Crowd sourcing and high resolution satellite imagery in public health

Crowd sourcing and high

resolution satellite

imagery in public health

Chris [email protected]

Improving health worldwide

www.lshtm.ac.uk

Screen grab of Google maps around LSHTM

Screen grab of Google maps in Tanzania

Crowd source mapping

Using people around the world to collect

and map features of interest into a

central location

• OpenStreetMap (OSM) & HOT

• Missing maps

Haiti earthquake 2010

Ebola: BRC online ebola map

Source: simonbjohnson.github.io

Benefits of OpenStreetMap

• Free, simple software to map area

• Shared workload

• Speed

• Meeting “open data” requirements

Satellite imagery

• Increasing in resolution

– Very high resolution imagery (VHR) now 30cm

• Costs reducing – free in emergencies

• Widely available

It is only with local knowledge and

previous experience that we can fully

generate datasets from satellite images.

What are the features in this image

Estimating populations using

satellite images

Am Timan, Chad, 2012

Stratum 1

Stratum 2

Stratum 3

Manual structure count

• Structures located by eye

• Type of structure determined by user

– Traditional hut

– Small building

– Large building

• Grid used to ensure

systematic counting

• Count checked

– Missed features / errors

Population estimates

Experience and being systematic are

vital when producing dataset from

satellite images.

Good image

Poor image

Dispersed population

Landing site

Population Density

Area

X

Population

density

Example of sensitivity analysis

Density 1 Density 2 Variable

Uganda 370,803 311,812 316,301

Kenya 152,128 138,575 139,767

Tanzania 555,177 429,689 473,575

Total 1,078,108 880,076 929,643

Density 1: 33,874 people per km

Density 2: 28,895 people per km

High density villages: 35,598 people per km

Low density Villages: 19,533 people per km

How to avoid main problems

• Know your software

• Experience counts

• Factor problems into proposal

• Two heads are better than one

• Be systematic

• Validate the method

Thank you.