🌐 Can gangs actually reduce crime?
PLUS: massive differences in land use datasets, CO2 emissions from private jets, and more.
Hey guys, here’s this week’s edition of Spatial Edge — a weekly round-up of geospatial news that you can digest quicker than you can say ‘modifiable areal unit problem’. The goal is to help make you a better geospatial data scientist in less than 5 minutes a week.
In today’s newsletter:
Crime and Social Order: Gangs' governance reduces neighbourhood crimes.
Urban Land Disagreements: Global datasets differ on urban land coverage.
Private Jets’ Impact: High emissions from short-haul luxury air travel.
Remote Sensing Advancements: E2DiffSR enhances satellite image resolution.
Geospatial Datasets Update: New forest and crop data improve analysis.
Research you should know about
1. Can gangs reduce crimes?
A new study published in Nature Cities explores how a local gang in Nottingham, England, influences community crime rates by acting as an unofficial governing body. The study looks into whether this governance-type organised crime can reduce ordinary crimes in a neighbourhood.
The researchers conducted in-depth interviews with local police officers and analysed a dataset of public phone calls to Nottingham police from 2012 to 2019, which included the time, location, and types of reported crimes. They used methods like correspondence analysis and k-nearest neighbours to compare the ward of Bestwood—home to an entrenched gang—with Bulwell, a similar ward without such a gang. This approach allowed them to map crime rates and assess the gang's impact on the community.
The study found that Bestwood had significantly lower rates of certain crimes, such as antisocial behaviour, burglary, and violence against persons, compared to Bulwell. I guess this suggests that the gang reduces ordinary crime by imposing its own form of social order. Of course, there are many other types of crimes that are unlocked by the presence of gangs.
2. Disagreements in urban land cover datasets
A new study published in Nature Communications looks into the significant differences between the main global land cover datasets. The bottom line is that differences in scale, urban definitions, and methodologies lead to differing estimates of urban land cover.
They analysed a bunch of high-res landcover datasets, including:
WSF 2019
Dynamic World
Esri Land Cover
ESA WorldCover
One thing I found pretty interesting is that each of these datasets has different definitions of ‘urban’. E.g. Dynamic World has a class called ‘built area’, which includes urban vegetation and green space. Whereas WorldCover’s class is called ‘built up’ and excludes urban vegetation.
The study found that while all datasets indicate global urban land nearly tripled between 1985 and 2015, the actual percentages vary widely—from 0.52% to 2.07% of the Earth's surface.
This comes back to one of my pet peeves in this space — many geospatial datasets tell us wildly different stories, while supposedly measuring the same thing.
3. Private jet contributions to climate change
A new study published in Communications Earth & Environment discusses the extent to which private jets contribute to climate change, with emissions growing rapidly in recent years. The researchers calculated that private jets emitted at least 15.6 million tonnes of CO₂ in 2023 alone.
The team analysed flight tracker data from the ADS-B Exchange platform, examining 18,655,789 individual flights made by 25,993 private aircraft between 2019 and 2023. They linked these flights to 72 aircraft models and their average fuel consumption to estimate CO₂ emissions. The data showed that almost half of all private flights are shorter than 500 kilometres, often associated with leisure travel and major events. They also found that 68.7% of private aircraft are registered in the USA, which indicates a bit of a geographical concentration.
Emissions from private aviation increased by 46% between 2019 and 2023, now accounting for about 1.7% to 1.8% of commercial aviation's CO₂ emissions.
4. Enhancing remote sensing images with E2DiffSR
A new study introduces E2DiffSR, a new super-resolution model specifically designed for remote sensing images. Super-resolution is something I’ve covered previously in this blog, and it’s something I’m spending a lot of time working on.
I find this paper interesting since it addresses some limitations in existing diffusion-based super-resolution (SR) methods, which often treat SR similarly to general image generation and are restricted to fixed integer scaling factors.
Traditional diffusion models generate images from scratch, which can be a bit inefficient for SR tasks that aim to enhance existing low-resolution (LR) images by adding missing high-frequency details. E2DiffSR tries to overcome this by using a two-stage latent diffusion approach:
Latent Space Pretraining: An autoencoder is trained to capture the differential priors—the differences between high-resolution (HR) and LR images. In other words, the encoder deliberately ignores the LR content it already knows. It just encodes the ‘extra details’ (i.e. the differential priors).
Conditional Diffusion in Latent Space: Once the model has learned how to represent the "missing details," it uses a fast process to predict those details for a given LR image. Instead of operating directly on the full image (which would be slow), this prediction happens in a compressed "latent space," making it much faster while still delivering high-quality results.
The results demonstrate that E2DiffSR outperforms state-of-the-art SR methods in both objective metrics and visual quality.
P.S. You can access the code here.
Geospatial datasets
1. GEDI and TanDEM-X dataset
The GEDI-TanDEM-X dataset combines GEDI lidar observations with TanDEM-X Synthetic Aperture Radar (SAR) data to produce high-resolution (25 m to 100m) global maps of forest canopy height and biomass, and includes uncertainty estimates. Data is available from 2019 onwards. You can access the data here.
2. River flows dataset
VegDischarge v1 tracks river flows across 64,000+ African river segments from 2001 to 2021. It provides runoff and discharge data.
3. Crops aerial image dataset
EcoCropsAID is a new dataset designed to classify land use for Thailand’s key crops. EcoCropsAID includes 5,400 Google Earth images (2014–2018) featuring rice, sugarcane, cassava, rubber, and longan at different growth stages, analysed using deep learning algorithms. You can access the data here.
4. New Digital Elevation Model data
The Global Digital Elevation Merged Model (GDEMM2024) combines the latest digital elevation models (DEMs) to provide high-resolution and accurate global topography, bathymetry, and ice thickness layers. GDEMM2024 merges data from sources like TanDEM-X, BedMachine, GEBCO2023, and ETOPO2022, while incorporating updated land-type masks and improved merging techniques. The result is a 30 arc second-resolution grid. You can access the dataset here.
Other useful bits
The October Crop Monitor updates are now available, featuring the latest reports from CM4AMIS, CM4EW, and the Global Crop Monitor. These reports provide detailed insights into crop conditions across the globe, helping to track agricultural trends and challenges.
A blog from Development Seed explores how Geospatial Foundation Models are transforming satellite imagery into actionable insights. ‘What do we mean by foundation models?’ I hear you ask. Luckily, I’ve got you covered here.
The World Bank has introduced a new poverty line of $6.85 per day to reflect the standards of upper-middle-income countries. They say they’ll monitor this new poverty line alongside the other $2.15 international poverty line.
Jobs
DAI is looking for a Specialist or Senior Specialist in Data Science/Monitoring, Evaluation, & Learning under their Development Innovations Department.
Carbon Mapper is looking for a Deputy Director for Impact who will help connect the organisation's data and insights to stakeholders.
Treeconomy is looking for a Geospatial Scientist who will be part of their Science team.
WFP is looking for a Geographic Information System Consultant who will support activities related to their Global Assurance project rollout.
ESA is looking for a Head to lead their Earth and Mission Sciences Division.
IMPACT Initiatives is looking for a Remote Sensing and Climate Specialist under their Global Emergency Unit.
Just for Fun
Charles Duke (former American astronaut) left a family photo on the Moon during the Apollo 16 mission, which remains sealed in a plastic bag.
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
Great research