Publicação

Soil Landscape Modelling – placing place in its place

Ver documento

Detalhes bibliográficos
Resumo:Landscape variables, which are also factors of soil formation, can be combined with existing soil map data to train Artificial Neural Networks (ANNs) in order to predict soil types in unmapped areas. In this study, the impact of location data and proximity of the training data on the performance of ANN models, for two catchments in northern Portugal, is evaluated. Results are largely concurrent between catchments, indicating that using latitude and longitude data produces more accurate models, whilst taking into account the spatial autocorrelative properties of input data makes ANN models converge for a better “local” rather than “global” solution. The conclusion is that hillslopes show some degree of connectivity which is passed onto soils, and conforms to the principles of the catena concept.
Autores principais:Fonseca, Inês
Outros Autores:Freire, Sérgio; Brasil, Ricardo; Rocha, Jorge; Tenedório, José A.
Assunto:Landscape Artificial Neural Networks Soil maps Geographical Information Systems
Ano:2013
País:Portugal
Tipo de documento:capítulo de livro
Tipo de acesso:acesso aberto
Instituição associada:Universidade de Lisboa
Idioma:inglês
Origem:Repositório da Universidade de Lisboa

Registos relacionados

Registos relacionados