Pamela Boelter Herrmann
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Pamela Boelter Herrmann has a total of 56 co-authors. Here are the top co-authors by number of shared publications:
- Camila Souza Silva
IBAMA
3 shared publications - Clódis de Oliveira Andrades Filho
Postgraduate Program in Remote Sensing, Institute of Geosciences, Federal University of Rio Grande do Sul (UFRGS), Porto Alegre, Brazil
3 shared publications - Deyvis Cano
Academic Program of Environmental Engineering, Universidad de Huánuco, Huánuco 10001, Peru
3 shared publications - Jack Endrick Pastrana Mojica
UFRGS
3 shared publications - Tássia Fraga Belloli
Universidade Federal do Rio Grande do Sul
3 shared publications - Victor Fernandez Nascimento
Universidade Federal do ABC image/svg+xml
3 shared publications
Latest Publications
- Open Access
- 24 October 2026
- Published by Springer Nature in Environmental Earth Sciences
- Vol. 85 (17) , 438
Topographic controls on vegetation recovery in mass-movement scars after an extreme rainfall event in Southern Brazil
Understanding vegetation recovery in landslide scars is essential for risk assessment and post‑disaster management. This study investigated how geomorphometric attributes influence vegetation regeneration in 85 mass‑movement scars triggered by an extreme precipitation event in January 2017 in southern Brazil. Annual time series (2018–2026) of four Sentinel‑2 spectral indices were integrated with topographic variables derived from the Copernicus GLO‑30 DEM. Principal component analysis (PCA) and k-means clustering classified scars into high, medium, and low regeneration groups, while linear mixed models evaluated topographic effects on recovery trajectories. Results demonstrated that regeneration follows a non-linear temporal trajectory, characterized by rapid initial recovery followed by progressive leveling off, which is partly tied to optical index saturation limits in maturing canopies. Slope negatively affected vegetation moisture levels, and elevation significantly modulated the bare soil trajectory, with higher altitudes showing faster soil recovery rates. Scar shape, measured by the circularity index, did not exert a significant effect on any spectral index. Curvature and aspect did not significantly influence regeneration trajectories. The high-regeneration class represented 42.3% of the scars analyzed, indicating relatively greater spectral recovery in most areas, while steep, low‑elevation sites require targeted restoration to reduce erosion and landslide reactivation risk. These results demonstrate that geomorphometric attributes are associated with different dimensions of the observed spectral recovery trajectories and that the approach integrating multiple spectral indices and non-linear modeling is effective for identifying priority areas for long-term ecological monitoring, thereby supporting landslide risk mitigation in subtropical environments.
- 24 October 2026
- Published by Springer Nature in Environmental Earth Sciences
- Vol. 85 (17) , 438
Open AccessTopographic controls on vegetation recovery in mass-movement scars after an extreme rainfall event in Southern Brazil
Understanding vegetation recovery in landslide scars is essential for risk assessment and post‑disaster management. This study investigated how geomorphometric attributes influence vegetation regeneration in 85 mass‑movement scars triggered by an extreme precipitation event in January 2017 in southern Brazil. Annual time series (2018–2026) of four Sentinel‑2 spectral indices were integrated with topographic variables derived from the Copernicus GLO‑30 DEM. Principal component analysis (PCA) and k-means clustering classified scars into high, medium, and low regeneration groups, while linear mixed models evaluated topographic effects on recovery trajectories. Results demonstrated that regeneration follows a non-linear temporal trajectory, characterized by rapid initial recovery followed by progressive leveling off, which is partly tied to optical index saturation limits in maturing canopies. Slope negatively affected vegetation moisture levels, and elevation significantly modulated the bare soil trajectory, with higher altitudes showing faster soil recovery rates. Scar shape, measured by the circularity index, did not exert a significant effect on any spectral index. Curvature and aspect did not significantly influence regeneration trajectories. The high-regeneration class represented 42.3% of the scars analyzed, indicating relatively greater spectral recovery in most areas, while steep, low‑elevation sites require targeted restoration to reduce erosion and landslide reactivation risk. These results demonstrate that geomorphometric attributes are associated with different dimensions of the observed spectral recovery trajectories and that the approach integrating multiple spectral indices and non-linear modeling is effective for identifying priority areas for long-term ecological monitoring, thereby supporting landslide risk mitigation in subtropical environments.
- Open Access
- 13 May 2026
- Published by Universitat Politecnica de Valencia in Revista de Teledetección
- No. 68
Transfer learning in convolutional neural networks for fire scene classification in the Pantanal Biome (Brazil)
During 2024, wildfires in the Pantanal, Brazil, posed a significant threat to the world’s largest wetland. This study presents a scene classification approach using Sentinel-2 imagery from fire-affected and non-affected areas, based on deep learning techniques. First, a model was experimentally trained using the EuroSAT dataset to classify grasslands and herbaceous vegetation. Subsequently, a transfer learning technique was applied to images of the fire-affected Pantanal. Finally, the VGG-19 architecture was trained from scratch, without using pretrained parameters. Considering these three experiments, training with the EuroSAT dataset achieved a loss of 0.202 after 50 epochs. The model with transfer learning, using the weights learned during the experimental training, obtained a validation loss of 0.15 and an accuracy of 95%, whereas training from scratch reached a validation loss of 0.07 and an accuracy of 97%. The results indicate similar performance between the two strategies, demonstrating that in-domain training can be considered an effective method for training CNNs applied to satellite imagery, without compromising classification accuracy, even in the presence of significant differences in the contextual information of the images. This approach is presented as a methodological alternative for environmental monitoring in complex tropical regions, characterized by high spatial and spectral heterogeneity.
- 13 May 2026
- Published by Universitat Politecnica de Valencia in Revista de Teledetección
- No. 68
Open AccessTransfer learning in convolutional neural networks for fire scene classification in the Pantanal Biome (Brazil)
During 2024, wildfires in the Pantanal, Brazil, posed a significant threat to the world’s largest wetland. This study presents a scene classification approach using Sentinel-2 imagery from fire-affected and non-affected areas, based on deep learning techniques. First, a model was experimentally trained using the EuroSAT dataset to classify grasslands and herbaceous vegetation. Subsequently, a transfer learning technique was applied to images of the fire-affected Pantanal. Finally, the VGG-19 architecture was trained from scratch, without using pretrained parameters. Considering these three experiments, training with the EuroSAT dataset achieved a loss of 0.202 after 50 epochs. The model with transfer learning, using the weights learned during the experimental training, obtained a validation loss of 0.15 and an accuracy of 95%, whereas training from scratch reached a validation loss of 0.07 and an accuracy of 97%. The results indicate similar performance between the two strategies, demonstrating that in-domain training can be considered an effective method for training CNNs applied to satellite imagery, without compromising classification accuracy, even in the presence of significant differences in the contextual information of the images. This approach is presented as a methodological alternative for environmental monitoring in complex tropical regions, characterized by high spatial and spectral heterogeneity.
- Open Access
- 24 January 2026
- Published by Reativar Ambiental in Revista Brasileira de Sensoriamento Remoto
- Vol. 7 (1)
Predição e tipificação do uso do fogo na região da Amazônia Legal no Brasil: Estudo com base nos eventos detectados pelo Painel do Fogo
Esta pesquisa tem como objetivo compreender os diferentes usos do fogo na gestão territorial, por meio da análise das queimas prescritas, controladas e de combate, cruzadas com dados do Painel de Fogo do CENSIPAM referentes ao ano de 2023. No âmbito desse estudo em específico, foi realizado uma investigação na Amazônia Legal brasileira, conforme delimitação administrativa oficial, abrangendo os estados do Acre, Amapá, Amazonas, Pará, Rondônia, Roraima, Tocantins, Mato Grosso e parte do Maranhão. Para isso, foram utilizados os eventos de fogo registrados em 2023, obtidos a partir da plataforma Painel do Fogo, que integrou uma base de dados com informações ambientais e socioespaciais, ampliando a capacidade de caracterização dos eventos. Esses dados foram cruzados com os registros das ocorrências de queimas prescritas, controladas e de combate realizadas pelos brigadistas do ICMBio no mesmo ano. A estratégia metodológica adotada considerou técnicas de aprendizado de máquina para realizar previsões e gerar insights sobre os impactos do uso do fogo nas Unidades Federativas (UFs). O modelo Random Forest apresentou a maior acurácia global (76%). As classes de queima de combate e prescrita alcançaram as maiores precisões (0,90 e 0,69, respectivamente), enquanto as curvas ROC e os valores de AUC indicaram boa capacidade de generalização para essas classes (0,84 e 0,79). Em contraste, a classe de queima controlada apresentou menor desempenho na generalização. As análises por Shapley Additive Explanations (SHAP) destacaram variáveis como biomassa e áreas sem fiscalização como fatores relevantes na modelagem da tipificação das ocorrências de fogo
- 24 January 2026
- Published by Reativar Ambiental in Revista Brasileira de Sensoriamento Remoto
- Vol. 7 (1)
Open AccessPredição e tipificação do uso do fogo na região da Amazônia Legal no Brasil: Estudo com base nos eventos detectados pelo Painel do Fogo
Esta pesquisa tem como objetivo compreender os diferentes usos do fogo na gestão territorial, por meio da análise das queimas prescritas, controladas e de combate, cruzadas com dados do Painel de Fogo do CENSIPAM referentes ao ano de 2023. No âmbito desse estudo em específico, foi realizado uma investigação na Amazônia Legal brasileira, conforme delimitação administrativa oficial, abrangendo os estados do Acre, Amapá, Amazonas, Pará, Rondônia, Roraima, Tocantins, Mato Grosso e parte do Maranhão. Para isso, foram utilizados os eventos de fogo registrados em 2023, obtidos a partir da plataforma Painel do Fogo, que integrou uma base de dados com informações ambientais e socioespaciais, ampliando a capacidade de caracterização dos eventos. Esses dados foram cruzados com os registros das ocorrências de queimas prescritas, controladas e de combate realizadas pelos brigadistas do ICMBio no mesmo ano. A estratégia metodológica adotada considerou técnicas de aprendizado de máquina para realizar previsões e gerar insights sobre os impactos do uso do fogo nas Unidades Federativas (UFs). O modelo Random Forest apresentou a maior acurácia global (76%). As classes de queima de combate e prescrita alcançaram as maiores precisões (0,90 e 0,69, respectivamente), enquanto as curvas ROC e os valores de AUC indicaram boa capacidade de generalização para essas classes (0,84 e 0,79). Em contraste, a classe de queima controlada apresentou menor desempenho na generalização. As análises por Shapley Additive Explanations (SHAP) destacaram variáveis como biomassa e áreas sem fiscalização como fatores relevantes na modelagem da tipificação das ocorrências de fogo
- Open Access
- 20 January 2026
- Published by FapUNIFESP (SciELO) in SciELO Preprints
Predição e tipificação do uso do fogo na região da Amazônia Legal no Brasil: Estudo com base nos eventos detectados pelo Painel do Fogo
Esta pesquisa tem como objetivo entender os diferentes usos do fogo na gestão, por meio das queimas prescrita, controlada e de combate, cruzadas com os dados do Painel de Fogo do CENSIPAM no ano de 2023. Foram aplicadas técnicas de aprendizado de máquina para realizar previsões. Foi possível gerar insights sobre os impactos do uso do fogo nas Unidades Federativas (UFs). O modelo Random Forest apresentou a maior acurácia global (76%). As queimas de combate e prescrita tiveram maior precisão (0,90% e 0,69%, respectivamente). As curvas ROC e os valores AUC indicaram boa capacidade de distinção entre essas classes (0,84 e 0,79), enquanto a queima controlada apresentou menor capacidade de generalização. Os valores de Shapley Additive Explanations (SHAP) destacaram as variáveis biomassa e áreas sem fiscalização como fatores que contribuem para a modelagem da tipificação de ocorrências de queimadas.
- 20 January 2026
- Published by FapUNIFESP (SciELO) in SciELO Preprints
Open AccessPredição e tipificação do uso do fogo na região da Amazônia Legal no Brasil: Estudo com base nos eventos detectados pelo Painel do Fogo
Esta pesquisa tem como objetivo entender os diferentes usos do fogo na gestão, por meio das queimas prescrita, controlada e de combate, cruzadas com os dados do Painel de Fogo do CENSIPAM no ano de 2023. Foram aplicadas técnicas de aprendizado de máquina para realizar previsões. Foi possível gerar insights sobre os impactos do uso do fogo nas Unidades Federativas (UFs). O modelo Random Forest apresentou a maior acurácia global (76%). As queimas de combate e prescrita tiveram maior precisão (0,90% e 0,69%, respectivamente). As curvas ROC e os valores AUC indicaram boa capacidade de distinção entre essas classes (0,84 e 0,79), enquanto a queima controlada apresentou menor capacidade de generalização. Os valores de Shapley Additive Explanations (SHAP) destacaram as variáveis biomassa e áreas sem fiscalização como fatores que contribuem para a modelagem da tipificação de ocorrências de queimadas.
- Open Access
- 14 October 2025
- Published by Portal de Periódicos UFU - Universidade Federal de Uberlandia in Caminhos de Geografia
- Vol. 26 (107) , 129-144
EVALUATING REMOTE SENSING TECHNIQUES FOR MONITORING GRASSLAND DEGRADATION
Ecosystem degradation is one of the most significant environmental challenges worldwide. In particular, grassland degradation poses serious threat to agricultural productivity, species diversity, and soil stability. This study aimed to compare the applicability of two remote sensing techniques – the Linear Spectral Mixture Model (LSM) and the Grassland Degradation Index (GDI) – in assessing and characterizing the degree of grassland degradation. The results revealed that the GDI exhibited higher overall accuracy than the LSM, with an accuracy rate of 73.4% compared to 63.1% for LSM. The GDI also yielded a higher Kappa coefficient (0.61), relative to the LSM (0.50). Furthermore, the GDI recorded higher F1 scores across all evaluated classes, indicating an enhanced capacity to identify true positives and minimize false positives and negatives. Both techniques demonstrated satisfactory performance and may serve as valuable tools to support grassland restoration and sustainable management studies.
- 14 October 2025
- Published by Portal de Periódicos UFU - Universidade Federal de Uberlandia in Caminhos de Geografia
- Vol. 26 (107) , 129-144
Open AccessEVALUATING REMOTE SENSING TECHNIQUES FOR MONITORING GRASSLAND DEGRADATION
Ecosystem degradation is one of the most significant environmental challenges worldwide. In particular, grassland degradation poses serious threat to agricultural productivity, species diversity, and soil stability. This study aimed to compare the applicability of two remote sensing techniques – the Linear Spectral Mixture Model (LSM) and the Grassland Degradation Index (GDI) – in assessing and characterizing the degree of grassland degradation. The results revealed that the GDI exhibited higher overall accuracy than the LSM, with an accuracy rate of 73.4% compared to 63.1% for LSM. The GDI also yielded a higher Kappa coefficient (0.61), relative to the LSM (0.50). Furthermore, the GDI recorded higher F1 scores across all evaluated classes, indicating an enhanced capacity to identify true positives and minimize false positives and negatives. Both techniques demonstrated satisfactory performance and may serve as valuable tools to support grassland restoration and sustainable management studies.
- Open Access
- 28 August 2025
- Published by Springer Nature in Landslides
- Vol. 22 (11) , 3615-3624
The biggest landslide event in Brazil: preliminary analysis of the Rio Grande do Sul mega disaster in May 2024
An unprecedented precipitation event in Rio Grande do Sul, Brazil, between April and May 2024 caused widespread floods and landslides. This extreme event impacted 96% of the state’s municipalities, affecting approximately 2.4 million people and resulting in 183 deaths. This article reports the magnitude of this extreme event, focusing on the preliminary results of a comprehensive landslide inventory. Our analysis revealed that this event triggered 15,376 landslides across approximately 18,000 km2, with 16,862 landslide initiation points. This event represents the largest landslide occurrence ever recorded in Brazil, measured by quantity. This study provides a scientific analysis of the event’s scale, spatial distribution and the associated topographic, environmental and infrastructural conditions.
- 28 August 2025
- Published by Springer Nature in Landslides
- Vol. 22 (11) , 3615-3624
Open AccessThe biggest landslide event in Brazil: preliminary analysis of the Rio Grande do Sul mega disaster in May 2024
An unprecedented precipitation event in Rio Grande do Sul, Brazil, between April and May 2024 caused widespread floods and landslides. This extreme event impacted 96% of the state’s municipalities, affecting approximately 2.4 million people and resulting in 183 deaths. This article reports the magnitude of this extreme event, focusing on the preliminary results of a comprehensive landslide inventory. Our analysis revealed that this event triggered 15,376 landslides across approximately 18,000 km2, with 16,862 landslide initiation points. This event represents the largest landslide occurrence ever recorded in Brazil, measured by quantity. This study provides a scientific analysis of the event’s scale, spatial distribution and the associated topographic, environmental and infrastructural conditions.
- Open Access
- 30 May 2022
- Published by Portal de Periódicos UFU - Universidade Federal de Uberlandia in Revista Brasileira de Cartografia
- Vol. 74 (2) , 437-458
Sensoriamento Remoto Aplicado à Análise de Fogo em Formações Campestres: Uma Re-visão Sistemática
Assessing the fire impact on grasslands requires understanding how the environmental and anthropic relationships affect the landscape dynamics. This study carries out a systematic literature review to understand fire behavior in grasslands through remote sensing techniques. To this end, the Scopus data was used employing the PRISMA method with the cluster mapping aid. Initially, 7,881 articles were found in the literature. The methodological steps applied to them resulted in 67 articles, which were used in the analysis. The results indicated increased interest in research in the area, with Brazil having the second-highest number of studies. Several publications utilized orbital images. However, there has been recent growth in the use of images obtained from UAV-mounted sensors. In addition to the NDVI and EVI indices, other indices have been used recently for analyzing burn severity and the vegetation recovery process. These subjects are primarily related to integrated fire management, which must consider conserving biodiversity and human use to reduce the fire's intensity and severity to make them more controllable and reduce their negative impacts. Therefore, remote sensing is essential for understanding the fire spatial-temporal behavior and, consequently, serves as a scientific aid to help decision-making in burn prescription cases, considering the ecosystems maintenance and the better grassland use.
- 30 May 2022
- Published by Portal de Periódicos UFU - Universidade Federal de Uberlandia in Revista Brasileira de Cartografia
- Vol. 74 (2) , 437-458
Open AccessSensoriamento Remoto Aplicado à Análise de Fogo em Formações Campestres: Uma Re-visão Sistemática
Assessing the fire impact on grasslands requires understanding how the environmental and anthropic relationships affect the landscape dynamics. This study carries out a systematic literature review to understand fire behavior in grasslands through remote sensing techniques. To this end, the Scopus data was used employing the PRISMA method with the cluster mapping aid. Initially, 7,881 articles were found in the literature. The methodological steps applied to them resulted in 67 articles, which were used in the analysis. The results indicated increased interest in research in the area, with Brazil having the second-highest number of studies. Several publications utilized orbital images. However, there has been recent growth in the use of images obtained from UAV-mounted sensors. In addition to the NDVI and EVI indices, other indices have been used recently for analyzing burn severity and the vegetation recovery process. These subjects are primarily related to integrated fire management, which must consider conserving biodiversity and human use to reduce the fire's intensity and severity to make them more controllable and reduce their negative impacts. Therefore, remote sensing is essential for understanding the fire spatial-temporal behavior and, consequently, serves as a scientific aid to help decision-making in burn prescription cases, considering the ecosystems maintenance and the better grassland use.
- Open Access
- 14 January 2022
- Published by Universidade Federal de Santa Maria in Ciência e Natura
- Vol. 43, e41
Ambientes de ocorrência do porífero Oncosclera jewelli no Parque Estadual Do Tainhas, Planalto Meridional do RS: uma análise geomorfométrica
O Parque Estadual do Tainhas abrange ambientes de ocorrência do porífero Oncosclera jewelli. Nosso objetivo é indicar as áreas de potencial ocorrência da espécie a partir de fatores relacionados à assinatura geomorfométrica dos pontos de ocorrência ao longo do Rio Tainhas no Parque e sua zona de amortecimento. Foram realizados cruzamentos e análises de dados a partir da construção e manipulação de um banco de dados geográficos, nos SIGs SPRING-5.4.3 e QGIS-2.18, contendo: a) MDEs oriundos das bases Topodata, Embrapa e Alos, obtidos por sensores remotos orbitais para toda a área de estudo, e MDE obtido por imagens aéreas geradas por Drone e as variáveis geomorfométricas produzidas: declividade, aspecto e índice de posição topográfica; b) mapa geológico do Rio Grande do Sul; c) mapa de uso e cobertura do solo, com base em imagens do satélite GeoEye. Os resultados demonstraram que a espécie ocorre exclusivamente sobre a unidade litológica da Formação Serra Geral. O relevo é marcado por terrenos de baixa declividade e orientação sul e leste e em relevo aplainado. Foi gerado o Mapa de Potencial Ocorrência da espécie, demonstrando que aproximadamente 4,5% do comprimento total dos trechos de cursos d’água na área de estudo reúnem as condicionantes geomorfométricas identificadas como de potencial ocorrência da espécie.
- 14 January 2022
- Published by Universidade Federal de Santa Maria in Ciência e Natura
- Vol. 43, e41
Open AccessAmbientes de ocorrência do porífero Oncosclera jewelli no Parque Estadual Do Tainhas, Planalto Meridional do RS: uma análise geomorfométrica
O Parque Estadual do Tainhas abrange ambientes de ocorrência do porífero Oncosclera jewelli. Nosso objetivo é indicar as áreas de potencial ocorrência da espécie a partir de fatores relacionados à assinatura geomorfométrica dos pontos de ocorrência ao longo do Rio Tainhas no Parque e sua zona de amortecimento. Foram realizados cruzamentos e análises de dados a partir da construção e manipulação de um banco de dados geográficos, nos SIGs SPRING-5.4.3 e QGIS-2.18, contendo: a) MDEs oriundos das bases Topodata, Embrapa e Alos, obtidos por sensores remotos orbitais para toda a área de estudo, e MDE obtido por imagens aéreas geradas por Drone e as variáveis geomorfométricas produzidas: declividade, aspecto e índice de posição topográfica; b) mapa geológico do Rio Grande do Sul; c) mapa de uso e cobertura do solo, com base em imagens do satélite GeoEye. Os resultados demonstraram que a espécie ocorre exclusivamente sobre a unidade litológica da Formação Serra Geral. O relevo é marcado por terrenos de baixa declividade e orientação sul e leste e em relevo aplainado. Foi gerado o Mapa de Potencial Ocorrência da espécie, demonstrando que aproximadamente 4,5% do comprimento total dos trechos de cursos d’água na área de estudo reúnem as condicionantes geomorfométricas identificadas como de potencial ocorrência da espécie.