iiinova — where your digital self keeps evolving

Yohan @yohan.so · 29/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

010

Yohan @yohan.so · 29/04/2025

See this image of the Pilbara region in Australia Here we see: 1. lights generated from mines being lit up at night (i.e. production-based economic activity), AND 2. lights generated by mining staff who are eating out at night (e.g. consumption-based economic activity).

100

Yohan @yohan.so · 29/04/2025

However, we need to be careful about double counting. E.g. combining production values with income and consumption figures without accounting for overlaps could distort things. Henderson et al., essentially view nightlights as a measure of nighttime consumption:

100

Yohan @yohan.so · 29/04/2025

1. Economic Activity It’s vague to say nightlights capture ‘economic activity’. What 𝙚𝙭𝙖𝙘𝙩𝙡𝙮 do we mean by economic activity? The most popular paper on nightlights and economic activity is Henderson et al. (2012). It uses nightlights as a proxy for real GDP growth.

100

Yohan @yohan.so · 29/04/2025

If you're using nightlights you need to know about two things: 1. What 𝙩𝙮𝙥𝙚 of economic activity it captures, and 2. 𝙎𝙥𝙖𝙩𝙞𝙖𝙡 𝙢𝙞𝙨𝙢𝙖𝙩𝙘𝙝𝙚𝙨 Here's the breakdown (in simple terms):

141

Yohan @yohan.so · 28/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

000

Yohan @yohan.so · 28/04/2025

• 𝗟𝗟𝗠-𝗖𝗮𝘁: Makes maps iteratively and improves them based on its own visual critique. • 𝗚𝗜𝗦 𝗖𝗼𝗽𝗶𝗹𝗼𝘁: Helps QGIS users do analysis more efficiently.

100

Yohan @yohan.so · 28/04/2025

𝗦𝗰𝗮𝗹𝗲𝘀 𝗼𝗳 𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻 There are three technical scales: 1. Local: Runs on a single machine 2. Centralized: Uses cloud computing to handle larger tasks. 3. Infrastructure-scale: Distributed systems for massive analysis, possibly run by governments or research institutions.

100

Yohan @yohan.so · 28/04/2025

There are 5 goals for autonomous GIS: 𝟭. 𝗦𝗲𝗹𝗳-𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 – It creates ideas, workflows, code, and insights on its own. 𝟮. 𝗦𝗲𝗹𝗳-𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗻𝗴 – It can run the tasks (e.g., calculating distances, drawing maps).

100

Yohan @yohan.so · 28/04/2025

A new paper has set out a roadmap towards Autonomous GIS. The breakdown is as follows:

100

Yohan @yohan.so · 28/04/2025

How far off are we from Autonomous GIS? A new paper provides the roadmap. Here's the key info you should know about:

110

Yohan @yohan.so · 25/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

000

Yohan @yohan.so · 25/04/2025

7. GPS data Finally, you can also purchase raw GPS data from mobile phones. This is available from companies such as: • Huq • Factori • Veraset • Quadrant • Foursquare

100

Yohan @yohan.so · 25/04/2025

6. Here's Traffic Data They provide real-time data on: • traffic flows (i.e. speed and congestion) • incidents (e.g. road closures) This is also a paid dataset. Link: here.com/platform/tra...

110

Yohan @yohan.so · 25/04/2025

5. @mapbox.com 's Traffic Data This covers live data for 30 billion road segments across the world. They also provide 'typical' traffic data. This is also a paid product. Link: mapbox.com/traffic-data

100

Yohan @yohan.so · 25/04/2025

4. TomTom's Traffic Stats This a paid product that provides granular data on traffic. It's available via API or web portal. You can get limited access through the 30-day free trial of the Move platform. Link: tomtom.com/products/tra...

100

Yohan @yohan.so · 25/04/2025

3. Meta's Movement Distribution data This shows shows the range of movement of people away from the area where they live on a daily basis. It's useful for measuring transportation, tourism, and displacement. This data is updated monthly and is available globally. Link: buff.ly/dxmAEPm

100

Yohan @yohan.so · 25/04/2025

2. UTD19 This is the largest open-source traffic dataset. It covers 40 cities (mainly in Europe), but its getting a bit out of date now.

100

Yohan @yohan.so · 25/04/2025

1. Google's Traffic Layer Google's Maps API allows you to access traffic data. Link: developers.google.com/maps/documen...

100

Yohan @yohan.so · 25/04/2025

By far, the data I'm asked most about is related to traffic and mobility. It's for good reason. This data is insanely hard to come by. So I've compiled a comprehensive list of datasets. These cover both paid and free options:

120

Yohan @yohan.so · 24/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

000

Yohan @yohan.so · 24/04/2025

7. Land use We can use land-use data to measure things like: • urban sprawl • deforestation • depletion of water • growth of agriculture My favourite two are: • Google's Dynamic World • ESRI's Living Atlas Links: • developers.google.com/earth-engine... • livingatlas.arcgis.com/en/home/

100

Yohan @yohan.so · 24/04/2025

6. Climate projections I've previously used climate projection data for temperature, precipitation and wind. These provide 𝙙𝙖𝙞𝙡𝙮 geospatial forecasts until 31 Dec 2100. They use Global Circulation Models (GCMs) for various scenarios. Link: data.isimip.org/datasets/24c...

100

Yohan @yohan.so · 24/04/2025

5. Flooding I've had experience with 3 flooding datasets: • Fathom • JBA Flood maps • The Global Flood Database N.b. the first 2 are commercial datasets. Link: developers.google.com/earth-engine...

100

Yohan @yohan.so · 24/04/2025

4. Temperature and precipitation There are many datasets that provide temperature data like MOD11A1 and GCOM-C. But my go-to resource is ERA5's Climate Reanalysis Dataset. It provides daily global data on: • wind • temp • precipitation Link: developers.google.com/earth-engine...

100

Yohan @yohan.so · 24/04/2025

3. Pollutants Sentinel-5p is the go-to resource for geospatial data on various pollutants such as: • Ozone • Methane • Sulfur dioxide • Nitrogen dioxide • Carbon monoxide Link: developers.google.com/earth-engine...

201

Yohan @yohan.so · 24/04/2025

2. Air Pollution (particulate matter) @ecmwf.int 's Copernicus Atmosphere Monitoring Service provides real-time data on PM2.5, PM10 and so on. It's a great resource for measuring air pollution. Link: developers.google.com/earth-engine...

101

Yohan @yohan.so · 24/04/2025

1. CO2 Emissions EU Joint Research Centre (JRC)'s EDGAR database provides global geospatial data on: • CO2 emissions • GHG emissions • pollutants data Link: edgar.jrc.ec.europa.eu/emissions_da...

101

Yohan @yohan.so · 24/04/2025

One of the best use cases for geospatial data is examining environmental factors. Here's a list of my favourite geospatial environmental datasets:

120

Yohan @yohan.so · 22/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

000

Yohan @yohan.so · 22/04/2025

For instance, ERA5 underestimated humidity in important areas, leading it to incorrectly predict lower rainfall. Further, ERA5 also struggles to capture local evaporation processes. E.g. when moisture is recycled back into the atmosphere from plants.

100

Yohan @yohan.so · 22/04/2025

𝙒𝙝𝙖𝙩'𝙨 𝙘𝙖𝙪𝙨𝙞𝙣𝙜 𝙩𝙝𝙞𝙨 𝙢𝙞𝙨𝙢𝙖𝙩𝙘𝙝? The main reason is how ERA5 handles moisture transport. This is how humidity moves through the air due to wind. ERA5 often miscalculates these atmospheric flows, especially where ground-based measurements are limited.

100

Yohan @yohan.so · 22/04/2025

They found a bunch of big differences: ERA5 shows big drying trends during dry seasons in certain areas. However, observational records show rainfall levels holding steady or slightly increasing.

100

Yohan @yohan.so · 22/04/2025

The authors compared ERA5 rainfall estimates with observational data (from weather stations and satellites) covering over 40 years (1980–2022).

100

Yohan @yohan.so · 22/04/2025

Reanalysis products like ERA5 blend: 1. real-world weather measurements, with 2. climate models to recreate past weather conditions. These are super helpful but, as the researchers discovered, they can be a bit off when it comes to rainfall.

100

Yohan @yohan.so · 22/04/2025

A new study published in Scientific Reports looks at how reliable these reanalysis datasets actually are (and the results aren’t amazing…).

100

Yohan @yohan.so · 22/04/2025

Some of the most commonly used climate datasets are 'reanalysis' datasets, like ERA5. These are hugely popular because they offer global daily (sometimes even hourly) data on: • rainfall • temperature • wind

100

Yohan @yohan.so · 22/04/2025

Climate datasets are the most commonly used geospatial data out there. But how accurate are they? A new study examines this (and the results are surprising):

100

Yohan @yohan.so · 21/04/2025

Instead, we wanted to adopt a SaaS-style model by: • Selling datasets (i.e. product) • Using digital marketing techniques • On-selling consulting services to provide analysis of that data

110

Yohan @yohan.so · 21/04/2025

Step 3: Assess your users and market size Conduct basic market sizing by: 1. listing out all of the potential customers for your product 2. estimating how much these customers could pay for your solution 3. deriving estimates of the ‘Total Addressable Market’ for your product.

100

Yohan @yohan.so · 21/04/2025

You can get €50k (equity-free) to launch a geospatial business. I’ve done it twice. Here are the exact steps I took to obtain funding from @esa.int 's Business Incubation Centre:

100

Yohan @yohan.so · 18/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

020

Yohan @yohan.so · 18/04/2025

How to get a job as a data scientist at the World Bank? This is something I was asked about a lot. But my experience was (very) unorthodox. It involved burnout, CrossFit and Kyiv:

120

Yohan @yohan.so · 15/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

000

Yohan @yohan.so · 15/04/2025

Once the model is trained, if it sees another forest patch somewhere else, it recognises it by placing it in that same patch-of-green corner of its internal “map.”

100

Yohan @yohan.so · 15/04/2025

We can then take this patch and create embeddings of it. In the graph below, we have two axes - “greenness” and “tree”. We can see that our patch is both pretty green and pretty tree-like. So it’s located in the top right.

100

Yohan @yohan.so · 15/04/2025

A quick embeddings example: Imagine dividing a satellite image into smaller “patches.” You take one patch—like the one in the orange square below:

100

Yohan @yohan.so · 15/04/2025

Geospatial Foundation Models are the geospatial space's answer to LLMs. But do they actually add value? Let's take a look:

110

Yohan @yohan.so · 14/04/2025

Interested in getting a short overview of the latest geospatial papers and datasets each week? Subscribe to the Spatial Edge newsletter: yohan.so

010

Yohan @yohan.so · 14/04/2025

𝟰. 𝗗𝗮𝘁𝗮 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 Once I've got data, I've been using Cursor to program in Python. My default is to use it with Claude Sonnet-3.7

100