Estimativa de biomassa e carbono em floresta estacional semidecidual utilizando índices de vegetação via satélites
Carregando...
Data
2021-09-01
Autores
Moraes, Felipe Góes de
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O maior desafio da humanidade no século XXI, sem dúvida, são as mudanças climáticas. Emissões de gases causadores do efeito estufa na atmosfera, em especial o dióxido de carbono (CO2), são sua principal causa e motor de aceleramento. Portanto, evitar que mais carbono atinja a atmosfera é a melhor opção no momento. As florestas nativas são estoques e sumidouros de carbono, então preservá-las é uma ferramenta poderosa de mitigação das mudanças climáticas. Biomas devastados por séculos de exploração e ocupação humana, como a Mata Atlântica, se encontram em situação crítica de conservação. O caminho mais efetivo para a manutenção da floresta em pé é oferecendo subsidio financeiro, seja por pagamentos de serviços ecossistêmicos ou pela negociação de créditos de carbono. Uma etapa obrigatória nesse processo é quantificar o estoque de carbono contido nesses fragmentos florestais, que pode ser mais rápido mediante uso do sensoriamento remoto. Essa técnica nos permite acessar essas informações com maior eficácia, até nos lugares remotos onde se localizam as florestas tropicais, sendo uma vantagem para monitoramentos periódicos dos estoques de carbono nestas florestas. Índices de vegetação podem ser correlacionados com características morfológicas e fenológicas das árvores, como a biomassa. Assim, estabelecer correlações, modelos preditivos e categorias entre estoques de carbono florestal acima do solo com índices de vegetação obtidos por meio de imagens de satélite é importante para ganhar escala e eficiência no processo de valoração das florestas nativas. Nessa pesquisa foram avaliados quatro índices de vegetação (NDVI, EVI, RVI e OSAVI) gerados por dois satélites (Landsat 8 sensor OLI e Sentinel-2 sensor MSI), e como se relacionam com a biomassa dos remanescentes de Floresta Estacional Semidecidual nas Fazendas Experimentais Edgárdia e Lageado da UNESP, no município de Botucatu-SP. As medições vieram de inventário florestal dos sete fragmentos, realizado entre setembro e novembro de 2017, com a alocação de 18 parcelas permanentes de 0,2 ha (20 m x 100 m) onde foram medidos o diâmetro e altura de todas as árvores com mais de 5 cm de DAP. Por meio do Google Earth Engine foram selecionadas seis cenas com pouca cobertura de nuvens (<10%) de ambos satélites, de setembro a outubro de 2017. Na plataforma foram calculados os valores médios dos índices de vegetação para cada parcela com um buffer de 60 m ao redor (um hectare de área). Análises estatísticas descritivas dos efeitos dos satélites na biomassa e área basal e regressões lineares indicam não-correlação entre os valores dos IVs com os medidos em campo. Alargando a abordagem, foi tentada a categorização das variáveis biomassa e área basal e sua posterior classificação com a técnica bootstrap, com resultados mostrando percentuais de acerto pouco superiores a 55% para categorias dos extremos. Concluiu-se que a utilização desses índices de vegetação para estimativas de biomassa acima do solo em áreas de Floresta Estacional Semidecidual não é eficiente, assim como modelos preditivos que incorporem esses IVs. A abordagem por categorias é promissora para sondagens prévias, mas necessita de refinamento. IVs podem ser mais indicados para monitoramento do sequestro de carbono do que do seu estoque.
Humanity's greatest challenge in the 21st century is undoubtedly climate change. Greenhouse gas emissions into the atmosphere, especially carbon dioxide (CO2), are its main cause and acceleration engine. Therefore, preventing more carbon from reaching the atmosphere is our best option at the moment. Native forests are carbon stocks and sinks, so preserving them is a powerful tool for mitigating climate change. Biomes devastated by centuries of human exploitation and occupation, such as the Atlantic Rainforest, are in a critical conservation state. The most effective way to maintain the forest standing is offering financial subsidy, either by ecosystem services payments or by negotiating carbon credits. A mandatory step for this financial compensation is to quantify the carbon stock contained in these forest fragments, which can be faster using remote sensing. This technique allows us to access this information more effectively, even in the remote places where tropical forests are located, being an advantage for periodic monitoring of these forests carbon stocks. Vegetation indices can be correlated with morphological and phenological characteristics of trees, such as biomass. Thus, establishing correlations, predictive models and categories between aboveground forest carbon stocks and vegetation indices obtained through satellite images is important to gain scale and efficiency in the remaining native forests valuing process. In this research, four vegetation indices (NDVI, EVI, RVI and OSAVI) generated by two satellites (Landsat 8 OLI and Sentinel-2 MSI) were evaluated on how they relate to the Semideciduous Seasonal Forest remnants’ biomass in Edgárdia and Lageado, UNESPs’ Experimental Farms, in Botucatu-SP. The measurement data came from a forest inventory of all seven fragments, carried out between September and November 2017, with 18 permanent plots of 0.2 ha (20 m x 100 m) allocation, of which the diameter and height of all trees larger than 5 cm DBH was measured. Through Google Earth Engine, six scenes with little cloud cover (<10%) from both satellites were selected, from September and October 2017. On the platform, the average values of vegetation indices for each plot were calculated with a 60 m buffer (a hectare of area). Descriptive statistical analyzes of the effects of satellites on biomass and basal area and linear regressions indicate non-correlation between VIs values and those measured in the field. Broadening the approach, an attempt was made to categorize and subsequently test the classification of variables biomass and basal area, with bootstrap technique, with results showing hit percentages just above 55% for extreme categories. It was concluded that the use of these vegetation indices for aboveground biomass estimates in Semideciduous Seasonal Forest areas is not efficient, as well as predictive models that incorporate these VIs. The category approach is promising for previous surveys, but needs refinement. VIs may be better suited for monitoring carbon sink than its’ stocks.
Humanity's greatest challenge in the 21st century is undoubtedly climate change. Greenhouse gas emissions into the atmosphere, especially carbon dioxide (CO2), are its main cause and acceleration engine. Therefore, preventing more carbon from reaching the atmosphere is our best option at the moment. Native forests are carbon stocks and sinks, so preserving them is a powerful tool for mitigating climate change. Biomes devastated by centuries of human exploitation and occupation, such as the Atlantic Rainforest, are in a critical conservation state. The most effective way to maintain the forest standing is offering financial subsidy, either by ecosystem services payments or by negotiating carbon credits. A mandatory step for this financial compensation is to quantify the carbon stock contained in these forest fragments, which can be faster using remote sensing. This technique allows us to access this information more effectively, even in the remote places where tropical forests are located, being an advantage for periodic monitoring of these forests carbon stocks. Vegetation indices can be correlated with morphological and phenological characteristics of trees, such as biomass. Thus, establishing correlations, predictive models and categories between aboveground forest carbon stocks and vegetation indices obtained through satellite images is important to gain scale and efficiency in the remaining native forests valuing process. In this research, four vegetation indices (NDVI, EVI, RVI and OSAVI) generated by two satellites (Landsat 8 OLI and Sentinel-2 MSI) were evaluated on how they relate to the Semideciduous Seasonal Forest remnants’ biomass in Edgárdia and Lageado, UNESPs’ Experimental Farms, in Botucatu-SP. The measurement data came from a forest inventory of all seven fragments, carried out between September and November 2017, with 18 permanent plots of 0.2 ha (20 m x 100 m) allocation, of which the diameter and height of all trees larger than 5 cm DBH was measured. Through Google Earth Engine, six scenes with little cloud cover (<10%) from both satellites were selected, from September and October 2017. On the platform, the average values of vegetation indices for each plot were calculated with a 60 m buffer (a hectare of area). Descriptive statistical analyzes of the effects of satellites on biomass and basal area and linear regressions indicate non-correlation between VIs values and those measured in the field. Broadening the approach, an attempt was made to categorize and subsequently test the classification of variables biomass and basal area, with bootstrap technique, with results showing hit percentages just above 55% for extreme categories. It was concluded that the use of these vegetation indices for aboveground biomass estimates in Semideciduous Seasonal Forest areas is not efficient, as well as predictive models that incorporate these VIs. The category approach is promising for previous surveys, but needs refinement. VIs may be better suited for monitoring carbon sink than its’ stocks.
Descrição
Palavras-chave
sensoriamento remoto, Landsat 8, Sentinel-2, Google earth engine, mata atlântica, remote sensing, atlantic rainforest