Search

Soil organic carbon (SOC) and total nitrogen (TN) influence physical, chemical and biological properties and process in soils that determine soil fertility and enhance or maintain agricultural productivity and food security. Knowledge on the spatial distribution of SOC and TN is necessary for sustainable soil resource management. In this study, we applied boosted regression trees (BRT) model to predict and map the spatial distribution of SOC and TN in the northeastern coastal areas of China, and C:N ratio map was generated from the predicted maps of SOC and TN. A total of 149 topsoil (0–20 cm) samples and 12 environmental variables that included topography, landuse and remote sensing indices, and climatic variables were selected. The performance of the models was evaluated based on cross-validation using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R 2 ). The BRT model was run for 100 iterations where the average of 100 SOC and TN maps was considered as final predicted maps and its standard deviation as prediction uncertainty. The BRT model showed a good predictive performance due to its higher R 2 and lower MAE, and RMSE indices. The model explained approximately 64% and 56% of the total SOC and TN variability. Topographic variables showed a maximum influence in the prediction of SOC and TN, followed by vegetation and climate. The SOC and TN contents were reduced from the northwest towards the southeast of the study area; average predicted SOC and TN contents were 15.4 g kg −1 and 1.01 g kg −1 , respectively. The spatial distribution of C:N ratio was closely related to landuse types in a decreasing order: woodland > orchard > cultivated land > grassland. We found topographic variables as main environmental indicators of SOC and TN distribution, and recommend to include them in future SOC and TN mapping studies in the coastal agroecosystems in China. [ABSTRACT FROM AUTHOR]

In digital soil mapping (DSM), soil maps are usually produced in a univariate manner, that is, each soil map is produced independently and therefore, when multiple soil properties are mapped the underlying dependence structure between these soil properties is ignored. This may lead to inconsistent predictions and simulations. For example, soil organic carbon (SOC) and total nitrogen (TN) maps produced independently may show unrealistic carbon–nitrogen (C:N) ratios. In the last decade the production of soil maps with machine learning models has become increasingly popular as these models are able to capture complex non-linear relationships between soil properties and environmental covariates. However, producing soil maps with multivariate machine learning models is still lacking and requires much investigation in DSM. In this paper we present the combined modelling of multiple soil properties with a multivariate random forest (MRF) model. We applied this model to mapping SOC and TN, and we compared it with results of two separate univariate random forest (RF) models. The comparison was done by means of stochastic simulations determined by sampling from the conditional distributions of the soil properties, given the covariates, as estimated by quantile regression forest. The results show that the MRF model is superior in terms of maintaining the dependence structure between SOC and TN, and consequently, is also able to produce more realistic C:N ratios. The models were also compared on the basis of prediction accuracy using commonly used accuracy metrics such as the root mean square error (RMSE). We found that the accuracy of the MRF model (RMSE-SOC = 40. 04 , RMSE-TN = 2. 26 , RMSE-CN = 3. 58) is comparable to that of the univariate RF models (RMSE-SOC = 39. 76 , RMSE-TN = 2. 26 , RMSE-CN = 3. 65). We performed the same comparisons between a regression co-kriging model and two separate regression kriging models, and made similar conclusions. • Multivariate RF accounts for the correlation structure between soil properties. • Simulations produced with multivariate RF are superior compared to those of RF. • In digital soil assessment it is vital to account for correlations in soil properties. • Multivariate RF C:N ratio simulations outperformed those of regression co-kriging. • Prediction accuracy of multivariate RF is comparable with that of RF. [ABSTRACT FROM AUTHOR]