Sharing insights from the intersection of geospatial data science and economics | PhD in Economic Geography from @lsenews. Views are my own.
4 subscribers
Sep 9 • 13 tweets • 4 min read
Everyone is talking about Zarr.
ESA is adopting it and others are testing it.
Does this mean the end of Cloud Optimized GeoTIFFs?
Here is what you need to know:
ESA recently announced Zarr as the new format for Sentinel-1, 2 and 3.
USGS has benchmarked it for Landsat’s archive.
But many in the community are asking: does this mean the end of COG?
Sep 3 • 13 tweets • 4 min read
Turns out there are some pretty big issues with DHS data.
A new study finds massive subnational differences in data quality across 35 African countries.
Here's the breakdown:
A new study in Nature Communications, analyses geocoded DHS data at a 5km resolution.
It highlights serious concerns for health and development policymaking:
Aug 29 • 13 tweets • 3 min read
Air pollution is usually blamed for lung and heart disease.
But new clinical data shows it may also drive diabetes.
Here’s what you need to know:
The researchers combined:
• Outpatient clinical records from the Italian Association of Diabetologists (AMD)
• Municipality-level pollution exposure data from ISPRA, Italy’s environmental protection agency
This gave them a unique dataset of pollution and diabetes at the local level.
Aug 20 • 15 tweets • 5 min read
Changing your map’s resolution can change your conclusions.
It’s called the Support Effect.
And it distorts everything from poverty estimates to climate models.
Here’s how it works:
In spatial analysis, “support” refers to the unit of measurement in space.
It could be:
• a point (e.g., GPS location)
• an area (e.g., census tract)
• a pixel (e.g., satellite image cell)
The support determines how and where a variable is measured.
Aug 12 • 10 tweets • 4 min read
We’ve been measuring HDI at the national level for decades.
But living standards can vary dramatically within a country.
A new dataset finally shows HDI at a much finer scale.
Here’s the breakdown:
The first sub-national HDI dataset was actually published in @ScientificData in 2019.
It was put together by @Globaldatalab.
Aug 2 • 22 tweets • 6 min read
Google DeepMind just released one of the most important tools in geospatial data science.
It’s called AlphaEarth Foundations.
I want to break it down for you in intuitive terms:
We have petabytes of satellite images.
But it’s still hard to answer questions like:
• What’s in this image?
• How has it changed?
• What kind of crop or forest is this?
AlphaEarth helps answer these questions, even in places with limited data.
Jul 28 • 12 tweets • 6 min read
Most countries don't publish official sub-national population data.
Luckily, there are several geospatial population datasets we can use instead.
Here's a list of them (I wish I knew about 5 years ago): 1. WorldPop (@WorldPopProject) provides data on:
• population counts
• population density
• population by age and sex
Data is from 2000-2020 and available at 100m or 1km resolution.
Here's what you need to know about it:
In simple terms, ecological fallacy is drawing conclusions about individuals from data that were aggregated over areas (e.g. counties, districts, grids).
Jun 6 • 13 tweets • 3 min read
What happens when a scientist pays $300 for 50 citations?
A study reveals a growing black market of citation mills, and why it threatens the credibility of science.
To test the Google Scholar system, researchers created a fake academic and paid $300 to buy 50 citations.
Within 40 days, citations started appearing in Google Scholar, published in journals with impact factors as high as 4.79.
Jun 3 • 12 tweets • 4 min read
We track poverty too slowly to respond effectively.
But a new open-source tool now lets us map wealth in near real-time using satellite imagery.
Here’s what you need to know:
Back in 2016, Stanford's AI Lab pioneered a new method of measuring economic activity using daytime satellite images.
Jun 2 • 13 tweets • 3 min read
If your regression results look great but perform poorly in some areas, there could be a cause:
Spatial nonstationarity.
It’s one of the most overlooked issues in spatial analysis.
Here’s what it is and how to detect it:
Imagine you’re studying how income affects house prices in a city.
You run a regression and get a single coefficient.
E.g., every $1,000 increase in income raises house prices by $5,000.
But is that true everywhere in the city?
May 16 • 11 tweets • 4 min read
Satellite data is being used to detect marine litter from space.
Here's everything you need to know about it:
A new paper in @NatureComms describes a novel way of detecting marine litter using satellite images.
They use data from Sentinel-2 to detect ‘litter windrows’.
May 12 • 12 tweets • 6 min read
Geospatial data science is all about having access to good data.
Here are a list of my favourite free geospatial resources: 1. Aiddata
Aiddata has an excellent list of geospatial datasets.
It's all available as a CSV.
So you don't even need to know GIS to access their data.
However, it's incredibly difficult to work out WHERE it comes from.
A new study has compiled over 34,000 measurements from 66 countries to trace its sources:
Researchers developed a blockchain-based isotopic database, compiling 34,815 isotopic fingerprints from 1,890 pollution events across the world.
It covers data from 1957 to 2023.
May 5 • 14 tweets • 3 min read
Rivers are shrinking.
This is happening because water is leaking into the ground.
Here's the breakdown:
A new study published in Nature Communications, analyses over 17,900 wells across Brazil.
It aims to understand where rivers may be losing water into aquifers.
May 1 • 13 tweets • 4 min read
Most people assume national household surveys are uniformly high quality.
But a new study finds massive subnational differences in data quality across 35 African countries.
Here's the breakdown:
A new study in Nature Communications, analyses geocoded DHS data (RIP😢) at a 5km resolution.
It highlights serious concerns for health and development policymaking:
Apr 14 • 14 tweets • 5 min read
We're constantly told how countries are run by the elites.
However, mapping the economic elite has been incredibly tough to do.
But now, a new dataset compiled by 70+ researchers provides data on over 3,500 elites in 16 countries:
In the 19th century, many countries — particularly in Europe, were run by the aristocracy.
Fast forward to today, most monarchs (if they’re still around), are mainly ceremonial figures.
Mar 24 • 22 tweets • 6 min read
To do good research, you need good tools.
Here are the (free) tools I can't live without:
𝟭. 𝗧𝗼𝗽𝗶𝗰 𝗜𝗱𝗲𝗮𝘁𝗶𝗼𝗻
The first category of tools I use is for ideating on a topic.
Nothing fancy here.
I use OpenAI models to brainstorm research ideas.
Mar 6 • 13 tweets • 4 min read
Meta is using Facebook and mobile phone data to produce super-granular household wealth estimates.
Here’s what you need to know about the Relative Wealth Index:
The Relative Wealth Index (RWI) is a geospatial measure, which shows wealth disparities within countries.
Main features:
• It's open source
• It measures 'asset-based' wealth
• It covers 135 countries in the world
• Data is provided at 2.4km resolution