Discriminação de Ipomoea cavalcantei, Ipomoea marabaensis e indivíduos híbridos por meio de traços foliares morfofisiológicos e hiperespectrais nas cangas da Serra de Carajás (PA-Brasil)
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
As cangas da Serra dos Carajás (PA-Brasil) são savanas metalófitas compostas de uma vegetação predominantemente herbácea-arbustivas que crescem sob afloramentos de rochas ferruginosas. Essa vegetação abriga uma flora com alto número de espécies endêmicas e com adaptações a condições extremas. Devido a sua associação ao afloramento de minério de ferro, atividades do setor minerador geram uma grande vulnerabilidade a esses ecossistemas. Considerando a importância da conservação da biodiversidade das cangas, especialmente das espécies raras, ameaçadas e em perigo de extinção, e as limitações dos métodos convencionais de identificação taxonômica em campo, este estudo investiga a utilização de dados de reflectância hiperespectral foliar in situ para discriminação de Ipomoea cavalcantei, Ipomoea marabaensis e híbridos ocorrentes nas cangas de Carajás, Pará, Brasil. O uso de plataformas com sensores hiperespectrais possibilita o monitoramento populacional de espécies vegetais em larga escala, diminuindo o tempo de aquisição de dados, os custos de trabalhos em campo e os riscos operacionais associados ao acesso a áreas de mineração. Dessa forma, este estudo pretende fornecer dados que contribuam com o monitoramento das populações de I. cavalcantei, I. marabaensis e híbridos por meio do uso de tecnologias de sensoriamento remoto. São objetivos específicos: 1. Avaliar a separabilidade espectral das espécies e dos híbridos estudados; 2. Identificar as melhores bandas espectrais para diferenciação das espécies e híbridos; e 3. Definir características morfofisiológicas que explicam a variabilidade espectral encontrada. Os dados de reflectância foliar dos alvos foram registrados na faixa do espectro eletromagnético óptico entre 350 e 2500 nm através do espectrorradiômetro hiperespectral portátil FieldSpec® 4.0NP acoplado ao acessório Leaf Clip. Após a obtenção dos dados espectrais, das mesmas folhas foram mensuradas características morfológicas e fisiológicas. O classificador Random Forest (RF) foi utilizado para avaliar a capacidade de separabilidade das classes estudadas. A seleção dos comprimentos de onda mais importantes para a separabilidade das classes foi calculada a partir do índice de Gini. Além disso, uma Análise de Redundância (RDA) das características morfológicas e fisiológicas foliares foi calculada para explicar a variabilidade espectral encontrada. Os resultados evidenciaram uma alta acurácia na separabilidade espectral de I. cavalcantei (0.84) e I. marabaensis (0.85). No entanto, a acurácia de classificação de híbridos foi inferior (0.47). As bandas espectrais 350 nm, 660 nm, 870 nm, 980 nm e 1240 nm foram mais importantes para a diferenciação espectral das classes analisadas. A variação na reflectância foliar encontrada foi explicada por parâmetros morfofisiológicos (largura, teor de água, nitrogênio, cálcio, ferro e fósforo). Os resultados alcançados neste estudo evidenciam a separabilidade de I. cavalcantei e I. marabaensis através de dados hiperespectrais de reflectância foliar. Por outro lado, os indivíduos híbridos não apresentaram uma boa discriminação, uma vez que esses indivíduos expressam fenótipos com características intermediárias de I. cavalcantei e I. marabaensis, a dificuldade na precisão de identificação poderia ser esperada. A abordagem aqui usada demonstra um grande potencial para o monitoramento das populações de I. cavalcantei e I. marabaensis nas Cangas de Carajás através de plataformas com sensores hiperespectrais. Palavras-chave: Espectros de reflectância. Cangas. Ipomoeas.
The cangas of Serra dos Carajás (PA-Brazil) are metallophytic savannas composed predominantly of herbaceous-shrub vegetation that grows under outcrops of ferruginous rocks. This vegetation harbors a flora with a high number of endemic species adapted to extreme conditions. Due to their association with iron ore outcrops, mining activities generate significant vulnerability to these ecosystems. Considering the importance of conserving the biodiversity of the cangas, especially rare, threatened, endangered, and vulnerable species, and the limitations of conventional methods of taxonomic identification and species mapping in the field, this study investigates the use of in situ leaf hyperspectral reflectance data to discriminate Ipomoea cavalcantei, Ipomoea marabaensis, and hybrids occurring in the cangas of Carajás, Pará, Brazil. The use of platforms with hyperspectral sensors enables population monitoring of plant species on a large scale, reducing data acquisition time, fieldwork costs, and operational risks associated with access to mining areas. Thus, this study aims to provide data that contribute to the monitoring of populations of I. cavalcantei, I. marabaensis, and hybrids through the use of remote sensing technologies. Specific objectives are: 1. To assess the spectral separability of the studied species and hybrids; 2. To identify the best spectral bands for differentiation of species and hybrids; and 3. To define morphophysiological characteristics that explain the spectral variability found. Leaf reflectance data from the targets were recorded in the optical electromagnetic spectrum range between 350 and 2500 nm using the portable hyperspectral spectroradiometer FieldSpec® 4.0NP coupled with the Leaf Clip accessory. After obtaining spectral data, the same leaves were measured for morphological and physiological characteristics. The Random Forest (RF) classifier was used to assess the separability capability of the studied classes. The selection of the most important wavelengths for classseparability was calculated from the Gini index. Additionally, a Redundancy Analysis (RDA) of leaf morphological and physiological characteristics was calculated to explain the spectral variability found. The results showed high accuracy in the spectral separability of I. cavalcantei (0.84) and I. marabaensis (0.85). However, the classification accuracy of the hybrid class was lower (0.47). The spectral bands 350 nm, 660 nm, 870 nm, 980 nm, and 1240 nm were most important for spectral differentiation of the classes. Leaf reflectance variation was explained by morphophysiological parameters (width, water content, nitrogen, calcium, iron, and phosphorus). The results achieved in this study demonstrate the separability of I. cavalcantei and I. marabaensis through hyperspectral leaf reflectance data. However, hybrid individuals did not show good discrimination, as these individuals express phenotypes with intermediate characteristics of I. cavalcantei and I. marabaensis, the difficulty in identification accuracy could be expected. The approach used here demonstrates great potential for monitoring populations of I. cavalcantei and I. marabaensis in the cangas of Carajás through platforms with hyperspectral sensors. Keywords: Reflectance spectra. Cangas. Ipomoeas.
The cangas of Serra dos Carajás (PA-Brazil) are metallophytic savannas composed predominantly of herbaceous-shrub vegetation that grows under outcrops of ferruginous rocks. This vegetation harbors a flora with a high number of endemic species adapted to extreme conditions. Due to their association with iron ore outcrops, mining activities generate significant vulnerability to these ecosystems. Considering the importance of conserving the biodiversity of the cangas, especially rare, threatened, endangered, and vulnerable species, and the limitations of conventional methods of taxonomic identification and species mapping in the field, this study investigates the use of in situ leaf hyperspectral reflectance data to discriminate Ipomoea cavalcantei, Ipomoea marabaensis, and hybrids occurring in the cangas of Carajás, Pará, Brazil. The use of platforms with hyperspectral sensors enables population monitoring of plant species on a large scale, reducing data acquisition time, fieldwork costs, and operational risks associated with access to mining areas. Thus, this study aims to provide data that contribute to the monitoring of populations of I. cavalcantei, I. marabaensis, and hybrids through the use of remote sensing technologies. Specific objectives are: 1. To assess the spectral separability of the studied species and hybrids; 2. To identify the best spectral bands for differentiation of species and hybrids; and 3. To define morphophysiological characteristics that explain the spectral variability found. Leaf reflectance data from the targets were recorded in the optical electromagnetic spectrum range between 350 and 2500 nm using the portable hyperspectral spectroradiometer FieldSpec® 4.0NP coupled with the Leaf Clip accessory. After obtaining spectral data, the same leaves were measured for morphological and physiological characteristics. The Random Forest (RF) classifier was used to assess the separability capability of the studied classes. The selection of the most important wavelengths for classseparability was calculated from the Gini index. Additionally, a Redundancy Analysis (RDA) of leaf morphological and physiological characteristics was calculated to explain the spectral variability found. The results showed high accuracy in the spectral separability of I. cavalcantei (0.84) and I. marabaensis (0.85). However, the classification accuracy of the hybrid class was lower (0.47). The spectral bands 350 nm, 660 nm, 870 nm, 980 nm, and 1240 nm were most important for spectral differentiation of the classes. Leaf reflectance variation was explained by morphophysiological parameters (width, water content, nitrogen, calcium, iron, and phosphorus). The results achieved in this study demonstrate the separability of I. cavalcantei and I. marabaensis through hyperspectral leaf reflectance data. However, hybrid individuals did not show good discrimination, as these individuals express phenotypes with intermediate characteristics of I. cavalcantei and I. marabaensis, the difficulty in identification accuracy could be expected. The approach used here demonstrates great potential for monitoring populations of I. cavalcantei and I. marabaensis in the cangas of Carajás through platforms with hyperspectral sensors. Keywords: Reflectance spectra. Cangas. Ipomoeas.
Description
Citation
SOARES, Nayara Donatti. Discriminação de Ipomoea cavalcantei, Ipomoea marabaensis e indivíduos híbridos por meio de traços foliares morfofisiológicos e hiperespectrais nas cangas da Serra de Carajás (PA-Brasil). 2024. 45 f. Dissertação (Mestrado em Botânica) - Universidade Federal de Viçosa, Viçosa. 2024.
