New granular Human Development Index dataset
PLUS: Using Facebook advertising data to map poverty, the drivers of Euroscepticism, new datasets from LinkedIn and more.
Hey guys, welcome to this week’s edition of the Spatial Edge — a weekly round-up of geospatial news that you can digest in less than 5 minutes. This newsletter is a bit like a warm beer on a hot summer’s day — it’s pretty niche, but some people are weirdly into it.
In today’s newsletter:
Granular HDI data: New estimates for 157 countries.
Facebook data: Estimating global poverty with big data.
Euroscepticism: Mapping discontent in EU regions.
Wildfire impact: Air pollution spikes in Spain.
Geospatial datasets: SIPRI arms transfers and LinkedIn Green Jobs data.
Research you should know about
1. New super-granular Human Development Index data for 157 countries
A working paper from the National Bureau of Economic Research presents a new granular Human Development Index (HDI) for 157 countries. HDI is one of the most popular metrics of living standards which captures: GDP per capita, life expectancy and education.
The HDI is published at the country level by the UN, which has been a limitation for us geospatial folk. A couple of years ago, Global Data Lab created a sub-national HDI index at an admin-1 level (e.g. provinces/stages). However, this paper creates even more granular HDI estimates for the admin-2 level (e.g. municipalities).
The authors essentially take the sub-national HDI data from Global Data Lab and, using daytime and nighttime satellite images, estimate a more granular HDI for 2018 (at roughly a 10km resolution) through a ridge regression model.
They use:
daytime images: from Planet’s Surface Reflectance Basemaps (to capture things like vegetation, built environment, and other features that could be correlated with human development).
nighttime images: (DMSP-OLS) as a proxy for economic activity.
P.S. You can find the replication code here.
2. Using Facebook advertising data to estimate poverty
An article in Scientific Reports has created a granular geospatial dataset on global poverty by combining a bunch of sources of big data, such as Facebook advertising data.
The researchers use a lot of different datasets such as:
Facebook advertising data
OpenStreetMap data (e.g. roads)
SAR from sentinel-1
daytime satellites
nightlights
climate data
pollution
The use of Facebook advertising data is pretty interesting — it provides insights into information like the types of phones people use, their interests, and demographics. The idea is that information like a higher percentage of people using expensive phones could indicate a wealthier area.
They used a wealth index from the DHS Program as reference data for 59 countries. To track changes in wealth over time, they used the wealth index from the most recent survey along with older surveys in 33 countries as reference data.
The authors experimented with a few different machine learning models, with XG Boost being the best-performing model for predicting current wealth levels. However, all models performed similarly when it came to estimating changes in wealth over time.
Their new wealth index is closely aligned with the reference data, with 54 out of 59 countries showing very high accuracy (i.e. an r² over 0.9).
PS the replication code is available on Github, plus they’ve also created an R package to help query the Facebook advertising package
3. Mapping Euroscepticism and the regional development trap
A study in Economic Geography explores why certain regions in the EU are experiencing rising discontent and increasing support for Eurosceptic parties, which are opposed to EU integration.
This paper explores the relationship between the development trap conditions and voting behaviour, using two main indices to quantify development traps:
Development Trap Index 1 (DT1): Measures the risk of a region being in a development trap by comparing its growth in GDP per capita, productivity, and employment against EU averages, national averages, and the region’s historical performance
Development Trap Index 2 (DT2): Assesses the intensity of the development trap by measuring the magnitude of the region's underperformance in these same economic indicators
The study found that regions in the EU trapped in prolonged economic stagnation are more likely to support Eurosceptic parties.
The longer and more intense the economic decline, the stronger the discontent and Eurosceptic voting. Surprisingly, even wealthier regions facing relative decline showed increased support for these parties.
This essentially suggests that discontent isn’t limited to traditionally less-wealthy areas but is also prevalent in regions where economic conditions have deteriorated relative to the past.
4. Estimating the impact of wildfire on air pollution
A new publication examines how wildfire smoke in Northwestern Spain contributes to levels of air pollution (i.e. fine particulate matter — PM2.5). I’m always interested in understanding how much fire contributes to air pollution — be it burning of crop residue, forest fires, or otherwise.
The research team set up a network of low-cost air quality monitors during the summer of 2022 when a bunch of large wildfires occurred. They used a method called Quasi-Empirical Orthogonal Functions (QEOF) to figure out where the air pollution was coming from. This essentially helped them understand how much of the PM2.5 pollution was caused by the wildfires.
They found that:
Wildfires in northwestern Spain caused massive spikes in air pollution.
Smoke occasionally crossed borders from Portugal.
The Folgoso do Courel fire led to PM2.5 levels exceeding 300 µg/m³.
The highest recorded PM2.5 concentration during a wildfire event was 435 µg/m³ in Vilela Seca.
Geospatial datasets
1. SIPRI arms transfers database
The SIPRI Arms Transfers Database tracks major conventional arms transfers from 1950 to the most recent full calendar year. It’s pretty useful for exploring questions like who supplies and receives weapons, the types of arms traded, sources of weapons in conflict zones, etc.
2. LinkedIn green skills data
LinkedIn has made green skills data available for 77 countries via the World Bank’s Prosperity Data360 portal. This data can help track the growing demand for green skills, highlight skill gaps, and support the transition to a green economy.
3. Sen4Map benchmark dataset
Sen4Map is a large-scale dataset with 64 x 64 patches from Sentinel-2 images, covering over 335,125 geo-tagged locations in the EU with detailed landcover and land-use data from the LUCAS survey 2018. The dataset is available as HDF5 files per country, along with training, validation, and testing files.
P.S. The models' code and data loaders are also available here.
4. SmartForest FOR-species20K dataset
Another week and another forest dataset. The FOR-species20K dataset from SmartForest is for benchmarking data for classifying tree species from proximally-sensed laser scanning data. It features over 20,000 trees from 33 species across various regions, providing a crucial foundation for developing and benchmarking deep learning models for tree species classification.
Other useful bits
ESRI has created a new tutorial to show you how to quickly create a coastal property map using ArcGIS Living Atlas. These properties are often in danger of flooding, so this is a pretty useful tool for doing disaster/flood risk assessments.
Apple Maps just got an upgrade with iOS 18 and has added a bunch of new features. Personally, I never shifted away from Google Maps, but it’s cool to see better topographic views, find popular hiking routes, etc.
The small satellite market is forecast to grow massively, potentially hitting $7.49 billion by 2030 with an annual growth rate of 12.77%. I hope all this competition means cheaper prices for us all.
Bahrain is leveraging Planet's satellite data and AI to better manage its smart cities.
Jobs
The European Centre for Medium-Range Weather Forecasts (ECMWF) has four positions open for Machine Learning Scientist, Data Specialist and High-performance Computing Engineer.
The International Methane Emissions Observatory (IMEO) is looking for a Data Engineer and Remote Sensing Junior Analyst.
UP42 is looking for a Senior Python Engineer under their Data Integration team.
ESA is looking for an Ecosystem Economic Modelling Specialist who will be part of its Copernicus Ground Segment and Data Management Division.
The Photogrammetry and Remote Sensing Group at ETH Zürich is looking for a PhD Candidate in Machine Learning for 3D Computer Vision.
Alcis Geo is looking for a GIS Technician with a passion for GIS and imagery analysis.
Maxar Intelligence is looking for a Satellite Operations Engineer.
The University of Maryland is looking for an Assistant or Associate Professor in Remote Sensing under the BSOS-Geography Unit.
The University of Illinois is after an Assistant/Associate Professor of Geography & GIS.
My team (the Data Division) at the Asian Development Bank is looking for a couple of interns. There are two internship spots:
Just for fun
An updated London tube map from a University of Essex lecturer went viral. The map uses circles to show colour-coded routes. Let’s be honest — who doesn’t love a good circle?
That’s it for this week.
I’m always keen to hear from you, so please let me know if you have:
new geospatial datasets
newly published papers
geospatial job opportunities
and I’ll do my best to showcase them here.
Yohan