Population cluster data to assess the urban-rural split and electrification in Sub-Saharan Africa

Abstract

Human settlements are usually nucleated around manmade central points or distinctive natural features, forming clusters that vary in shape and size. However, population distribution in geo-sciences is often represented in the form of pixelated rasters. Rasters indicate population density at predefined spatial resolutions, but are unable to capture the actual shape or size of settlements. Here we suggest a methodology that translates high-resolution raster population data into vector-based population clusters. We use open-source data and develop an open-access algorithm tailored for low and middle-income countries with data scarcity issues. Each cluster includes unique characteristics indicating population, electrification rate and urban-rural categorization. Results are validated against national electrification rates provided by the World Bank and data from selected Demographic and Health Surveys (DHS). We find that our modeled national electrification rates are consistent with the rates reported by the World Bank, while the modeled urban/rural classification has 88% accuracy. By delineating settlements, this dataset can complement existing raster population data in studies such as energy planning, urban planning and disease response.

Attention to detail – exploring effects on technology selection in geospatial electrification modelling

“…First, the population settlements that act as the basis of the analysis are generated. These clusters delineate each settlement in the area, and can be created from high-resolution raster data of population counts (see [37] for more details), or from building footprints, as further elaborated below.…”

“…A clustering tool was developed by Khavari et al to cluster high-resolution 1 population raster cells into settlement polygons based on the distance between cells [37], which can be used in the light version of OnSSET. Figure 2 displays the clusters developed by Khavari et al, overlaid on 1 km 2 raster data from WorldPop [41].…”

Engineering-Economic Evaluation of Diffractive NLOS Backhaul (e3nb): A Techno-economic Model for 3D Wireless Backhaul Assessment

“…This begins with identifying settlements, which in this analysis are defined as any area having more than 50 inhabitants per km 2 , with an overall total of 100 inhabitants. This is a common methodological step in telecom and energy strategy assessment [92]. Then it is necessary to find the nearest major settlement which has over 20,000 inhabitants [93], and therefore strong economics for existing digital connectivity, including a fiber PoP which can help route traffic to the Internet.…”

Accurate intercensal estimates of energy access to track Sustainable Development Goal 7

“…The existing works to understand energy access from EO data mostly employ generalized linear models as these models provide interpretability [ 7 , 22 , 24 ]. Some top performing machine learning methods for electrification prediction task are gradient boosting classifiers [ 25 , 26 ], logistic regression [ 27 ], Gaussian Process (GP) classification [ 12 , 27 ].…”