Fact-checking with open-source satellite images (Part 1)

Whether illegally operating coal plants, water conflicts along major rivers, or illegal gas flaring, it’s possible for journalists to reveal injustice by using open-source satellite images. An #OSINT satellite training guide to get you started.

Source: VIIRS Info from National Oceanic and Atmospheric Administration, Payne Institute of mines, Jan 2020 nighttime lights

Although there are arguably ‘unlimited opportunities’ for open-source satellite images to support journalism it’s often hard to use it in the field.

Low resolution can make it a hassle to prove or verify reports. Believe me, I've been there. Having access to update high-resolution images is great but that often might not be an option when time available is limited or when you don't have access to the right contacts.

If you work under pressure for a small investigative outfit you might have neither. In first post will try to convince you otherwise. There are several examples where low-resolution open-source satellite images excel and we will discuss each use case.

Incidentally, if you know nothing about #OSINT satellite investigations and how it can aid journalists, feel free to check out my previous tutorial on the subject HERE

1. Illegal operating coal power plants

Coal-fired power plants contribute a massive share to the global CO2 emission balance sheet. According to the IEA nearly a third of the total CO2 emissions are down to coal. The majority comes from Asia. China is the coal power plant construction powerhouse. In two years and often less, it can build major plants.

Despite earlier pledges to curb coal power and to increase the share of energy production from more sustainable sources, the latest plans speak of reopening commissioning for coal projects. Relaxing rules in China is a problem.

There are cases where constructions went ahead illegally — as a satellite analysis by Global Energy Monitor suggested, some time ago. Journalists trained in satellite #OSINT have an edge to find new evidence that either

  1. construction went ahead where they shouldn't have (by 2030, China is expected to retire at least 100GW of China’s capacity built before 2000) or
  2. locate signs that coal plants remain in use despite being officially retired or mothballed

Aside from the illegally continuing constructions (which has already been investigated by Global Energy Monitor) illicit operation beyond retirement and mothball status seems interesting enough.

Being mothballed means coal power stations are shut down temporarily to be able to start them up again at some point later.

The Global Coal Plant Tracker is our reference source. Coal power plants receive a status update should they be mothballed or retired. Ted Nace, expert at Global Energy Monitor pointed out to me that there arent many plats mothballed right now. Therefore, even if plants would keep running the impact on the climate is minimal (only eight units at three locations have this status).

Several coal power locations with mothballed status ( Global Coal Plant Tracker)

Let’s take a look at the Linyi Huasheng power station Unit 1,2, and 3. We weren't given the right coordinates but can figures it out on Google (34.956426, 118.306525).

Coordinates for the power station: 34.956426, 118.306525 (left on the Global Coal Plant Tracker)

Next, we head over to Sentinel Browser, the classic tool to access open-source satellite images with a neat interface for beginners. Zoom in on China and punch in your coordinates (here is the correct LINK).

We will use water vapor from the plants’ chimneys as an indication that the plants are still active. Let’s access all Sentinel 2 images between January 2019 and the latest available date, and generate a gif (you have to log in for that).

A potential start of a chain for evidence that mothballed coal power stations remain active

It does seem that chimneys at Linyi Huasheng started to relax for a while but then they smolder again around March/April.

Also perhaps interesting, a whole new section was built right next to the tent. At this stage, we don't know what it is. We could investigate more via Google Earth Pro which has high-resolution images (yet, may not have a frequent update). The latest images we get is from September 11, 2019. It shows two of the four chimneys in operation.

We can see the tent construction project on the right.

The next one is the Huaneng Beijing power station with it’s unit 2, 3, 4, 5. All mothballed. You can find the place under the coordinates 39.886253, 116.531779. In the past five months, the plant shows some activity. From here we could further investigate what’s going on.

Huaneng Beijing power station, December 2019 to May 2020 (39.886253, 116.531779)

Yaomeng power station (Coal tracker link here) was completed between 1975 and 2007 and unit 1 (320 MW) was retired in 2017. This left 2,160 MW remaining which must have been mothballed in 2020. The images suggest the place now up and running again.

From here it might be wise to check with commercial companies (such as Planet Labs or Maxar Digitalglobe) to check on high-resolution images to find out what’s going on. Also, it’s worth speaking to coal power analysts in China who know the situation well and can talk about how the mothballed status is enforced.

Power plant near Zhengzhou in the Henan province (Coordinates: 33.7379928, 113.2418822)

Retired plants

If we just pick a plant in Shangdong, a coastal province in China, it might not be surprising if some are still running.

In 2017 China’s National Energy Administration introduced new rules that saw over 100 planned and under-construction coal power projects suspended, including the one called Huadian Longkou Unit 7, in Yantai City in the Shandong province.

Sentinel 2 images: April 2019 — May 2020

We also have an image for April of this year on Google Earth. We can see two out of three Chimneys very much active. Now we would need to understand whether that's enough of an indication to classify it still operational.

April 4, 2020, Google Earth, Maxar

As with everything we do we need verifying evidence to make the claim that what’s happening is illegal. Remember, this is all just the early part of an investigation. Contracts, government announcements, and restrictions would all need to be carefully studied first before making any claims against the company or the Chinese provincial administration.

2. Riveting water conflicts

The situation is tense. China is being accused of restraining the amount of water that flows from the upper part of the Mekong river to the various countries downstream. This conflict could quickly turn into a diplomatic crisis between China and the other countries — notably Myanmar, Laos, Thailand, Cambodia, and Vietnam.

Last year downstream countries experienced the lowest Mekong water levels in half a century. That’s a problem when more than 60 million people feed and source their water from the river.

Images confirm low water levels downstream: Comparison between Mekong river 2018 and 2019.

Aerial video portraying the drought in the Mekong River in Thailand’s Nong Khai province (Oct. 28, 2019)
Mekong river near the Thai–Lao Friendship Bridge: Water levels are up in 2018 and start to drastically decrease throughout 2019

A study into the subject claims otherwise. A 2019 report by US government-funded Eyes on Earth Inc, says that in 2019 China’s part of the basin received more rain and snow than normal. The Chinese government fends off all these claims and says it was suffering from drought. So who should we believe?

“Had all that water flowed downstream, the river would have been between seven and eight metres deep as it entered Thailand — higher than usual for that time of year” (authors of the study).

From ‘A Dangerous Trajectory for the Mekong River UPDATE ON THE STATUS OF MEKONG MAINSTREAM DAMS’, 2017

Let’s check some open-source satellite images. The map on the right shows us the names of the major dams along the Mekong. The largest operating is in China. With 5,850 MW the Nuozhadu dam (22°39′22″N 100°25′06″E) is near the last dam before the border, the Jinhong dam (22°03′09″N 100°45′58″E). We head over to the link of the dams and copy the decimal lat/long coordinates.

Next, we head over to the Sentinel browser and use the Normalized Difference Water Index (NDWI) scrip for detecting water bodies. The aim of this exercise is to measure changes and then compare the values with those downstream.

Intro to NDWI index: It’s most appropriate for water body mapping. The water body has strong absorbability and low radiation in the range from visible to infrared wavelengths. The index uses the green and Near Infra-red bands of remote sensing images based on this phenomenon.Values of water bodies are larger than 0.5. Vegetation has smaller values. Built-up features have positive values between zero and 0.2.

Let’s see when in 2019 water levels hit rock bottom for the Nuozhadu dam.

Lowest levels are reached at around early May

The lowest level from the satellite images we can spot for May 10. The next thing we can do is to compare water levels from 2019 with those of the same day of 2018. There are nearly the same (a little smaller actually for 2018). It’s one sign that the extreme drought claims by China for 2019 may not hold up (Careful, this could have other reasons, we should keep looking. But it’s duly noted).

Comparison between May 10, 2018, and May 10 of 2019.

The water reservoir of the Jinghong dam behaves similarly. Lowest levels are hit on May 10 for 2019. Comparing this with 2018’s May 10, it's showing much lower levels.

Let’s look downstream and measure the size of a measurable river bed about 150 kilometers after the Mekong starts flowing alongside the Thai–Laotian border (Tha Bo District, Nong Khai, Thailand).

We check Sentinel 2 images with the NDWI filter for this location (17°50'4.27"N, 102°35'55.18"E). Let’s also quickly calculate how long it takes for the Mekong to reach this part. On average, the river flows at a speed of around 4 knots, which amounts to around 7.4 kilometers per hour. The distance from Nuozhadu Dam (22.656111, 100.418333) to this part of the river is around 900km, so it’s taking around 122 hours or around five days.

We saw before that the lowest point in the Nuozhadu Dam was reached on May 10. On May 12 we see the Mekong is only around 239m broad in the spot we measured. That all sounds about right. What is doesn't explain is the data from early February of 2019.

In February the water level in the Nuozhadu was higher, relatively speaking. It peaked in January and has fallen thereafter. But the distance between shores at our spot in Thailand only measured a meager 214m, some of the lowest levels in the series of 2019 satellite images for this part.

Left: Nuozhadu water levels only decline from early February, but downstream the Mekong already fell to its lowest level (river bed in 17.832306, 102.595730) downstream somewhere where the river divides Thailand and Laos

So, should we trust the Chinese government in their claims that a doubt affected them equally and there was not Mekong water tinkering involved?

It warrants saying that when new dams and water reservoirs are being built it takes time to fill them.

While it remained unclear what kind of tests China was conducting on dam equipment, water levels on the Mekong in Chiang Rai — where Thailand, Laos and Myanmar meet — have been reduced by about 40 percent, from three meters to less than one, Brian Eyler, director of the Stimson Center’s Southeast Asia program, told Al Jazeer (news report).

From the evidence we have here, we must question why earlier in the year 2019 levels dropped so dramatically downstream while at Chinese dams water reservoir levels were relatively high. From here we could look into measuring the debt of the Mekong with high-res satellite data.

Imminent Water conflict between Ethiopia and Egypt

Ethiopian and Egyptian governments stand at the brink of a water conflict, too. Ethiopia is building the Grand Renaissance Dam, Africa’s largest hydroelectric power plant. Egypt — which relies on the Nile for 90 per cent of its water — fears Ethiopia will control the volume of water in the Nile. Filling the water for the dam reservoir could majorly impact downstream nations, if too much of the river is being carved out for filling the reservoir.

If the fury between parties (which also affects Sudan and other countries) cannot be appeased this situation could turn into conflict. Once the dam is ready, a similar analysis as above could help to see how water levels are affected and whether it's intentional or not.

Other stories using Sentinel 2 scrips:

If you want to read an investigation where #OSINT satellite scripts are used in a similar fashion, the one below looked at lithium production sites in the Atacama salt flats.

A satellite analysis added evidence of how lithium production dries out important freshwater resources that provides food and water to the local population (Link)

3. Illegal flaring

January 2020, Tile 1, 75N 180W, VCMCFG (link); VIIRS Day/Night Band Nighttime Lights by the Earth Observations Group (EOG)

Flaring, the wasteful burning of superfluous gas from oil extraction sites, is one of the often undervalued drivers of global warming.

Methane, an extremely potent greenhouse gas is rarely discussed in the news but constitutes a considerable problem for the global environment. It’s now taking center stage as the sustainable development scenario dictate drastic cuts in flaring activities.

It’s a perfect opportunity for climate/environmental investigative journalists to take on the task by using open source tools to spot illegal flaring activities. Nighttime satellite images are perfect for this.

One word of warning before we getting started. Raw nightlight images have the potential to throw up false positives, en masse. It’s a common problem when we decide to acquire highlight raw data such as via the LAADS DAAC platform (here is an excellent tutorial by data visualization and cartography lead at @NASAEarth, Joshua Stevens).

The problem I and other journalists have with raw nightlight data is multifold. First, I am a Mac OS. It’s not really Mac friendly. Then, the output data you get is in HDF file format, which needs to be turned into a GeoTIFF file with GDAL that can then be readily be opened with open-source programs such as QGIS or other geo-data programs. In short, it’s complicated. Lastly, we need to perform complex adjustment methodologies. If you are new to this, I’d recommend a different data source.

The Payne Institute Colorado School of Mines tracks various kinds of stuff (night fires, boats and nightlight in time format) that can be viewed from satellite images at night. They rely on the VIIRS from the National Oceanic and Atmospheric Administration (NOAA). So, it’s provided by a trustworthy source and prepared for quick and easy access.

What’s VIIRS?

The Visible Infrared Imaging Radiometer Suite (VIIRS) is a sensor designed and manufactured by the Raytheon Company on board the Suomi National Polar-orbiting Partnership (Suomi NPP) and NOAA-20 weather satellites.

To track potentially illegal flaring we are interested in the night fire data, measuring VIIRS’s near-infrared and short-wave infrared data at night.

It’s published daily and output is via kmz file format — which comes in handy for upload to Google Earth Pro tool.

Go to the VIIRS Nightfire Nighttime Detection and Characterization of Combustion Sources page. Download the latest data, the KMZ file, as highlighted.

Latest data available for May 27

Upload this one to your Google Earth Pro desktop application (it's free, here).

An example: Nigeria’s flaring activities last night

We see a large number of red dots. They indicate the detecting of combustion sources of between 1,600 and 2,500 in temperature (the plank curve fitting is performed to estimate the temperature of background and hot sources).

Let’s look at some other known flaring sites in the US where flaring legislation could be tightening further(if the US gets a new president that cares about the global environment).

For this, we need earlier data. Let’s use the KMZ file from the beginning of May (VNF_npp_d20200501_noaa_v30.kmz).

Flaring sites on May 1, 2020, near Orla, TX, USA

Now, from here we could compare various oil producers in the Texan Permian basin. For example, we could track smaller oil producers such as Steward Energy or the Chinese-owned Surge, both notoriously bad in unneeded flaring, according to a study by the Environmental Defense Fund. It showed small private drillers flare the largest share of gas they produce in the Permian.

The Chinese-owned Surge Energy became one the five worst methane polluters in the Permian where it burned or vented nearly a quarter of its total natural gas production. Methane, the main component of natural gas, is a potent greenhouse gas that contributes to climate change (source).

Part of the problem is that these firms can do little with the gas they extract. It’s too expensive to supply it. So they burn it, which is bad for the environment.

Surge Energy operates multiple sites, one of them is called Moss Creek Asset. We can find a Moss Creek Lake in Texas. From the provided blueprint on the Surge website, we can see it's near the highway 87 that bends in the lower part of the image. This allows us to pinpoint it on Google maps.

Moss Creek Lake Texas, USA, Location: 32.233966, -101.324346

We try to find any flaring activities in this area and find a shit tonne (especially in spring of 2019). We find signs that there are fewer in April/May this year. This can be a good or a bad sign. Either, the flaring started to abate in late 2019, and 2020 (with Covid-19 not unthinkable) or gas is vented without being burned (that’s really bad). Either way, this gives journalists the power to investigate these companies.

If you want a tool to witness changes, a project called SkyTruth, which relies on the same data we just discussed (EOG, Payne Inst. for Public Policy, Colorado School of Mines), provides an easy access tool with daily activity over the past two years.

There are some caveats to this. The resolution for these map points can be poor. The points are usually pretty good but they have between a 50m-1km margin of error, experts say (thx to @ProjectGeoMedia for highlighting it). Their advice: “The more points the better to help narrow down the spots”.

Tracking boats/vessels at night

If the daily Nightfire data provided by Payne Institute Colorado School of Mines doesn't excite you, then maybe the nighttime boat data does the trick.

The nighttime boat tracking data can help your investigation as a last resort if ships decide to shut off their tracers at night.

This isn't unheard of. Last October one-third of the oil tanker fleet by COSCO (Dalian) shut off their ship-tracking transponders after the United States imposed sanctions on the company for allegedly shipping Iranian crude, according to this Reuters report.

At night we might not see them in visible light via commercial products like PlanetLabs or others. But the images that are updated daily we might have a chance.

There is one tiny caveat to this. Images are available and free when they are older than 45 days. But for many investigations, that’s fine (although, we will see that’s doesn't stop us to download the latest data).

Download the nightlight boat data in kmz format from here. If you want access to the daily data, click here (it’s the data repository, with kmz files).

I am choosing the shores of China and select the latest data for 2020–05–22 01:29 (VBD_npp_d20200331_chn_noaa_ops_v23.kmz). Throw this file into Google Earth and untick the boxes so you can only see ‘boats’ and the ‘coverage area).

The next steps would include to compare it with other open-source vessel tracking software/tools and see whether certain chips really turned off their tracking devices.

You can do more with nightlight:

If you fancy more investigative examples where nightlight helped to verify/disprove government statements, check out the piece on Chinese re-education camps in Xinjiang.

Full story here: Despite officials claiming camps closed, light radiance measured revealed campsites were still in operation.

Thanks for reading. DM me on Medium if you have ideas or comments






Investigative journalist with a technical edge, interested in open source investigations, satellite imgs, R, python, AI, data journalism and injustice