Mapping Imvepi Camp, Uganda
Geolocating Tweets with Tweetdeck, Spring 2017
This is a short blog about a technique we used to help source satellite images of Imvepi Camp in Uganda in 2017. We were able to locate the camp and start mapping it prior to the British Red Cross team’s arrival in the field, thereby helping them make an impact sooner. For more information about British Red Cross involvement in the Uganda Population Movement response, there are blogs from the BRC Communications team here and here.
Context
An escalation of violence in South Sudan in July 2016 led to a large population movement into the north of neighbouring Uganda.
One of the area’s first camps, Bidibidi, was originally anticipated to hold 40,000 people; however, at the start of 2017 it held 270,000 people, making it the largest refugee camp in the world at the time.
Elsewhere, camps at Palorinya and Rhino were similarly growing, and a new camp at Imvepi was under development, with displaced people already starting to arrive in the area.
Mapping with remote volunteers
In a rapidly developing situation like that, it is vital that accurate and up-to-date mapping information is available to allow those responding to be as effective and efficient as possible.
The British Red Cross were planning to deploy an Emergency Response Unit (ERU) team to Imvepi Camp to set up sanitation facilities and conduct hygiene promotion activities.
The team needed maps to enable sanitation facilities — including latrines, hand washing points and hygiene promotion activities — to be effectively provisioned for the people living in the camp.
OpenStreetMap (OSM) is our go-to source for geographic data, however in many parts of the world maps can be missing, and there was limited map data available in OSM for northern Uganda at the time.
There was also ambiguity about where Imvepi Camp was going to be formally located within northern Uganda, despite people already arriving in the locality and some sporadic/anecdotal data being sent from there by other aid agencies and journalists.
There was also a comparatively unique characteristic of these camps that made them harder to locate visually in satellite imagery: Ugandan Government policy provided a plot of land to each refugee household on which to construct a shelter and farm to support themselves.
Whilst this was a welcome improvement in the living conditions compared to other refugee camps globally, it meant that high density areas of tents or other semi-informal structures weren’t present in satellite images of northern Uganda. This made it more difficult to visually ascertain exactly where the camp actually was.
What we needed to do
What we wanted to do was narrow down an otherwise large area of northern Uganda where Imvepi camp might be, to an area that was small enough to be used as an Area of Interest (AOI) for a satellite image request.
We would then use that satellite imagery to map the camp on the HOT Tasking Manager, thereby helping the ERU team to do their work once they had arrived on the ground.
Technical
When someone takes a photo on a smartphone and has their GPS enabled, the location of the smartphone at the time the photo was taken is also captured by default.
The location (“geotag”) is stored in the accompanying metadata for that photo.
Twitter’s Tweetdeck application allows someone to search in the content of Tweets and also in the metadata of photos posted in Tweets.
That’s useful for two reasons:
- the location of the photo can be mapped from the metadata itself
- the content of the photo and Tweet can assist in working out other characteristics of the area the photo was taken that might be relevant to subsequent work (e.g. if the photo shows a densely populated place, we would need to make a slightly different task with different instructions on the HOT Tasking Manager compared to a more rural area)
What we did
Using the functionality in Tweetdeck, we set up searches for the camp names with location parameters that encompassed a large radius.
Each time we came across a geotagged Tweet with relevant content, we updated the search coordinates and/or reduced the size of the geographic search radius to narrow down the area.
Although the process was quite manual, we were able to narrow down the approximate location of the central reception area of Imvepi Camp to within a ~5km diameter area.
Using that approximate location, we created an Area of Interest (AOI) and used that to search for satellite imagery.
Using an AOI to search for relevant satellite images of the area on ApolloMapping’s ImageHunter
Outcome
The imagery we sourced allowed the global digital volunteer community to map the camp:
Once geographic information was added to OpenStreetMap, we made maps that were used by the ERU team on the ground when they arrived:
For example, we made maps that showed camp infrastructure combined with locations of latrines that the ERU team sent back to us from the field. We used these maps to analyse the distances people were living from latrines to see how they compared to Sphere Standards.
The data was and is available in OSM to anyone who wants to use it, including other agencies working in the area, who made other products like facilities and infrastructure maps.
Despite its relative simplicity, the geolocation technique we used in this case proved useful to enable us to source the satellite imagery we needed. If we had not been able to narrow down the area where Imvepi Camp was likely to be located, we would have needed to go through a time-consuming (and expensive!) process of manually downloading and looking through full resolution images to locate it ourselves, or waiting for the ERU team to be on the ground to send us coordinates back.
