2. Previously created data from the area

Knowing where people live is crucial to provide timely and effective relief in humanitarian crises. Based on work in the "Missing Maps" project, I developed a method to obtain cheap estimates of the population density, that combines automated processing of satellite imagery and a crowdsourced, manual classification collected through the MapSwipe app. The method produces high resolution population distribution data.

2.1. High population density area

The high population density areas are demarcated by settlement extents provided by processing Landsat images (Landsat 7, Landsat 8 - spatial resolution is 30m - or Sentinel-1 and Sentinel-2 - spatial resolution is 10m). The "urban surfaces" from the landcover raster data give us the "Settlements pattern".

Disadvantages:

• Spatial resolution - small settlements and farms remain hidden
• classification error - "bare soil” anomaly"

2.2. Low population density area

Low population density areas are demarcated using: VGI - Volunteered geographic information. The used MapSwipe application helps to determine the inhabited area in the field. What is the MapSwipe?

The international medical humanitarian organization Doctors Without Borders/Médecins Sans Frontières (MSF) released the MapSwipe application on 15th of July 2016, which allows users from anywhere to easily contribute to the Missing Map missions using their phone. MapSwipe uses satellite imagery divided into tiles. Currently each tile is approximately 256 x 256 pixels (24 000 m2), the spatial resolution is around 150 meters. The user may assign for each of tile the following:

• Not tapped: There is no house and/or road on this tile.
• Tapped once: Yes, there are houses and/or roads on this tile.
• Tapped twice: Maybe there are houses and/or roads on this tile.
• Tapped three times: Bad imagery

The received JSON data from the server contains the next information at the moment from each user:

{
"id":"18-135000-123489",
"user_id":"jd7qEV3E2KSzENl6yaXAI5q40Z12",
"project":6,
"timestamp":1473820644490,
"task_x":"135000","task_y":"123489","task_z":"18",
"decision":1,"yes_count":1,"maybe_count":0,"bad_imagery_count":0
}

During transformation the JavaScript collects the answers for each tile, creates a polygon for each tile and saves to a geojson file.

{
"type":"Feature",
"properties":{"1":2,"2":1,"3":0,"prj":5,"cat":"yes"},
"geometry":{"type":"Polygon","coordinates":[[[-11.317291259765625,9.86604011578547],[-11.317291259765625,9.867393094500857],[-11.31591796875,9.867393094500857],[-11.31591796875,9.86604011578547],[-11.317291259765625,9.86604011578547]]]}
}

The two data validate each other.