Artificial intelligence model can help Brazil earn money from carbon credits | AGÊNCIA FAPESP

Artificial intelligence model can help Brazil earn money from carbon credits Machine learning tool measures how environmental variables such as humidity and solar radiation affect the amount of carbon captured in a given area (image: Pixabay)

Artificial intelligence model can help Brazil earn money from carbon credits

March 15, 2023

FAPESP Innovative R&D* – The Amazon Rainforest takes 400 million metric tons of carbon out of the atmosphere every year, yet climate change and deforestation in the region could convert a carbon sink into a source of carbon emissions.

In this context, a study conducted under the aegis of the Integrated Environmental Analysis Graduate Program at the Federal University of São Paulo (UNIFESP) in Brazil, with the support of the Research Center for Greenhouse Gas Innovation (RCGI), developed an artificial intelligence (AI) model to measure how environmental variables such as humidity and sunlight affect the amount of carbon captured in the region.

RCGI is an Engineering Research Center (ERC) established by FAPESP and Shell at the University of São Paulo’s Engineering School (POLI-USP).

The study was part of the master’s research of environmental scientist Lucas Bauer, supervised by Luciana Rizzo, currently a professor at USP’s Physics Institute, and is reported in the article Neural Network model for classification of net CO2 fluxes scenarios in Tapajós Forest, in Amazon published as a preprint and not yet peer-reviewed.

“The study involved interdisciplinary research in two knowledge areas: atmospheric science and data science,” said Rizzo, a member of RCGI’s greenhouse gas research program. “We know the Amazon provides an important environmental service by removing carbon from the atmosphere, but how much does it vary, for example, in dry and wet years? The study set out to answer questions like this.”

To find the answers, Bauer had first to choose a manageable portion of the Amazon Rainforest, which occupies 7.2 million square kilometers in nine countries. The area he selected was the Tapajós National Forest in Pará state (North Brazil), where one of the monitoring towers is located for the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA), conducted since the 1990s by the National Institute for Research on Amazonia (INPA), an arm of the Brazilian government.

The data collected by the tower is public and served as a source for Bauer. “The numbers reflect local conditions in a radius of about 5 km, but our study was only the first step. We now plan to look at the region as a whole in order to quantify the amount of carbon removed from the atmosphere throughout the Amazon Rainforest,” Rizzo said.

The data used in the study was for the period 2002-05. “It’s important to note that deforestation decreased significantly after 2005 in response to the introduction of public policies to combat the problem. Unfortunately, however, it has rebounded in the last six years, as evidenced by statistics from the National Space Research Institute [INPE],” Rizzo said. “The data used in the study wasn’t recent because the data collected in recent years isn’t available to researchers yet.” 

The study also used data collected by two NASA satellites that have been surveying the atmosphere continuously for decades, encompassing aerosol optical depth, for example. “Aerosols are extremely small particles or droplets suspended in the atmosphere that interact with solar radiation and affect carbon removal, so inclusion of this information in the study was very important,” Rizzo said.

Using AI

After assembling the data, Bauer developed an AI model to estimate carbon fluxes in the study area. “He developed a machine learning model known as an artificial neural network [ANN], designed to capture the non-linearity between carbon removal, as response variable, and predictive variables such as relative humidity and solar radiation, for example,” Rizzo said. “The neural network simulates the processing of information by the human brain to obtain integrated knowledge about a specific scenario. The cells that do the processing are analogous to neurons that receive, process and transmit data to other cells in the system, creating an information network.”

Construction of the model was a challenge throughout the research project. “Carbon exchanges depend on a series of variables, and the model needed to capture this. In general, solar radiation levels are highest during the dry season, so that the forest performs more photosynthesis and captures more carbon from the atmosphere. However, other variables require attention in addition to photosynthesis. Leaf sprouting, for example, doesn’t depend only on sunlight. It peaks in July, whereas solar radiation peaks in September,” Rizzo explained.

The researchers wanted the AI model to be usable for the purpose of understanding other aspects of the Amazon Rainforest. “Our research is incipient but already points to highly promising results,” Rizzo said. “We’ve identified the predictive variables that most impact carbon sink conditions: the season of the year, heat flows and leaf area index. No one has ever used a neural network approach to analyze aspects of the Amazon before. Our research is groundbreaking in this respect.”

Quantification of carbon removal from the atmosphere by the Amazon Rainforest is needed because this ecosystem service is significant for Brazil and the entire planet. “As the carbon credit market develops, Brazil can monetize this service. The standing forest is very valuable,” she said.

The article “Neural Network model for classification of net CO2 fluxes scenarios in Tapajós Forest, in Amazon” is at: www.authorea.com/doi/full/10.1002/essoar.10512492.1

* With information from RCGI, an Engineering Research Center established by FAPESP and Shell.

 

  Republish
 

Republish

Agência FAPESP licenses news reports under Creative Commons license CC-BY-NC-ND so that they can be republished free of charge and in a straightforward manner by other digital media or by print media. The name of the author or reporter (when applied) must be cited, as must the source (Agência FAPESP). Using the button HTML below ensures compliance with the rules described in Agência FAPESP’s Digital Content Republication Policy.


Topics most popular