Summer Internship Report

on

Catastrophe Modelling
At
RMS INDIA

Submitted by
Tushar Gupta-11113114
Kaustubh Milind Kulkarni-1113052
Karan Tyagi-11113048
B. Tech. 4th year

Department of Civil Engineering

Indian Institute of Technology Roorkee

ACKNOWLEDGEMENT:
This internship proved to be a fruitful endeavor and we heartily thank the entire
Analytics and Exposure Development team and Spatial Modelling Group at Risk
1 | Page

Management Solutions (RMS), India who provided us with this opportunity. A special
acknowledgement is deserved by Farhat Rafique, Kunal Jain, Edida Rajesh, Alpana
Das and Nikhil Sharma who, throughout the internship period, not only helped us in
comprehending various problems but also in developing the skills required to tackle
them. It was because of their help that we could complete this entire task in eight
weeks.

Table of Contents
ACKNOWLEDGEMENT:........................................................................................................2
INTRODUCTION:.................................................................................................................3
2 | Page

Background.....................................................................................................................3
Catastrophe Modeling.....................................................................................................3
Software and frameworks:.................................................................................................3
EPA-SWMM:.....................................................................................................................3
inp.PINS:..........................................................................................................................4
inpFLOOD:....................................................................................................................5
HEC-Ras:.........................................................................................................................6
SWAT:..............................................................................................................................7
ArcGIS:............................................................................................................................7
Urban flood modeling using SWMM...................................................................................9
Urban Flooding................................................................................................................9
Urban Flood Modeling...................................................................................................10
Urban Flood Modeling Process:.....................................................................................10
STEPS:........................................................................................................................10
Sewer Networks Data:..................................................................................................11
Acquisition of Data.....................................................................................................11
Availability of Data.....................................................................................................12
Conclusion.................................................................................................................12
PANAMA CITY, FLORIDA:................................................................................................12
Storm Water Network Data........................................................................................13
Generating Outlets....................................................................................................14
Cleaning of Data........................................................................................................14
Delineation of Channels.............................................................................................15
Coorelation of sewer networks and delineated channels..........................................16
Corelation of sewer network and road network.........................................................16
Assigning flow direction to road networks.................................................................17
Conclusion.................................................................................................................18
BELLINGHAM CITY, WASHINGTON:................................................................................18
Sewer Network Data..................................................................................................19
SENSITIVITY ANALYSIS...............................................................................................20
IMPLEMENTATION.......................................................................................................23
INPFLOOD INVESTIGATION.........................................................................................23
ALGORITHM................................................................................................................24
MANUAL INCREASE OF WATER DEPTH.......................................................................25
3 | Page

NO RAIN CONDITION..................................................................................................27
DERIVING INVERT ELVEVATIONS FROM DEM..............................................................27
NO RAIN CONDITION FOR MODIFIED INVERTS...........................................................28
CONCLUSION..............................................................................................................29
SWAT................................................................................................................................29
Deliverables..................................................................................................................29
Working Concept...........................................................................................................30
Overview Of SWAT........................................................................................................30
Calculation of Sub basin Parameters.........................................................................34
HRU Analysis..............................................................................................................36
HRU Definition...........................................................................................................40
Writing Input Tables...................................................................................................43
Edit SWAT Input.........................................................................................................45
Edit Point Discharge Inputs........................................................................................46
Edit Inlet Dischargers Input.......................................................................................47
Edit Sub basins Data..................................................................................................47
SWAT Simulation Setup..............................................................................................47
Output Analysis..........................................................................................................48
Utility Network Analyst Tool.............................................................................................49
Creating Flow Direction for sewer network...................................................................50
Manually Assigning Direction:.......................................................................................50
Trace Options:...............................................................................................................53
Exposure Development:...................................................................................................55
Background...................................................................................................................55
Theory...........................................................................................................................56
PARAMETERS.................................................................................................................57
Building Inventory......................................................................................................57
Size of building..........................................................................................................57
Replacement cost value.............................................................................................57
EXPOSURE CALCULATION.............................................................................................57
EXPOSURE DOWNSCALING...........................................................................................57
OBJECTIVE..................................................................................................................58
METHODOLOGY..........................................................................................................58
Urban Atlas Classification..........................................................................................58
4 | Page

Land Use/ Land Cover Classification..........................................................................58

Soil Sealing data........................................................................................................59
Open Street Maps(OSM) Data....................................................................................59
PROCESS....................................................................................................................60
RESULT.......................................................................................................................60
CONCLUSION..............................................................................................................61
K-MEANS CLUSTERING FOR IDENTIFYING VEGETATION IN IMAGES...............................61
INTRODUCTION..........................................................................................................61
OBJECTIVE..................................................................................................................62
PROCESS....................................................................................................................62
AGRICULTURAL EXPOSURE DEVELOPMENT : BELGIUM.................................................66
Base data summary...................................................................................................66
BUILDING INVENTORY AND FLOOR AREA...................................................................66
Steps..........................................................................................................................66
Total Built-up Area Data:............................................................................................67
Construction costs.....................................................................................................68
Trending and Currency conversion............................................................................69
Exposure calculations................................................................................................70
RESULT.......................................................................................................................70
ArcGIS Work..................................................................................................................71
Key Takeaways:..........................................................................................................71
REFRENCES:.....................................................................................................................72

INTRODUCTION:
The following is an account of our work as interns for two months at the Noida office
of RMS India.
We were working in the Spatial Modeling Group and Model Development Group with
Farhat Rafique and Alpana Das as our mentor. We were working on developing
urban flooding models for estimating the flood inundation extents and flooding
models for river basins using SWAT.

5 | Page

Background
Risk Management Solutions is a provider of products and services for the
quantification and management of catastrophe risks. It offers technology and
services for the management of insurance catastrophe risk associated with natural
perils such as earthquakes, hurricanes, and windstorms and risk modeling for manmade disasters associated with acts of terrorism by analyzing the impact of
weapons of mass destruction on property. The company's objective is to help clients
improve financial performance by using products and services to gain the most
complete view of their risk portfolio. In a nutshell, the nature of work of RMS is
classified as catastrophe modeling.

Catastrophe Modeling
Catastrophe modeling is the process of using computer-assisted calculations to
estimate the losses that could be sustained due to a catastrophic event such as a
hurricane, earthquake, storm, flood and so on. The calculated losses are used by the
insurance industry to better analyze the risks associated with their portfolios and
helps in the pricing of premiums depending upon the susceptibility of the property
to any natural hazard. The models simulate natural phenomena by using the
engineering equations associated with the disaster.

Software and frameworks:

EPA-SWMM:
The EPA Storm Water Management Model (SWMM) is a dynamic rainfall-runoff
simulation model used for single event or long-term (continuous) simulation of
runoff quantity and quality from primarily urban areas.
6 | Page

The runoff component of SWMM operates on a collection of sub catchment areas

that receive precipitation and generate runoff and pollutant loads.
The routing portion of SWMM transports this runoff through a system of pipes,
channels, storage/treatment devices, pumps, and regulators.
SWMM tracks the quantity and quality of runoff generated within each sub
catchment, and the flow rate, flow depth, and quality of water in each pipe and
channel during a simulation period comprised of multiple time steps.

inp.PINS:
However, EPA SWMM does not have support for shp files and supports only inp files
which is a workspace for EPA SWMM. inp.PINS is an open source plug in for Map
Windows which converts data from shp files to inp files.

7 | Page

inpFLOOD:
inp.PINS has another module called inpFLOOD which uses the inp file for a particular
project containing geometric and other data about different elements, the rpt file
containing data about the results obtained during the run and the DEM of the region in
order to give the flood map for the region as a shp file.

8 | Page

HEC-Ras:
HEC-RAS is a program that models the hydraulics of water flow through natural
rivers and other channels. It considers only the one dimensional aspects of flow.
Although not the exact purpose for which it is meant, HEC-RAS can model the flood
extents by using the flows from junctions. It requires:

Centerline

Banks

Cross sections

Flows

HEC-GEORAS
Hec-GeoRAS is the plug-in for ArcGIS of HEC-RAS. The necessary layers for HECGeoRAS can be taken as:

Stream center lines Conduits

Stream banks A buffer functon on stream centre lines

Either a DEM or a TIN model

Stream cross sections.

9 | Page

Other layers are easily available but automated cross sections are difficult to obtain
from the DEM.

SWAT:
SWAT is a complex integrated river basin scale model which operates either on daily
or hourly time step. It quantifies the impact of land management practices in
hydrology, erosion and non-point source pollution.
The basic concept of SWAT involves water balance while accounting for weather,
surface runoff, return flow, percolation, evapotranspiration, transmission losses,
storage, crops, irrigation, groundwater flow, etc. For the calibration purposes, SWAT
uses LH-OAT which stands for Latin Hypercube One factor At a Time analysis. It
combines LH sampling and OAT design for simulation. It ensures parameter
reduction by filtering out the less influential ones at calibration stage. Thus SWAT
allows to calculate discharge over basins and simulate a model and compare it with
the original data.

10 | P a g e

ArcGIS:
ArcGIS is a geographic information system for working with maps and geographic
information. It is used for creating and using maps, compiling geographic data,
analyzing mapped information, sharing and discovering geographic information,
using maps and geographic information in a range of applications, and managing
geographic information in a database.
The system provides an infrastructure for making maps and geographic information
available throughout an organization, across a community, and openly on the Web.
ArcGIS includes the following Windows desktop software:

Arc Reader - It allows one to view and query maps created with the other
ArcGIS products;

ArcGIS for Desktop, which is licensed under three functionality levels: [6]

ArcGIS for Desktop Basic (formerly known as ArcView)- It allows one to

view spatial data, create layered maps, and perform basic spatial analysis;

ArcGIS for Desktop Standard (formerly known as ArcEditor) - in addition

to the functionality of ArcView, includes more advanced tools for
manipulation of shapefiles and geodatabases.

ArcGIS for Desktop Advanced (formerly known as ArcInfo) - It includes

capabilities for data manipulation, editing, and analysis.

ArcGIS is built around the geodatabase, which uses an object-relational

database approach for storing spatial data. A geodatabase is a "container" for
holding datasets, tying together the spatial features with attributes. The
geodatabase can also contain topology information, and can model behaviour of
features, such as road intersections, with rules on how features relate to one
another. When working with geodatabases, it is important to understand about
feature classes which are a set of features, represented with points, lines, or
polygons. With shapefiles, each file can only handle one type of feature. A
geodatabase can store multiple feature classes or type of features within one file.

Python Module for GIS:

ArcPy
ArcPy is a Python module for ArcGIS that can be used for scripting specific scripts
for processes that cannot be performed by the available tools. It provides a much

11 | P a g e

greater degree of customization as compared to the default tools or even the

available plug ins on the internet developed by developers.
Python scripts can be run in either a shell or in the field calculator for evaluation of
tables.

12 | P a g e

Urban flood modeling using SWMM

Urban Flooding
Urban flooding is the inundation of land or property in a built environment,
particularly in more densely populated areas, caused by rainfall overwhelming the
capacity of drainage systems, such as storm sewers. Urban flooding is a condition,
characterized by its impact on communities regardless of whether or not they are
located within formally designated floodplains or near any body of water. There are
several ways in which storm water enters properties: backup through sewer pipes,
toilets and sinks into buildings; seepage through building walls and floors; the
accumulation of water on property and in public rights-of-way; and the overflow
from water bodies such as rivers and lakes.
Urban flooding is specific in the fact that the cause is a lack of drainage in an urban
area. As there is little open soil that can be used for water storage nearly all the
precipitation needs to be transport to surface water or the sewage system. High
intensity rainfall can cause flooding when the city sewage system and draining
canals do not have the necessary capacity to drain away the amounts of rain that
are falling. Water may even enter the sewage system in one place and then get
deposited somewhere else in the city on the streets.

Urban Flood Modeling

The simulation of urban floods primarily involves the knowledge of phenomena
associated with hydrology and hydraulics and the associated engineering equations.
Different frameworks are available which use available sewer network data to
generate detailed simulations of water flow inside the sewers. The man-holes in the
sewer network from which water will overflow are highlighted and the volume of
water is then routed to the surrounding areas using the digital elevation model
(DEM) of the region. The height of water in the region is displayed as the flood
inundation map.

The primary assigned deliverables for the internship were:

1. Ideas and algorithms for the automatic cleansing of the data layer for major
conduits and their implementation.
2. Running the model to achieve the ponding at intersections.
13 | P a g e

Urban Flood Modeling Process:

The objective is to compare results obtained from simulations with the actual sewer
networks data for cities for which it is available and from channels delineated with
either DEM data or from any other utilities layer. The assignment of parameters
should be done so as to achieve good correlation between both results.

STEPS:

Review and understand

1. Review the study area city online.

Analyze storm network data

1. Check for flow pattern

2. Check available information
3. Check if cleaning by information is possible
4. Check if spatial cleaning is possible

Elevation data

1. Check 10 m DEM
2. Check availability of any other HD DEM
3. Interpolate 10 m DEM to 5m/2m

Delineation

1. Delineate channels from DEM

2. Compare delineated channels and storm network data
3. Derive correlation between both networks
4. Generate catchments for delineated channels
5. Clean catchments for pipes if available or else generate them

Attributization

1. Assign attributes required by urban flood model

14 | P a g e

Simulation

1. Run the model using both channels and pipes.

2. Derive the ponding with channels and pipes

Correlation of ponding

1. Compare the ponding of channels and pipes

2. Find out solution if no correlation

Sewer Networks Data:

Acquisition of Data
Data for various quantities is readily available in the United States as compared to
India. Information about utilities in urban areas is available in the form of shape files
or GIS layers. I was particularly interested in the sewer network data for regions in
the form of separate shape files for

Junctions

Conduits

Outfalls

Because SWMM uses these layers and corresponding information for simulation.
These layers are a prerequisite and simulations cannot be run in their absence.
Very detailed sewer networks for certain cities such as Los Angeles were available in
the form of shp files with all associated data.
The data for the city of Spokane in Washington was available as a kmz file for
Google Earth
New York however had limited data available only in the form of of sewer outfalls.
The vital sewer data for conduits and nodes wasnt available.

Availability of Data
The data for sewer networks, if available, is usually present on the website of the city or
region. For certain regions, although the sewer network data was not available, related
information such as

Minnesota - Sewer sheds and limited sewer interceptors.

15 | P a g e

Spokane - Wastewater data as a kml file.

Salem, Oregon - Entire sewer network metadata as html files

Austin, Texas Water and wastewater fee boundary.

So, all in all detailed sewer network data is not very readily available.

Conclusion
Detailed data for sewer networks is rarely available and even if its available, its
very rarely in a format suited to our purposes. So, instead of relying on sewer
network data for obtaining urban flood maps, the focus should be on using
delineated channels from DEM data or easily available networks like road maps by
TeleAtlas to act as substitute for conduits. The nodes and the outfalls can be
generated from the road network and the elevations can be assigned from the DEM.
The flooding from such an obtained network, and from the actual network for cities
for which data is available, should be compared. The parameters for the derived
network should be calibrated in order to get accurate results so that generating
urban flooding scenarios can be automated for the whole of the US.

PANAMA CITY, FLORIDA:

Panama City is a city in Bay County, Florida, United States. Panama City is located
at 301028N 853952W within the Florida Panhandle and along the Emerald
Coast. Panama City has a humid subtropical climate, with short, mild winters and
long, hot and humid summers.
The relatively flat terrain of Bay County and the large areas of wetlands and areas
with a high water-table combine to present unique challenges for managing the
frequent and sometimes intense rainfall events. Certain areas of Panama City are
low-lying and subject to flooding from rising water

16 | P a g e

Storm Water Network Data

The data for the sewer network for Panama City on the SMG server was pretty
detailed with layers for
Sewer gravity mains, force mains, manholes, lift stations, valves, meters
Water fittings, distribution lines, fittings, hydrants, sampling sites, valves, meters
Storm water pipes.

17 | P a g e

Investigation revealed that water layers indicated the water distribution network,
and the sewer and the storm water network were separate.
So the only layer useful for flood delineation was the storm water pipes. I found the
missing storm water inlet data on the official site of Bay County which had an
interactive GIS map.

Generating Outlets
Outlets are mandatory for running simulations with SWMM. The data for Panama
City did not have outlet data available.
I used a third party extension for ArcGIS called X Tools Pro for the generation of
outlets.
From nodes and to nodes were generated and then the removal of from nodes
from the to nodes left the outlets. Default parameters were assigned to the outlets.

Cleaning of Data
The storm water pipes are much cluttered and such networks are not necessary for
giving accurate flooding results. Rather, they make the computations for

18 | P a g e

simulations extremely intensive. Hence cleaning the pipe network for removing
obsolete pipes is necessary.

Spatial Cleaning:
The pipes and nodes are represented as a link node data structure and the lower
order pipes can be removed from the network.
For example, if, in an individual tree, the highest order pipe is 5, then all pipes with
order lesser than 2 can be removed and higher order pipes can be left to get
spatially cleaned networks.
Cleaning by attributes:
The information available about the attributes of the pipes can be used for removing
the smaller pipes at the beginning of the network. Attributes like type, diameter,
and material can be used.
Else a combination of spatial and attribute cleaning can be used to clean the
network and get good results.

Delineation of Channels
Delineated channels for a threshold limit of 1000 cells for the 3 m DEM.
Similarly channels were delineated for a threshold limit of 2000 and 4000 cells also.

19 | P a g e

Delineated channels for a threshold limit of 1000 cells for the 10 m DEM.
Similarly channels were delineated for a threshold limit of 5000, 2000 and 4000
cells also.

Correlation of sewer networks and delineated channels

Visual analysis indicated the best correlation between the channels delineated for
the 3m DEM at a threshold limit of 4000 or for the 10 m DEM at a threshold limit of
500 cells.
The network for the 3m DEM was obviously considerably denser as compared to the
10 m DEM for the same threshold value.
Possibly the 3 m DEM is giving better results because it can distinguish man made
artefacts like roads and sewer networks very often coincide with major roads.

20 | P a g e

Correlation of sewer network and road network

Very detailed road networks are available for the US and major roads very
frequently coincide with main sewer lines, which is what we are concerned with.
I took all the roads with Functional Road Class(FRC) up to 5 and found that they
coincided well with major sewers. Roads with FRC 6 make the network very
cluttered and will be very difficult to analyze.
I tried cleaning the road network using various other fields like type, length, class
but the best results were seen only with the condition FRC <= 5.

21 | P a g e

Assigning flow direction to road networks

The assigning of flow directions to road derived networks is essential.
I tried using various methods to assign directions.
Using DEM:
Taking points along a polyline and checking the flow directions of the polyline by
assigning flow from higher to lower elevations. Used Hawths Tools for the
sampling of data.
Using slope map:
Assigning flow directions using slope map of the region.
Using FDR:
Taking the clipped FDR map of the region as per the road network and assigning
flow directions to polylines.
Various checks will also have to be applied to the assigned flow directions of the
road network derived pipes. I was unable to get a concrete result but the matter can
be investigated further. On the other hand, using the DEM derived channels ought
to be a less cumbersome process.

22 | P a g e

Conclusion
I could not get flooding maps for Panama City because SWMM requires sub
catchment objects from which flow has to be assigned to a particular node. I had to
perform delineation of sub catchments. InfoSWMM Subcatchment Manager is a
tool that automates the process of delineation of sub catchments, their
characterization and the assignments of sub catchments to the appropriate nodes
and rain gages. I had yet to try out the tool
Back when I did was doing my internship after my first year in MWH Global, they
were working on developing flood maps for Mumbai. The delineation of sub
catchments and the assignment of nodes was done manually over there. However
an automated process would go a long way in deriving urban flood maps for the
whole of US and other countries.

BELLINGHAM CITY, WASHINGTON:

Bellingham is the largest city in, and the county seat of, Whatcom County in the
State of Washington.
The city is located at 4845N 12229W and it is situated on Bellingham Bay.
Flooding is an issue in the city with warnings for floods given on the official website
of Bellingham city.

23 | P a g e

Sewer Network Data

Detailed sewer network data is available for Bellingham City with separate layers for
junctions, conduits, outfalls and sub catchments.
The parameters needed for running simulations are mostly available or else, default
values are used.

24 | P a g e

SENSITIVITY ANALYSIS
COMPARTMENTS
EPA SWMM consists of different compartments which are used to model the
hydrological and hydraulic processes.
1. Atmosphere compartment
2. Surface compartment
3. Groundwater compartment
4. Transport Compartment

ATMOSPHERE COMPARTMENT
The atmosphere compartment consists of rain gauge objects which can be used to
assign precipitation characteristics. The rain gages can be linked to sub catchments
to model precipitation characteristics for that area. The data can be entered in the
form of a time series or a data file containing precipitation data for the station.

25 | P a g e

RAIN GAGE OBJECTS

The rain fall attributes are assigned to a rain gage object. The volume and intensity
of the rainfall can be varied and different inundation patterns are expected for
different storm events.
High intensity high volume events are expected to lead to worse flooding as
compared to low intensity high volume events.
Low volume events are obviously expected to lead to lesser flooding as compared to
high volume rainfall events.

SURFACE COMPARTMENT
The surface compartment consists of catchments and sub catchments with various
parameters that are used to distribute the incoming precipitation into surface runoff
and losses, which include evaporation and infiltration.
SUBCATCHMENTS
A large number of parameters are assigned to subcatchments which affect the
surface routing of rain water and hence also affect flooding. The different
parameters are :
1.

Area

2. Width
3. Assigned rain gage
4. Slope
26 | P a g e

5. Percent imperviousness
6. Manning's coefficient for impervious areas
7. Mannings coefficient for pervious areas
8. Surface storage for impervious areas
9. Surface storage for pervious areas

GROUNDWATER COMPARTMENT
The ground water compartments simulates the storage of water below the surface
of the ground using aquifer objects.
INFILTRATION
Different formulae are used to estimate infiltration of precipitation to aquifers.

Horton

Modified Horton

Green Ampt

Curve Number

TRANSPORT COMPARTMENT
The transport compartment models the flow of water over the surface of the land
and through pipes.
NODES
The transport compartment uses nodes and links to simulate the routing of water
over the surface of the land and through the sewer systems. The outfalls and the
junctions are represented as nodes,
Junctions:

Inflow

Invert elevation

Maximum depth

Initial depth of water

Surcharge depth

Ponded area

27 | P a g e

Outfalls:

Inflow

Invert elevation

Type

LINKS
Conduits:

Inlet node

Outlet node

Shape

Maximum depth

Length

Roughness

Initial flow

Entry, average and exit losses

Inlet and outlet offsets

IMPLEMENTATION
The plan was to change the parameters and see the corresponding change in the
flooding maps. The individual parameters for different objects were changed in
SWMM and the simulation run. Using the results obtained from the rpt file and the
data from the inp file and the DEM of the area, inp.PINS delineated the flood extents
for the flood.
The actual results were totally awry from what I had expected and pretty much
similar results were obtained for nearly each and every simulation with weird
changes not corresponding to the expected change as per the change in
parameters. And the results from SWMM were pretty good, so, most probably the
problem lied somewhere with inpFLOOD.
So I began to check as to how inp.PINS derived the flood extents from the data.

28 | P a g e

INPFLOOD INVESTIGATION
The main idea of the tool is to create flood plains, by analyzing max Node Depth on
the rpt file combined with transect data, if it exists and DEM grid. In case the
transect data is not available, inpFLOOD creates the transects from the DEM.
The transect data for the city of Bellingham was not available and hence inpFLOOD
could only perform DEM analysis.

ALGORITHM

The distance between the conduits is divided into units of distance "di",
called as the minimum breaking link distance.

The water depth in each section is calculated by linearly interpolating the

hydraulic grade line(HGL) from the upstream to the downstream node.

Then the conduit transect for irregular conduit shapes or DEM grid at that
section is analyzed.

The transect is placed along the conduit for each "didistance and its
elevation is changed linearly by the conduit upstream and downstream invert
level.

If these two options are selected inpFlood will give the smallest flood plain
between transect and DEM result.

29 | P a g e

Conduits are analyzed only at di distances and DEM analysis will take place
to define flood plains for non irregular shape conduit.

The max DEM transverse distance is the distance is the perpendicular

distance from the conduit for which the DEM is analyzed for flooding.

Inlet OR Outlet Node HGL >= Inv. level + Max Depth - inpFLOOD restricts
building of flood plains only to conduit results that follow this condition. Flood
extents are seen to decrease considerably when this condition is applied.

Obviously decreasing the distance and increasing the transverse distance

increases the complexity of the problem and greater time is required for
simulation

MANUAL INCREASE OF WATER DEPTH

I wanted to check whether the inpFLOOD actually follows the algorithm

mentioned and hence I manually changed the water depths in conduits.

I found the flood map for a particular region using a particular rain condition.

Then I manually modified the water depths in each node by using Notepad+
+ and Excel and added 2 meters to each depth to see how the flooding map
would be influenced.

30 | P a g e

The results were as expected and the map showed considerably more flood
extents as compared to the original map.

ORIGINAL

EDITED

31 | P a g e

NO RAIN CONDITION
In order to check whether inpFLOOD gave sensible results for Bellingham city, I used
the storm water network to run a simulation for a no rain condition and then
checked the results for flooding.
Funnily, a large part of the city was shown to be flooded!
It turns out that in the event of zero depth of water for a particular node inpFLOOD
takes the invert level of the node as the hydraulic grade line.

On checking by using the field calculator, the invert level of the many nodes
was actually found to be higher than the elevation at that point derived from
the DEM.

This discrepancy is possibly because the data for the DEM and the sewer
networks and the DEM comes from different sources.

In such a scenario using the invert elevations from the sewer networks data
makes no sense as the results are bound to be incorrect.

So, I instead decided to assign the invert elevations from the DEM data.

DERIVING INVERT ELVEVATIONS FROM DEM

I found out the value of the DEM rasters at each particular node by using the
Hawths tools for sampling of DEM
32 | P a g e

Then taking the difference between the sewer network invert and rim
elevations, the average difference was found to be 2.92 m

So I subtracted 3 m from DEM elevations for all the nodes and assigned them
as the invert elevations for the nodes.

Using this data, I again found out the flood extents.

NO RAIN CONDITION FOR MODIFIED INVERTS

Again running a no rain simulation for the modified network with the DEM derived
network, a similar result was obtained with flooding extents.
For example, between the junctions
J485 & J1220, the value of the DEM is about 1.20 m at a point between the two
junctions.
But the invert elevations and hence the HGL for J1220 is 1.88 m and HGL for J485 is
1.68. Interpolation will give a water depth above 1.2 m, hence incorrectly showing
flooding in the region.

33 | P a g e

CONCLUSION

Changing the invert elevations by deriving them from the DEM, changing the
minimum breaking link distance and the transverse distance for transects
and modifying the various parameters associated with various elements in
SWMM still could not give sensible results with inpFLOOD.

Possibly the algorithm for inpFLOOD is way to simplistic to work in a widely

varying terrain such as Bellingham City with the elevations ranging from a
minimum of 1 m to a maximum of 260 m.

inpFLOOD is certainly worth a try in flatter regions but on most cities it is

unlikely it will give the desired results.

34 | P a g e

SWAT
Deliverables

Calculate Discharge over basin on South Atlantic Gulf region of US.

Calibration with historical record/gauge and its result.
Inundation maps and validation with historical maps.

SWAT is the acronym for Soil Water and Assessment Tool, a river basin, or
watershed, scale model developed by Dr. Jeff Arnold for the USDA Agricultural
Research Service (ARS). SWAT was developed to predict the impact of land
management practices on water, sediment and agricultural chemical yields in large
complex watersheds with varying soils, land use and management conditions over
long periods of time.
The SWAT model is:

Is physically based. Rather than incorporating regression equations to

describe the relationship between input and output variables, SWAT requires
specific information about weather, soil properties, and topography,
vegetation, and land management practices occurring in the watershed. The
physical processes associated with water movement, sediment movement,
crop growth, nutrient cycling, etc. are directly modeled by SWAT using this
input data.
Uses readily available inputs. While SWAT can be used to study more
specialized processes such as bacteria transport, the minimum data required
to make a run are commonly available from government agencies.
Is computationally efficient. Simulation of very large basins or a variety of
management strategies can be performed without excessive investment of
time or money.
Enables users to study long-term impacts. Many of the problems currently
addressed by users involve the gradual buildup of pollutants and the impact
on downstream water bodies. To study these types of problems, results are
needed from runs with output spanning several decades.

Thus SWAT is a complex integrated river basin scale model which operates either on
daily or hourly time step and quantifies the impact of land management practices in
hydrology, erosion and non-point source pollution.

35 | P a g e

Working Concept
Hydrologic cycle is simulated by SWAT model, which is based on the following
balance equation:

SWt = SWo + (Rday Qsurf - wseep Ea Qgw)

where SWt is the humidity of the soil (mm H2O), SW o is the base humidity of the soil
(mm H2O), t is time (days), Rday is rainfall volume (mm H2O), Qsurf is the value of
surface runoff (mm H2O), Ea is the value of evapotranspiration (mm H2O), w seep is
the value of seepage of water from soil into deeper layers (mm H2O) and Q gw is the
value of underground runoff (mm H2O). SWAT model uses the following climate and
hydrologic inputs: rainfall, air temperature and solar radiation, wind speed, relative
air humidity, snow pack, snowmelt, elevation zones, water volume on plants,
infiltration, water seepage into deeper soil layers, evapotranspiration, subsurface
flow, surface flow, lakes, river network, underground flow and other inputs related to
vegetation growth and development, erosion on the catchment area, nutrients,
pesticides and land use.

Overview of SWAT
SWAT allows a number of different physical processes to be simulated in a
watershed. For modeling purposes, a watershed may be partitioned into a number
of sub watersheds or sub basins. The use of sub basins in a simulation is particularly
beneficial when different areas of the watershed are dominated by land uses or soils
dissimilar enough in properties to impact hydrology. By partitioning the watershed
into sub basins, the user is able to reference different areas of the watershed to one
another spatially.
SWAT comes with the option of Automatic Watershed Delineation which allows
delineating stream networks for a Digital Elevation Model data. Instead of
performing different operations individually as done in ArcGIS, it groups all the
options in a single task.
The tools functions are divided into five sections, namely: DEM setup, Stream
Definition, Outlet and Inlet Definition, Watershed Outlet(s) Selection and Definition.
This tool is used to create watershed delineations using a combination of DEM,
digitized network and other user inputs.

36 | P a g e

This is how the dialog box for Automatic Watershed Dileation looks like.
First of all we need to upload DEM on which the analysis has to be carried out.

37 | P a g e

After uploading the DEM, the Flow Direction and Flow Accumulation Raster are
created which gives the total area in acres by counting the number of cells and the
resolution of each cell. The projection system of the DEM also needs to be defined.
The DEM used by me had a resolution of 30m and the projection system used was
NAD_1983_ALBERS.
The other projection systems used are UTM
After creation of the FDR and FAC rasters, the next step is to create streams and
outlets. Streams and outlets are created based on the FAC rasters.

Here blue lines represent the stream network.

38 | P a g e

By specifying the threshold limit for Flow Accumulation rasters, the stream networks
can be selected based on the need of user.
I was supposed to work on South Atlantic Gulf Basin area and so the delineated
network for the area looked like this:

Catchment boundaries along with longest reach path and nodes are created.
After the creation of stream networks, inlets and outlets can be added manually in
the DEM based on the need of the user.
After the inlet and outlet definition, all the nodes are selected and based on those
nodes different watersheds are created.

Total 19 watersheds were created.

39 | P a g e

Calculation of Sub basin Parameters

Now the sub basin parameters are calculated.
It gives the length and the area of each watershed:

It also gives the reach and length of each watersheds and also the respective inlets
and outlets and also the X and Y co-ordinates of each node:

40 | P a g e

Similarly other data such as shape index, etc. is also calculated.

Now the watershed delineation is complete.
At the time of calculating sub basin parameters, we can also check whether longest
flow path calculation and geometry check are to be performed. It saves the
compilation time depending on the need of the user.

HRU Analysis
HRU stands for Hydrologic Response Units. It is the basic modeling unit-defined as
the network of elementary hydrologic areas with the selected discretization (grid
cell, representative hill slope, sub watershed), measure of which is dependent upon
the desired accuracy, as well as upon data accuracy.
Discretization techniques are classified as:
Sub Watershed: It divides the watershed into subbasins based on topographic
features of the watershed. It preserves the natural flow path, boundaries, and
channels required for realistic routing of water, sediment and chemicals.
Grid Cell: The routing methodology is same as that of sub watershed with the
difference being average size of the subbasin and method used to define
subbasin boundaries. It allows user to capture high level of heterogeneity or
variability in the simulation.
Representative hill slope: It allows overland flow from one subbasin to flow
onto the land area of another subbasin. It allows SWAT to model hillslope
processes.
HRUs are lumped land areas within the subbasin that are comprised of unique land
cover, soil, and management combinations.
41 | P a g e

HRU Definition requires assigning Land Use Cover, Soil Cover and Slope Cover data
for the concerned area.
The number of HRUs created are generally much more than the number of sub
basins.

Land Use, Soil and Slope Definition

The Land Use, Soil and Slope Definition option in the HRU Analysis menu
allows the user to specify the land use, soil and slope themes that will be used for
modeling using SWAT. These themes are then used to determine the hydrologic
response unit (HRU) distribution in each sub-watershed.
SWAT require land use data to determine the area of each land category to be
simulated within each subbasin. In addition to land use information, SWAT relies
on soil data to determine the range of hydrologic characteristics found within each
subbasin. Land Use, Soil and Slope Definition option guides the user through
the process of specifying the data to be used in the simulation and of ensuring that
those data are in the appropriate format. In particular, the option allows the user to
select land use or soil data that are in either shape or grid format. Shapefiles are
automatically converted to grid, the format required by ArcGIS to calculate land use
and soil distributions within the sub basins of interest.

Sources Of Data Sets:

42 | P a g e

USGS-US Geological Survey: classifies as 4 major levels, namely regions (21),

sub regions (221), accounting units and cataloging units.
NRCS-National Resource Conservation System
NLCD-National Land Cover Database: 16 class land cover classification
NOAA- National Oceanic and Atmospheric Administration

These are some of the agencies which provide datasets required for HRU Definition.

Application Of Different Datasets

LULC Map:
The Land Use and Land Cover (LULC) data files describe the vegetation, water,
natural surface, and cultural features on the land surface. The United States
Geological Survey (USGS) provides these data sets and associated maps as a part of
its National Mapping Program. The LULC mapping program is designed so that
standard topographic maps of a scale of 1:250,000 can be used for compilation and
organization of the land use and land cover data. In some cases, such as Hawaii,
1:100,000 scale maps are also used.

Slope Map:
Slope Map cannot be applied directly, rather slope classification map needs to be
created separately and the slope classes for HRU Definition are to determined
accordingly.
SWAT allows to classify slopes in 1 to 5 groups with the highest limit for the last
class being 9999.

43 | P a g e

Soil Map:
Soil Map is inbuilt in SWAT from US STASGO and can be used directly from there or it
can be uploaded manually.
This data set by STATSGO is a digital general soil association map developed by the
National Cooperative Soil Survey and distributed by the Natural Resources
Conservation Service of the U.S. Department of Agriculture. It consists of a broad
based inventory of soils and nonsoil areas that occur in a repeatable pattern on the
landscape and that can be cartographically shown at the scale mapped. The soil
maps for STATSGO are compiled by generalizing more detailed soil survey maps.
Where more detailed soil survey maps are not available, data on geology,
topography, vegetation, and climate are assembled, together with Land Remote
Sensing Satellite (LANDSAT) images.
Map unit composition for a STATSGO map is determined by transecting or sampling
areas on the more detailed maps and expanding the data statistically to
characterize the whole map unit.

44 | P a g e

Inbuilt Soil map

Manual Assignment of Soil Class:

Final Land Use, Soils, Slope Distribution Reports

HRU Definition
Before executing SWAT, the distribution of hydrologic response units (HRUs) within
the watershed must be determined based on the land use, soil and slope layers
specified in the previous step. The interface allows the user to specify criteria to be
45 | P a g e

used in determining the HRU distribution. One or more unique land use/soil/slope
combination or HRUs can be created for each sub basin. Subdividing the watershed
into areas having unique land use, soil and slope combinations enables the model to
reflect differences in evapotranspiration for various crops and soils. Runoff is
predicted separately for each HRU and routed to obtain the total runoff for the
watershed. This increases accuracy and gives a much better physical description of
the water balance.

HRU Thresholds:
This option allows to set threshold limits for Land Use, Soil and Slope Class
percentages. Based on the threshold limit, the classes with lower percentages can
be grouped together with a class having higher percentage.

The Dominant Land Use and Soil option will you allow to create only one HRU
for each sub basin defining the dominant landuse and soil class.
The Multiple Hydrologic Response Units option will allow you to create
multiple HRUs within each sub basin.

1. Land Use Percentage

2. Slope Class Percentage
3. Soil Class Percentage

For the analysis purposes I have considered a threshold limit of 10% for Land use,
15% for Soil class and 10% again for slope class.

46 | P a g e

Land Use Refinement

Land Use Refinement tab is used to specify more detailed criteria.
1. Split one land use type into two or several sub land use types.
2. Set one land use type exempt.

Final HRU Distribution Report

47 | P a g e

Writing Input Tables

This menu contains functions to build database files that include information
needed to generate default input for the SWAT model.
The commands on the menu need to be implemented only once for a project.
However, if the user modifies the HRU distribution after building the input database
files, these commands must be reprocessed again.

48 | P a g e

This dialog will allow the user to define the input data for rainfall, temperature and
other weather data. For weather data, you have the option of simulating the data in
the model or to read from data tables. If no observed weather data is available,
then information can be simulated using a weather generator. The weather
generator data must be defined before you can continue to define the other data,
like precipitation and temperature.
At this point you have the option to generate all the input data files using the
WRITE ALL option under the INPUT menu or generate each input file separately.
The input files needed are:

Watershed Configuration file (.fig)

Soil data (.sol)
Weather generator data (.wgn)
General HRU data (.hru)
Soil chemical input (.chm)
Stream water quality input (.swq)
Pond input (.pnd)
Management Input (.mgt)
Main channel data (.rte)
Ground water data (.gw)
Water use data (.wus)

For sub basin and main channel input files a prompt asks whether to change the
default Mannings n value. Click No to use 0.014. For the management input file
generation a message prompt verifies if the US weather database is sufficient to
estimate the Plant Heat Units. Click Yes if the study area is within US.
49 | P a g e

Edit SWAT Input

The commands listed under the Edit SWAT Input menu bring up dialog boxes that
allow you to alter default SWAT input data.
The Edit SWAT Input menu can be used to make input modifications during the
model calibration process.

By using a procedure similar to editing soils database, the Database option under
the Edit Input menu you can edit or add information to the weather, land
cover/plant growth, fertilizer, pesticide, tillage, and urban area databases.

50 | P a g e

Edit Point Discharge Inputs

The dialog box allows the input of point source data in one of four formats: constant
daily loadings, average annual loadings, average monthly loadings, and daily
loadings.

Edit Inlet Dischargers Input

The dialog box allows the input of inlet discharge data in one of four formats:
constant daily loadings, average annual loadings, average monthly loadings, and
daily loadings.

Edit Sub basins Data

This dialog box contains the list of sub basins, land uses, soil types and slope levels
within each sub basin and the input files corresponding to each sub basin/land
use/soil/slope combination. To select an input file, select the sub basin, land use, soil
type and slope that you would like to edit. When you select a sub basin, the combo
box of land uses, soil types, and slope levels will be activated in sequence. Specify
the sub basin/land use/soil combination of interest by selecting each category in the
combo box.
Similarly user can edit:

Soil physical data(.sol)-edit SNAME(soil name), NLAYERS(no. of layers),

HYDGRP(1-4 where 1 is well drained and 4 is poor drained),
ANION_EXCL(fraction of porosity from which anions are excluded),etc.
Weather generator data(.wgn)-WLATITUDE and WLONGITUDE( latitude and
longitude of weather station used to create statistical parameters, etc.

51 | P a g e

General HRU data(.hru)-HRU_FR(fraction of subbasin area contained in HRU),

SLSUBBSN-avg. slope length, HRU_SLP(Avg. slope steepness), etc.
Main channel input file(.rte)- CH_W(Avg. width of channel at top of tank),
CH_D(depth of main channel from top of tank to bottom), ICANAL(code for
irrigation canal), etc.

Similarly other input parameter files can be edited.

SWAT Simulation Setup

SWAT simulation menu contains commands that setup and run SWAT simulation.

The above dialog box lets you specify duration of simulation, type of rainfall
whether skewed normal or mixed exponential, whether the results should be printed
daily, weekly or monthly. It also lets you specify NYSKIP which is the warm up
period, that is the time taken for simulation to get adjusted to surroundings and the
results are printed after that particular warm up period.

Output Analysis
After setting up the SWAT simulation, the simulation is saved and the result files can
be imported to the folder Scenarios. The output files can be found inside TxtInOut
folder.

52 | P a g e

Simulated

Observed

The above graph shows the relation between the observed and simulated discharge
over a period of 10 years. As we can see that there is considerable difference
between the two results and hence the simulation needs calibration of different
parameters to match the results.

Utility Network Analyst Tool

As a part of my internship project, I was supposed to find a tool or extension in
ArcGIS using which the direction of stream or road (basically line) network can be
manually assigned as per the need of the user. To perform analysis of line networks,
Networlk Analyst extension and Utility Network Analyst tool were available but for
flow network data, Utility Network Analyst tool seemed better and more reasonable
choice. Here is an overview of Utility Network Analyst tool:
The Utility Network Analyst toolbar provides the interface for performing
analyses on geometric networks. The Utility Network Analyst toolbar in ArcMap
allows you to choose a geometric network, set and view flow direction, change
analysis settings, add flags and barriers to analyses, and perform various trace
tasks.
Two types of networks

Geometric datasets: River networks and utility networkslike electrical,

gas, sewer, and water linesallow travel on edges in only one direction at a
time. The agent in the networkfor instance, the oil flowing in a pipeline
can't choose which direction to travel; rather, the path it takes is determined
by external forces: gravity, electromagnetism, water pressure, and so on. An
engineer can control the flow of the agent by controlling how external forces
act on the agent.

Network datasets (transportation networks): Transportation networks

like street, pedestrian, and railroad networkscan allow travel on edges in
both directions. The agent on the networkfor instance, a truck driver
53 | P a g e

traveling on roadsis generally free to decide the direction of traversal as

well as the destination.

Geometric networks comprise two main elements: edges and junctions.

Edges An edge is a feature that has a length through which some commodity
flows. Edges are created from line feature classes in a feature dataset and
correspond to edge elements in a logical network.
Examples of edges: Water mains, electrical transmission lines, gas pipelines, and
telephone lines.
Junctions A junction is a feature that allows two or more edges to connect and
facilitates the transfer of flow between edges. Junctions are created from point
feature classes in a feature dataset and correspond to junction elements in the
logical network.
Examples of junctions: Fuses, switches, service taps, and valves.
Edges and junctions in a network are topologically connected to each otheredges
must connect to other edges at junctions; the flow from edges in the network is
transferred to other edges through junctions.

Creating Flow Direction for sewer network

Create Flow Direction is used to assign flow direction to a geometric network.
Geometric network along with flow option is taken as input. Flow Option indicates
method by which flow direction will be established; there is no default value.
There are four Flow Options namely:
WITH_DIGITIZED_DIRECTION Establish flow direction along the digitized
direction of edges.
AGAINST_DIGITIZED_DIRECTION Establish flow direction against the
digitized direction of edges.
WITH_SOURCES_SINKS Establish flow direction using sources and sinks.
RESET_FLOW_DIRECTION Reset flow direction on all edges to be
uninitialized.
54 | P a g e

Using Utility Network Analyst tool we can trace network elements upstream or
downstream from a point which can be used to determine which valves to shut off
when a pipe bursts.

Manually Assigning Direction:

To manually assign direction to a particular network Start Editing the layer which
contains source/sink features using Editor Tool. Select a particular junction point
according to which you want to modify your network. Click on the attributes table
button and go to Ancillary tab and from the drop down menu select source/sink as
per your needs. After that click on Set Flow Direction icon in Utility Network
Analyst toolbar. The flow direction for the entire network gets modified according
to that particular junction point. Save the edits.
The highlighted blue point serves as a sink for this flow direction.

Here the highlighted blue point serves as a source. Notice the change in flow
direction.

55 | P a g e

Types of flow directions: Depending upon the flow options, there are three types
of flow directions, namely:

Determinate flow direction: If the flow direction of an edge can be

uniquely determined from the connectivity of the network, the locations of
sources and sinks, and the enabled or disabled states of features, the feature
is said to have determinate flow. Determinate flow for an edge is specified as
either with or against the direction in which the feature was digitized.
Indeterminate flow direction: Indeterminate flow in a network occurs
when the flow direction cannot be uniquely determined from the topology of
the network, the locations of sources and sinks, or the enabled or disabled
states of the features. Indeterminate flow commonly occurs for edges that
form part of a loop, or closed circuit. It can also occur for an edge whose flow
is determined by multiple sources and sinks, where one source or sink is
driving the flow in one direction through the edge, but another source or sink
is driving it in the opposite direction.

56 | P a g e

For example, consider a geometric network with the sources and sinks
positioned like this:

In this case, the flow direction for edges 1 and 2 is set; however, edge 3 has
indeterminate flow. To understand why edge 3 has indeterminate flow, consider
the case where only the source is present.

This results in a flow direction of edge 3 to the right.

Now, consider the case where only the sink is present.

This results in a flow direction of edge 3 to the left. Due to the opposite potential
flow directions of edge 3, this results in a conflict.
For each edge, if the flow direction is in agreement between both the sourceonly and sink-only cases, the flow direction is set to that direction (as seen with
edges 1 and 2). However, if there is a conflict, as there is with edge 3, the flow
direction is set to indeterminate since there are two possible outcomes.
Another example resulting in indeterminate flow would be if an edge has a
source at both of its ends.
57 | P a g e

Uninitialized flow direction

Uninitialized flow direction in a network occurs in edges that are isolated from
the sources and sinks in the network. This can happen if the edge is not
topologically connected through the network to the sources and sinks or if the
edge is only connected to sources and sinks through disabled features.

Trace Options:
1. All network elements that lie upstream of a given point in your network

(traceupstream)

2. All network elements that lie downstream of a given point in your

network(trace downstream)

58 | P a g e

3. The total cost of all network elements that lie upstream of a given point in
your network (upstream accumulation)
4. An upstream path from a point in your network (find path upstream)
5. The common features that are upstream of a set of points in your network
(find common ancestors)

6. All the features that are connected to a given point through your network
(find connected)

59 | P a g e

7. All the features that are not connected to a given point through your network
(find disconnected)
8. Loops that can result in multiple paths between points in a network (find
loops)

9. A path between two points in the network (The path found can be just one of
a number of paths between these two points depending on whether or not
your network contains loops)(find path)

60 | P a g e

Exposure Development:
Background
Buildings or establishments are constantly under a risk of being razed down by
some catastrophe or the other. This directly results in a monetary risk to the owners
(either individual or a group) of that establishment. In order to reduce the burden on
these owners, the insurance companies agree to share the risk in return of a
comparatively small premium paid to the company. These companies rely heavily
on probability skills in order to evaluate the gain they might have in the entire
process. But, as history clearly demonstrates, there have been certain catastrophes
which have rendered the insurance companies bankrupt.
In order to reduce the risk on the insurance companies, and indirectly help
humanity, Risk Management Solutions (RMS) has been involved in developing
various methods of evaluating the loss these companies might have to bear in case
a catastrophe strikes a particular region insured by them. One of them is exposure
development.

Theory
Exposure can be defined as the amount of rebuild cost of all the buildings in a
region. The basic idea is to calculate the total amount of construction cost of all the
buildings in a region. The rebuild cost, which is the cost required to rebuild an
establishment once it is razed down in some calamity, is directly related to (some
fraction of) this total construction cost.

61 | P a g e

This entire procedure is an arduous one. It starts with the search for the correct data
sets which can be used in the development followed with formulation of a method
for exposure calculations and culminates in an exposure value.
The basic data required in the exposure calculation is the number of buildings or
establishments, floor area of each establishment and the construction cost per unit
area of the establishment. Combining them using correct methodology gives us the
exposure.

PARAMETERS
Building Inventory
It is the total number of buildings or establishments in a geographically well-defined
area.

Size of building
The average size of building, usually expressed in terms of the total floor area of the
buildings.

Replacement cost value

The Construction cost, or the present day replacement cost of the building,
expressed as construction cost per unit floor area.

62 | P a g e

EXPOSURE CALCULATION
If data that contains all the three parameters in the specified manner, then the
exposure calculation is easy and goes about in the following manner

EXPOSURE

TOTAL
=
FLOOR AREA
CONSTRUCTION
COST
*

Since, every country does not provide us the data in this form, hence each of them
require a different procedure to finally yield the above mentioned data.
On developing the exposure for two countries, South Korea and Vietnam, a direct
correlation was observed in between the exposure values and the economic
parameters like GDP per capita, which is what the intuition says. Similarly, the more
densely populated regions had more exposure values.
This exercise, at times, need elaborate calculation. For example, when the data for
construction cost is comprehensive then the cost per unit area is calculated using
the cost of items involved in construction (eg. Cement, steel bars, etc). This is quite
similar to the task of building materials estimation done by civil engineers.

EXPOSURE DOWNSCALING
The exposure calculated using the above mentioned steps, is often obtained at a
very coarse resolution, most often at country or county level. This exposure model
often needs to be downscaled to a finer resolution in order to obtain catastrophe
loss at a fine and detailed resolution. Thus the overall aim of downscaling is to
distribute the total building cost to a smaller resolution, such as municipality level or
a 50m grid level.

OBJECTIVE
The objective was to downscale exposure in Belgium to a 100m grid level

METHODOLOGY
In order to downscale Belgium exposure, parameters were to be selected that could
be used as predictors for assigning exposure at 100m grid level. By doing some
online research, I could come up with 3 such parameters that could be used as tools
for downscaling.

63 | P a g e

Urban Atlas Classification

The European Urban Atlas is part of the local component of the GMES/Copernicus
land monitoring services. It provides reliable, inter-comparable, high-resolution land
use maps for 305 Large Urban Zones and their surroundings (more than 100.000
inhabitants as defined by the Urban Audit) for the reference year 2006. The GIS
data can be downloaded together with a map for each urban area covered and a
report with the metadata.

Land Use/ Land Cover Classification

Land Use/Land Cover data refers to data that is a result of classifying raw satellite
data into "land use and land cover" (lulc) categories based on the return value of
the satellite image. There are not very many lulc datasets because a) satellite data
acquisition is usually very expensive, and b)the classification process is very labor
intensive. Most lulc data products are released several years after the satellite
images were taken, and thus out of date to a certain extent when they are released.
Nonetheless, lulc provides a very valuable method for determining the extents of
various land uses and cover types, such as urban, forested, shrubland, agriculture,
etc. Land use/land cover data are most commonly in a raster or grid data structure,
with each cell having a value that corresponds to a certain classification.

64 | P a g e

Soil sealing data

Soil Sealing is the loss of soil resources due to the covering of land for housing,
roads or other construction work. The covering of the soil surface with impervious
materials as a result of urban development and infrastructure construction is known
as soil sealing. The term is also used to describe a change in the nature of the soil
leading to impermeability (e.g. compaction by agricultural machinery). Sealed areas
are lost to uses such as agriculture or forestry while the ecological soil functions are
severely impaired or even prevented (e.g. soil working as a buffer and filter system
or as a carbon sink). In addition, surrounding soils may be influenced by change in
water flow patterns or the fragmentation of habitats. Current studies suggest that
soil sealing is nearly irreversible.

Open Street Maps (OSM) Data

OpenStreetMap (OSM) is a collaborative project to create a free editable map of the
world. Two major driving forces behind the establishment and growth of OSM have
been restrictions on use or availability of map information across much of the world
and the advent of inexpensive portable satellite navigation devices.

65 | P a g e

PROCESS

First Step: Downloaded the Open Street Map data set for Hungary containing
a whole inventory of more than 120,000 buildings classified into various
types by crowd-sourcing of data.
Created a mapping between the OSM classes and the required Lines of
business (Five in number) to classify the buildings in OSM dataset into
required LOBs.
Overlayed the LULC, UA and SS raster layers on the map and spatially joined
these classes to the OSM buildings dataset, to obtain a database which had,
for each building, all the three classes and the assigned LOB.
Divided the Open Street Map dataset for Hungary into two parts, one for
carrying out a logistic regression, and the other smaller dataset for checking
the regression results.
Ran a logistic regression between assigned LOBs and the LULC, SS and UA
classes to obtain probabilities of a building lying in each combination of the
three classes to fall in a particular LOB.
Carried out a logistic regression to assign probabilities of LOB occurrence for
each combination of the three classes .
Applied the regression results on the test sample to check whether assigned
LOBs and predicted LOBs match.

RESULT

66 | P a g e

CONCLUSION
As can be seen from the above table, the mapping between the assigned and
predicted LOBs can be seen, which suggests that only for SFD(Single family dwelling
buildings), the results were very good Almost 94% of the buildings in SFD were
predicted as SFD only. But in the case of other LOBs, the percentage of buildings
mapped into the correct LOB was poor(less than 30%). After discussing with senior
members of the team, we came to the conclusion that these classes(LULC, SS, UA)
were not good enough to be used as predictors for exposure downscaling.

K-MEANS CLUSTERING FOR IDENTIFYING

VEGETATION IN IMAGES
INTRODUCTION
K-means clustering is a method of vector quantization, originally from signal
processing, that is popular for cluster analysis in data mining. k-means clustering
aims to partition n observations into k clusters in which each observation belongs to
the cluster with the nearest mean, serving as a prototype of the cluster. This results
in a partitioning of the data space into Voronoi cells.
Given a set of observations (x1, x2, , xn), where each observation is a ddimensional real vector, k-means clustering aims to partition the n observations
into k sets (k n) S = {S1, S2, , Sk} so as to minimize the within-cluster sum of
squares (WCSS):

where i is the mean of points in Si.

In this manner, K-means clustering aims to partition n observations into k clusters in

which each observation belongs to a clusters whose mean has the smallest
euclidian distance to the observation.
67 | P a g e

OBJECTIVE
The aim was to develop a machine learning algorithm that could be used to
separate greenery from satellite real color images. This could be really helpful
especially in exposure downscaling as there would be no buildings in green parts of
an image. For this purpose, K-means clustering algorithm was adopted to filter
green pixels from an image.

PROCESS
A program was written which generates a sample of random points on a 2-d plane,
then obtains k values of cluster means and then classifies the data points into
clusters.

These points were divided into 7 clusters, and were color-coded according to their
clusters, so that if the color seperation is good enough, it can be implied that the
clustering algorithm was successful in classifying into the different clusters.

Trying the similar clustering on the pixels of an Image containing greenery, and then
showing the pixels that fell in the green cluster, the results were as follows:

68 | P a g e

Original Image

69 | P a g e

Now trying this method on a map that contains greenery, the result was as follows.

All the green pixels were apparently not separated, but this was surely a motivation
to continue further work on this method, which due to time limits of my vacations I
could not carry out.

70 | P a g e

AGRICULTURAL EXPOSURE DEVELOPMENT :

BELGIUM

Objective was to develop the exposure at country level for Belgium for Agricultural
line of business. i.e. Total building cost of all the agricultural buildings in the whole
of country. By doing some online research, I came to know that the buildings in
agricultural LOB contains mostly farm buildings such as small to large warehouses.

Base data summary

This is a summary of the data that was available from the websites of statistical
offices of Belgium, and of other international organizations which collect building
inventory data.

BUILDING INVENTORY AND FLOOR AREA

Steps
Overall aim was to obtain the total floor area at province level. For this the floor
area had also to be trended to 2014 vintage using the number of farms growth rate
data.

71 | P a g e

Total Built-up Area Data:

Since the total floor area data available was of vintage 2013, it had to be trended to
2015 vintage. For this the data containing number of farms at country level was
used, and the average annual growth rate was applied to trend the area to 2015
vintage.
Variation of number of farms in Belgium from 1997-2007
72 | P a g e

Trended total floor area : Province level

73 | P a g e

Construction costs
Steps

Given the steps, following is the summary of building cost per unit floor area that
was available from various online sources.

Construction costs : Belgium Agricultural LOB

As is apparent from the above graph, the construction cost are more or less similar
from all the three sources. I used the BKI data from the above three because in BKI,
type of buildings were specifically mentioned as agricultural buildings.
Sources : BKI 2013 , rlb.com, www.constructioncosts.eu

74 | P a g e

Trending and Currency conversion

Construction Cost Index
Construction Cost Index is an indicator of the average cost movement over time of a
fixed basket of representative goods and services related to Construction Industry.
CCI data is available for various countries from the Census websites.

CCI
CCI

Since the construction costs that were obtained from BKI were of vintage 2011, they
had to be trended to 2015, using CCI data available. Average growth rate of CCI was
used to trend these costs to 2015 vintage.
Now since the exposure values were to be obtained in USD, the costs were
converted from Euro to USD using the conversion rates in year 2011 available
online.

Euro to USD factors 2011

Euro to USD factors 2011

75 | P a g e

Using this data, the construction costs were converted to be used for 2015Q1.
Source : http://statbel.fgov.be/en/statistics/figures/economy/indicators/prix_prod_con/

Exposure calculations
Now having obtained the sufficient data for the purpose, the exposure calculation
was a simple step, just multiplying the total floor area and construction costs per
unit floor area.

EXPOSURE

= TOTAL FLOOR AREA * CONSTRUCTION COST

RESULT

PROVINCE LEVEL EXPOSURE

PROVINCE LEVEL EXPOSURE

ArcGIS Work

Calculation of distribution factors : Distribution factors means the fraction of a

particular building footprint that lies within a particular grid polygon, in my
case it was 50m grid level. For this the union and spatial join functions were
used.

Finding no. of buildings at county level : By finding the centroid of the

building footprint polygon, the number of buildings were assigned to different
counties.
76 | P a g e

Finding no. of buildings at metropolitan level : The method adopted was

same as above, except it was for municipality level.

Key Takeaways:

Exposure modelling concepts

Statistical concepts

Excel Skills

ArcGIS Skills

Coding in R

REFRENCES:

Influences of Potential Evapotranspiration Estimation Methods on SWATs

Hydrological Simulation In A Northwestern Minnesota Watershed by X. Wang,
A. M. Melesse, W. Yang.
SWAT-Based Runoff Modeling in Complex Catchment Areas Theoretical
Background and Numerical Procedures by Z. Simi, N. Milivojevi, D.
Prodanovi, V. Milivojevi, N. Perovi.
web.ics.purdue.edu/~vmerwade/education/arcswat.pdf- SWAT tutorial
prepared by Venkatesh Merwade and Adnan Rajib ,School of Civil
Engineering, Purdue University

77 | P a g e