We should be using SAR data more often in research
Everything you need to know about Synthetic Aperture Radar data.
Synthetic Aperture Radar (SAR) data will be game-changing in economics. It’s the most under-rated form of satellite data.
And I think we’ll be seeing more of it in being used in economics and environmental studies in 2024.
Economics is often pretty slow to adopt technological breakthroughs—particularly in the earth observation (EO) space. As an economist and geospatial data scientist, I’ve frequently used EO data to look at things like:
economic activity (night time luminosity),
deforestation (land use data),
air pollution, etc.
In most cases ‘optical imagery’—i.e. ‘photos from satellites’—are more than enough.
However, for two and a half years I was living in Hanoi, one of the most polluted cities in the world.
It’s a place where you could go weeks without having a cloud or smog-free day. In these sorts of places, optical imagery simply can’t work. You just get cloud/smog affected images like this:
When I was working for the World Bank in Hanoi, we’d develop algorithms to drop cloud/smog contaminated pixels and impute missing values. However, these imputations could be wildly off if whole districts were impacted by clouds for weeks on end.
At this point, we’d simply give up on the analysis.
However there’s a solution which I wish we’d used: Synthetic Aperture Radar (SAR).
It has the ability to ‘see’ through clouds, and provide imaging regardless of weather conditions.
Seeing through clouds with Synthetic Aperture Radar (SAR) data
SAR allows satellites to ‘see’ through cloud cover and smog. It can also penetrate canopy to better see crops or buildings. It can work day or night, and doesn’t require perfect lighting conditions.
SAR isn’t new—but it’s been massively underused in the economics space.
So what is Synthentic Aperture Radar data?
SAR builds on traditional radar technology.
It allows us to visualise the world in ways that the naked eye can’t.
Standard radar sends out radio signals and measures how long it takes for the signals to bounce back. This is how it determines where objects are located. However, traditional radar has a limited resolution that makes images a bit fuzzy.
SAR on the other hand sends pulses of microwave signals toward the ground from a satellite. When the signals hit a surface, it bounces back and are recorded by the SAR instrument. The strength of the return echo creates light and dark pixels in a 2D image.
Smooth surfaces like water reflect away from the radar, so they appear dark. Rough areas like forests reflect strongly, showing up bright.
The timing of the return signals determines where features are located. Advanced processing converts the raw data into high-res SAR images.
The 'aperture' in Synthetic Aperture Radar refers to the radar's antenna on the moving satellite. It's called a 'synthetic' aperture because the movement of the satellite combines with the radar signals to create a larger aperture, which helps to capture more detailed data. It’s sort of like having a wider lens on a camera, which allows you to get a more detailed shot.
Issues with SAR: Speckle
During the imaging process, you might notice a grainy, salt-and-pepper like pattern. This is 'speckle.' It's a form of noise that appears due to how the radar signals are processed.
However, there are ongoing advancements in reducing speckle. Two of the most common techniques of speckle reduction are:
filtering (which helps to suppress noise)
multi-looking (e.g. speckle-affected pixels are replace with the average values of adjacent pixels)
SAR Data Bands
To differentiate between various features on the ground, SAR uses different bands of radio waves.
These include X, C, L, and P-bands:
X-Band: With a shorter wavelength, it can’t penetrate forest canopy, but is useful in capturing detailed views of the Earth's surface.
C-Band: It has a moderate wavelength, it can slightly penetrate forest canopy, but is often used for studying soil moisture and vegetation.
L-Band: Penetrates deeper, providing insights into forest biomass and even the Earth's subsurface aspects.
By now, you’re probably thinking that this is all well and good. But what are the actual use-cases for this data?
Let’s take a look.
How can we use SAR in economics?
1. Monitoring agricultural productivity
SAR data can be used to track changes in crop density over time. This has been done by examining changes in the leaf area index (LAI).
The LAI tells us the density of leaves for a given area. A higher LAI indicates a dense canopy with potentially higher rates of photosynthesis.
Monitoring LAI changes provides insights into agricultural productivity, crop yield predictions, and forest management (see Jiao et al., 2021).
In short, it can be used to assess agricultural productivity.
2. Examining deforestation and natural capital stock
Analysing radar backscatter (i.e. the reflection of radar waves back to satellites), can be used to identify areas undergoing deforestation.
Dense forests usually have a high radar backscatter due to the complex structure of trees, while clear cut areas show a reduced backscatter because there are fewer structures to bounce back the radar waves.
By observing the changes in backscatter over time, we can identify areas where deforestation is happening.
3. Measuring economic activity through car parking occupancy
SAR data has been used to measure the occupancy of car parks.
This can be used as a proxy for consumption, economic activity and even population distribution.
Car filled areas appear brighter due to the reflective properties of vehicles. As a result, we can calculate the occupancy ratio of car parks (see Drouyer and de Franchis, 2020), and see how this changes over time (e.g. over days, weeks or even years).
This can be used to measure the impact of COVID lockdowns on economic activity. It can even help to inform urban economics and urban planning (e.g. informing parking pricing, or where new car parks need to be located).
4. Creating ‘poverty maps’ by looking at building material
Using SAR, we’re able to assess the material of rooftops.
And the quality of rooftops can be used as a general proxy for wealth (and poverty).
By gathering this kind of information on a large scale, experts can create estimates of the economic status at the household level.
Where can we download/interact with SAR data?
If you’re now a little SAR-curious, there are a few platforms where you can access free SAR data to play around with:
Google Earth Engine: has daily SAR (c-band data) from Sentinel-1 from October 2014 to the present day. The pixel resolution is down to 10m.
Alaska Satellite Facility: This website provides several sources of current and past SAR data, in addition to derived datasets.
Umbra: has an Open Data Program and they’ve also made a bunch of its data free on skyfi.com:
The takeaway
In short, SAR has the ability to overcome some of the issues we face with using optical satellite imagery—cloud cover, time of day, and adverse weather conditions.
It will help us build up a more robust time series, and help unlock new time-series analyses.