Global Gridded Relative Deprivation Index (GRDI), Version 1

The Story of a Data Set: Documenting the Development of GRDIv1

Introduction

The Global Gridded Relative Deprivation Index (GRDI), Version 1 (GRDIv1) data set characterizes the relative levels of multidimensional deprivation and poverty in each 30 arc-second (~1 km) pixel, where a value of 100 represents the highest level of deprivation and a value of 0 the lowest. GRDIv1 is built from sociodemographic and satellite data inputs that were spatially harmonized, indexed, and weighted into six main components to produce the final index raster. Inputs were selected from the best-available data that either continuously vary across space or have at least administrative level 1 (provincial/state) resolution, and which have global spatial coverage.

Global Gridded Relative Deprivation Index (GRDI), Version 1

The GRDI v1 represents the degree of deprivation of the population at a 1 km spatial resolution. The values range from 0 to 100, with 0 representing the lowest deprivation level and 100 representing the highest.

GRDI v1 defines areas of poverty and deprivation using non-traditional inputs at a higher resolution than previously possible, and for the entire world.

The GRDI v1 dataset is of use to Earth scientists, demographers, policy makers, government planners and leaders, NGOs (including humanitarian and development actors), climate adaptation planners, emergency responders, and the general public.

GRDIv1 can be integrated with Earth-observation, hazards and climate impact, socioeconomic, demographic, vulnerability, and poverty data for a variety of applications.

Explore GRDIv1

Explore the Map with GRDIv1:

povmap-grdi-v1 Map

GRDIv1 Zonal Mean for Administrative levels 0, 1, and 2:

Admin 0 Zones

Admin 1 Zones

Admin 2 Zones

Methodology

Input Components

GRDIv1 used nine input data sets to create the six components, along with one ancillary data set for alignment and resolution:

The input data was harmonized into the six components using Esri ArcGIS geoprocessing and R (R Studio) tools, and eventually aggregated them into the final GRDIv1 data set:

 Workflow process of GRDIv1 from top to bottom. Rasters (blue) and shapefiles (red) at the top of this figure were processed with specific geoprocessing tools (yellow) according to the type of input data provided. The resulting component CSV files (green) were aggregated to create the GRDIv1 product. Note: all processing tools used Gridded Population of the World, v4 (GPWv4) for cell size and spatial reference; processing tools that used the National Identifier Grid raster are denoted with a dagger (†), and the processing tools that used the fishnet feature class are denoted with an asterisk (*);  

All input data sets were resampled to match the resolution and format of Gridded Population of the World, v4, the ancillary dataset used for cell size and spatial reference. Next, the six components of GRDIv1—BUILT, CDR, IMR, SHDI, VNL 2020, and VNL slope—were produced and matched by cell ID to a fishnet feature class to then create rasters for each component.

Indexing and Weighting

An indexing method was applied using the Raster Calculator according to the  Vulnerability Hotspot Mapping method developed by CIESIN  as the following: Each input component was winsorized for the 5% and 95% quantiles. The winsorized input components were then indexed from 0-100 according to how the input was interpreted, i.e. a low value in BUILT and a high value in CDR both imply high deprivation (see below). The indexed components had a weight applied according to the original dataset resolution where a coarser resolution was assigned a lower weight:

The resulting input components of GRDIv1, how they are interpreted in the first indexing process, and the subsequent weight that was applied before the components were aggregated.

Degree of built up area (BUILT)

The BUILT component is the ratio of built up area to non-built up area in a square kilometer. Global rural populations are more likely to experience a higher degree of multidimensional poverty when compared to urban populations, other things being equal.

We assumed the degree of BUILT as a dimension where low values imply higher deprivation.

Child Dependency Ratio (CDR)

The CDR is defined as the ratio between the population of children (ages 0 to 14) to the working-age population (age 15 to 64) where a higher ratio implies a higher dependency on the working population.

We assumed CDR as a dimension where higher values imply higher deprivation.

Infant Mortality Rates (IMR)

The infant mortality rate (IMR) is defined as the number of deaths in children under 1 year of age per 1,000 live births in the same year. IMR is a common indicator of population health.

We assumed IMR as a dimension where high values imply higher deprivation.

Subnational Human Development Index (SHDI)

The Subnational Human Development Index (SHDI) attempts to assess human well-being through a combination of three dimensions: education, health, and standard of living.

We assumed SHDI as a dimension where lower values imply higher deprivation.

Subnational Human Development Index (SHDI)

The Subnational Human Development Index (SHDI) attempts to assess human well-being through a combination of three dimensions: education, health, and standard of living.

We assumed SHDI as a dimension where lower values imply higher deprivation.

VIIRS Night Lights 2020 (VNL_2020)

The intensity of nighttime lights is closely associated with anthropogenic activities, economic output, and infrastructure development.

We assumed VIIRS Night Lights (VNL) 2020 as a dimension where lower values imply higher deprivation.

VIIRS Night Lights 2012-2020 Slope (VNL_slope)

We calculated a linear regression from annual VNL data between 2012 and 2020 and considered the slope as a dimension where high values (increasing brightness) imply decreasing deprivation and low values (decreasing brightness) imply increasing deprivation.

To produce the final GRDI v1 raster, the weighted indices of all input components were summed for each fishnet cell and divided by six (the number of components). The resulting fishnet values were converted to raster, which was indexed from 0-100, where 100 represents the highest deprivation level and 0 the lowest.

Acknowledgments

 These data were produced with funding from Socioeconomic Data and Applications Center (SEDAC). The production of this product was led by Juan F. Martinez and supported by Susana Adamo, Carolynne Hultquist, Kytt MacManus, Alex de Sherbinin, Elisabeth Sydor, Cascade Tuholske, and Greg Yetman.  

Disclaimer

 CIESIN provides this data without any warranty of any kind whatsoever, either express or implied. CIESIN shall not be liable for incidental, consequential, or special damages arising out of the use of any data provided by CIESIN. No third-party distribution of all or parts of this dataset are permitted without permission. These data are for noncommercial use; commercial use is not permitted without explicit permission. Additionally, users of the data should acknowledge CIESIN as the source used in the creation of any reports, publications, new data sets, derived products, or services resulting from their use. CIESIN also requests reprints of any publications acknowledging CIESIN as the source and requests notification of any redistribution efforts. The Trustees of Columbia University in the City of New York hold the copyright on data created at CIESIN. CIESIN obtains permissions to disseminate data produced by others. Intellectual property rights and permissions associated with each particular data set are specified in the documentation of the data. 

Use Constraints

 This work is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0). Users are free to use, copy, distribute, transmit, and adapt the work for commercial and non-commercial purposes, without restriction, as long as clear attribution of the source is provided.  


CIESIN | Columbia Climate School, Columbia University

The Trustees of Columbia University in the City of New York.

 CIESIN is a center within the Columbia Climate School at Columbia University. Copyright© 2022 

 Workflow process of GRDIv1 from top to bottom. Rasters (blue) and shapefiles (red) at the top of this figure were processed with specific geoprocessing tools (yellow) according to the type of input data provided. The resulting component CSV files (green) were aggregated to create the GRDIv1 product. Note: all processing tools used Gridded Population of the World, v4 (GPWv4) for cell size and spatial reference; processing tools that used the National Identifier Grid raster are denoted with a dagger (†), and the processing tools that used the fishnet feature class are denoted with an asterisk (*);  

The resulting input components of GRDIv1, how they are interpreted in the first indexing process, and the subsequent weight that was applied before the components were aggregated.