Logo do repositório
 
A carregar...
Miniatura
Publicação

Modelação Preditiva da Vegetação Natural Potencial no Médio Tejo

Utilize este identificador para referenciar este registo.
Nome:Descrição:Tamanho:Formato: 
Pombo_Gonçalo_TM_2024.pdf2.8 MBAdobe PDF Ver/Abrir

Resumo(s)

Este trabalho teve como principal objetivo desenvolver e testar uma metodologia preditiva espacial para identificar a vegetação potencial da região do Médio Tejo e representá-la num Sistema de Informação Geográfica (SIG). Com base na identificação prévia das principais séries e geosséries de vegetação da área de estudo, foi possível criar modelos preditivos robustos, que integram observações de campo e variáveis ambientais como o pH do solo, textura do solo, altitude e índice ombrotérmico anual. A metodologia adotada utilizou o algoritmo de máxima entropia (MaxEnt), escolhido pela sua capacidade de lidar com variáveis contínuas e categóricas, bem como pela sua alta performance preditiva. Os resultados obtidos, validados por meio de uma validação cruzada (k-fold) com elevados valores de área abaixo da curva ROC, confirmaram a precisão dos modelos na estimativa da distribuição das séries de vegetação, como o azinhal, cercal, sobral e a geossérie ripícola. Além de identificar a distribuição potencial da vegetação, o estudo também forneceu informações importantes sobre os fatores ambientais que influenciam essa distribuição, contribuindo para projetos de conservação e recuperação ecológica. A metodologia desenvolvida oferece uma ferramenta eficaz para a gestão sustentável da biodiversidade e pode ser replicada noutros contextos biogeográficos.
The main objective of this work was to develop and test a spatial predictive methodology to identify the potential vegetation of the Médio Tejo region and represent it in a Geographic Information System (GIS). Based on the prior identification of the main vegetation series and geoseries in the study area, it was possible to create robust predictive models that integrate field observations and environmental variables such as soil pH, soil texture, altitude and the annual ombrothermal index. The methodology adopted used the maximum entropy algorithm (MaxEnt), chosen for its ability to deal with continuous and categorical variables, as well as its high predictive performance. The results obtained, validated by crossvalidation (k-fold) with high values for the area under the ROC curve, confirmed the accuracy of the models in estimating the distribution of vegetation series such as holm oak, hedgerow, cork oak and riparian geoseries. As well as identifying the potential distribution of vegetation, the study also provided important information on the environmental factors that influence this distribution, contributing to conservation and ecological recovery projects. The methodology developed offers an effective tool for the sustainable management of biodiversity and can be replicated in other biogeographical contexts.

Descrição

Palavras-chave

Vegetação Natural Potencial Clímax Modelação Médio Tejo Séries de vegetação

Contexto Educativo

Citação

Projetos de investigação

Unidades organizacionais

Fascículo

Editora

Licença CC