Artificial neural networks and adaptive neuro-fuzzy inference systems for prediction of soil respiration in forested areas southern Brazil.

The purpose of this study was to estimate the temporal variability of CO₂ emission (FCO₂) from O₂ influx into the soil (FO₂) in a reforested area with native vegetation in the Brazilian Cerrado, as well as to understand the dynamics of soil respiration in this ecosystem. The database is composed of soil respiration data, agroclimatic variables, improved vegetation index (EVI), and soil attributes used to train machine learning algorithms: artificial neural network (ANN) and an adaptive neuro-fuzzy inference system (ANFIS). The predictive performance was evaluated based on the mean absolute error (MEA), root mean square error (RMSE), mean absolute percentage error (MAPE), agreement index (d), confidence coefficient (c), and coefficient of determination (R²). The best estimation results for validation were FCO₂ with multilayer perceptron neural network (MLP) (R² = 0.53, RMSE = 0.967 µmol m⁻² s⁻¹) and radial basis function neural network (RBF) (R² = 0.54, RMSE = 0.884 µmol m⁻² s⁻¹) and FO₂ with MLP (R² = 0.45, RMSE = 0.093 mg m⁻² s⁻¹) and RBF (R² = 0.74, 0.079 mg m⁻² s⁻¹). Soil temperature and macroporosity are important predictors of FCO₂ and FO₂. The best combination of variables for training the ANFIS was selected based on trial and error. The results were as follows: FCO₂ (R² = 16) and FO₂ (R² = 29). In all models, FCO₂ outperformed FO₂. A primary factor analysis was performed, and FCO₂ and FO₂ correlated best with the weather and soil attributes, respectively. [ABSTRACT FROM AUTHOR]