VIEWS FROM ABOVE: Bełchatów Power Station, Poland - Climate TRACE

It’s no secret that Climate TRACE uses satellites to identify and observe source activity — and then correlate that activity with emissions estimates. But what do the satellites actually “see”? And how do we translate activity observations into emissions numbers? In this series, Views From Above, we’ll unpack these questions, one source at a time. In this edition, we spotlight Poland’s coal-fired Bełchatów Power Station.

Source: Wikimedia Commons, used with permission via CC BY-SA 2.0

What is it?

—Asset name: Bełchatów Power Station

—Sector / subsector: Power Sector – Electricity Generation

—Coalition sector lead: WattTime 

—Location: Poland

—Description: Since 1988, the Bełchatów Power Station has been burning coal to make electricity near Bełchatów, Poland. The power station is owned and operated by PGE GiEK Oddział Elektrownia Bełchatów, a subsidiary of the state-owned public power company Polska Grupa Energetyczna. As the largest thermal power station in Europe, the 5,030-MW coal power plant was the top emitter of greenhouse gases in the European Union’s Emissions Trading System in 2022 — and Climate TRACE estimates affirm this ranking, modeling that Bełchatów emitted 27.36 million tonnes of CO₂e100 in 2022 alone (making it the EU’s highest-emitting power plant). 

Observing emissions-causing activities

The Bełchatów Power Station comprises two distinct units: Bełchatów I, Units 1-12 (370 to 390 MW each) and Bełchatów II, Unit 1 (858 MW). Unit 1 was decommissioned in 2019, but the others are all still in operation, with planned retirements set for 2030–2036. 

To zoom in on these units’ GHG-emissions-causing activities, we assess a variety of satellite imagery from Landsat 8, Sentinel-2, and PlanetScope PSScene. We partner with coalition member Global Energy Monitor (GEM) to acquire important details about the plant including capacity, operating status, fuel type, and technology at the unit level. We also use OpenStreetMap (OSM), a free, public geographic database, to validate and augment plant-level data, including tagging parts of the power plant where we expect to see vapor plumes. These annotations can be used to focus machine learning (ML) models to observe only the most-pertinent parts of the plant. 

At the Bełchatów Power Station, the satellite imagery reveals highly visible vapor plumes coming from flue gas desulfurization stacks (FGDs) as well as from natural draft cooling towers (NDTs). For example, in this animated GIF of PlanetScope imagery, we can see vapor plumes from FGDs and NDTs captured across a range of dates.

Associating observed activity with relevant emissions factors

The sub-sources at Bełchatów equate to two common features of electricity plants, each of which has its own emissions factor:

—Natural draft cooling towers (NDTs): NDTs help manage the heat generated in power plants by transferring it to the atmosphere. They facilitate the cooling of water used in the plant's processes to ensure optimal operation of turbines and machinery and the efficient reuse of water. The more heat being produced, the more electricity the plant is generating at that time. Thus, heat expelled via NDTs — visible to the naked eye as steam and in visible spectrum satellite images — is a great proxy signal from which total electricity generation can be estimated, which in turn can be used to estimate CO₂ emissions by simply applying emissions factors. 

—Flue gas desulfurization stacks (FGDs): Sulfur dioxide (SO₂) is an air pollutant harmful both to human health and the environment. FGD is a technology used to remove SO₂ from the exhaust flue gasses produced by burning fossil fuels before being released into the atmosphere. When power plants use a wet FGD process, water vapor is emitted when the FGD process is running. While FGD stacks effectively reduce SO₂ emissions, they do nothing to change CO₂ emissions. Similar to cooling towers, FGD stack activity is a good proxy for electricity generation, and can therefore be used to model final CO₂ emissions estimates.

Translating reporting period activity into emissions estimates

WattTime combines multiple complementary approaches to generate source-level emissions estimates for a large number of fossil fuel power plants. For a given plant like Bełchatów Power Station, we first estimate its total electricity generation with a combination of the following approaches:

1. Apply machine learning (ML) models, previously trained on reported electricity generation data from the U.S. EPA, ENTSO-E from Europe, and the Australia National Electricity Market, to estimate electricity generation from the cooling tower and FGD proxy signals observed in Landsat 8, Sentinel-2, and PSScene satellite imagery.

2. Smooth ML predictions by ensembling them with region- and fuel-specific average capacity factors from country-level data to estimate generation. For plants lacking NDT or FGD technology (unlike Bełchatów), this region- and fuel-specific average is simply used on its own.

3. Infer CO₂ emissions from estimated electricity generation using region-, fuel-, and prime-mover-specific average carbon intensities to convert source-level generation estimates to emissions estimates, and report both facility-level estimates and aggregate facility-level estimates up to the country level.

Read more about WattTime's power sector methodology here. 

Cross-checking models for improved accuracy

Consider this sample of cross-validation predictions for Bełchatów from both NDT and FGD generation models, which predict the average capacity factor over the preceding 30 days using ENTSO-E generation data as ground truth. 

From these Climate TRACE data, we can see that:

—Bełchatów’s rolling average 30-day capacity factor never fell below 45% capacity during the period 2019–2022 and was maintained above 60% for the entirety of 2022. 

—The rolling 30-day-average hourly generation in MWh, inferred based on capacity factor, is displayed in the second plot. 

—The NDT model performs over two times better than the FGD model for this plant, a representative characteristic of the NDT task given its clearer signal.

Ready to dig into more asset-level data? Visit our website to learn more about Bełchatów, or to explore other assets of interest with Climate TRACE’s interactive Emissions Map.