Mapping of Threatened Vereda Wetlands in the Brazilian Midwest Using a Domain-Specific U-Net Jeaneth Machicao, Alexandre Augusto Barbosa, Leandro O. Salles, Peter Mann Toledo, Pedro Luiz P. Corrêa, et al. Remote Sensing, 2026 The palm swamp landscapes, particularly the Vereda wetlands and their associated swamp gallery forests (VED.SGF), comprise essential yet threatened ecosystems within the Brazilian Cerrado. In addition to supporting significant portions of biodiversity, they provide critical ecosystem services such as storing and filtering excess rainwater and serving as major carbon reservoirs in organic soils. These wetlands are directly linked to the drainage systems of the headwaters of the main Cerrado river basins, which together account for about two-thirds of Brazil’s hydrographic basins. Mapping and managing VED.SGF ecosystems through remote sensing present major challenges addressed in this first study. Their narrow, dendritic, and complex tabular spatial pattern, often elongated along watersheds on scales of hundreds of kilometers, suffering distortions due to human impact, and the limited amount of annotated data make segmentation particularly challenging. Existing deep learning (DL) methods, typically pre-trained on natural images, struggle to capture the spectral and spatial intricacies of these ecosystems. This study introduces a trained-from-scratch U-Net model supported by field-based experimental procedures to ensure high-quality wetland annotations. The resulting dataset covers approximately 7300 km2 in western Bahia and provides domain-specific weights tailored to remote sensing applications. Using high-resolution (4.6 m) RGB mosaics, the model was trained, validated, and tested to establish a reproducible and scalable pipeline. The proposed method achieved robust results in an independent test area of 8040 km2, with a mean IoU of 0.728, F1-score of 0.843, and Cohen’s Kappa of 0.837. These results demonstrate consistent performance and strong generalization to new areas, establishing a scientifically reliable baseline that situates the model competitively within the current state of the art. By releasing both the model weights and annotated dataset, this study provides valuable resources to advance future research on mapping and monitoring these unique and strategic wetland ecosystems.
Brazil nut journey under future climate change in Amazon Luciano J. S. Anjos, Gabriela S. R. Gonçalves, Vítor A. B. Dutra, Amanda G. Rosa, Lucyana B. Santos, et al. Plos One, 2024 Climate change is among the principal threats to global terrestrial biodiversity, especially to megadiverse ecosystems such as the Amazon rainforest. In this study, we investigate how it could affect an iconic forest species—Bertholletia excelsa—(the Brazil nut) which has values in multiple dimensions in an Amazonian context. We used an ensemble from various distribution modeling methods designed for four different climate scenarios from CMIP6 by the end of the century. Then, we simulate how spatial dynamics under climate change, including explicitly dispersal events, can affect the persistence, colonization, and potential extinction of Bertholletia excelsa in the future. Our results show that by the end of the century there would be a generalized loss of suitability on the Amazon biome, regardless of the climate scenario evaluated, which could promote a significant loss (up to 94%) of the area available for the species via extinction. Our results also show that, in the future, the species would colonize higher altitudes in search of favorable conditions for its survival. Finally, we detected that areas that had previously become unsuitable because of climate change would have favorable conditions by the end of the century. Such an outcome could be useful in fostering an active restoration agenda that can mitigate the negative effects of climate change on species in this study.
Potential of Multipolarization and Multifrequency Radar Images in Identification of Land Cover in the Eastern Amazon Fabrício Sousa da Silva, Laurent Polidori, Peter Man de Toledo, Aline Maria Meiguins de Lima Revista De Geociencias do Nordeste, 2024 O objetivo do presente trabalho é analisar o potencial de imagens de radar de abertura sintética para identificar as diferentes classes de cobertura de terras representativas para a Amazônia Oriental, em área de proteção. Foram analisadas imagens distribuídas gratuitamente referentes aos sensores ALOS PALSAR-2 (banda L, polarização HH e HV) e Sentinel 1A (banda C, polarizações VV e VH), referente ao mês de novembro de 2022, as quais foram avaliadas separadamente e integradas para identificação, através de classificação por aprendizagem de máquina, das seguintes classes: Cobertura Florestal, Cobertura Campestre, Campos úmidos e Cobertura Hídrica. Os resultados mostram que os alvos de interesse, em ambiente representativo da Amazônia, se apresentam melhor diferenciados através da integração multifrequencial e multipolarização entre as bandas C e L, permitindo alcançar acurácia de classificação excelente e todas as classes escolhidas apresentaram índice Kappa satisfatórios
Small Municipalities in the Amazon under the Risk of Future Climate Change Everaldo B. de Souza, Brenda C. S. Silva, Emilene M. F. Serra, Melgris J. Becerra Ruiz, Alan C. Cunha, et al. Climate, 2024 The focus of this work is on small municipalities (population below 50 thousand inhabitants) that cover around 87% of the territory of the Brazilian Legal Amazon (BLA). Based on a comprehensive integrated analysis approach using the three components hazard (climate extremes from CMIP6 future scenarios), exposure (directly affected population), and vulnerability (subdimensions of susceptibility and coping/adaptive capacity by using multidimensional indicators), the latter two using current datasets provided by the official Census IBGE 2022, we document a quantitative assessment of the risk R of natural disasters in the BLA region. We evidenced a worrying and imminent intensification of the curve of R in most Amazonian municipalities over the next two 25-year periods. The overall results of the highest proportions of R (total municipalities affected) pointed out the Amazonas, Roraima, Pará, and Maranhão as the main states, presenting projected categories of R high in the near future (2015 to 2039) and very high in the far future (2040 to 2064). The detailed assessment of the susceptibility and coping/adaptive capacity allowed us to elucidate the principal indicators that aggravate the degree of vulnerability: economy, the precariousness of urban infrastructure, medical services, communication, and urban mobility, whose combined factors, unfortunately, reveal a widespread poverty profile along the small Amazonian municipalities. Our scientific findings can assist decision makers in targeted strategies planning and public policies to minimize and mitigate ongoing and future climate change.
Landscape dynamics and ecosystem fragmentation in three river basins in the Eastern Amazon between 1985 and 2019 Vítor Abner Borges Dutra, Maurício Humberto Vancine, Aline Maria Meiguins de Lima, Peter Mann de Toledo Revista Brasileira De Geografia Fisica, 2023 O Nordeste Paraense sofreu profundas mudanças no decorrer dos seus ciclos de ocupação. Nessa região, a conversão de floresta para outras classes de uso e cobertura da terra ocorre há centenas de anos. Com o advento do Sensoriamento Remoto, é possível quantificar essas mudanças. Nesse sentido, o objetivo do estudo foi analisar a dinâmica da paisagem de três bacias hidrográficas no Nordeste Paraense entre 1985 e 2019 e as implicações dessas mudanças nas classes de formação florestal e mangue. A metodologia adotada abrangeu ferramentas associadas às geotecnologias no subsídio de uma análise integrada no âmbito das bacias hidrográficas. Os resultados evidenciaram uma conversão de floresta para pastagem proporcional à área do município de Castanhal/PA (aproximadamente 1.000 km²). A maior faixa de alteração ficou concentrada nas áreas de nascentes das bacias hidrográficas, como o expressivo aumento de 159% do número de manchas da classe formação florestal (de 2547 para 6604); a redução da média da área de mancha de formação florestal de 121,9 para 31,9 ha; e a redução do percentual de núcleo dessa classe de 33,7 para 16,2%. Por outro lado, não foram evidenciadas mudanças significativas na classe de mangue na área de estudo. Assim, as medidas de conservação e recuperação de áreas importantes para a manutenção dos ecossistemas costeiros e sistemas hídricos locais devem ser priorizadas pelos tomadores de decisão.Palavras-chave: MapBiomas; Formação florestal; Pastagem; Mangue; Métricas de paisagem. Landscape dynamics and ecosystem fragmentation in three river basins in the Eastern Amazon between 1985 and 2019 AbstractPará’s Northeast has undergone deep changes during its occupation cycles. In this region, the conversion of forest to other classes of land use and cover has been going on for hundreds of years. With the advent of Remote Sensing, it is possible to quantify these changes. In this sense, the objective of the study was to analyze the landscape dynamics of three watersheds in Pará’s Northeast between 1985 and 2019, and the implications of these changes in the classes of forest formation and mangrove. The methodology adopted included tools associated with geotechnologies to support an integrated analysis within the scope of watersheds. The results showed a conversion of forest to pasture proportional to the area of the municipality of Castanhal/PA (approximately 1,000 km²). The largest range of alteration was concentrated in the areas of headwaters of the hydrographic basins, such as the expressive increase of 159% in the number of patches of the forest formation class (from 2547 to 6604); the reduction of the average area of forest formation patch from 121.9 to 31.9 ha; and the reduction in the core percentage of this class from 33.7 to 16.2%. On the other hand, no significant changes were observed in the mangrove class in the study area. Thus, measures for the conservation and recovery of important areas for the maintenance of coastal ecosystems and local water systems must be prioritized by decision-makers.Keywords: MapBiomas; Forest formation; Pasture; Mangrove; Landscape metrics
Climate change could reduce and spatially reconfigure cocoa cultivation in the Brazilian Amazon by 2050 Tassio Koiti Igawa, Peter Mann de Toledo, Luciano J. S. Anjos Plos One, 2022 Cocoa is a plant with origins in northwestern South America with high relevance in the global economy. Evidence indicates that cocoa is sensitive to a dry climate, under which crop production is reduced. Projections for future climate change scenarios suggest a warmer and drier climate in the Amazon basin. In this paper, we quantify the potential effects in cocoa production due to its edaphoclimatic suitability changes to the Brazilian Amazon biome and account for regional differences in planning occupation territories. We modeled the suitability of cocoa’s geographical distribution using an ensemble of 10 correlative models that were run in the “biomod2” library and projected to two future climate scenarios (RCPs 4.5 and 8.5) by 2050. Combining information on climate and soil suitability and installed infrastructure in the macro-regions of the Brazilian Amazon. We defined a zoning system to indicate how cocoa production may respond to climate change according to the current and future suitability model. Our results suggest that a reduction in precipitation and an increase in temperature may promote a reduction in the suitability of cocoa production in the Brazilian Amazon biome. In addition of the areas suitable for cocoa plantation, we found a 37.05% and 73.15% decrease in the areas suitable for intensification and expansion zones under RCP 4.5 and 8.5, respectively, compared with the current scenario. We conclude that there may be a need to expand land to cocoa production in the future, or else it will be necessary to plant a cocoa variety resistant to new climatic conditions. Besides, we recommend procedures to combat illegal deforestation to prevent the most critical climate change scenarios from occurring.
