HOT Discussion Notes

Some quick stats on things we have mapped:



Place names:

The UN OCHA COD file of place names has ~700 towns in it for Mali.  All of these have been imported into the OSM DB with their OCHA unique identifier numbers.

The US GNS database ~14,000 place names of which ~2,000 have been imported into our database so far with residential areas traced for each of these.  Each GNS node has a unique ID number on it that identifies it in the GNS database.

All of these GNS places that are imported have been moved to their exact location with aerial imagery and where confusion exists about the name or which town it applies to, 'fixme' and 'note' tags have been added to describe the exact situation.

Where both OCHA and GNS nodes for the same place have been imported, we have de-duplicated them in our DB and kept the OCHA and GNS unique ID's allowing cross referencing from one database to the other.  See for example the town of Douentza:

http://www.openstreetmap.org/browse/node/495454624

At that link you can see the town name, an alternate spelling from one of the other databases, as well as the OCHA and NGA unique identifiers to allow objects that appear in both databases to be reconciled.

Here is an interactive map showing a live query with clickable points for all the places in north central Mali in our database (warning might be a farily big download).

You can change the bounding box coords to get other places in a similar fashion (or for the whole country).  The data can also be downloaded for offline use.

http://overpass-turbo.eu/map.html?Q=(%20node(14%2C-5%2C16%2C-1)%5Bplace%5D%3B%20)%3B%20out%3B



Water Wells:

We have mapped quite a few water wells in the country from aerial imagery.

Currently we have about 1,600 water wells mapped (although some of these are like 'false positives').

We would like to get some 'ground truth' checking on some of these to know how accurate our assessment of these wells is, both how many we are likely missing, and how many are false positives.

A map showing some of these wells:

http://overpass-turbo.eu/map.html?Q=(%20node(14%2C-5%2C16%2C-1)%5Bman_made%3Dwater_well%5D%3B%20)%3B%20out%3B



Cattle Chutes:

Probably less important for your work, but we did see a lot of special "pens" for putting cattle in with a long "neck" on the pen for lining the cattle up single file.

These are used for checking cattle one at a time and are important to the economics of many of these villages.

We currently have 121 of these mapped.  Here they are on a map:

http://overpass-turbo.eu/map.html?Q=(%20node%5Bman_made%3Dcattle_chute%5D%3B%20)%3B%20out%3B







We would like to know what towns/villages you have field workers in so that we can generate a list of things for you to ground truth in the area.

For example, we see a lot of water wells with odd looking "fans" around them in the aerial imagery.  Having on the ground pictures of these (and other features) would help us make better classifications.

Also, having your workers check them and add missing ones in a given "test area" will allow us to calculate false positive and false negative statistics, so we can give an indication of how "trustworthy" our well data is.






Due to the large size of the area we are mapping, and the empty nature of most of it.  The user PovAddict has put together a croudsourcing tool to allow non-technical people to examine imagery for us and mark features of interest.

This saves us a lot of time by not having to search largely empty areas ourselves and lets us focus on mapping.

Users are shown a small area of imagery and either click on the things we are looking for (which gets recorded in a database for our mappers to map later) or they click the "There is nothing here" button if the image is just empty desert.

We have used the tool to classify over 40,000 images just in the last two days, covering an area of ~6,000 square kilometers of area just south of Mopti.

The tool can be found online here:

http://stuff.povaddict.com.ar/mali-crowdsource/