It has been a while since I wrote a diary entry. However, I think now is a good time to inform the community about the progress that has been made in Epworth.

Context

Epworth is a settlement 12 km outside Harare, the capital of Zimbabwe. It has an official population of around 150 k but it is thought that the actual population could be three or four times greater.

Epworth is being mapped at building-level resolution for various humanitarian projects with Medecins sans Frontieres (MSF) being undertaken in the settlement. These include assessing access to water and sanitation facilities. The mapping was performed through Missing Maps.

We have created a wiki page for this project called Epworth Mapping Project.

Addressing system

The settlement is unusual in its addressing system. There are seven wards which are official boundaries, and within each ward there are plot numbers. These plot numbers do not relate to any roads like most addresses - they are simply ‘1201 Jacha’ rather than ‘14 Smith Road, Jacha’. The newer areas of Epworth seem to be numbered in a way so that their numbers are unique across all of Epworth - they are much larger (often over 10000).

To make things even more complicated, the local population does not commonly use the wards in descriptions of where they live - there is a whole unofficial hierarchy of placenames on top of the official wards.

Furthermore, the addresses relate to plots rather than buildings. When the settlement was founded around 100 years ago the plots were rather large. As the settlement has increased in population more buildings have been constructed on the existing plots, so that there many be several families living on one plot. The lack of definite boundaries makes determining which buildings are in which plot rather difficult at times.

Survey details

A survey of around 20-25% of Epworth has been performed and information pertaining to water facilities as well as official and unofficial addresses has been gathered. This is all valuable and so we have been working on inputting it into OpenStreetMap in a logical manner which is consistent with existing protocols and traditions.

The information was gathered on Field Papers with the information usually kept on a separate sheet which was cross-referenced to locations on the field papers. The surveyors were also local residents of Epworth, but they may not have been well-acquainted with the area of Epworth they were surveying. This has all been input to OSM by hand but as described below the manner in which we entered the information changed over the course of the project. I am currently trying to standardise it all.

Volunteer activities

This project used volunteer effort to put the data on the field papers into OSM. By the end, the process was written up into a tutorial for the volunteers that was edited based on feedback and also when the tags/procedure changed. The process involved choosing a field paper using a makeshift tasking manager in Googledocs, downloading the field paper into JOSM using the Field Papers plugin, downloading the existing OSM data and adding tags to features based on the data in the cross-referenced lists written by the surveyors. This information was then uploaded to the OSM server, and the tasking manager updated.

The HOT tasking manager could not be used because it is set up for base map creation tasks rather than survey tasks. There may be the capabilities for tasks such as this one but we weren’t aware of them.

Evolution of the tagging procedure

## House numbers

At first, we had a system where the house number would be input with the tag name=*. This was because it was thought that addr:housenumber was only useful with a corresponding addr:street tag. The streets in Epworth don’t have names, so this would not have been possible.

Further investigation found that addr:housenumber has been approved for use when used in conjunction with place=*. However, this is a rather vague tag because there are many different levels of place within OSM. My thanks go to @rab for the discussions and work he did on this.

Settlement hierarchy

The original hierarchy chosen for the unofficial settlement names was suburb > hamlet > neighbourhood, with Epworth being the suburb. However, this was not ideal because of the definition of a hamlet; you do not tend to get hamlets within larger urban areas or neighbourhoods within hamlets. However, this first approach worked to help us understand at which level of the hierarchy the various parts of each address entered on the field papers belonged to. It was often not apparent and since there are no official boundaries, it is partly down to the opinion of the surveyors and the residents.

Another problem with the hierarchy described above was that putting a house number and hamlet into the OSM search bar did not return any results. For this reason, the addr:hamlet was altered so that what were hamlets were now simply places - i.e. addr:place=*. This enabled searching and also eliminated hamlets from the settlement hierarchy.

However, this is still not perfect. My current plan is to alter the settlement hierarchy again. After checking the wiki, it appears that changing these addr:place tags to addr:quarter would be consistent with current OSM tagging rules as well as being a relatively good fit from a conceptual point of view. I hope that house numbers combined with quarters are searchable.

Thus, the final hierarchy would be suburb > quarter > neighbourhood.

Where to put the tags

Another issue we came across was where to put the tags. The addr:housenumber tag applies to the plot rather than the individual buildings, and so it makes sense to place a node in the middle of the plot with that information. However, there was also information about the buildings themselves, such as building material (see below in the section ‘Other tags’).

At first we weren’t sure about whether to tag the buildings themselves or create standalone nodes in the plots. It was a mish-mash as we tried to teach volunteers how to attach the information whilst trying to develop the protocol ourselves as we became more and more acquainted with the data. We opted for the standalone nodes.

However, I recently opened up JOSM to do some editing and it became apparent that these standalone nodes made editing quite difficult because the map was so messy. I resolved to attach the information from the node to the biggest building on the plot. This made the map much easier to observe and understand.

I could have attached the information to every building on the plot. However, there are two problems with this. Firstly, there would be many more numbers on the map without the conveyance of much more information. If you are looking for someone at a particular plot number, finding the biggest building is the greatest challenge. Secondly, as mentioned above it is not always clear where the boundary between the plots lies and therefore which building is in which plot.

Other tags

The building material of many of the buildings was recorded, and although the surveys had information about the type of brick used, this was not included in OSM because it does not fit into the building:material tag and is of limited use and relevance.

The ward in which the buildings are located was recorded in some cases. This has been included so that any errors in the current boundaries of the wards can be identified and corrected.

Next steps

We would like to complete the mapping of Epworth so that all house/plot numbers are included on the map. This is going to be quite expensive since it requires hiring surveyors again.

We are considering using smartphones rather than paper maps for the surveying, but this might take the volunteers longer to process than simply writing on field papers. However, there were issues around the legibility of the field papers cross-references which may have led to spelling mistakes in the data incorporated into OSM. This would likely be reduced in a direct inputting system. However, the existence of a set hierarchy and tagging system means that using apps such as opendatakit (ODK) would be possible and could lead to good quality data which would take less formatting by volunteers.

Any questions and comments are welcome!