Publicação

The importance of sampling for the efficiency of artificial neural networks in digital soil modelling

Ver documento

Detalhes bibliográficos
Resumo:In Portugal, soil mapping remains incomplete, and there are also significant problems with the existing cartography. Digital Soil Mapping uses advanced computerbased techniques such as Artificial Neural Networks (ANN) for mapping soil classes in a cheaper, more consistent and flexible way, using surrogate landscape data. This work used five different training sets to evaluate the impact that sampling has on the predictive accuracy of ANNs. The testes were carried out in IDRISI Taiga for two catchments in northern Portugal, using an ANN method known as multi-layer perceptron. Results show that sampling design is very important for the accuracy of soil mapping with ANNs.
Autores principais:Freire, Sérgio
Outros Autores:Fonseca, Inês; Brasil, Ricardo; Rocha, Jorge; Tenedório, José António
Assunto:Digital soil mapping AutoMAPticS IDRISI Taiga Mondim de Basto Vila Real
Ano:2012
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
Descrição
Resumo:In Portugal, soil mapping remains incomplete, and there are also significant problems with the existing cartography. Digital Soil Mapping uses advanced computerbased techniques such as Artificial Neural Networks (ANN) for mapping soil classes in a cheaper, more consistent and flexible way, using surrogate landscape data. This work used five different training sets to evaluate the impact that sampling has on the predictive accuracy of ANNs. The testes were carried out in IDRISI Taiga for two catchments in northern Portugal, using an ANN method known as multi-layer perceptron. Results show that sampling design is very important for the accuracy of soil mapping with ANNs.