At it's core,rapida is a python library and command line tool built on top of several curated geospatial datasets associated with specific geospatial risk's exposure components /variables capable to perform semi-automated assessments (zonal statistics) over a specific area of interest.
Essentially, rapida operates with a special folder or project that contains
A typical rapida session consists of following steps:
flowchart LR
i([🟢 Init]) --> a([🔐 Auth])
as{{🔍 Assess?}}
%% Switch to top-to-bottom (TB) direction for vertical alignment
subgraph vertical [ ]
direction TB
as <-->|🌥️cloud / 🖧remote| db[(🌍 Data)]
end
a --> vertical
vertical --> u{{📤 upload}}
u{{📤 upload}} --> p{{🌐 publish}}
- fetching/downloading and curating (h3id) administrative boundaries from OSM / OCHA@COD sources
- conducting geospatial assessments over and area of interest containing geospatial polygons
- buildings
- deprivation/poverty
- electrical grid
- GDP
- land use
- population
- roads
- integration with UNDP Azure cloud storage and UNDP GeoHub
- concept of project (create, list, delete, upload, download)
- integration with JupyterHub
- geospatial visualization
- notebooks
- rich UI/UX
- GDAL www.gdal.org
- rasterio https://github.com/rasterio/rasterio
- exactextract https://github.com/isciences/exactextract
- geopandas https://github.com/geopandas/geopandas
- rich https://github.com/Textualize/rich/tree/master
- click https://github.com/pallets/click
- tensorflow https://www.tensorflow.org/
... and others
Important
rapida operates with public datasets. However, some of these datasets rae currently being hosted in GeoHub. This is for two main reasons. First, in case a data source is available from one source only it was moved to Geohub to create a backup. Second, some data sources like population have been also curated and optimized as to facilitate the assessment process.
rapida is about conducting geospatial assessments on geo layers. Various properties/aspects of the layers are assessed or evaluated over an area of interest. The variables can be grouped into high level components that correspond to major areas of evaluation in the context of crises like : population, built environment (buildings, roads), natural environment (land use/cover, livestock), socio-economical environment (GDP, relative wealth index, deprivation,HDI).
🧑🤝🧑 Population
Sourced from WorldPop project, the components provides country based population data in raster format. The constrained (conditioned by buildings), UN adjusted version was selected as it was considered the best match. As the most recent population dataset was generated for 2020 rapida is forecasting the 2020 population using national data from World Bank that is available for several year in the past in respect to 2020. From there a coefficient is computed by dividing the target or requested year (ex 2025) to 2020 and the population statistics computed for 2020 or base year are multiplied with this coefficient.
🏙️ Built environment
The built environment refers to the human-made surroundings that provide the setting for human activity. This includes all physical spaces and infrastructure constructed or significantly modified by people.
- Buildings
This dataset merges Google's V3 Open Buildings, Microsoft's GlobalMLFootprints, and OpenStreetMap building footprints. It contains 2,705,459,584 footprints and is divided into 200 partitions. Each footprint is labelled with its respective source, either Google, Microsoft, or OpenStreetMap. It can be accessed in cloud-native geospatial formats such as GeoParquet, FlatGeobuf and PMTiles.
So far two spatial variables have been defined: the number of buildings per polygon to assess and the cumulative area of buildings per polygon to assess.
- Electrical grid
This dataset provides a global-scale, high-resolution predictive map of medium- and low-voltage electricity distribution networks. Developed using open-source geospatial data, satellite imagery, and machine learning models, it fills critical data gaps in power infrastructure—particularly in underserved or data-poor regions. The dataset supports energy access planning, infrastructure development, and policy-making.
The distribution network layout was estimated using predictive modeling based on spatial features such as population density, proximity to existing infrastructure, land use, and OpenStreetMap data. Models were trained and validated using empirical data from 14 countries, achieving ~75% accuracy. The resulting dataset includes both observed and predicted network lines and is suitable for global energy planning and accessibility/assessment studies.
- Roads
The GRIP dataset provides a globally harmonized vector map of road infrastructure, integrating over 21 million kilometers of roads from a variety of open and commercial sources. Designed to support global-scale environmental and accessibility modeling, the dataset offers standardized road classifications and extensive geographic coverage, especially in previously underrepresented regions.
Road data were compiled from national, regional, and global datasets, harmonized into a unified schema, and cleaned for spatial consistency. Next, they were classified into five standardized types (e.g., highways, primary, secondary, local, tracks) and validated using satellite imagery and supplementary datasets. The dataset is suitable for applications targeting land-use modeling, biodiversity impact assessments, and sustainable development planning.
🌾 Landuse
While conceptually simple, this layer features several characteristics that make its usage and applications difficult to interpret. Typically, land use layer is produced using middle to high resolution satellite imagery that is generated by taking snapshots of earth at particular instances of space and time. Next, imagery tiles that fulfill specific requirements are mosaicked together into a seamless, spatially contiguous dataset. This brings a large amount of heterogenity into the processing and interpretation of land use layer because various pixels that compose it have been acquired at and different times and under different conditions.
In the context of crisis resilience, the timing of satellite image acquisition is critical. Imagery captured before, after, or both before and after a specific event is typically used to assess conditions on the ground. In case of land use layer this is problematic because neighbour location could be acquired at different instances of time and conditions and bear little resemblance or positive spatial auto-correlation. As a result, rapida employes a different approach. Cloud accessible Sentinel 2 L1C imagery, available at global level is was fed into the Google dynamic world model to predict land use in close to real time for every image selected in a specific time interval with less than 5% cloud coverage.
The last step that is not yet implemented is to generate the cloud prediction and use the layer to mask/out filter cloudy/snowy pixels.
[!IMPORTANT] Predicting land use in close to real time is a computationally demanding task and should be treated with care
💰 Socio-economic environment
The socio-economic environment refers to the social and economic conditions that influence and shape the lives, behaviors, and development outcomes of individuals, communities, and societies.
- Deprivation
he Global Gridded Relative Deprivation Index (GRDI), Version 1, provides a detailed picture of relative deprivation and poverty worldwide, mapped at a high spatial resolution of approximately 1 km². The index ranges from 0 (least deprived) to 100 (most deprived). It is built using a combination of demographic and satellite data, carefully processed to ensure consistency across different regions. To create GRDI, six key factors were selected to represent different aspects of deprivation, such as economic activity, child mortality, and human development. The dataset covers the entire globe, integrating the best available data at either a fine-scale grid level or broader administrative boundaries.
GRDI combines six key indicators to assess deprivation levels:
- Built-up Area Ratio (BUILT): Measures the proportion of land covered by buildings compared to open land. Lower values indicate higher deprivation, as rural areas tend to have fewer economic opportunities.¹
- Child Dependency Ratio (CDR): The number of children (aged 0–14) per 100 working-age adults (15–64). A higher ratio suggests higher deprivation due to greater economic strain on households.²
- Infant Mortality Rate (IMR): The number of infant deaths (under one year old) per 1,000 live births. A higher IMR indicates poorer health conditions and higher deprivation.
- Subnational Human Development Index (SHDI): A local version of the Human Development Index (HDI), considering education, health, and living standards. Lower SHDI scores reflect higher deprivation.³
- Nighttime Lights (VNL, 2020): Measures light intensity at night, which is often linked to economic activity and infrastructure. Areas with less artificial light tend to have higher deprivation.⁴
- Nighttime Lights Trend (VNL Slope, 2012–2020): Tracks changes in nighttime lights over time. A decline in brightness suggests worsening deprivation, while an increase indicates economic growth.
The dataset can be used to:
- identify areas with high deprivation to guide poverty reduction efforts
- map socioeconomic inequalities at fine spatial scales
- support policy decisions and resource allocation for targeted interventions
- Relative Wealth Index
The Meta Relative Wealth Index is a high-resolution, machine learning–derived proxy for household wealth, developed by Meta’s Data for Good initiative. It estimates relative wealth scores at a 2.4 km grid level across low- and middle-income countries by analyzing de-identified Facebook connectivity data and satellite imagery. The RWI enables fine-scale economic analysis in data-scarce regions and supports humanitarian, development, and policy interventions.
- GDP
This dataset provides annual, global gridded estimates of Gross Domestic Product (GDP) at 0.1° spatial resolution (~10 km at the equator) from 2015 to 2100, fully aligned with the five Shared Socioeconomic Pathways (SSPs). It offers GDP values in both constant 2015 U.S. dollars and purchasing power parity (PPP), enabling spatially explicit long-term economic modeling under diverse socioeconomic scenarios. The dataset was up-sampled to 1km resolution and reprojected to EPSG:3857 to facilitate web usage.
To construct a spatially explicit and globally consistent GDP dataset aligned with the Shared Socioeconomic Pathways (SSPs), the authors began by collecting national-level GDP projections for each of the five SSP scenarios. These projections, expressed in both constant 2015 U.S. dollars and purchasing power parity (PPP), were sourced from authoritative institutions such as the Organisation for Economic Co-operation and Development (OECD) and the International Institute for Applied Systems Analysis (IIASA).
Next, population data at 0.1° resolution were obtained from the SSP Public Database. These gridded population distributions were used as the basis for disaggregating national GDP values. The key assumption underpinning the downscaling process was that GDP per capita remains spatially uniform within each country for a given year. Based on this, national GDP totals were allocated across grid cells in direct proportion to the local population count. This method produced a high-resolution, annual global GDP dataset spanning from 2015 to 2100, consistent with the spatial and temporal dynamics of each SSP scenario. The resulting data are well-suited for use in integrated assessment models, climate impact studies, land-use modeling, and other applications requiring detailed socioeconomic projections.
rapida builds on several open source geospatial packages and has a relative large dependency tree with several native ones like GDAL and this ads complexity to the installation procedure.
Important
We recommend installing into a virtual environment as opposed toi install into the default python interpreter/system
- ensure GDAL libs and binary tools are available on your system. ON linux this can be done using:
gdalinfo --version
-
create the virtual env suing tools/mean of your choice. Here there are two options
- create a virtual end that includes system packages
pipenv --python 3 --system-packages- create a clean virtual end and install GDAL
pipenv --python 3 GDAL_VERSION=$(gdalinfo --version | grep -oP 'GDAL \K[0-9.]+') echo $GDLA_VERSION pipenv run pip install --no-cache-dir --force-reinstall --no-binary=gdal gdal==$GDAL_VERSION pipenv run pip list Package Version ------- ------- GDAL 3.8.4 pip 25.0.1 -
install rapida from github
- with git available
pipenv run pip install git+https://github.com/UNDP-Data/rapida.git- without git
pipenv run pip install https://github.com/UNDP-Data/rapida/archive/refs/heads/main.zip
- with git available
-
test the installation
pipenv run rapida
Usage: rapida [OPTIONS] COMMAND [ARGS]...
UNDP Crisis Bureau Rapida tool.
This command line tool is designed to assess various geospatial variables
representing exposure and vulnerability aspects of geospatial risk induced
by natural hazards.
Options:
-h, --help Show this message and exit.
Commands:
init initialize RAPIDA tool
auth authenticate with UNDP account
admin fetch administrative boundaries at various levels from OSM/OCHA
create create a RAPIDA project in a new folder
assess assess/evaluate a specific geospatial exposure
components/variables
list list RAPIDA projects/folders located in default Azure file share
download download a RAPIDA project from Azure file share
upload upload a RAPIDA project to Azure file share
publish publish RAPIDA project results to Azure and GeoHub
delete delete a RAPIDA project from Azure file share
rapida can be deployed as a docker container and with some shell scripting be used effectively on a local desktop.
- ensure docker is installed and working.
Important
Avoid installing docker from snap even if it is convenient. snap docker has some specific restrictions (mount home folder, etc.)
docker --version
