Technische Universität München

TUM-School of Engineering and Design

Lehrstuhl für Computergestützte Modellierung und Simulation

Assessment and Mitigation of Flooding Scenarios for

the LMU Hospital, Munich through BIM-GIS
Integration

Scientific work to obtain the degree of Master of Science (M.Sc.)

in Civil Engineering

Author: Indraditya Bhattacharyya

Matrikelnummer: 03710605

1. Supervisor: Prof. Dr.-Ing. André Borrmann

2. Supervisor: Sebastian Esser

Registration Date: 01. September 2023

Submission Date: 01 March 2024

Preface II

Preface

As I embark on the journey of presenting this thesis, I am acutely aware of the vital
importance of effective flood risk management in our increasingly unpredictable cli-
mate. The decision to focus on the LMU hospital campus in Munich's Zimmensblock
area for this study was driven by the need to safeguard critical infrastructure and en-
sure public safety in the face of natural disasters.

In recent years, the spectre of climate change has loomed large, making extreme
weather events like heavy rainfall recently, in some other states of Germany and sub-
sequent flooding a more frequent reality.

This thesis is the culmination of extensive research, data analysis, and simulation
work, the goal of which is to not only understand the flood risks but also to provide a
blueprint for the development of robust flood management strategies. It reflects the
interdisciplinary nature of contemporary research, lying at the intersection of environ-
mental science, urban planning, and technological innovation.

The journey of crafting this thesis has been both challenging and enlightening. It has
been a privilege to work with experts from various fields, to learn from historical data,
and to employ advanced machine learning techniques. I am deeply grateful for the
guidance of my academic mentors, the support of my peers, and the collaboration with
the authorities of Munich and the LMU campus, all of whom have contributed signifi-
cantly to this work.

I would also like to extend my special thanks to Miss Pavla Rýzlerová, Mr. August Pries
from LBD, and Mr. Jonas Hürten from the Water Department. Your contributions, both
in expertise and support, have been invaluable. Your willingness to share your
knowledge and collaborate has greatly enhanced the quality and impact of this re-
search. With this preface, I invite you to delve into the pages that follow, which detail
the methodology, findings, and implications of this critical work in flood risk manage-
ment.
Abstract III

Abstract

Flood simulation and documentation are fundamentals in the realm of flood manage-
ment, offering indispensable tools for predicting, understanding, and effectively re-
sponding to potential flood events. By harnessing the capabilities of advanced tech-
nologies like Geographic Information Systems (GIS) and Building Information Model-
ing (BIM), flood simulation could allow us to envision and understand various flood
scenarios due to heavy precipitation, which can result in informed decision-making and
proactive measures.

Integrating precise terrain data, hydrological information, and building models, paved/
non-paved surfaces, soil characteristics, storm water drainage, these simulations may
offer comprehensive insights into flood extents, vulnerable areas, and potential infra-
structure impacts. Concurrently, flood documentation also contributes a major role in
recording historical flood events, their consequences on buildings and infrastructure
and the efficacy of response strategies. This documentation lays the foundation for
post-flood assessments, and continual improvements in flood management practices,
bolstering community resilience and mitigating the devastating effects of flooding.

This thesis aims to employ BIM and GIS modelling to calculate and simulate these
flood scenarios accurately to equip Munich's authorities with valuable insights to de-
velop robust flood management strategies and safeguard vital assets and public build-
ing at the LMU hospital campus.
Summary IV

Summary

This thesis presents a comprehensive approach to flood risk management for the LMU
hospital campus located in the Zimmensblock area of Munich. Employing advanced
Geographic Information Systems (GIS) and Building Information Modeling (BIM), the
study aims to simulate various flood scenarios precipitated by heavy rainfall and to
understand their potential impacts. Such simulations are crucial for equipping Munich's
authorities with the insights needed to develop effective flood management strategies
and to protect vital assets and public buildings.
Contents V

Contents

1 Introduction and Motivation 1

1.1 Aim of the Thesis ........................................................................................... 1
1.2 Area of Interest .............................................................................................. 2
1.3 Structure of Work .......................................................................................... 3

2 Literature Review 4
2.1 Towards an Integration of GIS and BIM data: What are the Geometric
and the Topological Issues? (Ohori et al., 2017) ........................................... 4
2.1.1 Methodology .................................................................................................. 4
2.1.2 Results .......................................................................................................... 6
2.2 An Integrated BIM–GIS Method for Planning of Water Distribution
System (Zhao,2019) ...................................................................................... 7
2.2.1 Methodology .................................................................................................. 7
2.2.2 Results ........................................................................................................ 10
2.3 A Preliminary Contribution towards a Risk-Based Model for Flood
Management Planning Using BIM: A Case Study of Lisbon (Del
Duca,2022) .................................................................................................. 10

3 Methodology 12
3.1 Data Collection ............................................................................................ 12
3.2 Data Preparation ......................................................................................... 19
3.3 Integration ................................................................................................... 27
3.3.1 Model Integration ......................................................................................... 27
3.3.2 Hydrological Data Integration ...................................................................... 29
3.4 Visualization ................................................................................................ 47
3.4.1 Weather API integration .............................................................................. 47
3.4.2 Web application ........................................................................................... 49
3.4.3 Setting Up Game Engine with Unreal Engines ............................................ 58
3.4.4 Sewage Network ......................................................................................... 60

4 Result 67
4.1 Simulation Based on Historical Data ........................................................... 67
Contents VI

4.2 Simulation results from Statistical Data ....................................................... 68

4.2.1 Result from 10-year return period ............................................................... 68
4.2.2 Result from 100-year of return period ......................................................... 69
4.3 Simulation results from ML prediction model .............................................. 70
4.4 Visualization from Unreal Engines .............................................................. 71

5 Conclusion 72

6 Discussion 74

7 Methods For Flood Mitigation 76

References 77

8 List of Figures 80

9 List of Tables 82

10 Declaration 84
List of Abbreviations I

List of Abbreviations

BIM Building Information Modelling

GIS Geographic Information Systems

LMU Ludwig-Maximilian-Universität München

IFC Industry Fondation Class

OGC Open Geospatial Consortium

LoD Level of Detail

UTM Universal Transverse Mercator

 1

1 Introduction and Motivation

The integration of Flood Simulation using Building Information Modeling (BIM) and
Geographic Information Systems (GIS) data heralds a novel advancement in flood
mitigation strategies. BIM can provide us with intricate geometric and semantic
information which offers a opportunity to enhance the accuracy of flood simulation and
prediction. By harnessing the power of BIM and GIS in flood management, we could
create a more refined and effective approach to deal with flooding events.The amount
of detail provided in BIM, is well-known for its capacity to provide exact and thorough
data about buildings and landscapes, greatly increases the accuracy of flood
simulations. This includes not only the geometric properties of buildings and
infrastructure but also the semantic information of the buildings which describes their
functions and materials. Additionally, GIS data complements BIM by providing
valuable geographic context. GIS integrates information about topography, hydrology,
land use, and more, allowing to create a holistic view of the environment in which
flooding occurs. The incorporation of this contextual information improves flood
prediction models by taking into account the larger environment and its relationship to
constructed structures.

In this thesis we delve into the intricate task of simulating various flood scenarios within
the historical Ziemmensblock area at LMU Hospital, a significant location in Munich.
Our approach involves the integration of multifaceted data sources, comprising terrain
information, hydrological data, detailed building models, surface material distinctions
(paved and non-paved), and soil attribute assessments to determine infiltration rates.
Additionally, we consider the intricate network of stormwater drainage systems, all of
which serve as crucial inputs to our comprehensive flood simulation model.

1.1 Aim of the Thesis

The primary aim of this thesis is to develop and present a methodological approach
that leverages Building Information Modelling (BIM) and Geographic Information Sys-
tems (GIS) for simulating various flood scenarios in the LMU hospital area. This ap-
proach is intended to facilitate a comprehensive understanding of the potential impacts
of flooding, enabling the identification of vulnerable locations and the assessment of
 2

effects on critical public infrastructure. By integrating sophisticated modelling tech-

niques with a diverse array of datasets, the thesis endeavours to provide a detailed
overview of the potential extent of flooding, contribute insights for flood management
strategies, and enhance preparedness measures. A significant focus is placed on en-
suring the protection of this historically significant and crucial site within Munich,
thereby contributing to the community's resilience against the catastrophic conse-
quences of flooding. Furthermore, the documentation outlines a basis for post-flood
evaluation and the continuous improvement of flood control measures. Ultimately, this
methodological framework aims to supply Munich's authorities with actionable infor-
mation to devise effective flood management plans and safeguard essential assets,
thus underpinning a resilient and well-prepared community in the face of flood risks.

1.2 Area of Interest

The study focuses on flooding scenarios for the LMU-hospital located in the historical
part of Munich, specifically in the urban district Ludwigsvorstadt/Isarvorstadt, within
the area of so called “Ziemssenblock” bounded by Nussbaumstraße, Ziemssenstraße,
Lindwurmstraße, and Goethestraße.

Figure 1 Innenstadt clinic LMU

 3

1.3 Structure of Work

The thesis is organized into the following structured components:

Figure 2 Overview of the developed method and workflow

• Data Collection and integration: We initiate our research by collecting data

from different public authorities in Munich. This data includes a digital terrain
model of the city, IFC models of various clinics at the LMU hospital, soil char-
acteristics, land use and the crucial step of georeferencing IFC models within
the broader geographical context and integrating everything into a single model.
• Hydrological Modelling: The next phase involves the development of a robust
hydrological model. This encompasses the calculation of rainfall-runoff
dynamics, an assessment of surface water flow patterns from historical precip-
itation data , and the estimation of water flow behaviours through statistical and
machine learning approaches. This phase is pivotal in understanding the
hydrological aspects of our study area.
• Visualization and Simulation: The final segment of our work revolves around
visualization and simulation. Here, we leverage the gathered data and the
hydrological model to simulate various flooding scenarios within our area of
interest. This step offers a practical and illustrative insight into potential flood
outcomes.
• This structured approach allows us to methodically progress from data
acquisition through hydrological analysis to the ultimate visualization and
simulation of flood scenarios in our designated study area.
 4

2 Literature Review

For the purpose of the thesis following literature review were made and detailed be-
low:

2.1 Towards an Integration of GIS and BIM data: What are the Geometric and
the Topological Issues? (Ohori et al., 2017)

The practical difficulties of combining Geographic Information Systems (GIS) and

Building Information Modelling (BIM) at the data level—a synergy thought to be ad-
vantageous for both domains—are examined in this paper. While GIS offers a more
comprehensive contextual perspective, BIM includes detailed information on buildings
and infrastructure. Prior research has mostly focused on the theoretical elements of
schema-level BIM-GIS integration. This paper’s focus, however, is on the practical
geometric and topological difficulties encountered in software development for this in-
tegration. The paper’s initial findings highlight several unresolved issues: managing
geometric and topological complexities in BIM models, addressing inaccuracies in
georeferencing, and determining the most effective methods for data conversion be-
tween IFC (Industry Foundation Classes) and CityGML (City Geography Markup Lan-
guage). These challenges underscore the need for further research and development
in this area.

2.1.1 Methodology
The literature review initiates by detailing the method of parsing and processing each
object within the IFC framework, employing a customized variant of IfcOpenShell. This
is integrated with a substitution of the original kernel, which is based on
openCASCADE, with a novel one that leverages the Computational Geometry
Algorithms Library (CGAL). Through this approach, the paper harmonizes the
geometrical data from GIS and BIM within a unified data environment. It also ensures
alignment between the instances of CityGML and IFC, maintaining a consistent link
between these datasets.The representations for different geometric forms include
points, polygonal curves, and polygons with holes. Implicit and parametric curves,
surfaces, and volumes from the IFC files are transformed into explicit boundary
representations, which are then discretized into polygonal curves and meshes.
 5

Additionally, all placements and transformations in IFC are translated into 3D affine
transformations, defined by matrices, and applied recursively to each object. This
results in a polyhedral representation of every volumetric object, stored as a CGAL
Polyhedron 3.

Furthermore, certain IFC entity types such as CSG(Constructive solid geometry) solid,
half-space, and Boolean result representations require extra processing steps. These
involve converting a polyhedron into a CGAL Nef polyhedron 3, followed by performing
Boolean set operations like union, intersection, or difference. Through these steps, the
methodology effectively brings together the detailed architectural data of BIM and the
expansive environmental context of GIS, allowing for intricate spatial analysis and
geometric operations within a cohesive framework.

In their study, the authors discovered that invalid objects are a common issue in the
IFC models they analyzed, aligning with previous experiences in BIM and GIS data.
The most frequent errors include self-intersections and intersections between objects,
surfaces that are uneven but meant to be planar, and disconnected objects modeled
as one. A particularly notable error involved objects that appeared valid and formed
topological 2-manifolds but actually contained self-intersections, a violation of the IFC
standard.

The strategies to counter the challenges are provided below :

Triangulation for Planarity: They triangulated non-triangular faces of objects to ensure

planarity, aiding the conversion from CGAL Polyhedron 3 to Nef polyhedron 3. This
approach guarantees that a plane passes through every face.

Validation Tests: A series of validation tests were performed on every object and its
openings. These tests, conducted at various stages of the object's construction,
checked for enclosed spaces, issues with CGAL's triangulation algorithm, self-
intersections, and failures during conversion to Nef polyhedron 3.

Discretization of Implicit Geometries: The implicit and curved geometries in IFC were
discretized into explicit b-rep linear geometries. Methods varied, like discretizing
ellipses into polygonal curves using equal-angle intervals and spheres into icosahedral
approximations.
 6

Challenges with CGAL Features: The authors noted that CGAL's features were
insufficient for modeling all complex features of IFC. They encountered compatibility
issues with CGAL Nef polyhedra and the extended kernel needed for half-space
representations. Challenges also arose in constructing meshes that don’t enclose a
volume or creating a Polyhedron 3 from a mesh in a Nef polyhedron.

Another challenge was regarding bad georeferenecing of IFC models, the IFC
standard includes entities and attributes specifically for georeferencing models,
allowing for the precise location of a model on a geographic coordinate system by
combining latitude, longitude, elevation attributes of the IfcSite, and the model's true
North orientation. However, the paper found that these georeferencing attributes are
often incorrectly filled by practitioners. This discrepancy is exemplified by a case where
the provided coordinates for a model in The Hague were significantly inaccurate,
placing the model at the center of a roundabout instead of its actual location.

Such inaccuracies can be substantial, with examples ranging from a few kilometers to
incorrect placements in entirely different cities or continents. These errors are
attributed to the lack of emphasis on georeferencing during the design phase by BIM
practitioners and the lack of strict rules or specifications mandating accurate
georeferencing information in IFC file generation.

2.1.2 Results
Geometrical and Topological Errors: BIM models often contain numerous
geometric and topological errors, which, while not problematic in a BIM context, pose
significant challenges for GIS applications that require precise spatial analyses and
operations.

Limited Scope of Transformation: Because of the complexity and large number of

IFC entities, the project has only created conversions for a portion of the IFC standard
thus far. It appears more practical to concentrate on the most often used and readily
convertible entities to polyhedral representations rather than trying to cover every
entity.
 7

Need for Standardisation: There is a pressing need for standardisation in the

transformation process from IFC to CityGML. Different interpretations and approaches
to this transformation can lead to varied outcomes, underscoring the need for a
uniform and standardised method, particularly for creating sustainable information
chains.

Detail Level in Integration: Given the higher detail in BIM data compared to GIS, a
fully detailed integration is unlikely. Instead, a generalised version of BIM models with
relevant GIS attributes is more practical.

The project's future direction of the paper involves formulating recommendations for
BIM practitioners and standardisation bodies to enhance integration possibilities, iden-
tifying key semantics to be preserved in the transformation process, and further devel-
oping the IFC/CityGML interface. These efforts will be grounded in practical use cases
to ensure alignment with real-world needs. This approach signifies a significant step
towards effective integration of BIM and GIS data, addressing both technical chal-
lenges and the need for standardisation and compatibility.

2.2 An Integrated BIM–GIS Method for Planning of Water Distribution System

(Zhao,2019)

The paper proposes an innovative method integrating Building Information Modelling

(BIM) and Geographic Information Systems (GIS) for planning water distribution sys-
tems (WDS). The integration aims to enhance planning efficiency, reduce conflicts dur-
ing construction, and improve sustainable development in urban areas.

2.2.1 Methodology
Integration of BIM and GIS Using Semantic Mapping

Industry Foundation Classes (IFC) and City Geography Markup Language (CityGML):
These are the chosen data schemas for BIM and GIS respectively, due to their neutrality
and wide acceptance. The integration process employs semantic mapping to harmonize
these models, ensuring efficient data transformation and sharing.
 8

Mapping IFC Data into CityGML System: This involves correlating IFC's comprehensive
structure for pipe networks with CityGML's model. The IFC schema details the relation-
ships among pipe networks, segments, and connections, while CityGML represents the
basic structure of various types of networks in a GIS environment. The mapping process
ensures that the intricate details of the pipe network in the BIM environment are accu-
rately represented in the GIS model.

3D Visualization Model:

Surface Model Development: Utilizes a Triangulated Irregular Network (TIN) surface

model in GIS to represent the site topography, providing a vivid 3D view of different
terrains like valleys and ridges. The topographic survey data, including coordinates and
elevations, are first imported into AutoCAD, and then transferred to ArcGIS to form a
geodatabase.

Modelling of Existing Buildings: Involves identifying building locations and footprints from
aerial images, assigning elevation values from the surface model, and developing a 3D
model in SketchUp. This model is then integrated back into GIS.

Modelling of Pipe Network Systems: Focuses on creating a 3D model of existing and

proposed pipe systems. CAD drawings of pipe systems are imported into Revit for 3D
modelling, which are followed by storing the information in IFC schema and incorporating
it into ArcGIS using CityGML/Utility Network.

Figure 3 Different Layers in the model

 9

Topological Validation Process

Identifying and Setting Topological Rules: A crucial step to examine spatial conflicts be-
tween the proposed WDS project and existing structures. The topology tool in GIS, with
a set of validation rules for points, lines, and areas (PLAs), is used to ensure the spatial
consistency of the network and prevent redundancy.

Process of Topological Analysis and Error Identification: Involves setting PLAs and top-
ological rules, determining cluster tolerance, and assigning topological rules that govern
spatial relationships. The topology validation process then identifies any violations or
conflicts, which are marked and reported for corrections.

Topology Editing and Revalidation: Utilizes ArcGIS's topology editing tools to fix identi-
fied errors. The changes are saved in topology, and the topological rules are revalidated
to ensure compliance.

Figure 4 Water distribution in ArcGIS

This comprehensive methodology demonstrates a detailed integration of BIM and GIS

for efficient and conflict-free planning of water distribution systems, showcasing an inno-
vative approach to urban planning and infrastructure development.
 10

2.2.2 Results
The application of this integrated method was tested on a real WDS project in Jinan,
Shandong Province, China enhanced understanding of the WDS project and its interac-
tion with the surroundings. The key benefits also include:

• Improved decision-making, considering on-site physical constraints and avoiding

unrealistic layouts.

• The development of a reliable WDS plan without conflicts, ensuring efficient use
of space and resources.

The paper concluded that the integrated BIM-GIS method is effective for planning water
distribution systems. The approach offers a comprehensive view of the pipeline and its
environment, leading to more informed and sustainable urban development. Future stud-
ies are suggested to extend the model's application to construction, operation, and
maintenance stages, and to explore further usability in the Architectural, Engineering,
and Construction (AEC) industry.

2.3 A Preliminary Contribution towards a Risk-Based Model for Flood Manage-

ment Planning Using BIM: A Case Study of Lisbon (Del Duca,2022)
The importance of multidisciplinary approaches in urban planning and design for reduc-
ing the effects of climate change is highlighted by this assessment of this literature, par-
ticularly for coastal cities. The amount of research in this field is increasing, but studies
that integrate land and sea surveys—a critical component of effective management of
coastline settlement—remain scarce. This disparity emphasizes the necessity of creative
approaches, including building information modelling (BIM), to effectively control coastal
risks.

The assessment delves into the possibilities of BIM as a game-changing instrument for
managing coastal risks. Through the utilization of crowdsourced data, BIM enables thor-
ough geographic and informational modelling of metropolitan regions, providing a dy-
namic and ever-changing framework for urban planning. This capacity is especially im-
portant for handling the changing and unpredictable characteristics of coastal areas.
 11

to evaluate the models and approaches currently in use of coastal city planning, a thor-
ough review is conducted.

The methodology section outlines the process adopted for risk assessment, following
the European Flood Directive. It involves assessing flood risk factors, identifying exposed
areas, defining possible hazards, and considering countermeasures. The study used a
non-probabilistic evaluation for risk analysis, based on readily available digital car-
tographies from public databases. The paper also details the process of classifying flood
vulnerability and the factors that influence it, such as proximity to the coastline, building
construction quality, and the presence of physical barriers.

Risk management is discussed in terms of identifying risks, assigning values of exposure

and vulnerability, and applying mitigation actions. The paper emphasizes the importance
of reducing vulnerability levels as the most effective flood risk mitigation action. Risk
prioritization is described as a process to identify severe risks and address them first,
and it includes evaluations of the importance of built, natural, and intangible heritage
before a flood occurs.

The paper then elaborates on City Information Modelling, focusing on interoperability

and resolution, and the need for accurate geometric models and standardized infor-
mation exchange requirements. It stresses the importance of setting models correctly
from the beginning to avoid rework and ensure interoperability between stakeholders.

The research also includes the development of various thematic maps, such as the Ex-
isting-Terrain Map, Minimum Units Map, Exposure Map, Vulnerability Map, Safe Risk
Map, and others, each serving a specific purpose in flood risk assessment and manage-
ment. These maps integrate spatial and thematic data to provide a comprehensive view
of the city's morphology and the prioritization of buildings and areas for action during a
flood. The Priority Map is specifically mentioned as a tool to coordinate flood damage
restoration operations, prioritizing residential buildings and those with high exposure and
vulnerability values.
 12

3 Methodology

3.1 Data Collection

The initial phase of the method is marked by a thorough and strategic collection of data,
essential for the various stages of the endeavour. This data, pivotal for the foundation
and success of the project.

The data for the thesis were collected from different governmental organisation of Mu-
nich and open-source data repositories from the internet. The major contribution of the
data is from Landesbaudirektion Bayern, who are responsible for maintaining and con-
struction of all the state buildings of Bayern, another major contribution comes from the
water department of Bavaria. The data from this organisation mainly include detailed
building models or IFC models, 2D CAD of the public buildings for detailed views of
building plans, elevations, and sections, CityGML of LoD2 for geospatial data model
represents urban objects at a specific Level of Detail (LoD2), providing crucial insights
into the urban landscape of Munich, and rainfall data from opensource public websites
and water department. This combination of precise architectural details and environmen-
tal data ensures a comprehensive dataset that is instrumental in driving the project for-
ward, allowing for a nuanced understanding of the urban and environmental dynamics
of the region. Each of the data and its types are explained in detail below:

Building Models : The building models are used and provided from the government
organisation, Landesbaudirektion Bayern which were both in IFC and Autocad
formats.The IFC format is a standardized, open file format used primarily in the building
and construction industry for sharing and exchanging information about building and
construction projects. Developed by BuildingSMART, a global organization working to
standardize processes within the construction industry, IFC is designed to facilitate
interoperability in the building industry, and it is commonly used in Building Information
Modeling (BIM) (Borrmann, A., Beetz, J., Koch, C., Liebich, T., Muhic, S, 2018). The
primary purpose of IFC is to enable the sharing and exchange of data between different
software applications used in building and construction. This facilitates collaboration
among various professionals like architects, engineers, and builders.
 13

Also, IFC files are designed to carry a vast amount of information, including geometrical
data, material properties, and spatial relationships. This makes them particularly useful
for detailed analysis and simulation. As a results these models played an important role
in our further simulation works.

Figure 5 IFC model of Psychiatry clinic

In addition to the IFC models, we also received comprehensive CAD (Computer-Aided

Design) models. These CAD models encompassed all the 2D drawings of each public
hospital building within the Zimmensblock area of the LMU Hospital. The significance of
these CAD drawings lies in their detailed representation of the buildings' layouts, offering
a granular view of each floor plan, section, and elevation.

The CAD models were particularly notable for including benchmark and elevation points
derived from local surveying data. This surveying data is crucial because it provides
accurate, real-world coordinates and elevational details, essential for precise mapping
and analysis. These benchmarks and elevation points are key for georeferencing – a
process where points on the map are aligned with geographic coordinates in the real
world. This is especially important because the provided IFC models lacked
georeferencing.However we discuss about the integration and
georeferencing procedure in the section 3.2.
 14

This georeferencing not only enhances the spatial accuracy of the models but also
makes them more useful for a variety of applications, in our case which is flood
simulation.

Overall, the project's dataset was greatly expanded with the addition of these intricate
CAD models that were enhanced with local surveying data. It made it possible to have a
more thorough understanding of the Zimmensblock hospital structures, which improved
modeling and design. This enhanced dataset turns into an indispensable resource for
planning and analysis, guaranteeing that the project is based on precise and contextually
relevant facts.

Figure 6 2D CAD model of the Zimmensblock area

Digital terrain models (DTM):

Digital Terrain Models or DTM are one of the most crucial data for our project as our end
goal was to make flood scenarios and, in the end, we had to deal with understanding
slopes and elevation points of the Zimmensblock area.

A Digital Terrain Model is a crucial component in the field of geospatial analysis and
environmental modelling. Essentially, it represents the bare ground surface, free of any
objects like vegetation and buildings. Unlike a Digital Elevation Model (DEM) which in-
cludes all surface features, a DTM specifically depicts the natural terrain and landform
shapes, providing a topographical representation of the Earth's surface.
 15

This model is generated through the collection of elevation data at various points across
a landscape, often through remote sensing methods like LiDAR (Light Detection and
Ranging) most likely from satellite. The DTM is then constructed by interpolating these
points to create a continuous surface. It's invaluable for a wide range of applications,
including watershed and drainage analysis, land-use planning, geological analysis, and
infrastructure development. DTMs serve as a foundational layer in Geographic Infor-
mation Systems. (Exploring Digital Elevation models—ArcGIS Pro | Documentation,
2023)

Figure 7 Different Layers in the DTM and DEM (Štular, Benjamin & Eichert, Stefan & Lozić, Edisa, 2021)

In the context of flood scenario modelling, the utilization of a Digital Terrain Model (DTM)
is paramount. This tool is critical for precisely mapping and analysing the topography of
an area, including slopes and elevation points. Understanding these topographical nu-
ances is vital for comprehending how water might navigate or pool in the event of a flood.
The intricacies of slopes and elevations are fundamental in dictating the direction and
velocity of water flow, which are crucial components in the realm of flood modelling. By
leveraging a DTM, it becomes possible to simulate a variety of flood scenarios, forecast
 16

potential flood trajectories, and pinpoint regions that are particularly vulnerable to flood-
ing. This advanced approach enhances our ability to prepare for and mitigate the impact
of flood events.

Furthermore, DTMs allow for the integration of hydrological data and the assessment of
how natural terrain channels water during heavy rainfall or similar events.

This level of analysis is crucial for developing effective flood mitigation strategies, urban
planning, and infrastructure development, ensuring that the area is better prepared for
potential flooding events. The use of DTMs, therefore, not only contributes to the creation
of realistic and accurate flood models but also aids in making informed decisions to en-
hance the resilience of the Zimmensblock area against future flood risks. In discussing
the project's objectives, we utilized two data sources to obtain the Digital Terrain Model
(DTM). The primary source is the data provided by the local public authority, which con-
sists of elevation details recorded by surveyors. These details are available in text files
containing coordinates (latitude and longitude) along with corresponding elevation fig-
ures, which can be used to generate a TIFF file encapsulating the elevation information.
The methodology for acquiring the DTM and the preparation of this data will be elabo-
rated upon in subsequent chapters. Additionally, open-source DTM data is accessible
on public websites, presented as 1-meter by 1-meter grid segments derived from satellite
imagery (OpenData, 2023). While this data is readily available and in a usable format, it
is important to note that DTMs based on satellite imagery may exhibit higher levels of
error when compared to those created from on-the-ground surveying methods.
 17

Figure 8 Elevations of the Zimmensblock area in QGIS

CityGML: The incorporation of CityGML models played a vital role in the project's suc-
cess. It was impractical to obtain highly detailed models for every building due to the
potential for excessive data consumption and increased security risks associated with
detailed renderings. Therefore, for our purposes, we selectively used Industry Founda-
tion Classes (IFC) for only essential buildings. For the remaining structures, we utilized
CityGML models, which provide merely the height information, thereby simplifying the
complexity of the models and mitigating any associated security concerns.

In short CityGML (Kolbe TH, Gröger G, 2003) is an open standardized data model and
exchange format that facilitates the storage and sharing of virtual 3D city models. Devel-
oped by the Open Geospatial Consortium (OGC) (Open Geospatial Consortium, 2023),
it's primarily used for urban and spatial planning, architectural and environmental simu-
lations, and infrastructural development. Unlike traditional 2D geographical data formats,
CityGML includes the geometry, topology, semantics, and appearance of city-related
objects, such as buildings, roads, rivers, and vegetation, providing a multi-dimensional
and multi-level representation of urban objects. It supports different Levels of Detail
(LoDs), ranging from simple block models to detailed architectural models, making it
 18

incredibly versatile for various applications. Its open format encourages interoperability,
enabling various stakeholders, including urban planners, architects, and government
bodies, to collaborate effectively using shared models. This facilitates more informed
decision-making and efficient management of urban environments.

These LoDs mentioned above range from basic, block-like shapes to highly detailed ar-
chitectural models, each serving different purposes and complexities in urban modelling.

LoD0: This is the most basic level, representing the city model as a 2.5D terrain model
(a flat surface) with footprints of buildings. It's typically used for regional planning and
large-scale urban studies where building heights are not critical.

LoD1: At this level, buildings are represented as block models with a uniform height. This
provides a rough but quick visualization of the urban landscape, useful for medium-scale
city modelling, such as urban zoning and sunlight studies.

LoD2: Here, buildings are modelled with differentiated roof structures, giving a more re-
alistic representation of the cityscape. LoD2 is often used for shadow analysis, noise
propagation studies, and basic architectural visualization.

LoD3: This level includes detailed architectural models with precise facades. It's suitable
for detailed urban and environmental analysis, like microclimate studies, and is often
used in detailed city planning and heritage conservation.

LoD4: The highest level of detail, LoD4, includes interior building structures like rooms,
stairs, and furniture. It's primarily used for indoor navigation, emergency response plan-
ning, and detailed building analysis.

Figure 9 Different Level of Detail as in CityGML (An Improved LOD Specification for 3D Building Models, 2016)
 19

In summary, the integration of CityGML models alongside selective IFC models for criti-
cal buildings in the project represented a strategic approach that balanced detail, data
management efficiency, and practicality. Specifically, the CityGML models were utilized
at Level of Detail 2 (LoD2), (as also provided by the Landesbaudirektion) which provided
a more realistic representation of the cityscape with differentiated roof structures. This
level of detail was sufficient to capture the essential aspects of the urban environment
without the complexity and security concerns of higher-detail models. This methodology
provided a robust framework for the project, ensuring that the urban model was both
detailed and functional, catering to the specific requirements of the flood analysis and
planning objectives.

Figure 10 CityGML LoD2 of the Zimmensblock area

3.2 Data Preparation

Upon completing our data collection, we proceeded to the data preparation phase. This
stage involved converting the collected data into a suitable format for subsequent anal-
ysis and simulation. The preparation process for each dataset is detailed individually in
the following sections. But first, let's introduce the software applications and program-
ming languages utilized during the data preparation phase:
 20

Autodesk AutoCAD: AutoCAD, a prominent software by Autodesk, is widely used in

architecture, engineering, and construction for its robust computer-aided design (CAD)
capabilities. It allows for the creation of precise 2D and 3D drawings, revolutionizing the
design process by replacing manual drafting with an efficient, automated method.

Known for its user-friendly interface, AutoCAD supports a wide range of design tasks,
from simple geometric shape creation to complex architectural plans. Its compatibility
with various file formats and integration with other Autodesk products make it an indis-
pensable tool in the design and drafting industry.

So, basically this software helped in understanding and cleaning the data we have got
from the public authorities and filter out the relevant items and layers which were relevant
to our simulation and analysis.

Autodesk Revit: Revit, developed by Autodesk, is a powerful Building Information Mod-

elling (BIM) software widely used in the architecture, engineering, and construction in-
dustries. It allows professionals to design buildings and structures in 3D, annotate the
model with 2D drafting elements, and access building information from the building mod-
el's database. Revit is highly valued for its collaborative features, enabling multiple team
members to work on a single project simultaneously and share updates in real time. Its
advanced tools for modelling, coordination, and simulation facilitate efficient planning,
design, and management of building projects. Revit's focus on BIM helps in creating
intelligent models that represent the functional characteristics of a building, offering a
holistic approach to building design and management.

In our approach Revit played an important role in cleaning the IFCs and as importing to
IFC and editing the same gets relatively easier in Revit so it was as the choice to choose
Revit over any other BIM software.

QGIS: QGIS, also known as Quantum GIS, is a sophisticated open-source Geographic

Information System (GIS) platform renowned for its comprehensive spatial analysis and
cartographic functionalities. The software is lauded for its intuitive interface, versatile
nature, and expansive suite of tools that empower users to efficiently analyse and depict
spatial information ("Welcome to the QGIS Project!", 2023).
 21

A pivotal feature of QGIS pertinent to our project is its georeferencing capability, which
is essential for calibrating spatial data with specific geographic coordinates.

In the context of integrating QGIS with BIM applications such as Revit, georeferencing
is of paramount importance. Our project primarily utilized data in IFC format; however,
these models frequently require precise positioning within a geographic framework.
QGIS facilitates this by enabling the georeferencing of Revit models, allowing for the
meticulous superimposition of intricate building designs onto specific geographic locales.
This meticulous alignment ensures that structures modelled in Revit are accurately
mapped out in relation to their actual spatial coordinates, bearing, and proportions.

Autodesk Civil 3D: A prominent civil engineering design software which is tailored for
infrastructure design and documentation. It is widely used for projects involving land de-
velopment, transportation, and environmental planning.

One of the key features of using Civil 3D in our thesis is its advanced georeferencing
capabilities. Georeferencing in Civil 3D involves aligning the design elements to real-
world coordinates, which is essential for accurate planning and AutoCAD models that
were provided from the government authorities. This feature allows users to import geo-
spatial data, such as topographical maps and satellite imagery (what we used here was
OSM) and use it as a reference. The software can handle various coordinate systems
and geospatial formats, ensuring compatibility and precision.

Lastly, 3D city database developed by the Geoinformatics chair of the Technical Univer-
sity of Munich is also used for visualising and preparing the CityGML data.

Step 1: Data preparation starts with filtering and keeping only data which are relevant to
our application. The AutoCAD is used in this process. In the process of cleaning, we
extract the elevation information from the AutoCAD file additionally we filter the model
based in layers and relevant buildings that we want to keep. Few of the items like vege-
tation and green infrastructures were removed from the drawing in order to reduce the
complexity as it was not relevant to our analysis. Lastly dimensions were noted, and
relevant building were made into a separate layer to use those later for better georefer-
encing.
 22

Step 2: This step involves in exporting the cleaned filtered AutoCAD model into Civil3D
for georeferencing. Georeferencing in Civil 3D is a process that ensures the accurate
alignment of CAD drawings with real-world geographic coordinates, using several key
concepts. It begins with the selection of an appropriate coordinate system, which lays
the foundation for defining locations on Earth's surface, for example to be WGS 84 for
our case. The process then utilizes control points - specific locations in the drawing with
known real-world coordinates - as anchors for alignment which is taken from the surveyor
data with known points and benchmarks. The software performs transformations, includ-
ing translation, rotation, and scaling, based on these control points to align the drawing
accurately. Further refinement is achieved through rubber sheeting, a method that
stretches or compresses parts of the drawing for a better fit, especially useful on irregular
terrains. Underpinning this process are the concepts of projection and datum; the pro-
jection translates the Earth's three-dimensional surface to a two-dimensional map, while
the datum defines the Earth's shape and size reference model. Ensuring that both the
CAD drawing and the base map use the same projection and datum is crucial for accu-
rate alignment. This combination of spatial reference components, including the coordi-
nate system, projection, and datum, allows Civil 3D to accurately place and orient the
CAD drawing in a true three-dimensional spatial context.

Figure 11 Georeferenced CAD model in Civil3D

 23

Conversely, our individual architectural models were procured in IFC format. These
models were imported into Autodesk Revit for refinement and optimization. Employing a
consistent methodology, we isolated only the pertinent architectural elements for
retention and excised the superfluous details. The received IFC models lacked
geospatial referencing, necessitating the implementation of georeferencing techniques.
To ensure precision, we subjected the IFC models to georeferencing through two distinct
methods, which are delineated below.

Method 1: This method includes the use of ArcGIS Pro and specialized plugins or tools
designed to facilitate the integration of BIM data into GIS. It is also possible to open the
IFC file in Revit. This could involve translating the BIM data into a feature class or a 3D
model that can be used within ArcGIS Pro.

If the IFC model does not come with inherent geospatial coordinates, one needs to
manually georeferenced it (BIM And GIS Integration | Models &Amp; Maps Working To-
gether, 2023). Initially an AOI is drawn on the map tile at ArcGIS pro, upon which the
IFC would be placed. This involves aligning the model with a base map or other geospa-
tial data within ArcGIS Pro. One might use reference points in both the IFC model and
the base map to align them accurately.

These reference points could be from the surveyor data or already know points from the
previous AutoCAD file. After the initial alignment, fine-tuning adjustments are sometimes
necessary toensure that the model precisely matches the real-world location. This could
involve moving, rotating, or scaling the model within the GIS environment.

Figure 12 Georeferencing IFC in ArcGISPro

 24

Figure 13 Georeferencing IFC in ArcGIS Pro

Method 2: The Second method includes the use of IFCopenshell (IfcOpenShell - the
Open Source IFC Toolkit and Geometry Engine, 2023). IfcOpenShell is open-source
library that provides tools for working with IFC data, such as viewing, editing, and con-
verting. This can include tasks like extracting information from an IFC file, modifying IFC
models, or converting IFC files into other formats. A major challenge in automating IFC
data processing is the lack of explicit georeferencing information in IFC exchanges.
There are two typical scenarios:

Local Reference System: Commonly in construction projects, a shared local reference

system is used, defined by a local origin and axis orientation. The local origin is usually
a prominent point near the site, and the axis orientation is selected for practical reasons,
like aligning with the main construction axis. Although the geographical location of this
local system is known among the project partners, it's often not included in the IFC file.

Geospatial Reference System: In some projects, IFC files are exchanged using a stand-
ard projected reference system recognized by land surveyors or in geospatial data, such
as EPSG:2169. This geospatial information, while known to project participants, is typi-
cally not explicitly mentioned in the IFC file, which hinders automated processing capa-
bilities.
 25

In each of these scenarios, the issue of missing georeferencing information can be ad-
dressed by utilizing the IfcProjectedCRS and IfcMapConversion entities from the IFC4
standard. The IfcProjectedCRS entity is used to define a known, target coordinate refer-
ence system (CRS). On the other hand, the IfcMapConversion entity is employed to out-
line the necessary coordinate transformation for converting coordinates from the pro-
ject's CRS to the designated target CRS. As illustrated in the following diagram from the
IFC standard, these specific IFC entities are directly linked to the IfcProject entity, which
is a subclass of IfcContext. (Stijngoedertier, 2023)

Figure 14 Diagram of georeferenced IFC in IFCopenshell

The next task involves in preparing the digital terrain model, which could also be
achieved by two methods. The first method involves the use of the available elevation
data from surveyors as found in the AutoCAD model. The elevation file provided was a
text file consisting of northing and eastings of projected coordinate system (like UTM -
Universal Transverse Mercator). So, after the coordinate system is found the file is con-
verted into relevant latitude and longitude with its subsequent elevation. Once the spatial
data is created it can be saved as a TIFF file, particularly a GeoTIFF, which is a popular
format for storing georeferenced raster imagery. GeoTIFF files contain metadata tags
(like ModelPixelScaleTag, ModelTiepointTag, and GeoKeyDirectoryTag) that store the
geospatial data, ensuring that the image aligns correctly with geographic maps.
 26

However, another approach is using already available DTM that are publicly available in
government sites, which includes this elevation information. However, it’s not much ac-
curate as the previous approach as in this method it is done by satellite imagery. Alt-
hough in our case we have used both methods, the first method is used for our main
area of interest where we need maximum accuracy and the opensource data is used for
the neighbouring areas. Lastly all these tiff files are stitched together in one final TIFF
file using QGIS.

Figure 15 Converting Elevation into TIFF images.

 27

3.3 Integration

3.3.1 Model Integration

The process of integrating all models begins by combining individual components, such
as the IFC models, CityGML, and digital terrain models, into a single, unified model,
which we refer to as Model X. The primary objective of creating Model X is to facilitate
the incorporation of additional hydrological models and data within the model. To simplify
this integration process, Autodesk InfraWorks was utilized as the software of choice.

Autodesk InfraWorks is a comprehensive planning and design platform that enables en-
gineers and designers to create, evaluate, and communicate infrastructure projects ef-
fectively. It offers a range of tools for 3D model generation, analysis, and visualization,
making it ideal for large-scale infrastructure projects such as roadways, bridges, airports,
and urban development. InfraWorks stands out for its ability to integrate with various
data sources, including GIS, topographical, and satellite data, enabling users to work
with real-world context. This integration allows for more accurate and realistic represen-
tations of proposed projects, facilitating better decision-making. Additionally, the soft-
ware's user-friendly interface and collaboration tools enable professionals from a variety
of disciplines to access it, fostering efficiency and teamwork in the creation of challenging
infrastructure projects. With Autodesk InfraWorks, users can anticipate project chal-
lenges, explore alternatives, and visualize the impact of their designs long before phys-
ical construction begins. By importing IFC models into InfraWorks, professionals can in-
tegrate detailed architectural and structural elements into their infrastructure projects,
ensuring a holistic approach to design and planning. This integration is particularly ben-
eficial in complex projects where the interaction between buildings and surrounding in-
frastructure is critical.

The combination of IFC models and DTM in InfraWorks enabled to visualize how pro-
posed infrastructure will interact with existing buildings and the natural terrain. This ap-
proach enhances decision-making, improves risk management and more accurate sim-
ulation, and ensures that the infrastructure is designed in harmony with its surrounding
environment, ultimately leading to more sustainable and efficient project outcomes.
 28

Figure 16 Model X in Infraworks

In order to align all our models which were already georeferenced from the above steps
we needed to import into a common project which are done by basic four steps:

Coordinate Systems and Projection: InfraWorks supports various coordinate systems

and projections (Import GIS Content into Autodesk InfraWorks, 2023) to accurately rep-
resent the earth's curved surface on a flat map. Users can select any appropriate coor-
dinate system based on the project's geographic location. This might be a global system
like WGS 84, which was our case.

Importing Geospatial Data: All the geospatial data from different sources like GIS tiff files,
IFCs, CAD files are imported. This data typically contains embedded spatial information,
including coordinates and projection details.

Aligning Data with Earth's Surface: When importing data, InfraWorks reads the embed-
ded geospatial metadata to place the data accurately on the earth’s surface. If the data
lacks geospatial metadata, we can also manually georeferenced it by aligning known
points in the dataset with corresponding points on a base map or by entering specific
coordinates, however in our case we already had it set.

Data Integration and Visualization: Once the data is georeferenced, InfraWorks inte-
grates it into a unified 3D model.
 29

This integration allows for the overlay of different data types (like terrain models, road
networks, and building models) in their accurate geographic positions and our model is
ready for simulation.

This model is further exported into object file which is later made ready for our simulation.
For the process of further visualisation blender is used.

3.3.2 Hydrological Data Integration

Hydrological data was collected from the water department of Bavaria, and from their
official website. These data were used for the flood simulation and ground water model-
ling as these data was used for making precipitation hydrograph from the previous rain-
fall data that would be explained in detail.

Table 1 Historical Rainfall data

The above chart explains the rainfall data for each day from 2013 to 2022 in mm which
was used to construct the unit hydrograph and understanding the worst rainfall scenario
for flood simulation.
 30

The trend of annual rainfall over the span of ten years, from 2013 to 2022. Initially, the
rainfall appears to decrease from 2013, reaching its lowest point in 2015. Following this,
there is a slight increase in 2016 before a significant drop occurs in 2019, marking the
lowest annual rainfall within the observed period. However, the subsequent years show
a recovery, with 2020 experiencing a modest rise followed by a sharp increase. The
trend continues upward, peaking in 2020. The graph concludes with a slight decrease in
rainfall in 2021, yet the levels remain high compared to the middle of the decade. This
visual data suggests a variable pattern of rainfall with significant fluctuations, highlighting
years of both drought and abundance.

Table 2 Annual Rainfall for Each Year

When the trend of day wise monthly rainfall patterns was studied over the course of ten
years, from 2013 to 2022. Each line in the below graph represents a different month,
identified by unique colours, plotting the variability of rainfall in millimetres (mm). The
data shows considerable fluctuation in rainfall from month to month and year to year,
with no clear consistent pattern across the years. Some months, such as those repre-
sented by the pink and grey lines, exhibit sharp peaks and troughs, indicating significant
variability in rainfall for those months over the years. In contrast, other months show
more modest variations. Overall, the graph depicts a complex interplay of rainfall
amounts, with some years experiencing higher peaks or lower troughs in certain months,
highlighting the unpredictable nature of monthly rainfall distribution over nearly a decade.
 31

Table 3 Year v/s Rainfall in each month

Software used in simulation:

HEC-RAS:

HEC-RAS, which stands for Hydrologic Engineering Centres River Analysis System, is
a software program developed by the U.S. Army Corps of Engineers for modelling the
hydraulics of water flow through natural rivers and other channels. The primary function
of HEC-RAS is to perform one-dimensional hydraulic calculations for a full network of
natural and constructed channels. The system is capable of handling both steady water
flow (constant flow over time) and unsteady water flow (changing flow over time), and it
can analyse conditions such as water surface profiles, sediment transport, and water
quality (Novrima, 2023).

HEC-RAS is widely used for floodplain management, supporting the design and analysis
of floodplain delineation, flood risk assessment, and for planning flood damage reduction
measures. It can simulate various hydraulic conditions, including the calculation of water
surface profiles under different flow rates, bridge and culvert effects, levee and floodplain
storage, and the impact of structures like weirs and gates on flow patterns.
 32

The software's flexibility and extensive capabilities make it a standard tool in hydraulic
engineering and a critical component in the development of flood risk management so-
lutions worldwide.

Figure 17 HEC-RAS Dashboard

HEC-RAS performs hydraulic calculations by utilizing established physical equations

and numerical methods to simulate the behaviour of water in channels and floodplains.
The calculation involved are explained below:

Energy Equation: HEC-RAS employs the energy equation, which is based on Bernoulli's
principle, for steady flow analysis. It compares the energy at different catchment area
cross-sections while considering variables like channel slope, frictional resistance (Man-
ning's equation), and elevation and velocity variations. Water surface profiles are calcu-
lated using the energy equation and are crucial for figuring out flood levels.

Momentum Equation: In unsteady flow analysis, both the conservation of momentum

and energy are applied to capture the dynamic behaviour of changing flows. The mo-
mentum equation considers the effects of friction, pressure, and gravity on the acceler-
ation and deceleration of water in a channel.

Continuity Equation: The volume of water entering a system must match the volume
leaving the system, minus any losses or gains inside the system. This equation illustrates
the conservation of mass. When calculating unsteady flow in HEC-RAS, the continuity
equation is utilized in conjunction with the momentum equation to ascertain the evolution
of flow depth and velocity.

To perform these calculations, HEC-RAS requires detailed input data such as geometry,
flow data, channel roughness coefficients, and boundary conditions. The software dis-
cretizes the catchment into a series of interconnected cross-sections for 1D analysis or
 33

into a grid for 2D analysis. Numerical methods such as finite difference or finite volume
methods are then used to solve the equations across these sections or grids, iterating
towards a solution that describes the flow conditions for each time step in the simulation.
HEC-RAS is extensively used in rainwater flood simulation to predict how rainfall events
might affect river systems and floodplains which is the motive of our project. By integrat-
ing rainfall data which are explained above, coupled though hydrologic model like HEC-
RAS can simulate the response of precipitation inputs. The process begins with the cre-
ation of a detailed geometric model of the channel and the surrounding floodplain, incor-
porating elements that can influence the flow pattern.

Once the geometric model is established, HEC-RAS uses rainfall data (typically in the
form of hydrographs) to simulate how rainwater will enter the river system and move
downstream. This may involve converting rainfall intensity into runoff using a separate
hydrologic model, which then provides input flows for the hydraulic model. HEC-RAS
takes these flows and computes water surface profiles for the river system, using its
steady or unsteady flow analysis capabilities.

For unsteady flow simulations, which are particularly relevant for rainwater flooding
events, HEC-RAS calculates the changes in water levels and flow velocities over time,
providing a dynamic picture of the flooding process. The model can simulate the timing,
depth, and extent of flooding across the floodplain, considering complex interactions be-
tween the inflow from rainfall, the channel geometry, and any existing flood control struc-
tures. By employing these simulations, HEC-RAS helps engineers and floodplain man-
agers to assess flood risks, design and evaluate flood mitigation measures, and plan
emergency responses. It also allows for the examination of different scenarios, such as
changes in land use or climate change impacts, which can alter the behaviour of rain-
water runoff and flood dynamics. The ability to accurately model rainwater flooding is
crucial for developing strategies to protect communities and reduce the economic and
environmental impact of flood events (Armain, 2021).

The methodology of the flood simulation involved the following steps:

• The terrain is loaded in the RAS mapper, for the terrain we have used the eleva-
tion data derived in the previous chapter. The elevations are provided in the left
 34

side of the provided map. From the terrain map we can understand the north-west
the greenish yellow path of our area of interest is having some lower elevation,
however the southwest region is having higher elevation.

Figure 18 Elevations in HEC-RAS

• The second step is importing the area of interest as a shape file.

• Proceeding to a crucial stage, this involves sketching out the flow regions, incor-
porating them with a boundary condition delineation. Initially, one outlines the pe-
riphery, closely matching our targeted zone, and then proceeds to establish the
mesh. A more refined mesh generally yields improved outcomes, albeit at the
expense of increased computational time. It's noteworthy to acknowledge the cre-
ation of a minor discharge point in the northwest segment of the target area, pred-
icated on the presumption that water will gravitate towards this specific zone due
to its being the lowest elevation across the plot.
 35

Figure 19 Catchment area for our area of interest

• The fourth step involves the setting of boundary conditions as shown in the UI of
HEC-RAS.

Figure 20 Boundary conditions as set in HEC-RAS

• The precipitation chart is also added into the HEC-RAS as shown in the previ-
ous chart and simulation time are also added.

The final step involves in setting values and coefficient which is important for the
accuracy of the simulation.
 36

Additional web software for Quick simulations:

For quick simulations Fast Flow App could be used. This app typically serves as a critical
resource for communities in flood-prone areas, offering real-time data on water levels,
weather forecasts, and flood warnings. The core functionality of Fast Flow centres
around its ability to collect and disseminate information quickly, allowing users to make
informed decisions during potential flood events.

Figure 21 Interface as in Fast Flow application

The DEM we use here is same as what we derived previously, and we make the rainfall
intensity from the unit hydrograph following the same procedure. Here we also consider
infiltration, soil characteristics and channel information to make our simulation more ac-
curate.

Estimating the Flow parameters:

Calculating rainfall intensity:

Rainfall intensity refers to the rate at which rain falls during a storm. It is usually meas-
ured in millimetres per hour (mm/hr).
 37

This data can be derived from the historical data that we got from the public authorities.
Rainfall intensity is an important factor in hydrology and water resource management,
as it affects runoff, infiltration, erosion, and flood potential.

Table 4 Rainfall intensity from historical data

The above chart explains the rainfall intensity of each day based on hourly readings, The
worst (highest) rainfall intensity over the years in the dataset occurred on June 14, 2015,
with an intensity of 3.66 mm/hr. This value represents the maximum hourly rainfall
intensity calculated from the daily data and we use this for our analysis.

In addition to the intensity of the rain, there are several other factors to consider, such
as the Rainfall Multiplier, the duration of the rainfall, and the shape of the rainfall. The
Rainfall Multiplier is a coefficient applied to modify the intensity for computational rea-
sons, and in this instance, it is set to "1". The Rain Shape pertains to how the intensity
of the rain is spread throughout the event, which is closely linked to its duration. This
distribution can either be consistent (flat) or exhibit peaks.

Rainfall intensities based on return periods:

Understanding rainfall intensities based on return periods is essential for a myriad of

reasons spanning infrastructure design, flood risk management, urban planning, agricul-
ture, climate change adaptation, and economic strategies. Such data enables engineers
to design resilient infrastructure capable of withstanding extreme weather, helps in iden-
tifying flood-prone areas for effective risk management, guides urban development in a
manner that minimizes flood risks, and is vital for agricultural planning and water re-
source management (Del Duca et al., 2022). Additionally, in the face of climate change,
it offers a basis for adapting infrastructure and planning strategies to cope with altered
 38

precipitation patterns. For the insurance sector and governmental economic planning, it
provides a critical tool for assessing risk and preparing financially for disaster response
and recovery. In sum, rainfall intensity and return period data are indispensable for safe-
guarding communities, ensuring sustainability, and enhancing resilience against ex-
treme weather events.

Table 5 Statistical data from KOSTRA

From the KOSTRA data we conclude the best is use 10-year return period data of 60
min duration, generally in sewage planning 5 years of return period is considered how-
ever in our case we consider 10-year return period considering climate changes.

Precipitation Hydrograph: A precipitation hydrograph is the graphical representation of

the rate of rainfall over a specific period. It plots the intensity of precipitation (usually in
millimetres per hour against time. The key aspects of a precipitation hydrograph are
given below.
 39

Time Axis: The horizontal axis (x-axis) represents time. The time scale can vary depend-
ing on the purpose of the hydrograph - it can span hours, days, or even months.

Rainfall Intensity Axis: The vertical axis (y-axis) shows the intensity of precipitation, typ-
ically measured in millimetres per hour (mm/hr).

Precipitation hydrographs are generally used to analyse the pattern of rainfall over a
period. They are essential in hydrology for understanding and managing water re-
sources, predicting potential flooding events, designing drainage systems.

Difference from Hydrographs: It's important to distinguish precipitation hydrographs from

discharge hydrographs, which plot the flow rate of water in a stream or river over time.
While a precipitation hydrograph deals with the input of water (rainfall), a discharge hy-
drograph deals with the output or movement of water within a river system.

Table 6 Precipitation Hydrograph

The above hydrograph demonstrates a series of varied intensity rainfall events over the
course of a decade. The data points are distributed across the graph, with several distinct
spikes indicating episodes of significant rainfall. These spikes are interspersed with
periods of low to no rainfall, which are made as flat lines iny-axis.

There does not appear to be a consistent annual pattern of rainfall intensity; rather, the
precipitation levels fluctuate without a clear repetitive seasonal trend.
 40

Some years show multiple high-intensity events, while others have fewer and less
pronounced peaks. This could suggest that the region experiences irregular rainfall or
that it is influenced by varying weather patterns year on year.

The data also do not indicate a progressive increase or decrease in overall rainfall
intensity throughout the years, which might rules out a linear trend. Instead, the
precipitation appears to be episodic, with some years experiencing more extreme
weather events than others. This kind of data is crucial for understanding the hydrological

characteristics of the region, as well as for managing water resources and preparing for
flood events.These results act as an input data for our simulation model.

Land Cover: Land cover data provides information on the types of vegetation present,
which can affect the soil's capacity to absorb water. Different plants and trees have
varying root depths and soil enhancement capabilities, which can significantly influence
the soil's ability to retain water and thus affect flood dynamics.

The land coverage data was taken from sentinel 2 satellite imagery to understand the
green blue infrastructure of the zimmensblock area. The data also does not indicate a
progressive increase or decrease in overall rainfall intensity throughout the years, which
might rule out a linear trend. Instead, the precipitation appears to be episodic, with some
years experiencing more extreme weather events than others. This kind of data is crucial
for understanding the hydrological characteristics of the region, as well as for managing
water resources and preparing for flood events. These results act as an input data for
our simulation model.
 41

Figure 22 Sentinel 2 imagery on Land Cover

The structure of the land cover can alter the paths that water takes across the landscape.
Bushes, for example, can intercept rainwater, while urban canyons created by building
can channelize and accelerate the flow of water. The above land cover imagery shows
the land cover area showing majorly streets in yellow, vegetation in green and water in
orange.

Here in the context of flood simulation the Manning coefficient, often denoted as n, is a
important parameter in hydraulic engineering used to estimate the roughness of a sur-
face which water is flowing over (Takata, 2024). It's a factor in the Manning's equation,
which is employed to calculate the flow velocity in open channels based on the channel's
shape, slope, and roughness. Different land cover types have different surface rough-
ness values. For example, a smooth concrete surface has a lower Manning coefficient
because it offers less resistance to flow compared to a rough forest floor with dense
under brush.

So, based on the land cover the different manning coefficients were taken accordingly:
 42

Figure 23 Manning’s coefficient with respect to land cover

Infiltration: Infiltration is the process by which water on the ground surface enters the
soil. When rainfall occurs, the ability of the soil to infiltrate water can greatly reduce the
volume of water that becomes surface runoff, the infiltration capacity of the soil changes
based on its saturation level, the dry soil can absorb water however when it’s get satu-
rated the ability of infiltration decreases which leads to surface runoffs. Urban areas with
impervious surfaces, such as roads and buildings, have low infiltration rates because of
the permissibility, leading to higher and faster runoff into drainage systems as seen in
our case we don’t have much vegetation in our area on interest.

The geology of Munich says most of it is covered by temperate brown and deep brown
soils. The soil type is a shallow and nutrient-poor harvestation soil with high water per-
meability and low filtration capacity. For our simulation, we have used the SoildGrid data
of Munich to download the infiltration zones (Bauer et al., 2005).

Figure 24 Munich clay content map from SoilGrid

 43

Along with the infiltration two additional data were also provided which included :

• Effective soil moisture : 0.5

• Average soil depth : 5 cm.

All this data were derived from soilGrid opensource data sources which were
further used in flood analysis.

Rainfall Prediction Model :

Machine learning is essential in rainfall prediction models due to its ability to process
complex and varied data sets, such as temperature, humidity, and wind patterns, more
efficiently than traditional methods.For our case we had data from the past for Munich
which we used as a training data.

It significantly improves predictive accuracy, particularly with the use of advanced ma-
chine learning techniques. These models are capable of real-time analysis, which is
crucial for timely and effective weather forecasting.From the predicted rainfall data we
use this predicted data to make assumption on future flood scenarios and generate flood
models out of it. (Liyew & Melese, 2021) We have chosen random forest algorithm for
our prediction which belongs to the family of ensemble learning methods, Random
Forest consists of a large number of individual decision trees that operate as an
ensemble. Each individual tree in the Random Forest spits out a class prediction or a
continuous output (depending on whether it's used for classification or regression), and
the class with the most votes becomes the model's prediction (Meenal et al., 2021).

Figure 25 Simplified Random Forest algorithm (Koehrsen, 2020)

 44

For our case in predicting flood scenarios the key strength of Random Forest is its ro-
bustness against overfitting, especially when dealing with large datasets. This is crucial
in weather forecasting where the model needs to generalize well to unseen data. In ad-
dition, Random Forest algorithm can handle missing values in the data, which is a com-
mon issue in meteorological datasets, however in our case we didn’t have much missing
rainfall data.

Implementation of Random Forest Model:

First step involved was to clean and make the data ready for training, the data used was
from the water department that consists of the rainfall data from 2013 to 2021, the data
is summarised below:

Table 7 Mean Annual Rainfall data.

Input for the Random Forest Algorithm:

Dataset: The input dataset is loaded from an Excel file named "Traingrain.xlsx"
using pandas. This dataset contains historical rainfall data, including year, month,
and average rainfall collected from the water department.

Features (X): The features used to train the Random Forest model are extracted
from the dataset columns 'Year' and 'Month', which are then converted to a numpy
 45

array and cast to integers. These features represent the time information of the
rainfall data.

Shape of X: It represents the number of samples and the number of features used for
training. Based on the code, X.shape would output the shape of this array.

Target (y): The target variable is the 'Avg_Rainfall' column from the dataset, which is
also converted to a numpy array and cast to integers. This column represents the amount
of rainfall, which the model will learn to predict.

Shape of y: It indicates the number of target values corresponding to the input features.
The shape is outputted to confirm the data's structure before training.

• Training and Testing Split: The dataset is split into training and testing sets, with
70% used for training and 30% for testing. This is done using the train_test_split
method from scikit-learn, ensuring that the model can be evaluated on unseen
data.

• Model Parameters: Specific parameters for the Random Forest Regressor are
defined, including max_depth, max_features, min_samples_leaf,
min_samples_split, and n_estimators. These parameters configure how the
Random Forest algorithm will be trained.

Output of the Random Forest Algorithm:

• Model Training: The Random Forest model is trained on the training data
(X_train, y_train) using the specified parameter

• Predictions: The trained model makes predictions on both the training set
(y_train_predict) and the testing set (y_test_predict), allowing for the evaluation
of model performance.

• Accuracy Scores: The model's accuracy on both the training and testing sets is
calculated and printed. These scores indicate how well the model has learned
from the training data and its performance on unseen data, respectively.
 46

Model Creation and Training:

The Random Forest Regressor is chosen for its ability to handle complex datasets with
non-linear relationships, which is common in weather data. The training process involves
the model learning to associate the input features (like temperature, humidity) with the
target variable (rainfall). The specific parameters set for the model define how it learns,
preventing issues like overfitting and underfitting to ensure a good balance between bias
and variance.

A RandomForestRegressor from sklearn.ensemble is instantiated with specific

parameters like max_depth, max_features, min_samples_leaf, min_samples_split, and
n_estimators.

The model is then trained using fit on the training dataset (X_train, y_train).

Prediction and Evaluation: The trained model is used to predict rainfall for both training
(y_train_predict) and testing datasets (y_test_predict).Evaluation metrics (Mean
Absolute Error - MAE, Mean Squared Error - MSE, Root Mean Squared Error - RMSE)
are calculated for both sets to assess the model's performance.

Both the training and testing datasets are used to evaluate the model's rainfall prediction
capability after it has been trained. This assessment is essential to determine the
correctness and generalization of the model. The selected metrics (MAE, MSE,
RMSE)aid in assessing the model's performance overall by offering insights into various
areas of the prediction errors.

Single Prediction: Making a prediction for a specific year and month demonstrates the
practical application of the model. It’s a way to see how the model works in a real-world
scenario, providing a single, specific prediction based on the learned patterns.
 47

Interactive Prediction Setup: Interactive widgets (IntText for year and month, Button
for prediction) are created using ipywidgets.A function predict_data is defined to make
predictions based on the input from these widgets.predict_button.on_click(predict_data)
binds the prediction function to the button click event.This part of the code creates a user
interface for making predictions. It allows users to input new data (year and month) and
receive predictions without delving into the code. This interactivity makes the model more
accessible and user-friendly, particularly for users who may not be familiar with Python
or machine learning

3.4 Visualization

In all the previous steps it is explained the procedure to collect, improve the data and
integrate it for the purpose of analysis and simulation. Now, we understand the repre-
sentation of all these data altogether.

3.4.1 Weather API integration

We have enhanced our model by incorporating data from a weather API, which provides
historical records of the most severe rainfall events for any given month. This integration
significantly refines the accuracy of our flood simulations and management strategies
and also provides a nice user experience.

The utilization of such precise meteorological data is essential for predicting the
probability and intensity of potential flood scenarios across different locales. This level
of detail is indispensable for emergency services and municipal authorities when
orchestrating preparations for extreme weather conditions.

Employing this data enables the strategic deployment of resources, the orchestration of
preventive evacuations, and the establishment of robust flood defenses. Moreover, this
 48

meteorological information is extremely valuable for the strategic planning and

development of infrastructure. It equips engineers and urban planners with the insights
needed to design and fortify structures—such as water drainage systems, dams, and
levees—to endure significant rainfall events, thereby enhancing their resilience against
severe weather.

However, in this paper the flood simulation is solely based on flood prediction on precip-
itation hydrographs from the historical rainfall data ad statistical analysis, however it this
API which could help us to know live data for enhanced user experience and as scope
for future improvement of the simulation model.

For this paper we have used the weather API that we used for fetching weather condition,
temperature, rain probability, rain amount.
Table 8 API service listings of weather API

Additionally, the model also checks the worst rainfall pattern and intensity over the pre-
sent month which might be helpful in future scope of the project.
 49

The picture above represents what we fetch and show in our web application. This is
one of the live data fetched in 12th Jan 2024.

3.4.2 Web application

The web application was made for the reason for visualisation, the objective was to have
a quick visualisation that could load all the semantics including all the IFC models and
other models from the city GML as discussed above in a simple and user-friendly way.
After the successful import of the model, we proceed with the simulation showing the
flood level within our area of interest.

The methods involved in building the web application is explained below:

Mapbox as a rastertile : Mapbox is a popular location data platform for mobile and web
applications. It provides developers with powerful tools to create custom, dynamic maps
and location-based experiences. Utilizing a combination of satellite imagery, vector tiles,
and a robust set of APIs, Mapbox enables the integration of detailed, interactive maps
into applications.Its usage here is for the purpose of a raster tile upon where we build
our model on.

Model Import : Model X, a comprehensive and versatile model, is specifically designed

for advanced urban and environmental simulations. The key aspect of Model X is its
ability to bridge the gap between detailed architectural models and geospatial data,
enabling a more realistic and accurate representation of urban environments inside our
visualisation.
 50

Alongside IFCs, Model X incorporates CityGML data. CityGML is an open standardized

data model and exchange format that defines the geometry, topology, and semantics of
3D city models. Elevation data is another crucial component of Model. All the procedure
of the integration are explained in the previous part of the paper. So, once the IFCs,
CityGML, and elevation data are compiled, they are converted into an FBX file format.
FBX (Filmbox) is a popular format known for its versatility in handling both 2D and 3D
data. This conversion is a crucial step as it makes the data compatible with a wide range
of 3D modelling software, including Blender. Blender is an open-source 3D creation
suite, widely used for its powerful modelling, texturing, and animation tools. The trans-
formation of Model X into an FBX file opens numerous possibilities. It allows for further
enhancements and manipulations in Blender or similar software, where additional de-
tails, textures, and animations can be added to the model.

This enhanced model can then be exported for use in various applications, including
web-based platforms for example gltf formats which is well renowned and used in web
applications due to its light weight and support with WebGL. In web applications, Model
X is used for a variety of purposes, for simulations.

Figure 26 Picture of Infraworks FBX export window

 51

Figure 27 Picture of Model X in blender

Web framework :
Table 9 Web application Architecture

The web application developed for flood simulation represents an intricate and well-
orchestrated integration of multiple web technologies, primarily focusing on HTML, CSS,
 52

and JavaScript,(the basic framework used for the web application) while leveraging the
capabilities of Mapbox GL JS and Three.js for enhanced geographical and 3D
visualizations, which supports georefernced models.

HTML Structure: The HTML layout used in this paper serves as the structural foundation.
It segments the web page into distinct sections, each with a dedicated purpose. The
modelInputContainer allows for the importation of 3D models, crucial for simulating the
flood in a realistic environment. The styleSwitcherContainer offers the functionality to
switch between different map styles, like 'Satellite Streets' and 'Standard', providing
flexibility in visual presentation. The weatherInfoContainer dynamically displays
weatherinformation, crucial for understanding the environmental conditions influencing
the flood simulation. Additional elements like the colorBar enhance the data visualization
aspect providing a gradient bar for an intuitive understanding of data metrics like water
depth or intensity relevant to the simulation.

CSS Styling: The CSS styling ensures that the application is not only functional but also
visually appealing and user-friendly. Elements like the alertContainer are styled to have
alert functionality in case of emergency, which would be explained in detailed in the fea-
ture section in the later parts. The use of absolute positioning and z-index ensures that
critical information overlays the map correctly and is easily accessible. The responsive
design of the map container (#map) ensures that the application is adaptable to different
screen sizes, enhancing user experience.

Figure 28 Model X in the web application

 53

JavaScript Functionality: JavaScript is at the core of our application’s interactivity and

functionality.Which consists the logic behind. It initializes the Mapbox map, a critical
component for rendering geographical data. The integration of the Mapbox GL JS library
enables the depiction of high-resolution maps, which are essential for accurately
simulating flood scenarios. The application also takes advantage of Three.js, a
JavaScript library that renders 3D models.

This is evident in the functionality provided to upload and display 3D models within the
map which include the entire model X consisting IFCs, City GML etc, allowing for a more
immersive simulation experience.

The script further enhances the application's functionality by incorporating external

weather data through API calls. This feature not only fetches current weather conditions
but also calculates and displays the worst rainfall data for the month, adding a layer of
depth to the simulation by providing historical context. The implementation of an alert
system, complete with visual and alarms, is a testament to the application's attention to
detail. This system is designed to simulate real-world emergency alerts, enhancing the
realism of the flood scenario.

Interaction Between Frontend and Backend:

The frontend sends requests to the backend (e.g., for weather data).

The backend processes these requests, interacts with the database or external APIs as
needed, and sends back the required data or results.

The frontend then uses this data to update the map, display weather information, render
3D models, or trigger alerts.

Data Flow:

User Interaction: Users interact with the frontend elements, triggering events.

Request Handling: JavaScript captures these events and sends requests to the
backend.

Data Processing: The backend processes these requests, accessing the database or
external APIs.
 54

Response: The backend sends back the processed data or results to the frontend.

Update UI: The frontend updates the user interface based on the received data, such as
updating the map, displaying weather info, or showing 3D models.

Alerts and Notifications: The alert system in the frontend gets activated based on specific
conditions or backend responses.

Integration of flood data :

The integration of flood data into our web application involves a meticulous process
utilizing Python's robust libraries—Numpy for numerical computations, Matplotlib for data
visualization, and Rasterio for handling geospatial raster data. Initially, we employ
Rasterio to open a GeoTIFF file, 'flood.tiff', which contains the flood simulation data from
our flood simulation software where we extract the flood data as a TIFF file. This file
format is ideal for storing raster graphics and is commonly used in geographic
information systems for its capability to embed georeferencing information.Upon
accessing the file, we extract the elevation data using Rasterio. Elevation is a critical
component in flood modeling as it influences how water flows and accumulates in
different areas. We then create a meshgrid using Numpy, which provides X and Y
coordinates corresponding to the longitude and latitude of the dataset. This meshgrid is
essential for plotting the elevation data in a two-dimensional space.

Figure 29 Rainfall data with elevation

 55

With the help of Matplotlib, a powerful plotting library, we visualize this data. We generate
a heatmap using the `pcolormesh` function, effectively displaying the elevation data. The
color map 'terrain' is chosen to represent different elevation levels visually, making the
data more accessible and understandable.

A color bar is added as a legend to relate colors to elevation values, as shown in the
above image.In a refined approach for a specific flood scenario, we further process the
elevation data by setting values of 0.05.

This step helps in emphasizing areas with significant flooding, enhancing the
visualization's accuracy and relevance to the flood simulation.

Finally, these visualizations are not just static images; they are translated into raster tiles
and superimposed onto the Mapbox map within the web application. This integration
provides users with an interactive and detailed view of potential flood impacts in the
context of real-world geography. Alongside this, Model X, which includes detailed urban
and architectural data, is also integrated, offering a comprehensive and multidimensional
perspective of flood scenarios in urban settings.

This method of flood data integration highlights the application's capability to combine
scientific data processing with sophisticated web mapping technologies, delivering a
powerful tool for visualizing, and understanding flood impacts.

Web features : All the features available in the model are explained below :

Interactive Map and ID Model Loading: The ‘#modelInputContainer’ allows users to

upload 3D models (.gltf or .glb files) for visualization on the map. This interactivity adds
a layer of customization, enabling users to see their own models in the context of the
flood simulation.This is what we call the model X. This actually opens up news scope of
the project in future when we try to integrate any other model in our map.

The modelInput.addEventListener function listens for changes in the model input field
and triggers the loading of the 3D model onto the map.

Map Style Switcher:The ‘#styleSwitcherContainer’ contains a dropdown (#styleSwitcher)

that lets users switch between different map styles, such as 'Satellite Streets' and
 56

'Standard'. This feature enhances user interaction by allowing them to choose their
preferred map visualization.

Dynamic Weather Information Display:The ‘#weatherInfoContainer’ is used to display

real-time weather information. This feature is particularly valuable in the context of a
flood simulation application, as it provides users with relevant environmental data.

Color Bar for Data Visualization:A color bar (‘#colorBar’) is included to provide a visual
guide to interpret data gradients, such as elevation levels or water depths for our flood.
This feature aids in understanding the simulation data more intuitively.

Figure 30 Flood elevation bar as in web application

Interactive Alert System:The ‘#alertContainer’ is a crucial UI element designed to display

emergency alerts. It becomes visible when certain conditions are met in the simulation,
such as reaching critical flood levels.It shows the user where exactly in the models are
at risk.

The showAlert function, along with an audio alert (‘#alertSound’), is set to trigger after a
specific timeout, drawing users' attention to urgent situations.
 57

Figure 31 Picture of alerts in vulnerable areas

Interactive Map Features:Clickable elements on the map are enhanced with popups. For
example, clicking on the 'room-extrusion' layer triggers a popup that displays detailed
information about the clicked area, like building name(which includes the type of import
either its an IFC or its a LoD2 building of cityGML), flood level, and status.

The application changes the cursor style when hovering over interactive map elements
(map.on('mouseenter', 'room-extrusion', function() {...})), improving user experience by
indicating clickable areas.

Figure 32 Clickable features inside web application

Custom Layer for 3D Models: The custom Layer configuration and its integration with
the Three.js library enables the realistic rendering of 3D models on the map. This layer
is dynamically rendered and responds to map movements and style changes.
 58

Responsive UI Design:

The application's UI elements are styled with CSS for a clean, uncluttered look and are
positioned fixedly over the map, ensuring they are always accessible regardless of
scrolling or map navigation.

3.4.3 Setting Up Game Engine with Unreal Engines

Unreal Engine 5 emerges as a groundbreaking tool in the realm of visualization,

particularly for simulations that demand high-fidelity and immersive experiences.
Renowned for its advanced graphics capabilities and real-time rendering power,
UnrealEngine 5 enables the creation of incredibly realistic and detailed environments
(Unreal Engine Is an Advanced Training and Simulation Software Platform, 2023).

This makes it an ideal platform for conducting various types of simulations, ranging from
architectural visualizations to complex environmental scenarios like flood simulations. Its
ability to handle intricate lighting, physics, and material properties allows for a level of
realism that is pivotal in accurately modeling and understanding real-world phenomena.
Furthermore, Unreal Engine 5's robust suite of tools, including its revolutionary Nanite
virtualized geometry and Lumen global illumination technologies, empowers developers
to craft rich, interactive 3D scenes with unprecedented ease and precision. In the context
of simulations, this means being able to recreate scenarios with a high degree of
accuracy and interactivity, providing valuable insights into how different factors interact
in dynamic environments.We use this game engine for the flood simulation which
consists of better and high end graphics which we would miss in our web application.

Figure 33 Zimmensbblock on unreal engines

 59

The steps involved in the simulation are explained below :

• Acquiring Google Earth Imagery: Google Earth provides high-resolution satellite

imagery which is beneficial for creating a realistic base layer for our simulation.
• We choose and select our area of interest and fix our center coordinates and
corresponding zoom levels.We also had to acquire the google cloud api in order
to use google earth inside unreal engines.
• Utilize Cesium for 3D Visualization: Importing the Google Earth imagery into
Cesium, a platform known for its capability to render accurate 3D global imagery.
Cesium helps in creating a detailed and interactive 3D scene of your chosen
area.Caesium can be found as a plug in to the unreal engine.We need to replace
the source of the caesium tile map into google tile map in order to use google
earth feature inside.We adjust the LoD according to our graphics and computation
power of the pc.
• Prepare the Flood Data: We ensure that the TIFF file, which holds the flood level
data, is in the correct format and contains all the necessary geospatial
information. This file should represent various flood levels across the area of
interest.
• Conversion for Compatibility: If necessary, convert the 'flood.tiff' file into a format
that is compatible with Cesium. Cesium generally supports a range of geospatial
data formats, but ensuring compatibility is crucial for seamless integration.
• Importing into Cesium: Import the 'flood.tiff' file into your Cesium project. This can
typically be done through Cesium's data import tools or APIs, which allow for the
integration of custom geospatial data.
• Overlaying Flood Data: Overlay the flood level data from the 'flood.tiff' file onto
the Google Earth imagery within Cesium. This step is critical as it combines the
realistic base imagery with the specific flood simulation data.
 60

3.4.4 Sewage Network

Stormwater drains are part of the sewage system, which is intended to divert water away
from populated areas. Understanding these systems' capacities and flow pathways is
crucial to flood modeling because it helps forecast how water will travel across a region
during periods of high rainfall or flooding. A restricted capacity of the sewage system
may cause water to back up and exacerbate flooding. Flood models need to account for
these artificial drainage paths to accurately simulate urban hydrology under flood
conditions.During extreme weather events, sewage systems can become overwhelmed
and overflow. This can lead to additional flooding, particularly in low-lying areas.
Accurately modeling the capacity and potential overflow points of sewage systems helps
in predicting areas at risk of inundation (Yang et al., 2020).

In my thesis, we employed the Storm Water Management Model (SWMM), an advanced

computational tool developed by the Environmental Protection Agency, to analyze and
simulate urban drainage systems and surface runoff in zimmensblock area. The primary
aim was to assess the effectiveness of existing stormwater infrastructure under various
rainfall scenarios and to propose enhancements for flood mitigation. Utilizing SWMM's
dynamic hydrology-hydraulic modeling capabilities, we meticulously configured the
model to replicate the local urban catchment, incorporating detailed data on land use,
soil characteristics, and existing drainage networks. We have used the KOSTRA
statistical data for the rainfall intensity. The outcomes of this research were significant;
the model not only demonstrated the current limitations and vulnerabilities of the existing
stormwater system under extreme weather conditions explained below (Son et al.,
2016). We begin with setting up the piping network in the swmm using a raster tile of the
Zimmensblock area collected from stBA Munich, who provided us CAD models
consisting every information on sewage network on that area.
 61

Figure 34 Pipe Network simplified of LBD from StBA München

The next step include getting the subcatchment areas shown in the above picture as
hatched line and also marked by S1 to S16, now we plot the junction node accordingly
the elevation of the nodes were provided by the StBA as mentioned above.

The junction are marked as J1 to J27 in the above picture.Each of the junctions are
connected by conduits labeled from C1 to C25.The shape of the conduits which are
circular are designed and also the depth of 1.5 m are assigned to each conduit.The next
steps involves setting the rain gauge as per the KOSTRA statistical 10 year return period
data.

The last step is to set the outlets , we consider two outlets for our entire area.The
assumption is made based on the direction of the flow and the junction with the main
sewage pipes.
 62

Results :

The summary of the results are shown in the charts below :

Table 10 Chart showing sub catchment runoff.

Table 11 Chart for Node flooding

 63

Table 12 Chart showing node depth.

Table 13 Chart for outfall loading

Table 14 Chart for conduit surcharge

 64

Table 15 Chart showing link flow.

Figure 35 Results from SWMM

 65

Map showing water depth, precipitation and flow at node, sub catchment and link re-
spectively.

Summary :

Subcatchment Runoff:

• All subcatchments (S1 to S16) experienced the same total precipitation of 158.20
mm.

• There was no surface runoff (0.0 mm) and the total evaporation was also zero,
suggesting either a very high infiltration rate, a simulation of a permeable surface.

• Each subcatchment had the same total infiltration (2.72 mm), impervious runoff
(39.59 mm), pervious runoff (112.81 mm), and total runoff (152.40 mm), indicating
uniform land use, soil type, and slope or an overly simplified model setup.

• The total runoff volume was 7.62 x 10^6 liters for each subcatchment.

• Peak runoff rate (CMS) and Runoff Coefficient were consistent across all
subcatchments (0.68 CMS and 0.963, respectively.

Node Depth :

• Nodes J1 through J27, presumably, show a mixture of average and maximum

depths, with a majority of nodes having some water depth,which indicates that
they are part of active drainage system.

• The maximum Hydraulic Grade Line (HGL) meters, which reflect the water
surface elevation, ranged from 17.00 to 22.55 meters across nodes, with the
maximum HGL typically occurring at the same hour as the maximum depth
reported.

• The day of maximum depth was zero for all nodes, again suggesting a simulation
conducted over a short period or a single storm event.

Conduit Surcharge:

• Conduits C3, C4, and C5 experienced both ends full for 5.13, 5.00, and 5.55
hours, respectively, which is consistent with the hours of upstream and
downstream full conditions.
 66

• The hours above normal flow and hours capacity limited are slightly lower than
the hours both ends full, indicating there may have been periods where the flow
was high but did not completely surcharge the conduit.

• Conduit C16 shows the least stress with only 3.20 hours of both ends full and
capacity limitation, while C5 shows the most with 5.55 hours.

Conclusion

The system demonstrates a high degree of uniformity in response to precipitation, which

could suggest either an overly simplified model(which is done by simplifying the existing
network and some assumptions), a highly consistent terrain and land use across the
modeled area.

The node flooding data point to specific areas prone to flooding, with node J15
experiencing the highest flow rate, which could indicate a localized low point or
bottleneck. Conduit surcharge information reveals that several conduits are operating at
or near their full capacity, suggesting that certain segments of the drainage network may
be undersized or facing blockage issues.

The outfall loading data reinforce the presence of a potentially overtaxed system,
especially at outfall O1, which handles a significant volume of water.
 67

4 Result

4.1 Simulation Based on Historical Data

The precipitation hydrograph analysis, spanning from 2013 to 2023, revealed that the
highest rainfall intensity was recorded on June 14, 2015, at a rate of 3.66 millimetres per
hour.

Relying solely on historical data for flood simulations can be problematic because his-
torical records may not encompass all possible flood scenarios, especially under chang-
ing conditions like urban development and climate change. Statistical return periods, on
the other hand, provide a broader and more adaptable perspective. They estimate the
probability of various flood magnitudes over time, integrating historical data with statisti-
cal models. This approach is more useful for future risk management and planning, as it
accommodates a wider range of scenarios and can be adjusted for ongoing environmen-
tal changes.

The results of the simulation are provided below:

Figure 36 Rainfall data from Historical data taken from Web application.

From the simulation we understand from the historical data of 3.66mm for 10-hour dura-
tion there is not much effect on the LMU hospital campus. We find at most 1 cm to 7 cm
of flood in northeast part of our area of interest which consists of the Portalklinik building.
 68

However interestingly we find some chance of 20 cm of flood at the Augenklinik der

Universität München.

Figure 37 Rainfall simulations results from Historical data taken from Web application.

4.2 Simulation results from Statistical Data

4.2.1 Result from 10-year return period

Statistical return periods are crucial in flood simulations as they offer a probabilistic as-
sessment of flood risks over time. This approach, unlike reliance on historical data alone,
accounts for the likelihood of various flood magnitudes occurring within specific time
frames, such as the infamous "10-year flood." This methodology integrates historical
flood data with advanced statistical models, allowing for a more comprehensive and for-
ward-looking analysis. It's particularly valuable in the context of climate change and ur-
ban development, where flood patterns and risks are evolving. By using statistical return
periods, planners and policymakers can better anticipate and prepare for future flood
events, ensuring more effective risk management, infrastructure planning, and emer-
gency response strategies. This approach is instrumental in mitigating potential flood
impacts in both short-term and long-term scenarios.
 69

The results shown below are from KOSTRA 10-year return period of data:

Figure 38 Rainfall simulation results for 10-year return period data taken from Web application.

The above results get interesting as we could find with this intensity of rainfall it would
be most likely to affect the northeast part of our area of interest with flood levels reaching
from 1 cm to 10 cm and at most 20 cm between the psychiatry building and portal clinic.

4.2.2 Result from 100-year of return period

A 100-year return period flood simulation which was asked from the water department,
though not mandatory, as it serves as a key benchmark in flood risk assessment and
urban planning. This type of simulation helps identify the potential impact of a severe
flood, guiding the development of resilient infrastructure and effective emergency re-
sponse plans. It is also often used to meet regulatory requirements and for insurance
purposes, as it represents a balance between extreme events and practical planning
needs. While not the only scenario to consider, the 100-year flood simulation plays a
crucial role in preparing for and mitigating significant flood risks.
 70

Figure 39 Rainfall simulation results for 100-year return period data taken from Web application.

The results show 0.1 m to 0.2 m of flooding in our area of interest mostly on the Psychi-
atry clinic. However, we see high vulnerability in the northeast part near Augenclinic with
flood levels reaching 0.4 to 0.5 m which could produce devastating results.

However, as these results are from a 100-year return period these results are likely to
occur.

4.3 Simulation results from ML prediction model

When we run our prediction of rainfall for the month of June 2024 as it’s the month with
the worse rainfall from the historical data it came around to be 96.09 mm of average,
which is well below the data we derived from KOSTRA statistical model. Thus, it’s obvi-
ous the results are good enough to have any effect on our area of interest.
As of now our machine learning model only takes the historical rainfall as a model input,
by feeding a neural network with not only rainfall patterns but also incorporating variables
such as humidity levels, wind velocity, and other climate change indicators, we can
create a more robust and accurate predictive model.

This approach allows the neural network to learn from the intricate interactions between
various weather elements and their cumulative effect on flood risks. For instance, high
humidity combined with specific wind patterns might increase the likelihood of heavy
rainfall, a nuance that our current model might overlook. Furthermore, including climate
change data can help adjust predictions based on long-term environmental shifts,
 71

offering a more forward-looking perspective on flood risks.Which might be an interesting

future scope of this project.

4.4 Visualization from Unreal Engines

From the results derived from above we make a visualisation with unreal engines con-
sidering the 100-year return period rainfall which caused almost 20 cm flood level near
the psychiatry building.

Figure 40 Rainfall simulation in Unreal Engines

 72

5 Conclusion

The analysis of flood risk at the LMU campus, specifically in the Zimmensblock area of
Munich, has demonstrated a minimal threat from heavy rainfall events. This conclusion
is supported by comprehensive studies that utilized historical data, machine learning
models, and statistical analysis.

Analysis of historical flood data, indicating a rainfall intensity of 3.66mm over a 10-hour
period, demonstrates minimal impact on the area of interest. This observation holds even
under the assumption that all sewage pipes are operating at full capacity, with flood lev-
els ranging merely from 2 to 7 cm—a magnitude of effect that can essentially be disre-
garded. The employment of a machine learning model, specifically utilizing the Random
Forest algorithm, yields simulation results that align with these observed low flood levels.
This concurrence is attributed to the model's training on past historical data alone, with-
out incorporating other significant factors that influence flooding. Nonetheless, as previ-
ously mentioned, there exists potential for enhancing the accuracy and predictive capa-
bility of this machine learning model in future analyses by integrating a broader spectrum
of influencing factors.

When considering the statistical data from KOSTRA, we concluded some interesting
results:

In the scenario of rainfall with a 10-year return period, flood levels ranging from 2 to 20
cm are anticipated at the Psychiatry building of the LMU campus. However, these levels
fall well below the building's threshold, rendering it safe from flooding in the foreseeable
future.

Conversely, for a 100-year return period, predicted flood levels span from 20 cm to 40
cm. In this instance, the primary concern shifts towards pedestrian safety, rather than
the integrity of critical infrastructure, which is expected to remain largely unscathed.

In summary, the collective evidence from historical records, machine learning

simulations, and statistical projections points to a minimal flood risk for the LMU campus
in Munich's Zimmensblock area. Yet, the drainage system shows signs of stress in cer-
tain areas for example J15, necessitating targeted improvements to ensure continued
 73

resilience against flooding, particularly as climate patterns evolve and the potential for
extreme weather events increases. Further studies and future and model refinements
could be made to address these localized concerns and to enhance the predictive capa-
bilities for future flood scenarios. While certain extreme weather events could present
challenges, the current infrastructure and building designs appear well-equipped to
mitigate these risks effectively.
 74

6 Discussion

The integration of sophisticated computational tools and methodologies significantly im-

proves our understanding and capability to forecast and manage environmental and
structural challenges in urban areas. This discussion compiles insights from the applica-
tion of ArcGIS and IFCopenshell for construction project data management, the study of
rainfall intensity and its impact on urban flooding, and predictive modelling of future rain-
fall events, offering perspectives on urban planning and infrastructure resilience. By uti-
lizing IfcProjectedCRS and IfcMapConversion entities, projects can now incorporate pre-
cise geographical data into IFC files, enabling more accurate and automated data han-
dling. This advancement is key for aligning project data with geospatial reference sys-
tems, thus boosting the interoperability and accuracy of construction project manage-
ment. Nonetheless, the adoption and comprehension of these capabilities remain a chal-
lenge, essential for enhancing project coordination and efficiency.

The highest recorded rainfall intensity serves as a crucial metric for assessing flood risk
and simulating worst-case scenarios. The specific flooding around node J15 and the
surcharge in several conduits emphasize the need for thorough drainage network
analysis in urban planning. These insights suggest that infrastructure might need
upgrades or redesigns to handle the increasing rainfall intensity, which climate change
may intensify. Additionally, the development of a web application for this project signifi-
cantly enhances data visualization and interaction. Built using JavaScript and HTML, this
application loads GLB models of the targeted area, displaying the elevation of floods
relative to IFC-structured buildings as GLB models with various flood prediction results.
This tool not only provides a dynamic platform for analyzing flood impacts but also facil-
itates a deeper understanding of flood elevations in relation to urban infrastructure, fur-
ther enriching the project's contributions to urban planning and resilience efforts.
Simulation results reveal significant flood level variations across the LMU hospital
campus, demonstrating a detailed understanding of how rainfall intensity affects specific
urban areas. Such findings are vital for targeted interventions and allocating resources
to highly vulnerable areas like the Augenklinik der Universität München.
 75

While the predictions indicate lower-than-expected rainfall, the application of machine

learning models holds promise for improving the accuracy of such forecasts. By incor-
porating a broader range of climatic variables, future models can offer more reliable pre-
dictions, thus better informing preparedness and mitigation strategies. However, this ap-
proach necessitates a continuous refinement of models and integration of up-to-date
data to reflect changing climate patterns accurately.

The findings from this thesis underscore the interdependence between technological ad-
vancements in data management, analytical precision in environmental assessment,
and the imperative for adaptive urban planning. As cities continue to grow and face the
multifaceted challenges posed by climate change, the integration of sophisticated com-
putational tools and methodologies will be paramount in crafting resilient and sustainable
urban environments. Future research should focus on expanding the datasets and vari-
ables involved in predictive modelling, alongside exploring the implications of these tech-
nologies in broader urban contexts. The journey towards resilient urban infrastructure is
iterative, requiring ongoing assessment, adaptation, and innovation.
 76

7 Methods For Flood Mitigation

The devastating flood disaster of July 2021, particularly in the states of Rhineland-Palat-
inate and North Rhine-Westphalia, underscored the catastrophic potential of natural ca-
lamities, resulting in the tragic loss of over 180 lives and leaving around 800 individuals
seriously injured. Thus, it gets very important for simulating and understand potential
flood threat levels in the historical location of Munich. Although, from the simulation re-
sults we didn’t find much vulnerability in our area of interest even from the 100 year of
return period if flood still it gets very important to prepare for unseen situations consid-
ering climate change.

These techniques include avoiding construction on floodplains, creating urban "sinks" to

hold water, implementing green infrastructure such as green roofs and permeable
pavements, utilizing underground systems like soakaways, and engaging households in
adopting private-side flood protection measures (Qin, 2020). Urban planning should be
adaptive to climate change, and a paradigm shift has occurred in flood management,
emphasizing softer approaches and the integration of natural systems into urban
planning to increase resilience to flooding. These techniques aim to reduce the risk to
houses, schools, hospitals, businesses, and other key infrastructure, ultimately
enhancing the city's ability to withstand and recover from flooding events.
References 77

References

Ohori, G. A., Biljecki, F., Diakité, A. A., Krijnen, T., Ledoux, H., & Stoter, J. (2017).
TOWARDS AN INTEGRATION OF GIS AND BIM DATA: WHAT ARE THE
GEOMETRIC AND TOPOLOGICAL ISSUES? ISPRS Annals of the
Photogrammetry, Remote Sensing and Spatial Information Sciences, IV-4/W5,
1–8. https://doi.org/10.5194/isprs-annals-iv-4-w5-1-2017

Zhao, L.; Liu, Z.; Mbachu, J. An Integrated BIM–GIS Method for Planning of Water
Distribution System. ISPRS Int. J. Geo-Inf. 2019, 8, 331.
https://doi.org/10.3390/ijgi8080331

Baldwin, B. M. (2018, February 4). What is IFC? Bimconnect.

https://bimconnect.org/en/software/what-is-ifc/

Borrmann, A., Beetz, J., Koch, C., Liebich, T., Muhic, S. (2018). Industry Foundation
Classes: A Standardized Data Model for the Vendor-Neutral Exchange of Digital
Building Models. In: Borrmann, A., König, M., Koch, C., Beetz, J. (eds) Building
Information Modeling. Springer, Cham. https://doi.org/10.1007/978-3-319-
92862-3_5

Exploring digital elevation models—ArcGIS Pro | Documentation. (n.d.).

https://pro.arcgis.com/en/pro-app/3.1/tool-reference/spatial-analyst/exploring-
digital-elevation-models.htm

Kolbe TH, Gröger G (2003) Towards unified 3D city models. In: Schiewe, J., Hahn, M,
Madden, M, Sester, M (eds): Challenges in Geospatial Analysis, Integration and
Visualization II. Proc. of Joint ISPRS Workshop, Stuttgart

OpenData. (2023). © Bayerische Vermessungsverwaltung.

https://geodaten.bayern.de/opengeodata/

Open Geospatial Consortium. (2023, June 29). CityGML - Open Geospatial

Consortium. https://www.ogc.org/standard/citygml/

Autodesk (2023): Education. Online verfügbar unter http://students.autodesk.com/, zu-

letzt geprüft am 07.09.2023.
References 78

Autodesk Revit (2023): Education. Online verfügbar unter http://students.auto-

desk.com/, zuletzt geprüft am 07.09.2023.

Welcome to the QGIS project! (n.d.). https://www.qgis.org/en/site/index.html

Autodesk Civil3D (2023): Education. Online verfügbar unter http://students.auto-

desk.com/, zuletzt geprüft am 07.09.2023.

BIM and GIS Integration | Models & Maps Working Together. (n.d.).
https://www.esri.com/en-us/industries/aec/overview/gis-and-bim

IfcOpenShell - The open source IFC toolkit and geometry engine. (n.d.).
https://ifcopenshell.org/

Stijngoedertier. (n.d.). GitHub - stijngoedertier/georeference-ifc. GitHub.

https://github.com/stijngoedertier/georeference-ifc

Autodesk Infraworks(2023): Education. Online verfügbar unter http://students.auto-

desk.com/, zuletzt geprüft am 07.09.2023.

Del Duca, G., Rocha, G., Orszt, M., & Mateus, L. (2022). A Preliminary Contribution
towards a Risk-Based Model for Flood Management Planning Using BIM: A
Case Study of Lisbon. Sensors, 22(19), 7456.
https://doi.org/10.3390/s22197456

Jackson, A. (2014, August 2). Discharge & hydrographs.

https://geographyas.info/rivers/discharge-and-hydrographs/

Novrima, A. (2023). Effectiveness analysis of East Sunter polder system for flood
control management using HEC-HMS & HEC-RAS software.
https://www.semanticscholar.org/paper/Effectiveness-analysis-of-East-Sunter-
polder-system-Novrima-
Anggraheni/bbe59ec0f76285fb2b0f1bd17583460dd0c5f087

Koehrsen, W. (2020, August 18). Random Forest Simple explanation - Will Koehrsen
- medium. Medium. https://williamkoehrsen.medium.com/random-forest-simple-
explanation-377895a60d2d

Armain, M. Z. S. (2021). Hydrodynamic modelling of historical flood event using one

dimensional HEC-RAS in Kelantan basin, Malaysia.
https://www.semanticscholar.org/paper/Hydrodynamic-modelling-of-historical-
References 79

flood-event-in-Armain-
Hassan/737873008d01b715ae10b50d0fc85d5e484ca899

https://www.oecd.org/climate-action/ipac/practices/germany-s-sponge-cities-to-
tackle-heat-and-flooding-7b6caa58/

Qin, Y. Urban Flooding Mitigation Techniques: A Systematic Review and Future Stud-
ies. Water 2020, 12, 3579. https://doi.org/10.3390/w12123579

Takata, H. (2024). Back-Calculation of Manning’s roughness coefficient by 2D flow

simulation and influence of In-Channel physical parameters in a mountain river,
Japan. https://www.semanticscholar.org/paper/Back-Calculation-of-
Manning%E2%80%99s-Roughness-Coefficient-Takata-
Obata/7ac1581cf4a057ba8edbe08230f96d6c54a4f6d0

Bauer, M., Thuro, K., Marcus, S., & Neumann, P. (2005). The geology of Munich
(Germany) and its significance for ground modelling in urban areas.
ResearchGate.
https://www.researchgate.net/publication/237692735_The_geology_of_Munich
_Germany_and_its_significance_for_ground_modelling_in_urban_areas

Liyew, C. M., & Melese, H. A. (2021). Machine learning techniques to predict daily
rainfall amount. Journal of Big Data, 8(1). https://doi.org/10.1186/s40537-021-
00545-4

Meenal, R., Michael, P. A., Pamela, D., & Ekambaram, R. (2021). Weather prediction
using random forest machine learning model. Indonesian Journal of Electrical
Engineering and Computer Science, 22(2), 1208.
https://doi.org/10.11591/ijeecs.v22.i2.pp1208-1215

Unreal Engine is an Advanced Training and Simulation Software Platform. (n.d.).

Unreal Engine. https://www.unrealengine.com/en-US/solutions/simulation

Yang, Y., Sun, L., Li, R., Yin, J., & Yu, D. (2020). Linking a storm Water Management
model to a novel Two-Dimensional Model for Urban Pluvial Flood modeling.
International Journal of Disaster Risk Science, 11(4), 508–518.
https://doi.org/10.1007/s13753-020-00278-7

Son, A., Kim, B., & Han, K. (2016). A simple and robust method for simultaneous
consideration of overland and underground space in urban flood modeling.
Water, 8(11), 494. https://doi.org/10.3390/w811049
8 List of Figures

Figure 1 Innenstadt clinic LMU ................................................................................................................................ 2

Figure 2 Overview of the developed method and workflow .................................................................................... 3
Figure 3 Different Layers in the model .................................................................................................................... 8
Figure 4 Water distribution in ArcGIS ...................................................................................................................... 9
Figure 5 IFC model of Psychiatry clinic .................................................................................................................. 13
Figure 6 2D CAD model of the Zimmensblock area ............................................................................................... 14
Figure 7 Different Layers in the DTM and DEM (Štular, Benjamin & Eichert, Stefan & Lozić, Edisa, 2021) ........... 15
Figure 8 Elevations of the Zimmensblock area in QGIS ........................................................................................ 17
Figure 9 Different Level of Detail as in CityGML (An Improved LOD Specification for 3D Building Models, 2016) 18
Figure 10 CityGML LoD2 of the Zimmensblock area.............................................................................................. 19
Figure 11 Georeferenced CAD model in Civil3D..................................................................................................... 22
Figure 12 Georeferncing IFC in ArcGISPro ............................................................................................................. 23
Figure 13 Georeferencing IFC in ArcGIS Pro .......................................................................................................... 24
Figure 14 Diagram of georeferenced IFC in IFCopenshell ...................................................................................... 25
Figure 15 Converting Elevation into TIFF images. ................................................................................................. 26
Figure 16 Model X in Infraworks ........................................................................................................................... 28
Figure 17 HEC-RAS Dashboard .............................................................................................................................. 32
Figure 18 Elevations in HEC-RAS............................................................................................................................ 34
Figure 19 Catchment area for our area of interest................................................................................................ 35
Figure 20 Boundary conditions as set in HEC-RAS ................................................................................................. 35
Figure 21 Interface as in Fast Flow application .................................................................................................... 36
Figure 22 Sentinel 2 imagery on Land Cover ......................................................................................................... 41
Figure 23 Manning’s coefficient with respect to land cover.................................................................................. 42
Figure 24 Munich clay content map from SoilGrid ................................................................................................ 42
Figure 25 Simplified Random Forest algorithm (Koehrsen, 2020) ........................................................................ 43
Figure 26 Picture of Infraworks FBX export window ............................................................................................. 50
Figure 27 Picture of Model X in blender ................................................................................................................ 51
Figure 28 Model X in the web application ............................................................................................................. 52
Figure 29 Rainfall data with elevation .................................................................................................................. 54
Figure 30 Flood elevation bar as in web application ............................................................................................. 56
Figure 31 Picture of alerts in vulnerable areas ...................................................................................................... 57
Figure 32 Clickable features inside web application.............................................................................................. 57
Figure 33 Zimmensbblock on unreal engines ........................................................................................................ 58
Figure 34 Pipe Network simplified of LBD from StBA München ............................................................................ 61
Figure 35 Results from SWMM.............................................................................................................................. 64
Figure 36 Rainfall data from Historical data taken from Web application. .......................................................... 67
Figure 37 Rainfall simulations results from Historical data taken from Web application. .................................... 68
Figure 38 Rainfall simulation results for 10-year return period data taken from Web application. ..................... 69
Figure 39 Rainfall simulation results for 100-year return period data taken from Web application. ................... 70
Figure 40 Rainfall simulation in Unreal Engines .................................................................................................... 71
9 List of Tables

Table 1 Historical Rainfall data 29

Table 2 Annual Rainfall for Each Year 30
Table 3 Year v/s Rainfall in each month 31
Table 4 Rainfall intensity from historical data 37
Table 5 Statistical data from KOSTRA 38
Table 6 Precipitation Hydrograph 39
Table 7 Mean Annual Rainfall data. 44
Table 8 API service listings of weather API 48
Table 9 Web application Architecture 51
Table 10 Chart showing sub catchment runoff. 62
Table 11 Chart for Node flooding 62
Table 12 Chart showing node depth. 63
Table 13 Chart for outfall loading 63
Table 14 Chart for conduit surcharge 63
Table 15 Chart showing link flow. 64
10 Declaration

I hereby declare that I wrote this Master’s thesis independently. Only the sources and
aids explicitly named in the work were used. I have identified ideas that have been
taken over literally or figuratively as such.

I also certify that this work has not yet been used as the basis for any other examination
procedure.

München, 29. February 2024 Indraditya Bhattacharyya

Vorname Nachname

Indraditya Bhattacharyya

Heiglhofstraße 66, 0625

D-81377 München

[email protected]