Map South Kivu tasking
Map South Kivu
South Kivu is a province located in the eastern area of the Democratic Republic of Congo. This region has been affected by decades of armed conflict and violence against civil population including women and children. Living conditions in South Kivu are characterized by pauperization, insufficient food and water supply, lack of health services and poor sanitation infrastructure. This has had negative impacts on the overall health levels of the population. As reported by HOT, humanitarian aid has been rendered difficult because of poor road connectivity and inadequate infrastructure. Moreover, indiscriminate violence, looting and arson has posed permanent threats to the security of civil society and international staff.
The Humanitarian OpenStreetMap Team (HOT) activation that followed the 2012 cholera epidemic focused on several areas of South Kivu. These were zones around Lac Kivu, areas along the road Goma - Bukavu, and neighbouring zones of Bweremana and Minova. The main purpose was to trace the tracks of displaced population, however mapping was not possible in some areas where aerial imagery was unavailable. High resolution Bing imagery covered only some areas dispersed in the rest of the region.
The extension of the area to be mapped and the challenges presented in several stages of the mapping process demanded a permanent support from the mapping community beyond a HOT activation.
Map South Kivu was therefore created as part of the Missing Maps project. It seeks to provide base maps of South Kivu, DRC in order to enable a more efficient response by international and local organizations. The inclusion of South Kivu as a priority area for Missing Maps has enabled to Organize mapping events to map the identified regions with missing geodata, however, the state of the map in South Kivu has still much room for improvement.
Identifying areas and tasking
South Kivu is located in the eastern area of DRC and has an extension of approximately 65.000 km2 . In other smaller areas, Missing Maps has adopted a conventional way of identifying areas for mapping, designing the tasks, uploading changes and drawing features, and validating the results. This mapping process has proved to be adequate when mapping small locations, however it is insufficient when tackling an extensive region such as the province of South Kivu.
Many challenges have to be overcome when coming up with a strategy for mapping South Kivu. The first obstacle, as mentioned above, is the extension of the total area. With several thousands of square kilometres, the tasks have to reconcile a sufficient amount of features and changes yet not surpassing the average surface of a mapping square. The second obstacle has to do with the characteristics of the territory and the distribution of population in the region. In past tasks, volunteers have spent vast amounts of time scanning bush, jungle, clouds or even areas with no imagery at all, when what they should be doing is using their skills and experience to digitise features.
This challenges have pushed for an evaluation of the mapping process and have highlighted the need for an adequate strategy for mapping South Kivu. So, whereas before we might have had a single task to map an area, we are now breaking the process down into many different parts, both to expediate the process and to provide mappers with tasking that reflects their skill level and ambition.
Missing Maps has followed a five stage iterative process for making specific South Kivu base maps, two of which involve large scale crowd sourced data activities.
- Imagery analysis
- Identification of features
Aerial imagery is a basic input when mapping South Kivu, however not all aerial imagery is a good base for mapping and drawing features for two reasons. Firstly, some areas in South Kivu are covered with a dense forest among a few villages and settlements appear sparcely. In other cases, villages were covered with dense clouds, which represented a huge obstacle for mapping. Secondly, some areas in South Kivu have no aerial imagery coverage, or if they do, the image resolution is insufficient for identifying features.
In order to avoid spending valuable time scanning unpopulated areas and areas with poor imagery, Disaster Mappers at Heidelberg University developed a crowd sourcing tool which enabled volunteers to analyse the coverage of Bing satellite imagery across the South Kivu province. Volunteers were asked to assess whether there are human settlements or major roads in the satellite imagery for a given bounding polygon. The tool can be found in the Missing Maps South Kivu Region imagery analysis task.
The quality and the results of this crowd sourced imagery analysis are being evaluated by Disaster Mappers. By distinguishing true positives (TP), true negatives (TN), false positives (FP) and false negatives (FN) and referring to these as “classification status”, the imagery analysis will be a valuable tool for discarding truly unuseful imagery.
Given the extent of the province of South Kivu, the conventional tasks designed for the South Kivu Tasking Manager have been split into different tasks with more specific focuses. The following table shows the different tasks designed for South Kivu, including a brief description and the status. Every task is based on a geojson file which can be visualized in this uMap.
Identification of features
The next stage decided on was to replace endless scrolling in JOSM and iD by crowdsourcing the identification of man made features within South Kivu. A simple pybossa crowdsourcing task has been developed and, it is hoped, that the results of this will mean the tracing / tagging of major roads and residential areas will be much quicker and easier.
Tracing roads and residential areas
This stage of tasking can be done by beginner / intermediate mappers. Need to understand highway tags and landuse tags. Ideal for new mappers at mapathons or people with a little mapping experience.
Once the roads and residential areas in a certain area have been mapped, we can then export the residential area shapes and re-import them in to the tasking manager as arbitrary task shapes, meaning that building mappers can then very quickly and efficiently trace buildings without any scanning at all. See Task 1053 for Unity State, South Sudan for an example of how this works.
These tasks are ideal for beginners and people who have just a few minutes to contribute. Worked great for a corporate mapping party involving 900 people in one day, where each participant only had 50 minutes.
Validation is done by more experienced mappers who have a trained eye and can evaluate the work of others. Throughout the mapping parties in London, a smaller group of validators checks the uploads made by the larger crowd of new volunteers. This stage is crucial as it verifies the quality of the map being produced.
As with identifying features, validation has a percentage of accomplishment. Ideally, no feature should be uploaded to OSM without counting first with an external validation.
Iteration of drawing and validating is important for building up on what has already been mapped. For example, when residential areas and main roads have been created then validated, it seems logical that the next step is to trace buildings, minor roads, and other features, then validate them. This improvement in the map can potentially also reflect an enhancement in the volunteers' mapping abilities. In fact, several mappers who attend their first mapping party come back the next time as JOSM users, demonstrating that there is an increase in the ability and, more importantly, in the interest of mapping.
A few important questions keep surging in the discussions among Missing Maps collaborators and volunteers.
Whilst there are many experienced mappers contributing, the majority of the Missing Maps mappers are new. New mappers are vital to the Missing Maps project and we are doing well in terms of recruitment. However, compared to the mapathons of old where you’d have much smaller, but more highly skilled groups, quality can suffer.
State of the map
The basic base map is not sufficient and pre-existing tasks need to be reworked and improved. When is a map finished? What is the final stage of mapping? How to operationalize this?
New mappers are obviously slower than those with experience. The Missing Maps has to take as many of these mappers as possible on the journey from being interested / intrigued to being full members of the HOT community. In many cases, this will inevitably come down to their experience of humanitarian mapping. The more effective and supported and encouraged they feel, the more likely they are to continue to map in non-mapathon time, meaning an increase in speed and quality.
Improving task design
The progress volunteers make and the quality of their work is down to two factors: the tasking and the training / support. The training and support we are working through in the mapping parties - getting better all the time. The tasking still needs a lot of work…
Should we, as in the case of Banagassou, do a two stage task - big squares for road networks, small squares with very precise AOIs for buildings? If so, what is the best way to produce those AOIs? For Bangassou, the task was designed by drawing the areas around the main roads. Is this the best way (essentially making it into a three step task)? Should the road network task also include mapping the boundaries of residential areas (making it a two stage task)? Is there a better way?
Mapping missing maps /https://www.mapbox.com/blog/mapping-missing-maps/
Secondary benefits: the social experiences of HOT contributors https://www.openstreetmap.org/user/dekstop/diary/37438