BRAZILIAN SOIL δ13C ISOSCAPE BASED ON MULTIPLE LINEAR REGRESSION ANALYSIS
Description
Carbon stable isotope ratios (δ13C) have been applied in several environmental contexts, such as feeding habits, origin of migratory species, and vegetation distribution patterns. Brazil presents a vast environmental diversity, which enables the development of studies to identify the soil δ13C distribution patterns. This work aimed to identify the environmental variables that influence the soil δ13C and develop a spatial model of Brazilian soil δ13C, based on multiple linear regression analysis. The model used 717 samples at a depth of 0-20 centimeters and a set of climate, soil, and vegetation variables. The model showed a range of soil δ13C values between -30‰ and -13‰, with the highest estimated values in the southeastern regions and the highest estimated values in northwestern Brazil. The results pointed out regional patterns in the spatial distribution of the soil δ13C and the relationships between the environmental variables incorporated in the model and the soil δ13C. Due to the high environmental diversity of Brazil and the local environmental characteristics, some of these variables presented opposite behavior to that reported in previous studies, which makes necessary the development of studies aimed at the better understanding of these relations. Nevertheless, the soil δ13C isoscape model presented a general panorama of the distribution of Brazilian soil δ13C with a more refined level of detail concerning the existing models for the region.
Files
Steps to reproduce
Data analysis: The environmental covariates in each group were tested against each other for multicollinearity using correlation matrices and Pearson's correlation coefficient. Pairs of variables with a high correlation coefficient (r > 0.70) were removed from the selection. Then, the procedure was repeated for all selected variables in each group to ensure that all variables entered into the multiple linear regression model were independent. Simple linear regression analyses were performed to describe the individual influence of some environmental variables on the soil δ13C. The dataset was randomly divided into training (75%, n = 537) and validation (25%, n = 180). The training dataset was used to fit the multiple regression model and generate the soil δ13C isoscape. The validation dataset was used in the isoscape validation step. The environmental covariates of the training dataset were standardized for z-score using the scale () function so that it was possible to evaluate the weight of each variable in the regression model. The multiple linear regression model was adjusted from a stepwise procedure. First, a regression model was fitted with all independent variables. Then, an F-test was performed to evaluate the significance of the model. The final model was fitted with the significant variables from the first model and chosen based on the Akaike Information Criterion (AIC) (Akaike, 1974). The Shapiro Wilk test and the Moran Index assessed the normality and spatial autocorrelation of the residuals of the adjusted model. Isoscape fit and model validation: The estimated coefficient from the multiple linear regression was used to generate the soil δ13C isoscapes. The raster data of the selected environmental covariates were resampled to a spatial resolution of 1 km and standardized for z-score using the scale () function before applying the regression equation. For model validation, a simple linear regression was performed between the predicted and measured values of the validation dataset. All analyses were performed in R version 4.0.5 ( R Core Team, 2018).