaged or undamaged track sections. This may correlate with the similar water depths of the flash flood hazard for damaged and undamaged track sections. For further research, the determination of flow velocities is strongly advised, especially in combination with the water depths.

Further analysis regarding the distance to a water body showed that 80% to 90% of the located damages occurred within a 50 m distance to the Urft river, whereas areas with no visible damage are located equally distributed with a maximum of up to 100 m to the Urft river. This underlines a higher damage potential of the rail infrastructure in close proximity to a waterbody. Furthermore, the washout of ballast and the deposit of gravel mainly occur at a smaller distance to a river cliff of a water body. The deposit of wood appeared within 50 m distance of a slip-off slope of the Urft river. As erosion is to be expected on the river cliff and deposition on the slip-off slope [22], higher washout is to be expected at a smaller distance to a river cliff. This correlates with the results in this study. Furthermore, the extensive deposition of wood near a slip-off slope confirms this assumption. The deposit of gravel and stones at a river cliff can be linked to the density of the material since the density of stones is higher than that of wood. It can be assumed that wood was transported for a longer distance than gravel and therefore deposited in close proximity of a slip-off slope as it was flooded.

Overall, the comparison of the located damages with the flooded areas and the correlating water depth is an easy-to-implement method using QGIS (versions 3.28 and 3.40) tools to identify potentially affected sections of the railroad track. Flash flood and flood hazard maps can be used to determine the damage potential of the rail infrastructure. Especially of note, the combination of water depth and flow velocities is used in the literature to classify the damage potential to people, cars or buildings, as summarized by Smith et al. [63]. Therefore, the analysis of these two parameters is strongly proposed. The identified hot-spots could be analyzed more detailed in a next step.

The results from this study can be used as a basis for future studies. However, extensive real-time flood data should be analyzed and compared to the location of railroad damages, if available. True correlations of damaged track sections, water levels and flow velocities can only be conducted with real flood data. However, if extensive flood data is not available and the water depth of a river gauge in the vicinity of the affected area is similar to the water depth from the hazard maps, hazard maps can be used to establish relations.

5. Conclusions and Outlook

The extensive heavy-rainfall-induced flood damage potential of the railroad infrastructure and the predicted increase in the frequency and intensity of these events by climate models highlights the need for a more flood-resilient rail infrastructure. The first step to increase the resilience of the system and to develop technical or organizational measures is knowledge about the most frequent railroad flood damages. Therefore, this study proposed a multi-criteria flood damage assessment framework for the railroad infrastructure containing the development and analysis of object groups, damage processes and damage classes. The key remarks identified in this study are as follows:

• Extensive damage documentation should be conducted during a damaging event (e.g., using EO), immediately after the event (e.g., photo documentation), and after initial clean-up work is conducted (e.g., photo documentation and analysis of affected structures) to record both direct and indirect damage. Additionally, an App could be developed for uniform in-field damage classification based on the proposed multi-criteria framework.
• A multi-criteria flash flood damage assessment framework for the rail infrastructure was newly developed.
• Damages to the Eifel route in Western Germany after the heavy rainfall and flood event in July 2021 were classified and analyzed by applying the framework.
• The class-based analysis shows that the track superstructure, the track substructure, surrounding areas and technical infrastructure were particularly affected by erosion, transport and deposit of debris. The washout of ballast, gravel and soil, and the deposit of wood, branches and trees or and stones, were present. To reduce the flood damage potential of the railroad infrastructure, technical measures could address these railroad elements and damage processes.
• Flash flood and flood hazard maps can be used to identify potentially flooded areas and therefore potentially damaged railroad sections. Therefore, a (flash) flood damage hazard map for the rail infrastructure could be created using this data.
• The flood damage exposure analysis of the railroad using QGIS shows that railroad track sections within a 50 m distance to a river are likely to be damaged by a flood. The location of the rail infrastructure at the river cliff of the Urft river increased the damage potential of a washed-out track bed.

These results provide a valuable basis for future rail road flood damage research. However, some limitations apply to this study. Damage classifications were only based on observable characteristics and limitations of the documented damages possibly lead to the underrepresentation of affected elements in this study. Furthermore, only a simplified comparison of damaged track sections and other parameters was conducted. For a more detailed understanding of the underlying damage processes, it is necessary to correlate the location of damaged track sections with further data, like topographical features and flow velocities. Additionally, financial metrics were not considered. Nevertheless, if financial data becomes available, the financial and operational aspects of reconstruction could be linked to the damages to identify easy to implement and cost-effective measures. Moreover, the developed damage classification system could be used as a guideline to prioritize the reconstruction of the rail infrastructure, e.g., with a larger number of damages or higher damage classes correlating to a higher priority.

For future research, this framework should be validated with events of different magnitudes and applied to railroad routes in other topographical settings. Technical and organizational measures to reduce the flood damage potential to the rail infrastructure should be determined based on these results. Furthermore, adaptations to the proposed multi-criteria framework could be conducted to describe and classify damages caused by other natural hazards such as snowfall, heat or storms.