Saltar para o conteúdo principal

FCCNFCT Digital Services

Access the national information on scientific and technological funding in Portugal

Activity Funded

A Deforestation Prediction System in Amazon Forest based on Neuro-Symbolic AI

Um Sistema de Predição de Desmatamento na Floresta Amazônica baseado em IA Neuro-Simbólica

Reference
2024.14198.PEX
Project Start Date
2026-02-23
Project End Date
2027-08-22
Principal Investigator
Scientific Area
Exact sciences
Funding Program
Concurso para Projetos de Investigação de caráter Exploratório em Todos os Domínios Científicos 2024

Abstract

Tropical forests, especially those in the Brazilian Legal Amazon (BLA) region, play a crucial role in maintaining global climate balance, biodiversity, and environmental sustainability. However, in 2021, forest loss in the Brazilian Amazon reached alarming proportions, marking its highest increase in 13 years. In 2023, a degradation of approximately 2.5 X 10^6 km^2 (~38% of all remaining forests) was recorded, caused by various factors/actors such as fires, timber extraction, illegal mining, and extreme droughts. The constant loss of forest cover results in significant environmental, social, and economic repercussions, both globally and for local communities that rely on the forest for their livelihoods. Furthermore, the destruction of the largest tropical forest (~40% of the total existing) is a major concern for the world (UN Agenda 2030) and especially for Amazonian countries, due to its potential impacts on global climate. In response to the growing threats to these critical ecosystems, various initiatives have been adopted to mitigate the impacts of deforestation, including accurate monitoring of forest degradation activity. Among the monitoring programs (governmental, non-governmental organizations, and for-profit institutions) in place to measure and prevent deforestation in the BLA region are: PRODES and DETER from INPE (National Institute for Space Research), MapBiomas (a multi-institutional initiative), SAD from Imazon, and GLAD from Global Forest Watch. The analysis of deforestation patterns and trends in the BLA can help to understand the dynamics of land occupation and deforestation in the region, allowing for predictions of future land occupation pathways. This understanding is crucial for identifying deforestation expansion zones and formulating public policies aimed at halting deforestation in the area. In this context, the overall goal of this scientific project is to develop a monitoring system based on intelligent approaches that will issue alerts for new areas of degradation and deforestation in the BLA region. To achieve this, environmental experts’ knowledge of tropical forest deforestation behavior patterns will be transformed into symbolic representations understandable by machines. Thus, the organic knowledge of these professionals will be represented through logical rules and/or graph structures (actor and process relationships), and can be used to accelerate the training of sub-symbolic techniques (e.g., neural networks and deep learning), as well as enhance the robustness of intelligent approaches in the task of predicting, either semi-automatically or fully automatically, new deforestation areas in the Amazon rainforest. In the literature, one research field thathas yielded very important results in different knowledge domains (medicine, militar, and finance) for inference tasks is called neuro-symbolic AI. Neuro-symbolic AI can build more powerful reasoning and learning systems by combining the strengths of deep learning-based methods and symbolic reasoning techniques. In the final stage of this scientific project, the expected contribution is a prediction system based on neuro-symbolic AI that will perform inferences for new deforestation in the BLA, providing greater agility to the existing monitoring governmental programs (e.g., DETER), which are often carried out manually by scarce experts. Finally, to ensure the success of this scientific project, an extremely qualified team of professionals has been formed, and partnerships have been established with key research institutions in Brazil (INPE, UNIFESP, and IEAv) and Portugal (IST/ULisboa). These professionals have many years of experience in the fields of artificial intelligence, image processing, remote sensing, ecology, and geoscience, with a primary focus on monitoring degradation and deforestation activities in Brazilian tropical forests.

Institutions

Main Institutions

  • Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa (INESC-ID)

Other Institutions

  • Instituto de Estudos Avançados (IEAv)
  • Instituto Nacional de Pesquisas Espaciais (INPE)
  • Universidade Federal de São Paulo (UNIFESP)

Funding 59.859,41 €

Fundação para a Ciência e a Tecnologia (FCT) - Portugal

59.859,41 €