Comparing urban methods

Map comparison is a complicated process, as there are many tools and approaches. In this project, we are comparing two methdologies for quantifying urban, both of which produce multiple binary rasters of urban areas:

Degree of Urbanization

Urban area	Min Pop Density	Min Settlement Pop
Urban areas	300 people/km2	5000 people
High density urban areas	1500 people/km2	50000 people

Dartboard

Urban area	Definition
Urban areas	contiguous pixels for which the density is above the 95th percentile of the counterfactual
Cores	contiguous pixels within urban areas that are above the 95th percentile of the counterfactual within the urban core
Cities	urban areas that have a core

Any attempt to compare these methods has to start with a question of what do we compare: the DoU method produces two layers, and the DB method produces three. After investigating the data it is clear that the DoU urban area is equivalent to the DB urban areas. However, the comparison of the DoU high density to the DB could be to either the Cores or the cities. In the figures below you can see the comparison to both, however, let’s look at the nature of the comparison:

The code to compare the data can be found here (LINK FORTHCOMING), for both comparisons, we use the table below to define agreement

	DB Core/City	DB Urban Area	DB Rural
DoU High Density	High Density	Disagree Urban	Disagree Rural
DoU Urban Area	Disagree Urban	Urban	Disagree Rural
DoU Rural	Disagree Rural	Disagree Rural	Rural

Kenya comparison

Using these categories the maps below were created to compare the DB classes. Based on these results we will focus on the comparison of the DB cores to DoU high density areas.

DB Cores - Nairobi

Fig. 1 Nairobi zoom: comparing DB cores to DoU high density areas

DB Cities - Nairobi

Fig. 2 Nairboi zoom: comparing DB Cities to DoU high density areas

DB Cores

Fig. 3 Comparing DB cores to DoU high density areas

DB Cities

Fig. 4 Comparing DB Cities to DoU high density areas