Site icon Conservation news

New satellite data documents deforestation across ecosystems worldwide

Forest clearing for soy in Parecis in the state of Rondonia

Forest clearing for soy in Parecis in the state of Rondonia. Image by Microsoft Zoom Earth.

  • New satellite data documents the loss of vegetation in all types of ecosystems, not just tropical rainforests, around the world.
  • The OPERA DIST-ALERT was developed by the makers of popular forest monitoring platform Global Forest Watch.
  • The monitoring system incorporates managed forests, grasslands, shrublands and croplands along with tropical forests.
  • The data obtained from two of NASA’s Landsat satellites and two of the European Space Agency’s Sentinel-2 satellites reflected lost vegetation around the world due to various factors — ranging from fires in Canada to logging in the Republic of Congo and cyclones in Malawi.

When it comes to monitoring deforestation, tropical rainforests rightfully get the lion’s share of attention. However, as climate change-induced natural disasters and conversion of natural lands for agriculture increase in frequency and intensity, it becomes even more imperative to track vegetation across ecosystems.

A new product by the makers of Global Forest Watch, the popular forest monitoring platform, aims to resolve this gap. Developed as a collaboration between the University of Maryland, nonprofit organization World Resources Institute and NASA, the OPERA Land Surface Disturbance Alert (OPERA DIST-ALERT) is a monitoring system that uses satellite data to provide near-real-time detection of disturbances in all types of vegetation around the world.

“It serves as a first order of information,” Matthew Hansen, a remote sensing scientist and professor at the University of Maryland who was involved in developing the platform, told Mongabay in a video interview. “It alerts you that something has changed on the ground, and that the vegetation has decreased in some way that’s markedly different than in the recent past.”

Deforestation monitoring tools that alert users about reduction in forest cover aren’t new. In recent years, several such platforms have played a vital role in documenting deforestation around the world. They have also helped authorities and conservation organizations in taking immediate action on the ground. However, more often than not, these alerts only pick up disturbances in tropical forests.

The OPERA DIST-ALERTs incorporate managed forests, grasslands, shrublands as well as croplands. “It’s not tuned just for trees but for any vegetation,” Sarah Carter, resource associate at the World Resources Institute, told Mongabay in a video interview.

The product is also designed to pick up disturbances that aren’t necessarily caused by deforestation. In most other forest monitoring platforms, once a disturbance is detected in a forest, it is labeled as a disturbed forest. “The algorithm doesn’t monitor it again, and it excludes the disturbed pixel from the layer,” Carter said. “The new alerts are continuously monitoring, and so repeated disturbances can be picked up.”

Savanna and forest mosiac in the Congo Basin, including recently burned areas, in 2018.
Savanna and forest mosiac in the Congo Basin, including recently burned areas, in 2018. Image by Rhett A. Butler/Mongabay.

For the OPERA DIST-ALERTs, data is obtained from two of NASA’s Landsat satellites along with two of the European Space Agency’s Sentinel-2 satellites. The University of Maryland worked with a team at NASA to develop the algorithm, which can identify and monitor vegetation at a resolution of 30 meters (98 feet). To calibrate the model, the team collected data from drones that they flew over different types of vegetation across many countries, including the U.S., Sudan, Paraguay and the Republic of Congo.

“It’s basically an anomaly detector,” Hansen said. “We take advantage of the time series to say that repeated anomalies are an indication of something of large magnitude.”

Since January 2023, the team has used the data to track vegetation cover and disturbances based on information acquired every 2-4 days. They were also able to analyze the data to find patterns of vegetation loss around the world.

For instance, the data documented the expansion of soy farmlands in the Brazilian Cerrado savanna. “When we overlaid the crop maps, we could see that the alerts happened in field-shaped patterns,” Carter said.

In other locations around the world, too, the data detected reduction in vegetation due to a plethora of factors — fires in Canada, logging in the Republic of Congo, mining in Ghana, urban expansion in Texas and cyclones in Malawi.

“What really impressed me is when we started getting the stories around these alerts and how it showed climate-driven anomalies that are unprecedented,” Hansen said. “If we ran these alerts in, say, 1980, we wouldn’t see a lot of that stuff.”

Once the data was gathered, Hansen said, “packaging it to ensure that users can really make sense of it” was the trickiest part. While currently attempting to make the product more user-friendly, the team is also working to integrate the new alerts with other deforestation alerts.

They are also working to figure out how to better filter the alerts, which inevitably pick up and analyze more information than other deforestation alerts that focus just on tropical forests. “If we don’t filter the alerts, it’s too much information and you get speckles everywhere due to all kinds of different things,” Carter said. “That’s not a product that is easily usable by many stakeholders who might need a bit more guidance on what’s happening.”

This will require labeling the alerts in order to attribute vegetation loss to a specific factor, something for which the platform currently doesn’t provide information. Certain drivers of vegetation loss are easy to detect. “For example, fires have a certain special trajectory that is very different from other things,” he said. “For mechanical clearings, we could probably apply a deep learning algorithm.”

However, challenges are likely to arise while trying to factor in other drivers of vegetation loss.

“What does cyclone-induced forest loss look like? Or forest disease or droughts?” Hansen said. “How do you make that spatially explicit? Our brains might be able to identify these things, but we want to automate these kinds of labeling.”

Banner image: Forest clearing for soy in Parecis in the state of Rondonia. Image by Microsoft Zoom Earth. 

Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.


Exit mobile version