The document presents a comprehensive analysis of the aid relief response to the 2015 Nepal earthquake, highlighting the need for effective data collection and distribution strategies among humanitarian agencies. It discusses the estimation of aid needs based on severity indices and the effectiveness of aid delivery in various village districts. Key findings indicate complexities in aid coverage and the importance of pre-existing conditions in determining aid distribution efficiency.

1. Georgi D. Gospodinov, PhD
Sr. Data Scientist, Wal-Mart Stores Inc.
Chris Shannon
Sr. Data Scientist, Tata Consultancy Services
Aidan McGuire
CEO at ScraperWiki Ltd and HDX Product
Manager at UN OCHA
Javier Teran
Statistician at United Nations OCHA
Andrej Verity
Information Management Officer UN OCHA
and
an international group of data science
volunteers
Nepal Earthquake Analysis
Aid Relief Coverage, Severity
Index, and Aid Capacity Index
September-December, 2015
United Nations OCHA
http://www.inform-index.org
http://www.osgeo.org
http://www.worldpop.org.uk
https://data.hdx.rwlabs.org

2. Nepal Earthquake Analysis: Table of Contents
Table of Contents:
 Introduction and Background
• Humanitarian Data Exchange: Data Collection Process
• Humanitarian Systems Interaction
• 2015 Nepal Earthquake Disaster at a Glance
• Casualties and Injuries in the Nepal Earthquake
 Aid Agencies and Disaster Relief for Nepal
• Nepal Earthquake Disaster Relief Analysis
• Agency-VDC Aid Network Construction
• Properties of the Agency-VDC Aid Network
 Definition and Estimation of Need for Aid
• Definition of Need for Aid and Initial Model
• Severity Model Background: Hazard and Exposure
• Severity Model Background: Vulnerability
 Aid Coverage Analysis
• Need for Aid Analysis at VDC Level: Heatmap
• Need for Aid Analysis at VDC Level: Coverage
 Agency-VDC Aid Capacity Index
• Agency-VDC Aid Capacity Index Construction
• Agency-VDC Aid Capacity Index Analysis
 Conclusions

3. • 600 indicatorsfrom over 30 data
sources
• Includesdata from acrossthe
programmecycle:
• Country context (languages,
currency, of icelocations, etc)
• Preparednessdata
• Operational data
• Humanitarian financing
• Geospatial data
• Accessdata
The Common
Humanitarian Dataset
Introduction and BackgroundHumanitarian Data Exchange: Data Collection Process
https://data.hdx.rwlabs.org
Towardsreal-time, granular data

4. Humanitarian Systems Interaction
http://digitalhumanitarians.com/content/what-we-do-
Introduction and Background

5. 2015 Nepal Earthquake Disaster at a Glance Introduction and Background
http://www.icimod.org/nepalearthquake2015

6. Casualties and Injuries in the Nepal Earthquake
http://data.unh cr.org/nepal/
Introduction and Background

7. Nepal Earthquake Disaster Relief Analysis
UN’s Initial appeal for aid to assist with Nepal disaster: $415 million
• The Nepalese government lead the response through the National Emergency
Operations Centre (NEOC) with support from the United Nations and international
community
• The $415 million was just for the first three months of operations in Nepal
• Food ($128 million), health ($75 million) and shelter ($50 million) would make up the
majority of the spending
• The main aims were to combat outbreaks of communicable diseases, to ensure the food
needs were being met and to build basic shelters
• This $415 million was to address immediate need and did not include much money for
rebuilding
Aid Agencies and Disaster Relief for Nepal
http://www.un.org/apps/news/story.asp?NewsID=50725#.Vxx6wmPC5GJ

8. Agency-VDC Aid Network Construction
Overview:
• The Nepal disaster prompted
nearly 80 major agencies to send
some type of aid (in the form of
supplies, food, assistance, money,
etc.) to about 800 distinct Village
Development Committee (VDC)
districts. There were many
channels of aid delivery, and the
distribution of aid varied in
amount, diversity, and number of
different contributing agencies.
• We depict the agency providing the
aid and the particular VDC district
receiving the aid, with the heatmap
showing the districts with most aid
received
• We used data from all sources on
the first page (HDX, WorldPop, and
OSGeo/qGIS, see title page) to
represent the VDC map accurately
• We used aid data to overlay the
abstract aid agency representation
Aid Agencies and Disaster Relief for Nepal

9. Properties of the Agency-VDC Aid Network Aid Agencies and Disaster Relief for Nepal
Aid Agency Network:
• Distinct aid instances:
Min. 1st Qu. Median Mean 3rd Qu. Max.
1.00 5.50 18.00 40.64 40.00 256.00
• Distinct targeted VDCs with aid:
Min. 1st Qu. Median Mean 3rd Qu. Max.
1.00 4.00 9.00 17.67 22.50 127.00
• Shared VDCs with a given agency:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0 5 11 13 20 53
• Weighted Shared VDCs:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00 12.00 32.00 51.89 76.00 364.00
VDC Aid Network:
• VDCs with common aid source:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00 24.00 68.00 75.86 113.00 309.00
• Weighted VDCs with common source:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.0 48.0 136.0 157.5 226.0 662.0

10. DISTRIBUTION OF NEED FOR ALL VDCs
VDC NEED FOR AID RANK
FREQUENCYOFNEEDRANKOCCURENCE
0 10 20 30 40
0200400600800
LOW
NEED
64.1%
MED
NEED
33.0%
HIGH NEED 2.9%
Definition and Estimation of Need for AidDefinition of Need for Aid and Initial Model
Severity = (Hazard x Exposure × Vulnerability)^ (1/3)
Assume Need is proportional to Severity
Assume Need is inversely proportional to (Distance to Epicenter) ^(1/3)
Assume Need is inversely proportional to (Health Care Facilities) ^(1/3)
Need Stats:
Min. : 0.000
1st Qu. : 0.570
Median : 1.860
Mean : 4.197
3rd Qu. : 5.595
Max. : 85.530
• Estimating how well the supplied aid has met the need
for each VDC district is a difficult task because it is hard to
quantify need. The effort in this work is to outline a
measure of relative need for aid among VDC districts and
to compare how well aid resources were distributed
accordingly.
• We define need using the severity index which
incorporates measures such as hazard level, exposure
level, and vulnerability: all quantities characterizing the
pre-existing conditions for the population, the distance to
epicenter, and the weighted density of health centers.
http://www.inform-index.org
http://www.inform-index.org/InDepth/Methodology

11. Severity Model Background: Hazard and Exposure
Severity Model Background: Hazard & Exposure
The hazard & exposure dimension reflects the probability of physical exposure associated with
specific hazards. There is no risk if there is no physical exposure, no matter how severe the
hazard event is. Therefore, the hazard and exposure dimensions are merged into hazard &
exposure dimension. As such it represents the load that the community has to deal with when
exposed to a hazard event. The dimension comprises two categories: natural hazards and
human-induced hazards, aggregated with the geometric mean, where both indexes carry
equal weight within the dimension.
Definition and Estimation of Need for Aid
http://www.inform-index.org
http://www.inform-index.org/InDepth/Methodology

12. Severity Model Background: Vulnerability
Severity Model Background: Vulnerability
The impact of disasters on people in terms of number of people killed, injured, and made homeless is
predominantly felt in developing countries while the economic costs of disasters are concentrated in the
industrialized world. The Vulnerability dimension addresses the intrinsic predispositions of an exposed
population to be affected, or to be susceptible to the damaging effects of a hazard, even though the
assessment is made through hazard independent indicators. So, the vulnerability dimension represents
economic, political and social characteristics of the community that can be destabilized in case of a hazard
event. Physical vulnerability, which is a hazard dependent characteristic, is dealt with separately in the
hazard & exposure dimension.
Definition and Estimation of Need for Aid
http://www.inform-index.org
http://www.inform-index.org/InDepth/Methodology

13. • This model is a
measure of relative
need as absolute need
values are exceedingly
challenging to
estimate.
• The primary
conclusion is that
need is in a large part
determined by pre-
existing conditions
and risks and thus can
be anticipated and
prepared for.
Aid Coverage AnalysisNeed for Aid Analysis at VDC Level: Coverage
% VDCs that Received Aid From
Each Group: Low Need, Medium
Need, High Need :
• Low Need: 64.1%
• Medium Need: 33.0%
• High Need: 2.9%

14. Aid Coverage AnalysisNeed for Aid Analysis at VDC Level: Coverage
• Our basic findings are that the
distribution of aid agencies
shows that some VDCs
received aid from as many as
12 agencies, with the average
number of agencies supplying
aid per VDC equal to 2.
• From the Medium to High
Need VDCs, only about 70%
received some type of aid,
while from the Low Need
VDCs, only about 40%
received some type of aid,
according to the data record.
• This raises the question of
coordinating aid distribution
and creating a consolidated
aid coverage of a region.
Especially for areas with high
estimated need and a single
agency aid source.
• The figure depicts all the
VDC regions where aid was
not distributed or was
distributed in very limited
quantities thus creating a
deficit of disaster relief.
• Comparison with the
available resources reveals
opportunities for a more
efficient and effective aid
delivery, which is a subject of
ongoing study.

15. Agency-VDC Aid Capacity Index Construction
Aid Capacity Index is a function of the weighted
sum of the VDC Degree in the Aid-Agency
Network, weighted by each Agency’s Hub score
Agency-VDC Aid Capacity Index
Agency-VDC Aid Capacity Index
The need distribution and aid coverage analysis raise
the question of coordinating aid distribution and
creating a consolidated aid coverage of a region.
Especially for areas with high estimated need and a
single agency aid source. The problem is creating
adequate aid coverage of a region prior to disaster,
using pre-existing conditions to help inform us of the
expected level of need in case of a disaster.

16. Agency-VDC Aid Capacity IndexAgency-VDC Aid Capacity Index Analysis
Agency-VDC Aid Capacity Index Analysis
• Near liner distribution agency distribution overlap
• Highly non-linear aid distribution density
• Pre-existing factors such as vulnerability and severity
index good indicators for aid capacity estimation since
they capture secondary effects of infrastructure,
access to medical aid, food, and supplies
• Aid capacity can largely be impacted prior to a
disaster, with a global map for aid delivery channels

17. Nepal Earthquake Analysis: Conclusions
Conclusions:
• Humanitarian data efforts enable the global impact
of data for humanitarian purposes
• Aid agencies and disaster relief coordinators have
many opportunities to improve the process to
address need for aid more effectively
• The need for aid distribution is largely dependent
on pre-existing conditions
• Aid coverage is a highly complex non-linear process
• Agency-VDC Aid Capacity Index is a way to capture
the potential ability for appropriate aid coverage to
a region should disaster occur