Research backed by the RCGI could contribute to Brazil profiting in the future from the carbon dioxide market and to reducing global warming.
The Amazon rain forest is responsible for removing 400 million tons of carbon dioxide from the atmosphere annually. However, climate change and deforestation in the region could transform areas that absorb CO2 into sources of carbon dioxide emissions. The development of an artificial intelligence model for measuring how these environmental variables, like humidity and solar radiation, impact the quantity of carbon dioxide captured in the region, was the main objective of the Master’s research study of environmental scientist Lucas Bauer.
The study thesis was defended last November and carried out within the Integrated Environmental Analysis program of the Federal University of São Paulo (UNIFESP), with the backing of the Research Centre for Innovation in Greenhouse Gases (RCGI), sponsored by the Research Support Fund of the State of São Paulo (FAPESP) in a partnership with Shell of Brazil. A summary of the work can be found in the recently published article Neural Network model for classification of net CO2 flux scenarios in Tapajós Forest, in the Amazon.
The article is signed by Bauer and other researchers, like Luciana Rizzo, who guided the dissertation and is now a Professor at the Physics Institute of the University of São Paulo (USP), as well as a co-guide for the research study, Pedro Luiz Pizzigatti Corrêa, Professor in the Computation and Digital Systems Department of USP’s Polytechnic School. “This is an interdisciplinary work that unites two knowledge areas: the atmospheric sciences and data science,” explains Rizzo, who is a member of the RCGI’s Greenhouse Gases (GHG) research program. “We know that the Amazon rain forest provides an important environmental service by removing carbon dioxide from the atmosphere, but what would be its degree of variability, for example, between dry and wet years? These are the types of questions that guided the study.”
In his search for answers, Bauer did a study of the immense Amazon rain forest that covers 7.2 million square kilometers (km²) spread over nine countries. In this case, the focus of the researcher was the Tapajós National Forest, in the State of Pará, where one of the monitoring towers of the Large-scale Biosphere-Atmosphere Experiment of the Amazon (LBA) is located, and has been operated since the 1990s by the Federal government and linked to the National Institute of Amazon Studies (INPA).
The data recorded by the tower, and available for public consultation, served as Bauer’s information source. “The numbers reflect a local reality, within a radius of about 5 km. But our research is a first step that should, in the long-term, be extrapolated to a regional scale, in order to quantify how the Amazon rain forest, as a whole, is able to remove carbon dioxide from the atmosphere,” says Rizzo.
The data utilized by Bauer cover the period from 2002 to 2005. “We must remember that as of 2005 the deforestation dropped significantly due to the development of public policies for combating this problem,” Rizzo adds. “Unfortunately, over the past six years, the deforestation rate started climbing again, as shown by data of the INPE (National Institute of Space Research). However, the study did not use recent data, because they had not yet been made available for consultation.”
In the study, the researcher also included data from two satellites of the North American Space Agency (NASA), which have continuously gathered data on the atmosphere, since the first decade of 2000, and made that information available to the public. The NASA records inform, for example, the optical thickness of aerosols. “Aerosols are particles in suspension in the atmosphere that interact with solar radiation and interfere with the removal of carbon dioxide. Thus the importance of including this information in the research study,” Rizzo explains.
After gathering these data, Bauer developed an Artificial Intelligence model for estimating the exchange of carbon dioxide in that part of the forest. “In the study, he used the Artificial Neural Network (ANN), which is a machine learning technique that is able to capture the non-linear aspect between the response variable, which would be the removal of carbon dioxide, and the predictive variables, like humidity and solar radiation, for example,” Rizzo explains. “The neural network simulates the processing of information by the human brain to obtain integrated knowledge regarding a given scenario. The processing cells are species of neurons that receive, process, and transmit data to other cells in the system, thus creating an information network.”
Constructing the model presented challenges throughout the study. “The exchanges of carbon dioxide depend on a series of variables and the model needed to capture that,” Rizzo points out. “Generally speaking, the greatest incidence of solar radiation takes place during the dry season. Thus the forest is able to make more photosynthesis occur and, consequently, capture more carbon dioxide from the atmosphere. But, besides photosynthesis, there are other variables that demand attention. Leaf buds, for example, do not depend exclusively on solar radiation: the peak of the budding season takes place in July, while the peak of solar radiation usually comes in September.”
Rizzo says that the plan is that the Artificial Intelligence model developed by Bauer be used to understand other contexts within the Amazon rain forest. “Our study is just beginning, but it points to highly promising results. We identified the predictive variables having the greatest impact on the absorption of carbon dioxide: season of the year, heat flows, and area foliage indices,” Rizzo says. “We can say that the ANN had never before been applied to understanding the Amazon context. We are pioneers in this sense.”
“Brazil needs to, in fact, quantify how the Amazon removes carbon dioxide because this is a fundamental environmental service not only for Brazil but for the entire planet,” notes Luciana Rizzo. “With the evolution of the carbon dioxide market, our country could profit financially due to this service. That is, a standing forest is highly valuable,” she concludes.