Some of Dr. Nishant Sinha's Research Papers

Volume 2, Issue 3, March 2012 ISSN: 2277 128X

International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com

Development of SDSS for Ensuring Insurers Nishant

* Neena Priyanka

P. K. Joshi

Natural Resources, TERI University Natural Resources, TERI University Natural Resources, TERI University

Pitney Bowes Software, NOIDA Pitney Bowes Software, NOIDA

[email protected]

Abstract— In arena of catastrophe management in India, managing risk at varied levels along with timely and effective decision making

by Insures/Reinsurers is a complex task. This unique dynamic system makes the assessment and management of enterprise-wide risk

much more multidimensional and uncertain resulting in failure of connection between lines of business. Geospatial technology viz. remote

sensing, GIS and SDSS has emerged as powerful aid to assist risk managers and decision makers to manage risk for several years.

However, if used alone, it has limited functionality. This paper presents the conceptual design and development of remote sensing and

GIS-assisted Spatial Decision Support System (SDSS) to improve property insurance underwritings that involves procedural and

declarative knowledge. SDSS, coined as Insurance Profiler (InsPro), integrates geocoder, multi-criteria risk evaluation techniques and

state-of-art web interface framework which is applied at three phases viz. geospatial visualization and querying of insured points, multi-

criteria comprehensive evaluation of risk and report generation. It is flexible in that it can be adapted in evaluation of any property type. It

is scalable because the system can be designed at local, regional, national or international level as being data driven .The system is

integrative because it incorporates a number of different data types and sources (e.g., multispectral remote sensor data, numerous

thematic information on hazard and vulnerability), and geo-statistical tools and techniques, and human expert knowledge of the seismic

region. The system is designed to be flexible, scalable and integrative. Thus, this SDSS tends to cater the needs of users at all levels viz.

risk analyst, insurer, brokers, reinsurers etc. to manage share and interact effectively and reliably.

Keywords— Catastrophe, SDSS, Insurers, GIS, Real Estates, InsPro

I. INTRODUCTION

Risk analysis is a complex task that entails consideration of

complex parameters which are difficult to interpret and

quantify ([1]–[3]). In addition, risk analysis involves a

comprehensive database to model uncertainty and vagueness.

As a consequence, insurers/reinsurers fail to evaluate and

underwrite actual risk. In addition, there are other

shortcomings, such as poor visualization of insured points and

risk zones ([4]–[7]) slow model based update of information

that further contributes to complexity and underestimation of

potential loss from natural hazards and even failure and

insolvency of some insurance companies.

Fig. 1 Country-wise total natural disaster events: 1976-2005 (Source: EM-DAT)

The catastrophe imposed risk in India can be described as

worst as being high on number of events and intensity as

depicted in Figure 1, owing to an elevated probability of

hazard occurrences and high exposure due to geographical,

topographical and socio-economic settings [8]. This trend is

expected to continue as higher concentration of populations

and built-ups continue to develop in areas susceptible to

natural hazards. India’s vulnerability to natural catastrophes

coupled with rapid growth and transformation of the insurance

market, it is crucial to address this high level vulnerability in

order to avoid the present scale of losses and damage. Despite

leveraging such transfer of risk through integrated product

choices and schemes, there are very limited sections of

population (0.5%) in India those have any kind of property

insurance [9].

There are various other inadequacies such as poor location

identification of insured exposures on paper maps, primitive

modelling assumptions and slow update of information that

add to complexity of insurers/reinsurers. Such limitations aid

to underestimation of severe nature of disaster and associated

potential loss resulting in unexpected significant drop in

surplus and bankruptcy of some insurance companies. Beside

these, there are other reasons which could be attributed for

such low profiling. This include a general lack of awareness

about insurance practices, two-dimensional nature of

spreadsheets and reports which requires skill set for

understanding, lack of spatial database that could provide easy

visualization and data querying, absence of scientifically

designed enterprise solutions focused for insurance

underwriters to promote faster and effective decision making.

Against the above-stated deficiencies of current systems,

http://www.ijarcsse.com/

Volume 2, Issue 3, March 2012 www.ijarcsse.com

© 2012, IJARCSSE All Rights Reserved Page | 332

adoption of geospatial technology for niche areas such as

actuarial underwritings, claims management, risk based

pricing, could be very useful as much of the data required

within these domains contain geographic component ([11]–

[14]).

II. GEOSPATIAL TOOL & TECHNOLOGY IN INSURANCE

Remote Sensing (RS) and GIS together have emerged as

useful tool for insurers/reinsurers because of spatio-temporal

component involved within [15] and its ability to integrate

large volume of information through repertoire of analytical

tools for disaster risk management ([16], [17]).The system is

further aided by development of modelling approaches such as

catastrophe models with basic components including hazard,

exposure, vulnerability, and loss ([18], [1]). The derived

models tend to quantify the likelihood of disasters occurring

and estimate the extent of incurred losses, both from single

event and multiple events and eventually help in development

of spatial decision support system (SDSS). In the basic

framework of risk management, a combination of RS, GIS

and SDSS; RS & GIS can be used in potential hazard zonation,

inventory preparation, whereas SDSS for vulnerability

assessment, loss estimation and in decision processes by key

stakeholders [16]. Insurance companies are increasingly using

SDSS as an essential business tool [15] to mitigate exposure

to risk by ensuring a wide spatial distribution of policyholders.

III. MULTI-CRITERIA DECISION BASED RISK ASSESSMENT

Solving problems and taking rational decisions in a

complex domain such as risk assessment needs integration of

information, knowledge and expertise from a wide range of

disciplines. It also needs some kind of support mechanism (i.e.

tools) that can assist planners and decision makers in informed

and rational decision making. Risk assessment being a

problem of multiple dimensions; involving multiple criteria,

conflicting objectives, and its planning is considered as a

multi-criteria decision making (MCDM) problem that needs

specialized tools and techniques that can support a systematic

approach of decision analysis. MCDM is characterized by the

need to evaluate a finite set of alternatives on the basis of

conflicting and incommensurable criteria of quantitative,

qualitative or both in nature and based on preference values of

the alternatives on permissible scale measure the overall

preference values ([19], [20]). For this reason, there has been

a growing interest in applying GIS and spatial MCDM to risk

analysis which is very much evidenced by an increasing

number of published articles on this topic. Entrenched in a

GIS milieu, MCDM technique provide the framework of a

SDSS which improves the effectiveness of decision making

process by incorporating decision maker’s judgments and

computer based programs ([19]-[22]). In the domain of risk

planning, MCDM approach is considered essential because of

its demonstrated ability to integrate multiple criteria,

preferences of different groups, expert’s knowledge, and with-

standing spatial; non-spatial and inexplicit data from various

sources. The most significant characteristics of this

methodology are that they are transparent to the participants.

Such methodologies make it possible to integrate risk

assessment information in knowledge structures and networks,

and opens prospects for improved risk mitigation and planning

to investigate a number of multiple objectives (criteria).

With this backdrop, it is obvious that the deductive, well-

structured problem-solving methodologies are inadequate

when it comes to the analysis of urban area risk assessment as

there are multiple representations or understandings on this

concept. Therefore, identifying an appropriate design structure

for assessment procedure among competing options is perhaps

the most important part of analysis. The design must

recognize divergent perspectives of urban morphology and

hence associated risk. In this paper, we deduce that one of the

useful alternatives to design risk assessment procedure is to

adopt an inductive approach based on spatial MCDM. We

have chosen earthquakes as a subject of this research not only

because of their severe impacts on urban area, but also

because they have provided the basis for some of the

fundamental physical, technological and social research in

field of natural hazards: work that has often been a model for

studies of other hazardous natural agents. The objectives

formulated for current study focuses on development of a

geospatial and web analytics based actuarial solution for

insurers/reinsurers which would minimize uncertainty and

cater to their needs for profiling overall scenario of property

risk.

IV. RESEARCH NEEDS

A small region of capital city Delhi, India is taken up to

demonstrate this concept of risk assessment using web based

solutions. The study area is characterized as susceptible to

earthquake and as majority of the population dwell in urban

areas and even the slightest structural and physical damages

will affect lives immensely. The prime objective of present

research is to develop a generic methodology which is

applicable to any study area. Nevertheless for initial

development, a test site is required. One of the main

considerations of selecting study site was availability of

several experts from different discipline who are well

acquainted with study area. Also, being the metropolis and

capital city, spatial and non-spatial data for several themes

were readily available. Most importantly, study area

characterizes a typical urban landform with socio-economic

activities revolving around risk planning and mitigation. Such

characteristics suit selection of area for case study to

demonstrate applicability of methodology.

Despite of gaining importance and widespread acceptance

of multi-criteria analysis based decision-making in risk

assessment and regional planning; it is still in its infancy stage

in India. In this respect, this study will have a significant

contribution to explore potentials of this approach to address

issue of risk management.

V. SDSS ARCHITECTURE - INSURANCE PROFILER (INSPRO)

The development of SDSS solutions for actuarial Industry,

coined as InsPro – Insurance Profiler, involved four (4)



critical areas of development (Figure 2), as illustrated

stepwise in subsequent section:

Development of geocoding engine

Creation of hazard, vulnerability score maps

GIS database integration

Development of Web-based solution - InsPro

Fig. 2 Schematic architecture of Insurance Profiler (InsPro)

A. Development of geocoding engine

Geocoding is the process of assigning geographic

coordinates to data that contain addresses. The coordinates

assigned to each address turn each record into a geographic

object that can be potentially displayed on a map. This was

designed as first ―gateway‖ into InsPro application. The

development of geocoder involved:

1) Electronic Research work and use of local knowledge:

The first and foremost step in this was identification of

administrative hierarchy. In administrative structure of India it

was seen that it is composed of states which are further

divided into districts (zila) and districts are further split into

into sub-districts, locally known as Tehsils/Talukas. The block

is the next level of administrative division following tehsil.

Villages are often the lowest level of administrative divisions

in India. Thus, these datasets were captured to be used for

address identification. In addition, building footprint data was

created with house numbers/name for urban centres so that

geocoder was capable of approximately locating houses.

These datasets were used for tagging POIs, streets, localities

and town datasets as these are building block for geocoder.

Second step was determination of postal formats

determination. In India, there are 8 PIN regions and the first

digit indicates one of these regions. Postcode is however six

(6) digits long where the first 2 digits together indicate sub

region or one of postal circles, first 3 digits together indicate a

sorting / revenue district and last 3 digits refer to delivery post

office. Thus, recognition of postal hierarchy helped in creating

an approximate postal reference data for India. This was

another milestone in development of Geocoder. Third step

was to extract address patterns/ formats. General pattern of

address followed in India includes writing of recipient’s name

in first line followed by house number/street name, locality

name, district, postal code and state name. The address pattern

identification helped in development of various permutations

and combinations of address being entered by user and hence

further enhancement of geocoder to fetch correct results or

nearest match on hits being made by user by using these

permutations and combinations of address pattern. The fourth

and last step was determination of thoroughfare types: In India,

different thoroughfare types identified include motorized ways,

non-motorized ways and waterways. The local terms used for

these thoroughfares such as highway, flyover, expressway,

lane, way, avenue, gali, path, road, marg, sadak, walk, street,

channel were added to the geocoder configuration files in

order to determine best possible match. Also, prefix and

suffixes used with road names such as NH4, directional words

viz. north etc were incorporated which further assisted in

enhancement of Geocoder.

2) Data build: Geospatial files viz. GeoInfo, PostInfo,

POIs (Point of Interests) and StreetRef files were created to be

used as input for Geocoder. GeoInfo files were point data

containing information on capitals, cities, towns, villages

whereas PostInfo file were polygon data with information on

postal codes. StreetRef file were polyline data with details on

streets names, their types, pre-post fix, house number ranges.

POIRef was point file with information on landmarks viz.,

business hubs, commercial centres, stations, scenic places,

shopping centres. All these spatial files were tagged with

administrative level information.

3) Component build:

Based on electronic database searches, data build, and local

knowledge, configuration files, to be used by MapMarker

geocoding engine, was created which contained following

information besides spatial data files:

Coordinate precision information: This was to

determine number of decimal places of coordinate

values should be used to precise the results. This was

set as 6.

Word dictionaries: Created with words generally being

used by locals in writing address. Minimum quality of

words used for searching areas, streets and postcodes

with values assigned between 0.0 and 1.0, with a value

of 1.0 indicating the words have to be perfect matches.

Pre filtering information –This allowed showing up of

results with candidates having matching search area

words thereby reducing false positives and speed up

matching process due to reduced number of candidates.

Searches based on alternate key: This was used to

determine the use of alternates keys based on

transposed characters, missing characters, incorrect

characters, extra characters etc.

Soundex parameterization: This involved grouping of

characters or group of characters to get best possible

match based on sound property. For example -

soundex_replace_1=C,ts; soundex_replace_2 = A,aw.

Weights assigning: The street information and post

address were assigned with scores for obtaining better

results during reverse geocoding. For example: POIs

data (such as landmarks) was given high scores while

matching data as in Indian context these POIs are taken

as identifier such as near XYZ place. Hence, better

geocoding precision.



Assigning precision code: Coordinates to an address

based on how well it matched in address dictionary was

assigned a code based on precision of matched results.

The code represents success or failure of geocoding

operation and conveys information about quality of

match. Each character of code provides information on

how precisely geocoded results matched each address

component. The code is an alphanumeric code of 1–10

characters and falls into the categories such as single

unique match; postcode centroid match and geographic

match. Each category is further subdivided into sub-

categories. Table I enlist geocoded precision code

results and their accuracy description.

TABLE I: GEOCODING PRECISION CODE DESCRIPTION

Single unique match (S category): This implies record was

matched to single address candidate. First character (S) reflects

that geocode component found street address that matched

record. First two characters of S result code indicate type of

match found.

Results Accuracy Level

S5 match located at street address position

S4 match located on the street centroid

SX match located at street intersection

Street level geocode result, codes S4 and S5 are followed by

additional characters, indicating details of match precision.

These result characters appear in order, immediately after S4 or

S5.

H Exact match on house number

P Street prefix direction

S Street suffix direction

C Exact match on town name

Z Exact match on postcode name

A or U

A if address is returned from the address

dictionary. U if address is returned from the

user dictionary

- If any field does not have an exact match, then

its position will be replaced by a dash

Example of geo-coding results explained below:

S4-PSCZA Street centroid match (S4) with exact match

on all other criteria except house number

S5HPS—A

Street centroid match (S5).Exact match on

house number, but no exact match on town

name or postcode

SX Street intersection match

Geographic centroid matches (G category): The matches

under this category indicate that a match was made at

geographic (town or locality) level. This may be because no

street match was possible and geocoder results fell back to

geographic area.

Results Accuracy Level

G3 geographic match with town centroid -

areaname3

G4 geographic match with locality -

areaname4

If Areaname3 input matches both town and locality names,

then G3 candidates appear at top of candidate list followed by

G4 candidates. When both town and locality is provided as

input, highest scoring candidates are listed at top. Exception is

when geographic input matches both town and locality.

Thus, configuration and data build binary files were used in

geocoder engine to create geocoding components which was

capable of handling Single/Multiline input, address correction,

reverse geocoding and bulk/batch geocoding.

However challenges faced during the development of Delhi

Geocoder were non-availability of street names, unorganized

addresses, house numbers etc. and variation in address pattern.

Thus, this aroused difficulty in geocoding at street level as

most of addresses do not include street names and hence

geocoded at geographic levels than street level. Besides, street

interpolation can’t be done because of non- standard house

numbers. Also, address search precision is poor due to above

stated deficiencies. Thus, with these limitations, geocoder

works on the hierarchy of identifying pincode and locality,

identifying the street (as already segmented), identifying

POI/Landmark, and identifying administrative boundaries for

getting precise results.

B. Creation of hazard, vulnerability score maps

The integrated system designed here, is divided into two

phases of risk score generation: static and dynamic phase. a)

Construction of composite hazard and vulnerability layer

score map which was preset in SDSS formed static component

and run-time generation of risk maps formed dynamic

component based on user’s permutation and combination of

vulnerability classes. Hence, an aggregate risk score map was

developed for a particular property under insurability

consideration.

For seismic hazard score map generation, Saaty’s (2000)

analytical hierarchy process, a MCDM methodology, in a

participatory decision-making framework was used to rank

and develop seismic hazard and vulnerability layer score map

of study area [23]. Nine experts (two academic researchers,

three from government organizations, and four from

nongovernment organizations who work closely in seismic

risk assessment areas) were engaged to perform pair-wise

comparison of criteria and weights were determined at two

levels of hierarchy i.e attribute values of the map layers and

map layers to generate hazard and vulnerability layer score

maps.

Pair-wise comparisons were carried out based on Saaty’s

semantic nine-point scale which relates numbers to judgments

(Table II).

TABLE II: PAIR-WISE COMPARISON SCALE

Intensity of

Importance Definition Explanation

1 Equally

important

Two elements contributes

equally to the property

3 Moderately

important

Experience and judgment

moderately favor one

element over other

5 Strongly

important

Experience and judgment

strongly favor one element

over other

7 Very Strongly

important

An element is strongly

favored and its dominance is

demonstrated in practice

9 Extremely The evidence favoring one



important element over another is of

the extremely highest order

of affirmation

2, 4, 6, 8 Intermediate

values

Compromise is needed

between two judgments

Reciprocal

of above

numbers

If an activity has one of

the above numbers assigned

to it when compared with a

second activity has the

reciprocal value when

compared to the first.

In this way different criterion were weighted with

homogenous measurement scale. Through this method, the

weight assigned to each single criterion reflected the

importance which every expert involved in the project

attached to objectives. Once experts were through with

comparative analysis, weights and consistency ratios besides

calculating eigenvalue, Consistency Index (CI) and Random

Index (RI) were calculated (Refer Saaty and Vargas, 1993 for

calculation steps). The pair-wise comparison matrices for

each expert that met the consistency ratio (CR) i.e CR < 0.1

were then aggregated using geometric mean (Saaty, 2000). A

geometric mean was used instead of arithmetic mean when

comparing different criteria and finding a single "figure of

merit" for these criteria as geometric mean "normalizes" the

ranges being averaged and hence no range dominates the

weighting, and a given percentage change in any of the

properties has the same effect on the geometric mean. Thus,

through MCDM approach of pair-wise comparisons weights

of criterions were determined for below enlisted criteria and

final hazards score maps was generated.

1) Seismic zones: A seismic zone is a region in which the rate

of seismic activity remains fairly consistent. There are five

seismic zones named as I to V as details given below:

Zone V - Covers the areas liable to seismic intensity IX

and above on Modified Mercalli Intensity Scale. This is

the most severe seismic zone and is referred here as

Very High Damage Risk Zone.

Zone IV - Gives the area liable to MM VIII. This, zone

is second in severity to zone V. This is referred here as

High Damage Risk Zone.

Zone III - The associated intensity is MM VII. This is

termed here as Moderate Damage Risk Zone.

Zone II - The probable intensity is MM VI. This zone is

referred to as Low Damage Risk Zone.

Zone I - Here the maximum intensity is estimated as

MM V or less. This zone is termed here as Very Low

Damage Risk Zone.

2) Peak Ground Acceleration (PGA): Peak ground

acceleration is the maximum value observed from an

accelerograph recording in an earthquake. Because it is a

value derived readily from ground motion records, there is a

much larger global dataset of PGA available.

3) Soil characteristics: The soil parameter controls relative

amplification of ground motion. The soil value is actually an

index related to the shear-wave velocity (Vs) of the top 30

meters at a site. This material property has been shown to

correlate well with shaking amplification; lower Vs generally

result in a larger ground motion than hard materials with a

high velocity.

4) Liquefaction: Liquefaction is form of ground failure that

can be triggered by strong shaking. It is the temporary

transformation of a solid soil into a liquid state. It can occur

when certain types of saturated, unconsolidated soils are

subjected to repeated, cyclical vibration and therefore most

commonly occurs during earthquakes.

5) Geology: Geology is the study of the Earth, the materials

of which it is made, the structure of those materials, and the

processes acting upon them. Geology plays an important role

in determining seismic hazard as regional geology enables in

assessment of sources and patterns of earthquake occurrence,

both in depth and at the at the surface.

6) Land use: Most of the Delhi area has changed land use

from the forest to agricultural areas to urban centres to

business hubs especially in the central portion. This has

actually led to increase in the urban population, decrease in

open spaces and forested areas. Delhi has also experienced a

large population in growth in the last decades and this

combined with rapid infrastructure development has

intensified the seismic vulnerability in the area.

7) Proximity to the fault: A fault is a break in the earth's crust

along which movement can take place causing an earthquake.

When an earthquake occurs on one of these faults, the rock on

one side of the fault slips with respect to the other. Faults can

be centimeters to thousands of kilometers (fractions of an inch

to thousands of miles) long. The fault surface can be vertical,

horizontal, or at some angle to the surface of the earth. Faults

can extend deep into the earth and may or may not extend up

to the earth's surface. Faults with evidence of Holocene (about

10,000 years ago to present) movement are the main concern

because they are most likely to generate future earthquakes. If

the earthquake is large enough, surface fault rupture can occur.

8) Proximity to the epicenters: The epicenter is the point on

the Earth's surface that is directly above the hypocenter or

focus, the point where an earthquake or underground

explosion originates. In the case of earthquakes, the epicenter

is directly above the point where the fault begins to rupture,

and in most cases, it is the area of greatest damage. However,

in larger events, the length of the fault rupture is much longer,

and damage can be spread across the rupture zone

The weight maps were standardized by applying a linear

function. Linearity was chosen to limit discussion with

stakeholders for selecting other membership functions. The

composite hazard score map generated herein formed the

static framework of risk analysis in InsPro.

Multi-criteria evaluation (MCE) technique was adopted for

creation of vulnerability score map. MCE was applied with

following factor maps:

Building height

Year built

Construction type

Building area (square footage)



Each vulnerability score map generated herein also formed

the static component ofInsPro.

C. GIS database integration

Workspace was created which is a simple text based

scripting resource containing commands to open tables, create

and position the necessary map, browser and other windows,

define layer style and thematic settings. The layers included

in workspace were: administrative, gazetteer, point of interests

(POIs), streets layers. This workspace was used as base map

and formed the front end visualization component of the

InsPro on which analysed results were to be depicted.

Microsoft Access® 2007 was used for data storage and

queries and contained non-spatial data assembly including

hazard and vulnerability score tables. Hazard score table

contained pre generated composite hazard score at pincode

level derived using MCDM techniques. MCDM was too

applied to obtain vulnerability score tables with individual

score tables of building height, year built, construction type,

and building area (square footage) of case study region which.

The advantages of storing it in access lies in the fact that these,

tables are not static and can be updated, revised at any given

point of time by the administrator of Ins Pro.

Ahlers and Boll (2008) introduced five classes in terms of

spatial granularity (country, region, city, street, and building)

for geocoder development. For the current study street level

geocoder engine was developed and integrated into InsPro

using C# language. The geocoder was incorporated enabling

geocoding functionality to fetch and show search results to the

user on web interface.

D. Development of Web-based solution - InsPro

Cognizant of the need for a risk assessment tool for better

underwriting and actuarial engineering, and to provide a

system that can generate and manage risk information for

acquisition of insurance/reinsurance facilities and catastrophe

cover, GIS assisted SDSS wascalled as Insurance Profiler

(InsPro) was developed. The codes were developed for

integrating geocoding components and multi-criteria score

mapsand layers to create sync between them. Besides for

visualization of results of search, query, geocoder, statistical

analysis web interface was created using MapXtreme

framework. The SDSS were built with basic functionalities

such as zoom, pan, search and locate, address validate, bulk

geocode, on the fly risk score computations, report generation,

print and save. Computation of risk score involved using

multiplicative function of hazard potential and vulnerability

i.e. Risk = Hazard potential x Vulnerabilitywhich is also the

definition of risk. To be able to portray the risk of region, the

risk scores/map is based on an aggregated hazard map and an

integrated vulnerability map, and it enables us to see the level

of risk related to a region. This concept was applied in InsPro

where hazard score map was pre-computed and stored in

database, vulnerability criteria classes were selected by user

and dynamic risk score was computed as output using

multiplicative function. However, such simplification doesn’t

devalue flexibility and usefulness of the SDSS tools in

disaster insurance underwriting. In support of the robust

expert-system shell, more use can further populate the

knowledge bases of hazard, vulnerability and risk assessment

making them more complete, more sophisticated and easily

adjustable by satisfying demands for decision-making.

Besides these, in InsPro, the flexibility for calibrating data,

parameters and even risk computation logic and limits, as per

user’s requirement were provided. The better visuals and array

of the applications has capability to draw more acute

fascination of customer toward insurance

underwritings/pricing. The application will tend to bring in

uninsured segment of population into insured segment by

giving a logical view of where and why asset should be

insured.

VI. RESULTS AND DISCUSSION

A. Multi-criteria evaluation based Hazard Score Map for

SDSS

In the present study the Spatial-MCDM method was used

in which different hazard criteria were appraised in order to

establish their validity and usefulness, and eventually

amalgamation of the different factors were provided in form

of a composite hazard score map for study region. Following a

multi-criteria decision making - analytical hierarchical process

(AHP) (Saaty, 1980), each theme and features were assigned

weights and rankings respectively according to their perceived

relative significances to seismic hazard (Refer Tables III -IX).

TABLE III: SCORES OF SEISMIC ZONES

Seismic Zone Risk Zone Weight

Seismic Zone-1 Very Low Damage Risk Zone 0.009793

Seismic Zone-2 Low Damage Risk Zone 0.009838

Seismic Zone-3 Moderate Damage Risk Zone 0.424964

Seismic Zone-4 High Damage Risk Zone 0.546386

Seismic Zone-5 Very High Damage Risk Zone 0.009019

TABLE IV: SCORES OF PEAK GROUND ACCELERATION

Peak Ground acceleration

(PGA, in g) Susceptibility Weight

0 – 0.12 Very Low 0.091047

0.12– 0.14 Low 0.128602

0.14 – 0.16 Moderate 0.189169

0.16 – 0.18 High 0.266218

0.18 – 0.20 Very High 0.324964

TABLE V: SCORES OF SOIL CHARACTERISTICS

Soil characteristics Soil

Susceptibility Weight

Very Hard to Hard Rock Very Low 0.091047



Loamy Sand Low 0.128602

Soft Rock to Older Alluvium Moderate 0.189169

Younger Alluvium High 0.266218

Fill to Shallow Bay Mud Very High 0.324964

TABLE VI: SCORES OF LIQUEFACTION CHARACTERISTICS

Liquefaction characteristics Liquefaction

Susceptibility Weight

Rock very stiff or cohesive clays,

sediments older than Pleistocene

(>1.6 Ma); sites with deep water

table

Low 0.101103

Holocene to Pleistocene (11Ka to

1.6Ma) alluvial fan deposits Very Low 0.127001

Modern alluvial fan deposits Moderate 0.2673707

Modern floodplain or beach ridge

deposits High 0.504526

TABLE VII: SCORES OF GEOLOGICAL CHARACTERISTICS

Geological characteristics Susceptibility Weight

Polycyclic sequence of brown silt-

clay with kankar and brown to grey

fine to medium grained sand 1 0.0421679

Yellowish fine to medium grained

sand with minor silt and siliceous

kankar 2 0.109563

Quartzite with interbanded schit

and phyllite 3 0.113868

Multiple fill alternate sequence of

grey micaeous fine to medium

grained sand 4 0.212850

Grey micaeous fine to coarse

grained sand and overbank silt 5 0.521552

TABLE VIII: SCORES OF LAND USE

Land

use Risk Zone Weight

Group 1 High Density Vegetation, Waterbodies 0.068837

Group 2 Low Density Vegetation, Open, Quasi

open area 0.112326

Group 3 Industrial area, Residential/village,

Agriculture 0.225349

Group 4 Skyscrapers, Urban low density ,

Urban High Density, Airport 0.593488

TABLE IX: SCORES OF PROXIMITY TO FAULTS

Proximity to Faults

(Neotectonic, Subsurface) No. of faults Weight

0-20 km 1 0.006543

21-40 km 1 0.005479

41-60 km 4 0.168103

61-80 km 7 0.286638

81-100 km 12 0.545258

TABLE X: PAIR-WISE COMPARISON SCALE

Proximity to Epicenter No. of epicenter Weight

0-25 km 0 0.021693

26-50 km 4 0.255663

51-75 km 6 0.4091855

76-100 km 2 0.191668

101-125 km 1 0.121790

The composite seismic hazard score map of Delhi region

involved evaluation of different seismic hazard components

namely seismic zones, peak ground acceleration at seismic

bedrock, soil characteristics, liquefaction potential, land use,

geological characteristics, proximity to faults and epicenter.

These layers were, thereafter, integrated through MCDM

techniques to obtain composite seismic score map addressing

site specific hazard scores for seismic micro-zonation. A

composite hazard score map was generated with indices value

from 0.17 to 0.89 (Figure 3).

Fig. 3 Seismic hazard score map of Delhi

The hazard scores were set into five categories, negligible (0),

low (0.01 - 0.25), moderate (0.26 – 0.50), high (0.51 – 0.75)

and very high (0.76 – 1.00). The map depicted that seismic

susceptibility of Delhi region follows the order: east > north >

west > south areas. East regions of Delhi are considerably

high vulnerable area because it is positioned in high seismic

zone and greater liquefaction potential. Overall, parts of

central Delhi are also subjected to greater seismic scores due

to social-economic assets accumulation. Accumulation of

people and their assets seemingly become major cause of the

hazard risk. The generated composite hazard score map was

integrated in InsPro.



Vulnerability score maps were too integrated into the InsPro

application so that in the fly final risk computation can be

made using hazard and vulnerability layers.

B. SDSS – Insurance Profiler

Insurance underwritings, risk pricing and claims management

are a number of objectives by which an insurer can reduce the

volatility and liquidity in characteristics of risk to

'homogenize' it, and make it fall in to the basket of 'risk pools'.

The booming geospatial technological have made possible to

build geo-analytical custom insurance solutions that leapfrog

capabilities of traditional offerings. InsPro – Insurance

Profiler is an upshot of such offerings, coming out with

hitherto hard-to-obtain location data with integrated risk

scores. Delhi region is used as case study site to showcase

few of the functionality of InsPro. The details of further

offerings of InsPro are explained in Table 3.

1) Mapping: Insurers/reinsurers/Risk managers can locate

address and visualize spatially (Refer Figure 4). The mapping

solutions incorporated in InsPro enables to depict myriad of

themes. All visualizations that user can see is rendered from

workspace. The series of standard procedures were involved

from conversion of data from OSL (Oracle) format to

MapInfo *.TAB format and forms the background for

geospatial results visualization based on functionality

executed in InsPro. For example, locating address Central

Cottage Industrial Corporation, Delhi. InsPro was able fetch

this result by making use of geocoding engine. Reverse was

also possible i.e. on entering Latitude/Longitude values,

address could be returned.

Fig. 4 Insurance Profiler (InsPro) - Web based mapping solution for

Insurers/Re-insurers

2) Risk Assessment: InsPro generates comprehensive

assessment of location under consideration by

insurer/reinsurers to produce more objective patterns of risk

assessment in lieu support of the expert knowledge base

(Figure 5). Besides these, InsPro has inherent functionality of

data analytics. For example, if a zone presents an

unacceptable risk for insuring new property then such risk can

be pre-screened by underwriters by varying the vulnerability

parameters. If the new risk falls in the alarming range of score,

it means there is already a concentration of risks, and they

should be careful while writing risk based on the actuarial

guidelines.

Thus, its very well evident from the above case study that

close association of geospatial technology with insurance

decision making processes, InsPro application is perfectly

suited for insurance domain to address its deficiencies.

Fig. 5 Insurance Profiler (InsPro) - Web based mapping solution for

Insurers/Re-insurers

TABLE XI: INSPRO FUNCTIONALITY AND USAGE

Functionality Usage

Assessment of

spatio-temporal

hazard risk

patterns

The spatial decision support system can

provide comprehensive analysis of hazard

based risk score in addition to building

parameters to produce more objective

patterns of risk assessment in support of the

expert knowledge base

Evaluation of

spatio-temporal

variation of

exposure

Discrepant insured buildings have differential

spatial variation of loss risk and thereby have

their respective vulnerability and loss curve.

Further, it is necessary to correctly estimate

the regional total loss at risk from all kinds of

properties so as to classify the insurance

portfolios

The past claims

and their

correlation by

different policies

Although the past claims data alone can’t

provide enough accurate information

concerning the occurrence patterns of natural

hazards, they are an available indicator for

the vulnerability and loss curve of exposures

and contain important information for pricing

and for determining insurance rates.

Mapping Underwriters can examine specific regions on

a digital map to see, how much of the current

book of business is concentrated within a



given radius or is proximal to historical claim

records? This would give them a clear picture

of the potential risk of the specific building/

pincode/ regions.

Analytics Trend analysis with historical data can be

performed to determine if a zone presents an

unacceptable risk for insuring new industry in

the area. Or by varying parameters of

building contents, a new risk can be pre-

screened by the underwriter. If the new risk

falls in the alarming range of score, it means

there is already a concentration of risks, and

they should be careful while writing risk

based on the current guidelines

Risk Search A simple query interface for Risk Portfolio

Manager to ease out the process of extracting

information from the database using pre-

stored policy and claims database.

Running an event

footprint on the

policy database

A geospatial footprint of any disaster viz.

earthquake, flood or cyclone can be overlaid

on insurance company’s current portfolio

using and this would help in estimating extent

of losses due to event

Thematic Risk

Reports

Map a portfolio and then determine its

exposure to various risks or intensity of risk.

These maps can be exported from Risk

Portfolio Manager in an image format.

Accumulate the insured value, premium,

PML, net retention, treaty limits and limit

either by a geographical point of reference

like building, pincode, or various admin

boundaries, risk zones, proximity to a

selected location. Even the accumulation by

risk parameters of a peril can be carried out.

For example, in case of the earthquake peril,

risk accumulation can be monitored by

building characteristics like occupancy,

construction type, construction quality,

number of inhabitants by day / night

Tabular

Accumulation

Reports

Tabular reports generated on the fly

incorporating some of the following key

aspects of the portfolio such as Premium

Distribution, Claims Distribution, Loss Cost

etc.

Insurance Profiler (InsPro) developed herein in this paper,

overcome some of the deficiencies of traditional actuarial

assessment in India such as inadequate understanding of the

geographical settings and its relationship to historical events

[24], analysis based on anticipation and correlation of

evidences ([25], [26]), small coverage and non-homogeneous

information ([11], [27]), fixed scale analysis. In the current

knowledge-based system shell, the geocoding engine,

supporting multi-criteria evaluations and visual display

facilitate insurers with the spatial visualizations, database

management, data analysis, querying, trend analysis,

estimating loss cost, avenue for new business expansion and

underwriting risks. This system will even allow, risk managers

to assess hazard concentration, determine degree of

vulnerability and anticipate damage, in case of occurrence of

catastrophe.

VII. CONCLUSION

Natural disaster and its vagaries contribute to complexity of

the risk analysis. Insurance pricing of these involves manifold

factors and interdisciplinary cooperation between disaster

experts, meteorologists and actuaries ([29]–[31]). From the

initial phase of hazard simulation, vulnerability and risk

analysis to rate-making and premium-making, there is no

clear-cut method or model that can give a comprehensive

answer ([29], [32], [33]). The location based knowledge

system designed specifically to deal with situation involving

procedural and declarative knowledge is thus an appropriate

choice of technology ([34],[35]). The SDSS - InsPro

developed in this study incorporates the advanced expert-

system shell, sophisticated visual GIS and robust spatial

multi-criteria statistics components into a coherent and

integral system using the industry standard interface protocol.

Such a system is flexible, portable, extendable, low-cost and

effective to provide a solid base for more accurate risk

analysis and pricing of insurance policies. The application

based on insurance guiding principles and scientific risk

assessment considerations, has the potential to basically

transform the lifecycle of most of the insurance business

processes as known in present day. Because of its flexibility,

scalability, user-friendliness GUI, despite some shortcomings

(hazard assessment, unsystematic uncertainty analysis), InsPro

can be easily enhanced and become more powerful with

continuous update of knowledge bases, enhancement of data

to support geocoding and incorporation of developed risk

models. This suite of product developed as prototype for

insurance sector, can also replicated for various domains.

REFERENCES

[1] V. D. H. Voet, and W. Slob, ―Integration of probabilistic exposure

assessment and probabilistic hazard characterization‖, Risk Analysis,

vol.27, pp. 351–371, 2007. [2] T. Q. Zeng, and Q. Zhou, ―Optimal spatial decision making using GIS:

a prototype of a real estate geographic information system (REGIS)‖,

International Journal of Geographical Information Science, vol. 15, pp. 307–321, 2001.

[3] M. R. Zolfaghari, and K. W. Campbell, ―A new insurance loss model

to promote catastrophe insurance market in India and Pakistan‖, Earthquake Engineering, vol. 2, pp. 1-8, 2008.

[4] R. T. Kozlowski and S. B. Mathewson, ―Measuring and managing

catastrophe risk‖, Journal of Actuarial Practice, vol. 3, pp. 211–232, 1995.

[5] R. T. Musulin, ―Issues in the regulatory acceptance of computer

modeling for property insurance ratemaking‖, Journal of Insurance regulation, vol.15, pp. 342–359, 1997.

[6] L. Lianfa, J. Wang, and C. Wang, ―Typhoon insurance pricing with

spatial decision support tools‖, International Journal of Geographical Information Science, vol.19 (3), pp. 363–384, 2005.

[7] B. Rabkin, and D. Sonnen, ―Frameworks to Develop Spatial

Perspectives of the Insurance Value Chain‖, IDC Technology Spotlight, IDC 1043, pp. 1-8, 2010.

[8] (2011) The EM-DAT website. [Online]. Available: http://

www.emdat.be / [9] K. Nagesh, ―GIS as Decision Making Tool for Insurer‖, Bimaquest, vol.

4(1), pp. 48-59, 2004. [10] R. I. Mehr, ―Insurance, Risk Management, and Public Policy‖,

Huebner International Series on Risk. Insurance and Economic

Security, vol. 18, pp. 182-200, 2000. [11] M. Goodchild, R. Haining, and S. Wise, ―Integrating GIS and spatial

data analysis: problems and possibilities‖, International Journal of

Geographical Information Systems, vol. 6, pp. 407–423, 1992.



[12] A. Amendola, Y. Ermoliev, T. Y. Ermolieva, V. Gitis, G. Koff, and J. Linneroothbayer, ―A systems approach to modeling catastrophic risk

and insurability‖, Natural Hazards, vol. 21, pp. 381–393, 2000.

[13] Z. Zhang, and D. A. Griffith, ―Integrating GIS components and spatial statistics analysis‖, International Journal of Geographical Information

Science, vol. 14, pp. 543–566, 2000.

[14] Y. Ding, and P. Shi, ―Fuzzy risk assessment model of typhoon hazard‖, Journal of Natural Disasters, vol. 11, pp. 34–43, 2002.

[15] R. Thomas, ―Insurance pricing wit GIS: It’s all about business‖,

Geospatial Solutions, vol. 20, 30-45, 2000. [16] P. Grossi, H. Kunreuther, C. C. Patel, ―Catastrophe modeling: a new

approach to managing risk‖. Huebner International Series on Risk.

Insurance and Economic Security, vol. 25, pp. 252-270, 2005. [17] G. Carpenter, ―The Catastrophe Bond Market at Year-End 2007: The

Market Goes Mainstream‖. GC Securities, vol. 1, 2008.

[18] P. Peduzzi, ―The Disaster Risk Index: Overview of a quantitative approach‖. In: Birkmann, J. ed. Measuring Vulnerability to Natural

hazards – Towards Disaster Resilient Societies, New York: United

Nations University: pp. 502-524, 2006.

[19] I. Vertinsky, S. Brown, H. Schreier, W. A. Thompson, and G. C.

Vankooten, ―A hierarchical-GIS-based decision model for forest

management: the systems approach‖, Interfaces, vol. 24, pp. 38–53, 1994.

[20] H. Jiang, and J. R. Easterman, ―Application of fuzzy measures in

multi-criteria evaluation in GIS‖, International Journal of Geographical Information Science, vol. 14(2), pp. 173–184, 2000.

[21] Q. Wu, S. Ye., X. Wu, and P. Chen, ―Risk assessment of earth

fractures by constructing an intrinsic vulnerability map, a specific vulnerability map, and a hazard map using Yuci City, Shanxi, China as

an example‖, Environmental Geology, vol. 46, pp. 104–112, 2004.

[22] A. Sakamoto, and H. Fukui, ―Development and application of a livable environment evaluation support system using Web GIS‖, Journal of

Geographical Systems, vol. 6, pp. 175–195, 2004.

[23] T. L. Saaty, ―A scaling method for priorities in hierarchical structures‖, Journal of Mathematical Psychology, vol. 15, pp. 234–281, 1977.

[24] R. Klostermann, ―Planning Support Systems: A New Perspective on

Computer-Aided Planning‖, In: R. Klostermann, ed. Planning Support Systems, Integrating GIS, Models, and Visualizations Tools, Redlands:

ESRI Press, pp. 1-24, 2001. [25] Y. Leung, and K.S. Leung, ―An intelligent expert system shell for

knowledge-based Geographical Information Systems: the tools‖

International Journal of Geographical Information Systems, vol. 7, pp. 201–214, 1993.

[26] K. A. Knut, ―A Markov model for the pricing of catastrophe insurance

future and spreads‖, Journal of Risk and Insurance, vol. 68, pp. 25–50, 2001.

[27] J. R. Easterman, ―Uncertainty management in GIS: decision support

tools for effective use of spatial data‖ In: C. Hunsaker, and M. Goodchild, eds. Spatial Uncertainty in Ecology: Implications for

Remote Sensing and GIS Applications, New York: Springer- Verlag,

2001, vol. 14. [28] R. Leigh, and I. Kuhnel, ―Hailstorm loss modeling and risk assessment

in the Sydney region, Australia‖, Natural Hazards, pp. 171–185, 2001.

[29] J. F. Wang, and L. F. Li, ―Improving Tsunami Warning Systems with Remote Sensing and Geographical Information System Input‖, Society

for Risk Analysis, 2008. doi: 10.1111/j.1539-6924.2008.01112.x.

[30] S. T. Algermissen, and K. V. Steinbrugge, ―Seismic hazard and risk assessment. Some case Studies‖, The Geneva Papers on Risk and

Insurance, vol. 9 (30), pp. 84-123, 1984.

[31] S. Mansor, M. A. Shariah, L. Billa, I. Setiawan, and F. Jabar, ―Spatial technology for natural risk management‖, Disaster Prevention and

Management, vol. 13(5), pp. 364-373, 2004.

[32] M. T. Pareschi, L. Cavarrs, M. Favalli, F. Giannini, and A. Meriggi, ―GIS and volcanic risk management‖, Natural Hazards, vol. 24, pp.

187–196, 2000.

[33] D. Sommer, ―The Impact of Firm Risk on Property-Liability Insurance Prices‖, Journal of Risk and Insurance, vol. 29, pp. 501-514, 1996.

[34] D. C. M. Dickson, ―Insurance Risk and Ruins‖ International Series on

Actuarial Science. Cambridge: University Press, 2005. [35] Y. Ermolieva, and V. V Norkin, ―Spatial Stochastic Model for

Optimization Capability of Insurance Networks under Dependent

Catastrophic Risks: Numerical Experiments‖. IIASA Interim Report, IR-97-028, 1997.

2012 AARS, All rights reserved.* Corresponding author

Landscape Characterization to Assess Impact and Magnitude of Roads on the Urban Spaces of Delhi, India

Nishant1, 2*, Neena Priyanak1, 2 and P K Joshi1

1TERI University, New Delhi 110070, India

2Pitney Bowes Software, Noida 201301, India

Abstract

Quantification of landscape pattern and its transformation is crucial for assessment and monitoring of environmental consequences of urban infrastructure development. In the present study, geospatial tools and landscape metrics have been coalesced to quantify impacts of roads on spatial pattern of urbanization in Delhi using Quickbird (0.6m) dataset by varying grain size and across the transects 1 (Roads and urban class as individual entity) and 2 (roads and urban class treated as aggregate entity). Landscape metrics were computed along a 31 km long and 6 km wide transect (West to East direction) using standard and moving window analysis. The results of transect analysis showed that urbanization together with infrastructure development have resulted in increased patch density (PD), patch and landscape shape complexity (LSI), while a spectacular decrease in the largest and mean patch size (LPI) and landscape connectivity or increased fragmentation have been observed. The changes in landscape pattern along the transect have important ecological implications, and quantifying it at varied grain size, as illustrated in this paper, is an important first step to link patterns with processes in urban environs.

Key words: Urbanization, landscape metrics, patch, remote sensing, roads

1. Introduction

Urbanization and rapid infrastructure developments are considered key factors of land transformation profoundly influencing microclimatic conditions, green spaces and human life. Ecological consequences of urbanization and developmental plans are interesting and important to monitor and assess. Landscape analysis is one such attempt that can be used to quantify these. It further assists to understand concept of urban-rural gradient (McDonnell et al. 1997, Miller and Pillsbury 2008), which enhance variety of ecological issues in urban areas (Harshberger 1923). Still there exists a great gap in understanding these ecosystems (Collins et al. 2000, Wu 2000). Various methods such as gradient analysis (Godron and Forman 1983, McDonnell and Pickett, 1990) and landscape pattern metrics have been

employed to understand the spatial pattern of urbanization with its ecological processes and thereby providing means to relate urban environment and people spatially and location of urbanization center with multiple indices (Alberti and Botsford 2000, Alberti 2001). Geographers and social scientist have carried out spatial pattern and urban dynamics of urban-rural areas with little or only superficial consideration of ecology in and around cities (Forrester 1969, Berry and Kasarda 1977, Batty and Longley 1994, Schneider and Woodcock 2008). By uncovering such characteristics of urban fragmentation along the gradient of land use zones, spatial distribution of urban fragmentation can be understood.

Rapid developmental activities in transportation sector due to increasing urban demand have resulted in alteration of


land use land cover (LULC) pattern. Construction of roads is one such activity that has brought important effects to landscape. Depending on the adjacency to nearby LULC the impacts of roads vary such as some are easily identifiable and some show effect with time. For example, impacts of forest roads such as dissecting the land, leading to habitat fragmentation, shrinkage, and attrition have been spatially viewed and quantified at several times (Reed et al. 1996), however, ecological impacts of roads in urban landscape have rarely been reported. The integrated urban ecosystems need new and integrative perspectives (Pickett et al. 1997, Grimm et al. 2000, Zipperer et al. 2000).

Over the year, the urban development of Delhi have been on the fringes and in radial pattern with reliance on road infrastructure. The development envisaged by previous plans were polynodal with hierarchy of Commercial Centers located on either ring or radial roads. The MRTS network, underpass, overpass, metro networking have brought connectivity thereby having impact on the existing structure of city and consequently its development. This changed scenario has provided opportunities for city restructuring and alterations in LULC pattern. The present study was taken up with multifold objectives. We aim to assess changes of road patches to urban landscape pattern and relationship between the two. It has been accomplished while analyzing landscape in transect where road patches merge with urban areas. In this paper, the theoretical basis and general structure of landscape pattern metrics and effect of grain size have been used to address impact of road development on urban landscape.

2. Study Area

Delhi lies between latitudinal parallels of 28°40' N and 28°67' N and longitudinal parallels of 77°14' E and 77°22' E and occupies northern region of India (216 meters above sea level). With an area of 1483 sq.km it corresponds to a typical patch of tropical region, completely engrossed with residential, commercial and urban centers. The region is undergoing rapid urban sprawl because of unprecedented developmental activities and population growth. A 31 km × 6 km study area located in central region traversing from West to East of Delhi is chosen in this case study (Figure. 1).

3. Material and Methods

The Quickbird images acquired on 2008-03-15 (pan sharpened-0.6m) were georeferenced using a polynomial approach. Five LULC classes were extracted using interactive approach of both visual and digital interpretation with the aid of ancillary data (e.g., pre-classified maps, topographic maps). Two transects were subset from imagery. Transect 1 comprises five LULC classes viz. open area, green spaces, urban area, roads and water body and Transect 2 contains four classes in which roads were merged with the urban area. The LULC raster was resampled to varying pixel size (1.2m, 3m, 5m, 15m, 30m and 60m) from the original data of 0.6m

Figure 1. Location of study area

grain size using nearest neighborhood technique.

The derived resampled files were exported to GRID file format. Class properties file was prepared to set the run parameterization using Fragstats v.3.3. A series of landscape metrics at class and landscape level were calculated using 8-neighbors patch delineation rule. Standard and moving window analysis were performed each for Transect 1 and Transect 2. Landscape metrics at class and landscape level with variable pixel size were analyzed with regard to dynamic information of landscape and to determine the optimal grain size for impact analysis study of roads and characteristics of landscape dynamics.

4. Results and Discussion

The major LULC classes are open area, green spaces, urban area, roads and water body. The open area refers to agricultural fields, scrub, riverbed and vacant lands in and around the city. The green spaces are ridge forest, biodiversity part of city and all urban green spaces along roadsides and settlements. Because of high resolution data linear green spaces could be mapped very conveniently. Urban area refers to all settlements in and around city. No attempt has been taken to classify the type of settlement and define any part of the settlement as rural, which is very difficult in Delhi. Roads are linear feature in and around settlements. Roads were also mapped in open areas and green spaces. Visual interpretation

Asian Journal of Geoinformatics, Vol.12,No.1 (2012)

technique was used to delineate road network. River, canal network and small water bodies are classified as water. The overall accuracy of LULC interpretations exceeds 85% for all classes based on validation using the random points selected from original images.

Class area (CA) is a measure of landscape composition i.e. how much of landscape is composed of particular patch type. CA of transect 1 (Figure. 2a) shows that open areas composition is highest and rest follows the sequence as urban> vegetation> roads> water body. However in Transect 2 (Figure. 2b), when urban and roads are merged, CA is still higher for open areas but value of urban and road class area has increased which is not additive in nature. Rest classes

0

20

40

60

80

100

Open Urban Vegeation Road Waterbody

LULC

Cla

ss A

rea

(sqk

m)

Figure 2a: LULC and class area for Transect 1

0102030405060708090

Open Urban Vegeation Waterbody

LULC

Clas

s Are

a (s

qkm

)

Figure 2a: LULC and class area for Transect 2

Figure 2a. LULC and class area for Transect 1

Figure 2b. LULC and class area for Transect 2

does follows similar trend as in transect 1 such as urban + roads> vegetation> water body. Thus this shows that road and urban area when combined together exert a greater influence on landscape pattern and alters the landscape composition.

PLAND reveals the most important information about landscape composition because quantitatively different LULC types generally would have different landscape pattern attributes. The PLAND of open area is considerably higher followed by urban structures. Vegetation class is slightly lower in occupancy. Road though being a linear feature does show greater occupancy thus showing its impact in landscape composition. Percentage occupancy of land shows similar trend as class area.

PLAND of all class in Transect 1 decreased at varying grain size thus suggesting that grain size play a key role in determining composition of landscape classes (Figure. 3a). Up to 6m grain size, change in PLAND is quite significant but as the grain size increased from 6m to 15m, 30m and 60m, grain size does tend to show saturation and hence change in PLAND is not quite significant. This suggests that up to 6m resolution urban landscape composition at local scale can be evaluated for studying identified classes as at coarser resolution, PLAND value gradually saturates and hence is degree of differentiation reduces at coarser resolution datasets. The similar trend was observed in transect 2 (Figure. 3b). This is in concordance because width of roads in Delhi does not exceed more than 15m and streets are much

0.00

10.00

20.00

30.00

40.00

50.00

60.00

1.2 3 4 15 30 60

Pixel Size (m)

PLA

ND

Open Road Vegetation Urban Waterbody

Figure 3a: PLAND for Transect 1

0.00

10.00

20.00

30.00

40.00

50.00

60.00

1.2 3 4 15 30 60

Pixel Size (m)

PLA

ND

Open Urban + Road Vegetation Waterbody

Figure 3b: PLAND for Transect 2

Figure 3a. PLAND for Transect 1

0.00

10.00

20.00

30.00

40.00

50.00

60.00

1.2 3 4 15 30 60

Pixel Size (m)

PLA

ND


Figure 3a: PLAND for Transect 1

0.00

10.00

20.00

30.00

40.00

50.00

60.00

1.2 3 4 15 30 60

Pixel Size (m)

PLA

ND

Open Urban + Road Vegetation Waterbody

Figure 3b: PLAND for Transect 2

Figure 3b. PLAND for Transect 2


narrower than this and thus coarser resolution datasets may fail to capture landscape pattern beyond 15m grain size.

The results from class level metric moving window analysis along transect are shown in Figure 4a and 4b. The diagram shows spatial changes of landscape pattern. On each diagram the horizontal axis represents rural-urban-rural gradient from West to East and vertical axis represents the metric value. The major reason behind the following interpretation is how the changes in landscape pattern are related to process of urbanization. This also identifies the impact of grain size and visualizing impact of road network on the adjacent features and landscape patterns (patch density primarily). The V shape curve indicates that metric having an inverted V-shape distribution is positively correlated to the degree of urbanization and others (representing V-shape distribution) are negatively correlated to degree of urbanization.

Patch Density (PD) is highest value in the urban core indicating a highly fragmented landscape and decreasing on both sides of the urban axis consisting of regions of sub-urban and rural areas. The central region being the city zone

0

20

40

60

80

100

120

140

Rural Sub-Urban Urban Sub-Urban Rural

Patc

h D

ensit

y (n

o. o

f pat

ches

/ 100

ha)

Urban Open Roads Vegetation Waterbody

Figure 4a: Patch density for Transect 1

020406080

100120140

Rural Sub-Urban Urban Sub-Urban RuralPatc

h D

ensit

y (n

o. o

f pat

ches

/ 100

ha)

Urban+Roads Open Vegetation Waterbody

Figure 4a: Patch Density for Transect 2

0

20

40

60

80

100

120

140

Rural Sub-Urban Urban Sub-Urban Rural

Patc

h D

ensit

y (n

o. o

f pat

ches

/ 100

ha)

Urban Open Roads Vegetation Waterbody

Figure 4a: Patch density for Transect 1

020406080

100120140

Rural Sub-Urban Urban Sub-Urban RuralPatc

h D

ensit

y (n

o. o

f pat

ches

/ 100

ha)

Urban+Roads Open Vegetation Waterbody

Figure 4a: Patch Density for Transect 2

Figure 4a. Patch density for Transect 1

Figure 4b. Patch Density for Transect 2

area, higher PD and hence higher fragmentation is obvious across all classes. Figure 4a and 4b show the trend of PD in both transects. The fragmentation in urban class is higher in transect 2 than that of 1 as road patches have been merged. This suggests that road developmental activities together with urbanization tend to have influence on landscape composition and structure and the developmental activities inappropriately planned would influence it to greater extent.

Grain size also plays important role in determining the PD as it determines the maximum number of patches per unit area. An inverse relationship is observed between PD and grain size (Figure. 5a). The graph shows that PD tend to decrease across all classes as the grain size decreases from 1.2m to 3m, 6m, 15m, 30m respectively. However at coarser resolution PD of all classes are very low and differentiation of classes is not much significant. However in transect 2 the PD is comparatively higher for urban region thus exhibiting fragmentation characteristics even at coarser resolution (Figure. 5b). The graph also depicts sensitivity of roads to varying grain size and saturation being achieved beyond 15m. This is also attributed to small width of roads which could be picked up only because of unique spatial signature and only using spatial resolution less than 15 m.

Mean patch size (MPS) is lowest at urban core region indicating a fragmented landscape that is composed of many small patches. For the Western and Eastern half of the transect MPS increases gradually with distance from city

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

1.2 3 6 15 30 60

Grain Size (m)

Patc

h D

ensi

ty (

patc

hes/1

00 h

a)


Figure 5a: Patch Density for Transect 1 in different grain size

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

1.2 3 6 15 30 60

Grain Size (m)

Patc

h D

ensi

ty (

patc

hes/1

00 h

a)

Open Urban+Roads Vegetation Waterbody

Figure 5b: Patch Density for Transect 2 in different grain size

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

1.2 3 6 15 30 60

Grain Size (m)

Patc

h D

ensi

ty (

patc

hes/1

00 h

a)


Figure 5a: Patch Density for Transect 1 in different grain size

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

1.2 3 6 15 30 60

Grain Size (m)

Patc

h D

ensi

ty (

patc

hes/1

00 h

a)

Open Urban+Roads Vegetation Waterbody

Figure 5b: Patch Density for Transect 2 in different grain size

Figure 5a. Patch Density for Transect 1 in different grain size

Figure 5b. Patch Density for Transect 2 in different grain size

center, indicating an increase in land parcel size. However MPS of urban class is comparatively higher in Transect 2 than 1 which illustrates road patches when merged with urban class tends to exert greater influence than urban and road class individually. Landscape Patches Index (LPI) showed monotonically increasing trend with increasing pixel/grain size indicating dominance value of class increases with increasing resolution. Moreover, LPI saturated beyond a resolution value of 30m and does not have appreciable


effect on landscape class. Thus LPI measures should be used carefully when comparing landscapes at varied grain size. Landscape Shape Index (LSI) showed a gradual declining trend. There are apparent effects to respond to variable grain sizes in class-level and landscape-level.

Perimeter-Area Fractal Dimension (PAFRAC) approaches 1 for shape with very simple perimeters such as squares and approaches 2 as the patch complexity increases. It’s an indicative of shape complexity across a range of spatial scales. PFRAC of all classes is considerably higher at resolution of 1.2m and it tends to decrease with increasing grain size (Figure. 6a). Thus it suggests that at high-resolution shape complexity is much greater and this trend gradually diminishes as spatial scale varies from finer to coarser resolution. The increased complexity of merged urban and roads landscape in transect 2 suggests that it tends to have greater influence on urban landscape structure than roads and urban classes alone (Figure. 6b). Thus, road indeed tend to increase complexity of landscape which is identifiable at finer scale thus its developmental planning should be taken with greater concern and contemplation.

Clumpiness Index (CLUMPY) was calculated for determining the focal patch type disaggregation/aggregation and degree of disaggregation/ aggregation. CLUMPY equals -1 when maximally disaggregated, 0 distributed randomly, and approaches 1 as maximally aggregated. Among all classes in Transect 1 road patches showed maximum disaggregation and degree of disaggregation increased as resolution of grain size increased. This is quite evident with transect 2 study too. As the transect grain size increased road patches appear maximally disaggregated and the degree of disaggregation was highest among all class patch type. Water body showed highest aggregation at all grain sizes however the degree of

1.00

1.20

1.40

1.60

1.80

2.00

1.2 3 6 15 30 60Grain Size (m)

PAFR

AC

Open Road Vegetation Urban Waterbody Figure 6a: PAFRAC for Transect 1 in different grain size

1.00

1.20

1.40

1.60

1.80

2.00

1.2 3 6 15 30 60Grain Size (m)

PAFR

AC

Open Vegetation Urban+Roads Waterbody

Figure 6b: PAFRAC for Transect 1 in different grain size

Figure 6a. PAFRAC for Transect 1 in different grain size

1.00

1.20

1.40

1.60

1.80

2.00

1.2 3 6 15 30 60Grain Size (m)

PAFR

AC

Open Road Vegetation Urban Waterbody Figure 6a: PAFRAC for Transect 1 in different grain size

1.00

1.20

1.40

1.60

1.80

2.00

1.2 3 6 15 30 60Grain Size (m)

PAFR

AC

Open Vegetation Urban+Roads Waterbody

Figure 6b: PAFRAC for Transect 1 in different grain size

Figure 6b. PAFRAC for Transect 2 in different grain size

Figure 7. Effect of grain size on road (red colored) patches

association weakened at resolution greater than 30m. The aggregation index of class patch type followed the sequence as water body followed by vegetation, open, urban and road class patches (Figure.7 and 8).

However scale of disaggregation follows the reverse sequence. In the Transect 2 CLUMPY of merged Urban and road class is higher than the vegetation patch showing greater


Figure 8. Effect of grain size on water body (cyan colored) patches

aggregation measure than Transect 1 (Figure. 9a and 9b). Road tends to increase aggregation measure of urban area and hence tend to transform landscape pattern and processes.

Patch cohesion index (PCI) measures the physical connectedness of the corresponding patch type. In present analysis, PCI is highest for open areas and least for road thus typifying that open areas is maximum aggregated and road is least aggregated in its distribution and hence more physical connectivity among open areas than roads. The degree of

0

0.2

0.4

0.6

0.8

1

1.2 3 6 15 30 60Grain Size (m)

CLU

MPY

Open Road Vegetation Urban Waterbody Figure 9a: Clumpiness index for Transect 1 in different grain size

0

0.2

0.4

0.6

0.8

1

1.2 3 6 15 30 60Grain Size (m)

CLU

MPY

Open Vegetation Urban+Roads Waterbody Figure 9b: Clumpiness index for Transect 2 in different grain size

0

0.2

0.4

0.6

0.8

1

1.2 3 6 15 30 60Grain Size (m)

CLU

MPY

Open Road Vegetation Urban Waterbody Figure 9a: Clumpiness index for Transect 1 in different grain size

0

0.2

0.4

0.6

0.8

1

1.2 3 6 15 30 60Grain Size (m)

CLU

MPY

Open Vegetation Urban+Roads Waterbody Figure 9b: Clumpiness index for Transect 2 in different grain size

Figure 9a. Clumpiness index for Transect 1 in different grain size

Figure 9b. Clumpiness index for Transect 2 in different grain size

PCI decreased with increasing grain size but changes were more evident in road class only. In Transect 2, degree of physical connectedness is still higher than Transect 1 (Figure. 10a and 10b). Merged urban and road classes show comparatively higher PCI than urban and road structures alone indicating more clumped or aggregation in its distribution, hence more physically connected.

The above interpretations conclude that selection of appropriate grain size is the first parameter to be established.

2030405060708090

100

1.2 3 6 15 30 60

Grain Size (m)

PC

I

Open Road Vegetation Urban Waterbody Figure 10a: Patch Cohesion Index for Transect 1 in different grain size

20

40

60

80

100

1.2 3 6 15 30 60

Grain Size (m)

PCI

Open Vegetation Urban+Roads Waterbody Figure 10b: Patch Cohesion Index for Transect 2 in different grain size

2030405060708090

100

1.2 3 6 15 30 60

Grain Size (m)

PC

I

Open Road Vegetation Urban Waterbody Figure 10a: Patch Cohesion Index for Transect 1 in different grain size

20

40

60

80

100

1.2 3 6 15 30 60

Grain Size (m)

PCI

Open Vegetation Urban+Roads Waterbody Figure 10b: Patch Cohesion Index for Transect 2 in different grain size

Figure 10a. Patch Cohesion Index for Transect 1 in different grain size

Figure 10b. Patch Cohesion Index for Transect 2 in different grain size


The appropriate range of grain for landscape indices of Delhi transect was 1.2m to 15m. The above results show that urbanization has resulted in dramatic structural changes of metropolitan landscape. For example, as urbanization progressed large and contiguous patches were broken up with an increasing number of patch types (LULC types) occurring in landscape. The density of patches of various types and thus PLAND composition increased exponentially. The overall LPT increased steadily mainly due to increasingly even proportions of dominant LULC types whereas geometric shapes of patches in landscape as a whole became more and more irregular. As a result, urbanization has brought about increased structural fragmentation and complexity of landscape in Delhi region. In present analysis most critical points occur within 1.2-15m which is width range of some landscape elements, such as roads, branches of rivers. When grain size increases over this range, these elements shrink to small patches or are masked by other dominating elements, thus inflexions occur. Satellite imageries such as IKONOS, Quickbird, Worldview-1/2, SPOT PAN, XS, ASTER with 1.2-15m resolution are adequate for assessing impact of road on urban landscape research. Roads were thus sensitive to grain size of 15×15m2 because most of the roads in the study area were 10-20m wide. High percent coverage of roads indicated high patch density of landscape. A major ecological impact of roads in process of urban land transformation was leading to habitat fragmentation.

5. Conclusion

The present research work adopted a combined method of landscape metrics analysis and sensitivity of metrics to varying grain size to analyze impact of road dynamics on landscape pattern Delhi, India. For this, degree of urbanisation and infrastructure developmental (roads) were considered focal factors. The research design helped to answer research objectives such as how changes of road patches alters urban landscape pattern and what is the degree of changes at varying grain size. The major findings include (i) landscape compositional diversity and degree of fragmentation is positively correlated to degree of urbanization both along rural-urban –rural gradient, (ii) road patch type has unique spatial signature as compared with other LULC types, which differ with varying grain size, (iii) different patch type have differential and distinguishable landscape pattern attributes along transect and across various grain resolution, and (iv) changes in pattern of road structures shows positive correlation to degree of urbanization and developmental activities. This study is a step in direction of better understanding of impacts of road on landscape pattern and processes both of which would tend to have severe ecological consequences. The study also substantiated that urban landscape is more heterogeneous in composition and are mostly fragmented.

Landscape metrics quantify pattern of landscape within designated landscape boundary and facilitates differential

scenario based planning. Grain size is one important parameter in such analysis and provides insights to regional planning scales. Consequently, through the interpretation of these metrics and ecological significance of grain size an acute awareness of the landscape context and openness of landscape relative to phenomenon under consideration can be determined. The concept can be applied to identify indicators to mitigate negative effect of urbanization and sustainable LULC planning in urban landscapes.

References

Alberti, M. (2001). Quantifying the Urban Gradient: Linking Urban Planning and Ecology. In: Avian Ecology in an Urbanizing World. J. M. Marzluff, R. Bowman, R. McGowan and R. Donnelly. New York, Kluwer.

Alberti, M. and E. Botsford (2000). Behavior of land use and land cover metrics along an urban-rural gradient. Working Paper, Urban Ecology Research Laboratory, Department of Urban Design and Planning, University of Washington. Seattle.

Batty, M., and P. Longley (1994). Fractal cities: A geometry of form and function. San Diego: Academic Press.

Berry, B. J. L., and J. D. Kasarda (1977). Contemporary urban ecology. New York: Macmillan.

Collins, J. P., A. Kinzig, N. B. Grimm, W. F. Fagan, D. Hope, J. Wu, and E. T. Borer (2000). A new urban ecology. American Scientist 88: 416–425.

Forrester, J. W. (1969). Urban dynamics. Cambridge: The M.I.T. Press.

Godron, M., and R.T.T. Forman (1983). Landscape modification and changing ecological characteristics. In: H.A. Mooney and M. Godron (eds.). In: Disturbance and ecosystems: components of response. Springer-Verlag, N.Y. P. 12-18.

Grimm, N.B., J.M. Grove, C.L. Redman, and S.T.A. Pickett (2000). Integrated approaches to long-term studies of urban ecological systems. BioScience 50:571–584.

Harshberger, J.W. (1923). Hemerecology: The ecology of cultivated fields, parks, and gardens. Ecology 4:297–306.

Luck, M., and J. Wu (2002). A gradient analysis of urban landscape pattern: a case study from the Phoenix metropolitan region, Arizona, USA. Landscape Ecology 17:327-339.


McDonnell M. J., S. T. A. Pickett, P. Groffman, P. Bohlen, R.V. Pouyat, W. C. Zipperer, R. W. Parmelee, M. M. Carreiro, and Medley K. (1997). Ecosystem processes along an urban-to-rural gradient. Urban Ecosystems 1: 21–36.

McDonnell, M. J. and S.T.A Pickett (1990). The study of ecosystem structure and function along urban-rural gradients: an unexploited opportunity for ecology. Ecology 71: 1231–1237.

Pickett, S., W. R. Burch, S. E. Dalton, T. W. Foresman, J. M. Grove, and R. Rowntree (1997). A conceptual framework for the study of human ecosystems in urban areas. Urban Ecosystems 1: 185–199.

Pillsbury, F.C. and J. R. Miller (2008). Habitat and landscape characteristics underlying anuran community structure along an urban-rural gradient. Ecol Appl. 18(5): 1107-18.

Reed, R.A., J. Johnson-Barnard, and W.A Baker (1996). Contribution of Roads to Forest Fragmentation in the Rocky Mountains. Conservation Biology 10: 1098-1106.

Schneider, A. and C. Woodcock (2008). Compact, Dispersed, Fragmented, Extensive? A Comparison of Urban Growth in Twenty-five Global Cities using Remotely Sensed Data, Pattern Metrics and Census Information. Urban Studies 45(3) 659–692.

Wu, J. (2000). Landscape ecology: Concepts and theory. Chinese Journal of Ecology 19(1):42-52.

Zipperer, W. C., J. Wu, R. V. Pouyat, and S. T. A. Pickett (2000). The application of ecological principles to urban and urbanizing landscapes. Ecological Applications 10: 685–688.

Exploring non-conventional options of rain water harvesting – responding to climate change impacts

using geospatial tools

Nishant* 1, 2

, N. Priyanka1, 2

and P. K. Joshi2

1Pitney Bowes Business Insight (MapInfo), Logix Cyber Park, 5

th Floor, Tower - B, C-28&29, Sector-62, Noida-

201301 U.P. India, www.pb.com, www.mapinfo.com 2 TERI University, Vasant Kunj, New Delhi, www.teriuniversity.ac.in

* Corresponding Author: mailto:[email protected]?subject=Climate Change Workshop

Abstract

Global climate change analysis has indicated variations in the temperature and precipitation regimes

and as a result water resources are likely to come under increasing pressure. This coupled with

anthropogenic activities is increasing the ecological footprint and thereby trampling the fragile

hydrological systems. For supplementing the ever-increasing needs, techniques such as rainwater

harvesting schemes are one of the adaptation options in the climate change scenario. These are

employed to intercept additional water to minimize run-off loss (which is ~ 45% of average annual

rainfall). However, the conventional rainwater harvesting methods are inadequate to address water

demands as vertical expansion is far exceeding horizontal expansion in populated cities. In the wide

expanding cities, road network is the most planned and developed infrastructures that could be explored

for the water harvesting, if surveyed, planned and executed appropriately.

In view of this, we present a conceptual framework for supplementing water supply through a prototype

study in Delhi region. It has been the endeavor of this study to identify the options to harvest the

rainwater using geospatial tools. It recognizes roadside in the sprawling cities providing an additional

source of water to harvest. The collected water can be put to use for groundwater recharge or made

potable or variously exploited employing bioremediation techniques. The prudent and integrated water

resources development for sustainable water utilization is important even in absence of climate change

impact. Such conceptual studies could gauge the extent of problems that the cities are likely to envisage.

Introduction

Climate change refers to variations in the mean state of climate or variability of its properties in its rate,

range and magnitude that extends for a long period usually decades or longer (IPCC, 2007). Theses long-

term changes are in unequivocal agreement between climate models which points towards increasing

warming trends globally (IPCC, 2007). The Intergovernmental Panel on Climate Change (IPCC) Fourth

Assessment Report (IPCC, 2007; Baines et al., 2007), reports that extending from 1956 to 2005 the

global surface warming increased at a rate of 0.13°C per decade which was nearly double that

experienced in 100 years from 1906 to 2005 and is further likely to increase by 1.1-6.4°C towards the

end of the 21st century without showing any sign of ceasing (Figure1). The relationship between

hydrological system and climate change is even more complex, and the scientific consensus has

broadened that climate impacts on water resources are already appearing worldwide.

http://www.pb.com/

http://www.mapinfo.com/

http://www.teriuniversity.ac.in/

mailto:[email protected]?subject=Climate%20Change%20Workshop

Figure 1: Temperature projections to the year 2100, based on a range of emission scenarios and global climate models. Scenarios that assume the highest growth in greenhouse gas emissions provide the estimates in the top end of the temperature range. The orange line (“constant CO2”) projects global temperatures with greenhouse gas concentrations stabilized at year 2000 levels. Source: NASA Earth Observatory, based on IPCC Fourth Assessment Report (2007)

In India, studies by various authors illustrates that there is escalating trend in surface temperature

(Hingane et al., 1985; Pant et al., 1999, Goswami et al., 1992; Chylek et al., 2007) no significant drift in

rainfall pattern (Pant et al., 1999) on all-India basis, but decreasing/increasing trends in rainfall pattern

(Mall et al., 2006; Dhar and Majumdar, 2009) at regional levels. However, little work has been done on

hydrological impacts of possible climate change for Indian regions/basins. Average annual rainfall over

India is about 117 cm which is highly variable in spatio-temporal scales and is mostly concentrated in

four months viz. June, July, August and September (southwest monsoon season) (IMD, 2009). Therefore,

the variation in seasonal monsoon rainfall may be considered a measure that the climate change is

exacerbating the spatial and temporal variations in water availability over the Indian domain in the

context of global warming.

The demand for water has already increased manifold over the years due to urbanization (Rodell et al.,

2009), agriculture expansion (Umapathi and Ramashesha, 2001), increasing population (Census of India,

2006), rapid industrialization and economic development (Alam et al., 2007) and is projected that

towards 2025, the freshwater demand globally will rise by 25% or more (Kundzewicz et al., 2007). The

situation is getting worse in the cities where mounting demographics along with space crunch is regular

phenomenon. In a study by Rodell, it is illustrated that more than 26 cubic miles of groundwater

vanished from aquifers in the states of Haryana, Punjab, Rajasthan and the National Capital Territory of

Delhi since 2002. The capital city of India, Delhi is an exemplary region where as per the projections of

Census of India, the population is expected to be over 24 million by 2021 and touch 28 million by 2026

(Department of Urban Development, 2010) and demand for water will increase manifold. Thus, catering

to such populations, an assessment of the availability of water resources in the context of requirements

and expected impacts of climate change and its variability is critical for long-term development

strategies and sustainable development of the city (Lean et al., 2008).

In purview of the above, we present a conceptual framework for supplementing water supply through a

prototype study in Delhi region to identify the options to harvest the rainwater along roadside using

geospatial tools. It recognizes that, the similar experimental design, if successful, can be replicated for

other populated for sustainable development of surface water and groundwater resources within the

constraints imposed by climate change. It is therefore essential that individuals, societies and

institutions are made aware of the likely changes and have strategies in place to mitigate or adapt to a

changing climate for sustainable water resource management.

Study area

New Delhi is located in the centre of Northern India within co-ordinates of latitudes 28°27’15” to

28°34’44” N and longitudes 77°10’9” to 77°18’41”E (Figure 2). It is positioned with the Great Indian

Desert of Rajasthan to the west and south west, central hot plains to the south and Gangetic plains of

Uttar Pradesh to the east while cooler hilly regions of Uttarakhand to the north. It covers an area of

1,483 sq km with a population estimate of 16 million (Census, 2006).

Figure 2: Location of the Study Area - Delhi, India

Experimental framework

The overall experimental design for current study was divided into two components as illustrated below:

Vulnerability Assessment Adaptation Assessment

Analysis of current rainfall variability

Projected future rainfall variability

Anthropogenic Impacts

Evaluation of Rain water harvesting method

Proposition of road area as alternative for harvesting

Proposition of cleansing harvested water

Vulnerability Assessment

The IPCC Special Report on Emission Scenarios (SRES) suggests that climate changes is mainly driven by

increased atmospheric concentration of Greenhouse Gases (GHG’s) and are intricate processes that are

highly uncertain to manifest (IPCC, 2007). Thus, climate scenarios viz. A1, A2, B1 and B2, exploring

alternative development pathways covering a wide range of demographic, economic, technological, and

environmental and policies driving forces, have been developed (Refer Figure 3 and 4) as tool of

assessing plausible alternatives of how the future emission may unfurl.

Figure 3: Illustration of the four SRES scenario families Figure 4: Projected global average temperature

increases for different SRES scenarios (IPCC, 2007) Source: IPCC Fourth Assessment Report (2007)

Two regional scenarios (Refer Figure 3) were evaluated for current study viz. A2 sceanrio (describes a

very heterogeneous world with high population growth, slow economic development and slow

technological change) B2 scenario (which stipulates a world with intermediate population and economic

growth, emphasizing on local solutions to economic, social, and environmental sustainability). The

Hadley Centre Coupled Model version-3 (HADCM3) data at 30 arc-seconds resolution developed by

Hadley Centre were processed and used, as it is one of the AOGCM’s participating in IPCC’s Fourth

Assessment Report (AR4) (IPCC, 2007) and does not need flux adjustment (additional artificial heat and

freshwater fluxes at the ocean surface) to produce good simulations (Gordon et al., 2000 and Pope et

al., 2000). These data were tailored for Delhi region and statistical and graphical methods were applied

to detect changes in the rainfall regime over Delhi. Also, projected rainfall trend for year 2020, 2050 and

2080 were analyzed by employing visual and statistical techniques. Rainfall pattern in current and future

scenarios were analyzed only for four months viz. June, July, August and September as almost 80% (88

cm ± 10 SD) of the long term average annual rainfall comes down in these months (Refer Figure 5)

through southwest monsoon (Mall et al., 2006; IMD, 2008).

Figure 5: 25 year (1981 - 2006) average rainfall of Delhi

Source: Adapted from National Data Centre, India Meteorological Department, Pune (2008)

Moreover, for analyzing anthropogenic impacts on water resources two proxy measures such as landuse

change pattern and demographic trends were analyzed. For landuse change pattern analysis, Landsat

satellite images available with GLCF were processed to find out the indicators of urban sprawl and

development of infrastructures such as roads resulting in loss of open areas. The datasets of October

(1977, 1989, 1999 & 2006) were taken for sprawl assessment which were available with the required

preprocessing (radiometric and geometric corrections). The band information was used to compute

Normalized Difference Vegetation Index [NDVI = (NIR - RED) / (NIR + RED)] for detecting greenness. The

raw bands were put to digital classification with classification scheme viz. (i) Settlements (ii) Vegetation

(iii) Open area and (iv) Water aimed to assess the changes in land use pattern. The local areas were

visited with paper print of satellite data, topographic sheets and GPS to collect the ground truth and

field verification. The demographic trends were statistically analyzed using Census data and its

projections for future.

Adaptation Assessment

Literature review was carried to unveil available adaptation measures and coping strategies for water

resources especially in context of rain water harvesting methods. Geospatial analysis such as spatial

correlation, statistics and mathematical calculations were formulated to propose road as an alternate

method of rain water harvesting and its feasibility for making additional water available for groundwater

recharge or making it potable for drinking. Techniques to purge off the oil pollutants from harvested

water collected from roadside were reviewed herein.

Results and Discussions

Vulnerability Assessment: Analysis of current and projected rainfall variability

A number of studies have been carried out to assess the impact of climate-change scenarios on

hydrology of various basins and regions in India (Gosain and Rao, 2003, Mall et al., 2006) and it is

projected that increasing temperature and decline in rainfall has been observed which tends to reduce

net recharge and affect groundwater levels (Gosain and Rao, 2003, Mall et al., 2006). However, little

work has been done on hydrological impacts of possible climate change for Delhi regions/basins. In

purview of the above, the rainfall regime shift analysis was performed using current and future climate

change under two scenarios. It was observed that under A2 sceanrio the shift in the rainfall was

compartively more than in B2 scenario which stipulates a world with relatively better economic, social,

and environmental sustainability (Refer Table 1). Also, the shift in the mean value of rainfall was

observed higher as one graduated from current to future climate years viz. 2020, 2050 and 2080 in

descending fashion (Refer Figure 6 -Shift anlysed for the month of June).

Figure 6: Climate change and induced variability in rainfall (Source: Prepared from Worldclim data, 2011)

Table 1: Precipitation trends under current and future climate years for Delhi region (Source: Prepared from Worldclim data, 2011)

Climate Scenarios

Month Current Climate 2020 A2a 2050 A2a 2080A2a

June

Current Climate

2020 B2a 2050 B2a 2080 B2a

Climate Scenarios


July

Current Climate

2020 B2a 2050 B2a 2080 B2a

Climate Scenarios


August

Current Climate

2020 B2a 2050 B2a 2080 B2a

Climate Scenarios


September

Current Climate

2020 B2a 2050 B2a 2080 B2a

It is therfore observed spatially and statitically that the rainfall regime contraction is bound to happen in the years to come. This is very much

evident with the shift in the mean average value of rainfall from current to future years (Figure 6). Thus, its imperative to explore new avenues

for capturing maximum amount of rainfall to sustain the climate change imapct on water resources in urban centres such as Delhi.

Anthropogenic Impacts

Beside observed climate shifts, anthropogenic activities are exerting great pressure on water resources

(Mckenzie and Ray, 2009) from rising human population and sprawling concrete structures, particularly

growing concentrations in urban areas (Mookherjee and Hoerauf, 2004). Urban agglomerations

magnetize various sectors such as manufacturing, construction, trade and service of all kinds (Lall and

Mengistae, 2005) thereby opening avenues of employment and have become the pull factor for the

ever-increasing migration (Iyer and Kulkarni, 2007), employment opportunities (Lopez et al., 2003) and

population growth (Mookherjee and Larvey, 2000). Figure 6 shows the impact of expected population

growth on water usage by 2025, based on the UN mid-range population projection and the current rate

of per capita water use (UNEP and Earthscan, 1999; Min. of water resources, 2003, Alexander et al.,

2006). This clearly indicates the ‘two-sided’ effect on water resources – the rise in population will

increase the demand for water leading to faster withdrawal of water and this in turn would reduce the

recharging time of the water-tables (Mall et al., 2006; IWMI, 2008). As a result, availability of water is

bound to reach critical levels sooner or later. Secondly, the sprawling concrete structures will disrupt the

permeability of water in the ground by forming impermeable surface thus making scenario even worse.

Figure 7: Observed and projected decline in per capita average annual freshwater availability and growth of population from 1951 to 2050 (Source: Mall et al., 2006)

Landuse Pattern - Land use pattern change analysis of Delhi very well highlights the urban sprawl and

increase in the concrete forest in a very random pattern. The classified output reflects a rapid increase

in the Settlements and observed decrease in green cover and open area over the period, as most of the

areas have been converted to Settlement (Refer Table2). The new Settlements have been seen in the

fringes and outskirts of the city. Green cover though increased in selected areas, but the net assessment

shows a decreasing trend. While assessing the direction of urban sprawl, it’s generally from core to

peripheral and extending outwards leading to dramatic changes in the eastern, south western and

western fringes of the Delhi resulting in prime sub-cities of the National Capital Reserve (NCR) namely

(a) Gurgaon township (as one of the IT and communication hub); (b) Dwarka sub-city (primarily designed

for the human settlements and amenities). Some of the changes have also been observed in the western

part named as Noida Township (a complicated mosaic of small scale industries, infrastructure and

human settlements).

Table 2: Distribution of land use classes depicting change over the year (Area in sq. km.)

Year/ Class

1977 1989 1999 2006

Ne

t A

rea

Ch

ange

Year / Class

1977 to

1989

1989 To

1999

1999 to

2006

Urban 419.67 444.83 478.55 545.37 Urban 25.16 (+ve) 33.72(+ve) 66.82 (+ve)

Vegetation 223.65 204.91 187.97 155.65 Vegetation -18.74(-ve) -16.94(-ve) -32.32(-ve)

Open Area 752.19 738.58 720.23 688.19 Open Area -13.61(-ve) -18.35 (-ve) -32.04(-ve)

Waterbody 78.45 85.64 87.21 84.75 Waterbody 7.19 (+ve) 1.57 (+ve) -2.46 (-ve)

Year/ Class

1977 1989 1999 2006

Pe

rce

nt

chan

ge Year/

Class 1977

to 1989

1989 To

1999

1999 to

2006

Urban 419.67 444.83 478.55 545.37 Urban 1.71 2.29 4.53

Vegetation 223.65 204.91 187.97 155.65 Vegetation -1.27 -1.15 -2.19

Open Area 752.19 738.58 720.23 688.19 Open Area -0.92 -1.24 -2.17

Waterbody 78.45 85.64 87.21 84.75 Waterbody 0.49 0.11 -0.17

Besides rapid gain in urban agglomerations, the associated development of road infrastructures is huge

and expanding at an alarming rate. This was analyzed using time series road vector data extracted from

Quickbird imagery for 2006 and 2010 vintage. Figure 7 shows the observed increase in road from 2006

to 2010. This clearly indicates that as the urban agglomerations will increase leading to more avenues of

employment and migration, the demand for connectivity will increase proportionately.

Figure 8: Pictorial representation of road infrastructure sprawl in Delhi [Color scheme: Red – Delhi roads in year 2006; Grey – Delhi roads in year 2010]

Demographic trends - Delhi has witnessed a phenomenal population growth during past few decades. A

population of 405,809 in 1901 has grown to 13,782,976 in 2001. Between 1999 and 2001, population in

the region grew by about 54% (Census of India, 2006) while the amount of developed land increased by

about 146%, or nearly three times the rate of population growth. There is tremendous increase in the

urban population as compared to rural population of the state. This conforms to the land use pattern

analysis illustrated above. As per the projections of Census of India, the population of Delhi is expected

to be over 24 million by 2021 and touch 28 million by 2026 (Department of Urban Development, 2010).

Figure 9: Population trend in Delhi (1901-2016)

Source: Population Projections for India and States 1996-2016, Department of Planning, Govt. of NCT of Delhi)

Coping and Adaptation – A Geospatial approach

Rationalization on various means of capturing and storing water is the need of the hour in order to fulfill

water demands in the future (IWMI, 2008). Harvesting of rainwater is one alternative which could

contribute in meeting water requirements sustainability (Varis et al., 2005). Therefore, efforts are

needed for harnessing excess monsoon in more efficient manner through identification, adoption of

rainwater harvesting technological alternatives to not only boost the availability of water to meet the

growing demand, but also help in controlling damages from extreme events such as floods.

Presently, rain water harvesting is undertaken through a) Capturing runoff from rooftops b) Capturing

runoff from local catchments c) Capturing seasonal floodwaters from local streams and d) Conserving

water through watershed management (CSE, 2003; GoI, 2003, Che- Ani et al., 2009). These measures of

collection are inadequate in meeting the needs of urban centres such as Delhi. This is because the urban

expansion being more on vertical front of expansion than horizontal, has led to engulfment of open

lands and unhabitated lands for the urban development process (Mookherjee and Hoerauf, 2004).

These areas are turned into planned/unplanned residential or commercial areas which are marked with

high density of population, roads and flyovers, lack of proper sanitation and drinking water, low cost

business activities, informal or unregistered business and manufacturing activities, etc (Lall and

Mengistae, 2005). Thus, there are less open areas for water seepage to groundwater for its recharging

as concrete jungles forms an impermeable layer. Besides these, the rooftop area is also inadequate to

capture the rainwater in the long run as the vertical expansion exceeds the horizontal expansion with

increasing population. Thus to substantiate the water harvesting, roads are proposed as an alternative

for capturing rainwater to supplement the additional need of water besides other measures of rain

water harvesting. Geospatial statistics and comparative analysis was performed to evaluate the

potential of rainwater harvesting using road in conjugation with urban areas (roof top harvesting) which

is described below:

Calculations: How much water can additionally be harvested using road area? (Approach adopted from

Centre for Science and Environment. 2003)

0

50

100

150

200

250

1901 1921 1941 1961 1981 2001 2003 2005 2007 2009 2011 2013 2015

Year

To

tal

Po

pu

lati

on

(in

lacs)

The total amount of water that is received in the form of rainfall over an area is called the rainwater

endowment of the area. Out of this, the amount that can be effectively harvested is called the water

harvesting potential.

Water harvesting potential = Rainfall (mm) x Collection efficiency

The collection efficiency accounts for the fact that all the rainwater falling over an area cannot be

effectively harvested, because of evaporation, spillage etc. Factors like runoff coefficient and the first-

flush wastage are taken into account when estimated the collection efficiency (Refer Table 1 for details).

Table 3: Runoff coefficients for various catchment surfaces (Source: Pacey, Arnold and Cullis, Adrian 1989)

Type of Catchment Coefficients

Roof Catchments o Tiles o Corrugated metal sheets

o 0.8 - 0.9 o 0.7- 0.9

Ground surface coverings o Concrete o Brick pavement

o 0.6 - 0.8 o 0.5 - 0.6

Untreated ground catchments

o Soil on slopes less than 10 per cent o Rocky natural catchments

o 0.0 - 0.3 o 0.2 - 0.5

Untreated ground catchments

o Soil on slopes less than 10 per cent o Rocky natural catchments

o 1.0 - 0.3 o 0.2 - 0.5

The following is an illustrative theoretical calculation that highlights the enormous potential for water

harvesting using road as an additional source. Assumptions taken while calculating rain water harvesting

potential are:

o The road or rooftop area is impermeable and all the rain that falls on it is retained without

evaporation

o Factors like runoff coefficient and the first-flush wastage are taken to be constant for both

urban area and road

o Only 60 per cent of the total rainfall is effectively harvested

Table 4: Calculations for Rain water harvesting potential

Calculations Rooftop Area (in 2006) Road Area (in 2006)

Total Area = 37% of 1483 sq. km. = 548710000sq. m.

= 8% of 1483 sq. km. = 118640000sq. m.

Height of the rainfall ((June - 83.7mm; July – 184mm; August - 227.2mm, September - 119.6mm)

= 614.5 ~ 615mm = .615m

= 614.5 ~ 615mm = .615m

Volume of rainfall (Area x Height of rainfall) = 337456650m3 = 72963600m3

Volume of water harvested (assuming that only 60 per cent of the total rainfall is effectively harvested) (1 m3 = 1000 litres)

= 202473990 m3 = 202473990000litres

= 43778160 m3 = 43778160000litres

No. of person that can be catered with this water where water requirement per person is 10 litres (IWMI, 2010)

= 20247399000 = 4377816000

No. of families with average 5 members that can be catered with this water

= 4,049,479,800 = 875,563,200

Thus its very well indicative that road area if utilized for rainwater harvesting in addition to roof top, it

can cater the need of additional 875,563,200 families. Seeing the rising trend in area of road from 8% (in

2006) to 12% (in 2010), the potential for road as source for harvesting rainwater will be even more.

Since the more of the vertical expansion rather than horizontal expansion is happening these days in the

urban centers such as Delhi, exploring road area as potential rainwater harvesting measures will be a

boon for mitigating water resources crunch.

The major concern that can be foreseen with road side harvested water is the presence of pollutants

such as oil products along with rainwater. This problem can be overcome by utilizing microbial-based

product that degrades oil contaminants very fast such as Oilzapper, Oilivorous-S developed by the

Microbial Biotechnology laboratory at TERI or through development of other strains through

experimental work, once the feasibility study of roadside rain water harvesting measures are conducted

and brought into system.

Conclusion

This research work is a conceptual study about calculating and assessing additional sources of

intercepting rain water in lieu of climate change caused water stress. Three frameworks have been

presented here – a) Schematic framework for vulnerability characterization by evaluating current and

future scenarios of change in rainfall regime over Delhi region; b) Representation of the proxy

relationship of anthropogenic induced changes such as urbanization and demographic rise & the water

resources stress and c) Computation of harvested water using road area in addition to rooftop area

method of rain water harvesting and assessing its importance in climate mediated water crunch.

Using these one can proceed stepwise to understand the interrelationship of all the components

responsible and affected by the possible change in water resources due to projected climate change in

Delhi. The typical character of the tropical monsoon climate, unique geography, enormous population

and rising infrastructural development are the characteristic features that define the face of the future

environment and the status of water resources in Delhi (Status Report on Delhi , 2001; ). The need of the

hour is to explore the alternatives for harnessing water to avoid potential threats and risks to the human

system as well as the environment. Thus, for the current assessment employing road area as an

additional method for water harvesting was carried out involving the use of temporal and spatial data. It

was seen through this conceptual study that, road area harvested water has additional capacity to cater

large no. of families and hence not letting it to become the limiting factor and thus jeopardize the

progress in social-economic development of the region. It is recognized that if road network if surveyed,

planned and executed appropriately for the water harvesting, it will strengthen the existing water

resources development and management for optimum and sustainable water utilization even without

the occurrence of climate change impact.

References

Alam, M., Rahman, A., Rashid, M., Rabbani, G., Bhandary, P., Bhadwal, S., Lal, M. and Soejachmoen, M.

2007. Impacts, vulnerability and adaptation to climate change in Asia. Background Paper.

Produced for the UNFCCC Asia Regional Workshop on Climate Change Adaptation, Beijing China

11-13 April 2007.

Alexander, L. V., Zhang, X., Peterson, T. C., Caesar, J., Gleason, B., Klein Tank, A. M. G., Haylock, M.,

Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A., Rupa Kumar, K., Revadekar, J., Griffiths, G.,

Vincent, L., Stephenson, D. B., Burn, J., Aguilar, E., Brunet, M., Taylor, M., New, M. Zhai, P.,

Rusticucci, M. and Vazquez-Aguirre, J. L. 2006. Changes in daily temperature and precipitation

extremes in central and south Asia. Journal of Geophysical Research. 111. D16105pp.

doi:10.1029/2005JD006316.

Arnold, P. and Adrian, C. 1989. Rainwater Harvesting: The collection of rainfall and runoff in rural areas.

Intermediate Technology Publications, London: 94pp

Baines, P. G. and Folland, C. K. 2007. Evidence for rapid global climate shift across the late 1960s. Journal

of Climate. 20: 2721 – 2744pp. doi:10.1175/JCLI4177.1

Census of India. 2006. Population projections for India and States 2001-2026. Report of the technical

group on population projections constituted by the national commission on population May 2006.

Available at http://www.jsk.gov.in/projection_report_december2006.pdf (Accessed 10 Feb

2010)

Centre for Science and Environment. 2003. Site dedicated to Rainwater Harvesting. Accessed on various

dates at http://www.rainwaterharvesting.org/

Che-Ani, A. I., Shaari, N., Sairi, A. and Tahir, M.M. 2009. Rainwater Harvesting as an Alternative Water

Supply in the Future. European Journal of Scientific Research. 34 (1): 132-140pp

Chylek, P., Lohmann, U., Dubey, M., Mishchenko, M., Kahn, R. and Ohmura, A. 2007. Limits on climate

sensitivity derived from recent satellite and surface observations. Journal of Geophysical Research.

112: D24S04pp. doi: 10.1029/2007JD008740.

Delhi Water Situation. 2003. New Sources of Water Vital for Delhi’s Health.

http://www.teri.res.in/teriin/news/terivsn/issue35/water.htm

Delhi Government. 2002. 4 Years: Working Report Delhi Government December 3, 1998- December 3,

2002.

Department of Urban Development .2010. Environment and Social management framework. A report by

Ministry of Urban Development, Govt. of India. September 2010. 58pp

Dhar, S. and Mazumdar, A. 2009. Hydrological modelling of the Kangsabati river under changed climate

scenario: case study in India. Hydrological Processes. 23(16): 2394–2406pp. doi: 10.1002/hyp.7351

Freshwater Year 2003. 2003. Vision for Integrated Water Resources Development and Management.

Report by Ministry of Water Resources, Govt. of India, New Delhi: 53pp

Gosain, A. K. and Rao, S. 2003. Climate Change and India: Vulnerability Assessment and Adaptation (eds

Shukla, P. R. et al.), Universities Press (India) Pvt Ltd, Hyderabad: 462pp

Goswami , B. N., Venugopal, V., Sengupta, D., Madhusoodan, M. S. and Xavier, P. K. 2006. Increasing

trend of extreme rain events over India in a warming environment. Science. 314. 1410-1442pp. doi

10.1126/science.1132027

http://www.jsk.gov.in/projection_report_december2006.pdf

http://www.rainwaterharvesting.org/

http://www.teri.res.in/teriin/news/terivsn/issue35/water.htm

Government of India. 2003. Ground Water in Delhi: Improving the sustainability through Rainwater

Harvesting, Central Ground Water Board, Ministry of Water Resources. Environment Centre for

Civil Society: 438pp

Government of India. 2003. Rainwater Harvesting: A necessity in South and Southwest Districts of NCT,

Delhi. State Unit Office, Delhi, Central ground Water Board, Ministry of Water Resources.

Government of India. 2003. Details on Water Harvesting. Accessed on various dates at

http://www.cgwaindia.com

Hingane, L. S., Kumar, K. R. and Murty, B. V. R. 1985. Long term trends of surface air temperature in

India. Journal of Climatology. 5: 521–528pp

IMD. 2009. New Statistical models for long-range forecasting of southwest monsoon rainfall over India.

Climatological Charts of the Indian Monsoon Area: 139pp

Inter-governmental Panel on Climate Change. 2007. Climate Change 2007: Synthesis Report.

Contribution of Working Groups I, II and III to the Fourth Assessment Report of the

Intergovernmental Panel on Climate Change. IPCC, Geneva, Switzerland: 104 pp. Available at

www.ipcc.ch/ipccreports/ar4-syr.htm. (Accessed 9 Jan 2010)

International Water Management Institute. 2008. IWMI Research in Southeast Asia.

www.iwmi.cgiar.org/Publications/Other/PDF/IWMI_South_East_Asia_Brochure.pdf (Accessed 2

Jan 2010).

Iyer, N. K. and Kulkarni, S. 2007. Economy, population and urban sprawl a comparative study of urban

agglomerations of bangalore and hyderabad, india using remote sensing and GIS techniques.

Paper presented at PRIPODE workshop on Urban Population, Development and Environment

Dynamics in Developing Countries Jointly organized by CICRED, PERN and CIESIN. 11-13 June 2007

Kundzewicz, Z.W., Mata, L.J., Arnell, N.W., Döll, P., Kabat, P., Jiménez, B., Miller, K.A., Oki, T., Sen, Z.

and Shiklomanov, I.A. 2007. Freshwater resources and their management. In Climate Change 2007:

Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F. Canziani, J.P. Palutikof,

P.J. van der Linden and C.E. Hanson, Eds., Cambridge University Press, Cambridge, UK: 173-210pp

Lall, S. V. and Mengistae, T. 2005. Business environment, clustering, and industry location: evidence from

Indian cities. World Bank Policy Research Working Paper November 3675: 2065pp

Lean, J. L. and Rind, D. H. 2008. How natural and anthropogenic influences alter global and regional

surface temperatures: 1889 to 2006. Geophysical Research Letter. 35: L18701pp.

doi:10.1029/2008GL034864.

Lopez, R. and Hynes, P.H. 2003. Sprawl in the 1990’s: Measurement, Distribution and Trends. Urban

Affairs Review. 38: 325-355pp

Mall, R. K., Gupta, A., Singh, R., Singh, R. S. and Rathore, L. S. 2006. Water resources and climate change:

An Indian perspective. Current Science. 90(12): 1610 – 1626pp

Mckenzie, D. and Ray, I. 2009. Urban water supply in India: Status, reform options and possible lessons.

Water Policy. 11(4): 442-60pp

Ministry of Urban Development, Govt. of India. Annual Report 2006-07.

Mookherjee, D. and Larvie, K. A. 2000. Urban Agglomerations in India, 1981-1991: Occupational

Correlates in Two Mega City Agglomerations. Ianos, I., Pumain, D., Racine, J.B. (Editors). Integrated

Urban Systems and Sustainability of Urban Life. Bucuresti, Editura Tehnica: 487-497pp

http://www.ipcc.ch/ipccreports/ar4-syr.htm%20%20Accessed%209%20Jan%202010

http://www.iwmi.cgiar.org/Publications/Other/PDF/IWMI_South_East_Asia_Brochure.pdf

Mookherjee, D. and Hoerauf, E. 2004. Cities in transition: monitoring growth trends in Delhi urban

agglomeration 1991 – 2001. Dela. 21: 195-203pp

Pant, G. B. and Kumar, R. K. 1999. Climates of South Asia. Chichester: John Wiley & Sons: 320pp (ISBN 0-

471- 94948-5)

Rodell, M., Velicogna, I. and Famiglietti, J. S. 2009. Satellite-based estimates of groundwater depletion in

India. Nature. 460: 999-1002pp. doi:10.1038/nature08238

Status Report for Delhi 21: Delhi Urban Environment and Infrastructure Improvement Project

(DUEIIP).2001. Government of India, Ministry of Environment and Forests and Government of

National Capital Territory of Delhi Planning Department. 7: 3-4pp

United Nations Environmental Program (UNEP) and Earthscan. 1999. Global Environment Outlook, UNEP

and Earthscan, Nairobi and London: 100pp

Varis, Olli. 2005. Externalities of Integrated Water Resource Management in South and Southeast Asia. In

A. Biswas et al. (eds.) Integrated Water Resources Management in South and South-East Asia. New

Delhi and Oxford. Oxford University Press: 1-38pp