A carregar...
Publicação
Integrating portable x-ray fluorescence and machine learning for soil texture prediction in mangrove ecosystems
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.49 MB | Adobe PDF |
Orientador(es)
Resumo(s)
This study investigated the application of portable X-ray fluorescence (pXRF) spectrometry integrated with machine learning (ML) algorithms to predictict soil texture in mangrove ecosystems. This method is a rapid and environmentally friendly alternative to traditional wet chemistry techniques. A total of 360 soil samples were collected from six mangrove sites in Baía de Todos os Santos, Brazil, which including areas influenced by both marine and riverine processes. Particle size distribution (PSD) was analyzed using laser diffraction, and pXRF was employed to quantify the elemental composition of the samples. Nine ML models were evaluated, with Neural Boosted, Boosted Tree, and K-Nearest Neighbors exhibiting superior predictive performance, characterized by a ratio of performance to deviation (RPD) greater than 2.0, a root mean square error (RMSE) below a specified threshold, and a coefficient of determination (R2) reaching up to 0.99. The critical predictors identified included rubidium (Rb), iron (Fe), zinc (Zn), and molybdenum (Mo), which correspond to the unique geochemical processes present in mangrove soils. This study underscores the effectiveness of integrating pXRF with ML for precise soil texture prediction, facilitating the large-scale monitoring of mangrove ecosystems in terms of carbon storage, nutrient cycling, and conservation initiatives. The proposed approach overcomes the limitations associated with conventional methods, offering a cost-effective and non-destructive solution for environmental assessments, which is particularly vital for quantifying ecosystem services in developing countries.
Descrição
Palavras-chave
Soil texture Green chemistry Ex-site analysis Machine learning pXRF