The Jameel Observatory combines the local knowledge and concerns of communities facing on-the-ground threats of hunger with innovations in data science and humanitarian action;  teaming up to devise solutions that can help predict, prepare for and overcome climate-related food security and nutrition challenges in dryland areas.

This post is one of a series illustrating expertise that partners bring to the Observatory. It introduces ways that geospatial data is used to help assess and target where to allocate resources. It draws from work at the University of Edinburgh and its partners

Why is estimating poverty from geospatial data important in East Africa?

Over 35% of the East African population live on less than $1.90 per day and poverty is strongly related to food insecurity. There is evidence to suggest that the most successful food security assistance programmes are those that target chronic poverty rather than target feeding people. Thus, food insecurity could be helped enormously with up-to-date fine spatial-resolution poverty maps.

However, data on socioeconomic conditions collected through household surveys and censuses are time consuming to collect and therefore expensive, meaning that this type of data is collected less frequently than required. It is not possible to commission more household surveys or more frequent censuses because the costs would be prohibitive.

Research by Gary Watmough and colleagues at the university of Edinburgh explores how geospatial data such as remotely sensed satellite imagery can be used to predict socioeconomic conditions, filling the gaps between household surverys and census data collection.

Multi-resolution Earth Observation estimates of socioeconomic conditions

Using Earth observation to support sustainable rural development. Source:

In rural areas, environmental resources are a key component of people’s livelihoods. Therefore, thematic models of how households pursue their livelihoods can be built to identify metrics that could be derived from earth observation satellites.

The image illustrates 1) at the household level, how poverty can be proxied using earth observation classifications of building roof type, building size, agricultural field size; (2) at the community level, how poverty can be characterised by the types and amounts of roads connecting the community, and evidence of services such as schools and markets; (3) at a regional scale, how poverty can be partly characterised by access to larger towns and cities, agricultural productivity, and evidence of environmental shocks including droughts and floods.

In Assam in India, for example, we found that Earth observation metrics were able to predict which were the poorest villages with up to 61% accuracy, using travel time to market towns, percentage of a village covered with woodland and percentage of a village covered with winter crop significantly related to poverty.

In Western Kenya we found that Earth observation data were able to predict the poorest households with 62% accuracy with the size of individual building footprints being the most important predictive metric. More recently, we have been looking at how geospatial data can be used to predict changes in poverty over time in Mozambique. We are also examining how mobile phone call data records and social media data could be incorporated into our models.

A more global example is our work with the Data for Children Collaborative and UNICEF to produce a ‘Children’s Climate Risk Index.’ Utilizing high-resolution geographical data, this report provides new global evidence on how many children are currently exposed to a variety of climate and environmental hazards, shocks and stresses.

Relevance for the Jameel Observatory

Resilience to food insecurity has been found to be strongly linked to poverty and so any early warning system needs to consider the local livelihoods. Given the expense and time to collect household surveys, earth observation, call record and other geospatial data is likely to be required to provide uptodate estimates of socioeconomic conditions. We are working with colleagues in the University’s School of Infomatics to build models that can be run at fine spatial scales to provide updated information on livelihoods that can feed into estimations of resilience.

More information

Contact Gary Watmough

Read some related reports:

Satellite Earth observation to support sustainable rural development

Towards achieving the UNs data revolution: combining earth observation and socioeconomic data for geographic targeting of resources for the sustainable development goals

Using open-source data to construct 20 metre resolution maps of children’s travel time to the nearest health facility

Contact us for more information or to partner with us: