ObjectVision / NetworkModel_World

GNU General Public License v3.0
0 stars 0 forks source link

Merge settlement layers #9

Closed jipclaassens closed 10 months ago

cjacobscrisioni commented 11 months ago

... in order to define settlement hierarchies / local and regional centres using the networkmodel_world config.

Methodology as in https://rsaiconnect.onlinelibrary.wiley.com/doi/abs/10.1111/pirs.12727

cjacobscrisioni commented 11 months ago

... ought to be followed by a comparison with mozaiked EU settlement hierarchies.

cjacobscrisioni commented 11 months ago

Note - currently integrating 1km grid points immediately in the network definition. Need to revisit the grids later as I will have to vectorize the grids at a late stage.

@jipclaassens my plan is to take the mollweide grid, turn every grid cell into an intermediate union(x4) of the grid corners, reproject the corners to WGS84, and turn the resulting lat/lon grids into a polygon. Just to check, am I missing a more efficient hidden GeoDMS operator to turn grids into vectors?

jipclaassens commented 11 months ago

I just discussed this with @MaartenHilferink. He pointed out that he had already proposed a solution to a similar problem for Bas in 2up. You'll get distorted results if you use connect to points in wgs84. He, therefore, added functionality to transform the roads and points to Mercator and connect it there.

For your case, where you need an approximate 1km grid pointset, you'll also get a distortion towards the poles. Since we need to analyse at a local level, he suggests making horizontal banded study areas, e.g., every 5 degrees. In that way, if you use Mercator, you'll limit the distortion. Also, when using convolution, on the equator, you'll go ten cells west and east and ten cells north and south, whereas in Norway, you'll still go ten cells north and south but 100 cells west and east.

Implementation in 2UP, is now 2BURP: /Input/Socioeconomic/Infrastructure/road

cjacobscrisioni commented 10 months ago

Status: work underway. More complicated than anticipated. Ferry routes are typically not encoded in OSM, so that continents with many islands (ie Oceania) typically have many unconnected networks (eg Australia and New Zealand). To ensure that we maintain relevant unconnected networks, I am extending the connected network routine: 1) split mulitpolygons into separate entities; 2) select all roads that are the largest subnetwork from at least one split polygon.

cjacobscrisioni commented 10 months ago

Ferries downloaded manually through https://overpass-turbo.eu/ . I will add them as a separate datasource and include them in all continental networks

cjacobscrisioni commented 10 months ago

Settlements merged successfully. Network issues resolved (and off-topic anyway)