Update Types_of_location_graphs.md

docs/Types_of_location_graphs.md

@@ -3,6 +3,7 @@
 ![](images/network.png)
 
 The location graph is an essential part of the Flee simulation, and can be constructed in three different ways:
+
 * Location-based
 * Region-based
 * Grid-based
@@ -37,10 +38,12 @@ Typically all relevant regions are included in the graph, irrespective of their
 
 ### Obtaining conflict events
 In this approach, we first filter ACLED data based on the desired admin level. ACLED data usually has four columns for locations levels: 
+
 * admin1    
 * admin2    
 * admin3    
 * Location
+
 For the first three levels, we filter the dataset and select the first date these locations had conflict with over 1 fatality. The result will be the list of locations at the desired admin level. To find the coordinate of these locations, we use the coordinates of the largest cities in the region. In terms of the population, we use the total number of population of each admin level.
 For each region (Admin level), if we had a town with the population more than a threshold, first we subtract that population from its region population and then add it to the locations.csv as a town location type.
 For route construction, we use the ORS tools plugin in QGIS and then once we have the results, we prune the list of the routes.
@@ -60,6 +63,29 @@ If a conflict occurs in a region, then the conflict value should be set higher t
 * Regions may have arbitrary sizes and shapes, introducing erratic properties in simulations.
 * Still unclear how to define conflict intensities, e.g. for small events.
 
-## Grid-based location graph construction
+## Grid-based location graph construction (draft)
+
+In this approach, we start out with a population database that provides estimates across a country in a grid-like fashion. Grid cells may be perfect squares (equal in height or width in kilometers), or be formed accoding to world coordinate offsets (which gives them a slightly curved shape, depending on the distance from the equator).
+
+A grid-based location graph requires several steps to construct:
+
+* Obtain grid-based population data information.
+* Establish a population threshold per grid cell: if the population is higher than the threshold in the cell, then it should be a vertex in the graph. If not, then it should be omitted.
+* Obtain conflict data from ACLED and maps the conflict events to the original population data grid.
+* Establish a (lower) population threshold per grid cell that has a conflict: if the population is higher than the threshold in the cell, then it should be added as a vertex in the graph. If not, then it should be omitted unless it has been added already in an earlier step.
+* Map the conflict events to the grid cell IDs in conflicts.csv.
+* Generate the routes by calculating the distance between the centre points of each grid cell. This could be done using a route planner if the grid is sparse, or using a straight line distance (if the grid is dense and/or there are very few obstacles).
+
+### Advantages
+
+* All populous areas are included.
+* Grid-based population data is relatively reliable.
+* Flexiblity in the level of detail.
+* Consistency in the level of detail.
+* Output data is not tightly linked to individual roads.
+
+### Disadvantages
 
-To be written.
+* Existing Flee default rule sets are unlikely to work with grid-based graphs, and it is likely we will need to develop new rules altogether to make this accurate.
+* Ambiguity when a major obstacle resides *within* a grid cell.
+* Ambiguity when routes between grid cells should be omitted due to the presence of obstacles.