Spatially resolved estimates of change in soil organic carbon (SOC) stocks are necessary for supporting national and international policies aimed at achieving land degradation neutrality and climate change mitigation. In this work we report on the development, implementation and application of a data‐driven, statistical method for mapping SOC stocks in space and time, using Argentina as a pilot. We used quantile regression forest machine learning to predict annual SOC stock at 0–30 cm depth at 250 m resolution for Argentina between 1982 and 2017. The model was calibrated using over 5,000 SOC stock values from the 36‐year time period and 35 environmental covariates. We preprocessed normalized difference vegetation index (NDVI) dynamic covariates using a temporal low‐pass filter to allow the SOC stock for a given year to depend on the NDVI of the current as well as preceding years. Predictions had modest temporal variation, with an average decrease for the entire country from 2.55 to 2.48 kg C m ⁻² over the 36‐year period (equivalent to a decline of 211 Gg C, 3.0% of the total 0–30 cm SOC stock in Argentina). The Pampa region had a larger estimated SOC stock decrease from 4.62 to 4.34 kg C m ⁻² (5.9%) during the same period. For the 2001–2015 period, predicted temporal variation was seven‐fold larger than that obtained using the Tier 1 approach of the Intergovernmental Panel on Climate Change and United Nations Convention to Combat Desertification. Prediction uncertainties turned out to be substantial, mainly due to the limited number and poor spatial and temporal distribution of the calibration data, and the limited explanatory power of the covariates. Cross‐validation confirmed that SOC stock prediction accuracy was limited, with a mean error of 0.03 kg C m ⁻² and a root mean squared error of 2.04 kg C m ⁻² . In spite of the large uncertainties, this work showed that machine learning methods can be used for space–time SOC mapping and may yield valuable information to land managers and policymakers, provided that SOC observation density in space and time is sufficiently large. Highlights: We tested the use of machine learning for space–time mapping of soil organic carbon (SOC) stock. Predictions for Argentina from 1982 to 2017 showed a 3% decrease of the topsoil SOC stock over time. The machine learning model predicted a greater temporal variation than the IPCC Tier 1 approach. Accurate machine learning SOC stock prediction requires dense soil sampling in space and time.

Country-specific soil organic carbon (SOC) estimates are the baseline for the Global SOC Map of the Global Soil Partnership (GSOCmap-GSP). This endeavor is key to explaining the uncertainty of global SOC estimates but requires harmonizing heterogeneous datasets and building country-specific capacities for digital soil mapping (DSM). We identified country-specific predictors for SOC and tested the performance of five predictive algorithms for mapping SOC across Latin America. The algorithms included support vector machines (SVMs), random forest (RF), kernel-weighted nearest neighbors (KK), partial least squares regression (PL), and regression kriging based on stepwise multiple linear models (RK). Country-specific training data and SOC predictors (5 × 5 km pixel resolution) were obtained from ISRIC – World Soil Information. Temperature, soil type, vegetation indices, and topographic constraints were the best predictors for SOC, but country-specific predictors and their respective weights varied across Latin America. We compared a large diversity of country-specific datasets and models, and were able to explain SOC variability in a range between ∼ 1 and ∼ 60 %, with no universal predictive algorithm among countries. A regional (n = 11 268 SOC estimates) ensemble of these five algorithms was able to explain ∼ 39 % of SOC variability from repeated 5-fold cross-validation. We report a combined SOC stock of 77.8 ± 43.6 Pg (uncertainty represented by the full conditional response of independent model residuals) across Latin America. SOC stocks were higher in tropical forests (30 ± 16.5 Pg) and croplands (13 ± 8.1 Pg). Country-specific and regional ensembles revealed spatial discrepancies across geopolitical borders, higher elevations, and coastal plains, but provided similar regional stocks (77.8 ± 42.2 and 76.8 ± 45.1 Pg, respectively). These results are conservative compared to global estimates (e.g., SoilGrids250m 185.8 Pg, the Harmonized World Soil Database 138.4 Pg, or the GSOCmap-GSP 99.7 Pg). Countries with large area (i.e., Brazil, Bolivia, Mexico, Peru) and large spatial SOC heterogeneity had lower SOC stocks per unit area and larger uncertainty in their predictions. We highlight that expert opinion is needed to set boundary prediction limits to avoid unrealistically high modeling estimates. For maximizing explained variance while minimizing prediction bias, the selection of predictive algorithms for SOC mapping should consider density of available data and variability of country-specific environmental gradients. This study highlights the large degree of spatial uncertainty in SOC estimates across Latin America. We provide a framework for improving country-specific mapping efforts and reducing current discrepancy of global, regional, and country-specific SOC estimates.

The use of Unmanned Aerial Vehicles (UAVs) in viticulture permits the capture of aerial Red-Green-Blue (RGB) images with an ultra-high spatial resolution. Recent studies have demonstrated that RGB images can be used to monitor spatial variability of vine biophysical parameters. However, for estimating these parameters, accurate and automated segmentation methods are required to extract relevant information from RGB images. Manual segmentation of aerial images is a laborious and time-consuming process. Traditional classification methods have shown satisfactory results in the segmentation of RGB images for diverse applications and surfaces, however, in the case of commercial vineyards, it is necessary to consider some particularities inherent to canopy size in the vertical trellis systems (VSP) such as shadow effect and different soil conditions in inter-rows (mixed information of soil and weeds). Therefore, the objective of this study was to compare the performance of four classification methods (K-means, Artificial Neural Networks (ANN), Random Forest (RForest) and Spectral Indices (SI)) to detect canopy in a vineyard trained on VSP. Six flights were carried out from post-flowering to harvest in a commercial vineyard cv. Carménère using a low-cost UAV equipped with a conventional RGB camera. The results show that the ANN and the simple SI method complemented with the Otsu method for thresholding presented the best performance for the detection of the vine canopy with high overall accuracy values for all study days. Spectral indices presented the best performance in the detection of Plant class (Vine canopy) with an overall accuracy of around 0.99. However, considering the performance pixel by pixel, the Spectral indices are not able to discriminate between Soil and Shadow class. The best performance in the classification of three classes (Plant, Soil, and Shadow) of vineyard RGB images, was obtained when the SI values were used as input data in trained methods (ANN and RForest), reaching overall accuracy values around 0.98 with high sensitivity values for the three classes.

A critical aspect of predicting soil organic carbon (SOC) concentrations is the lack of available soil information; where information on soil characteristics is available, it is usually focused on regions of high agricultural interest. To date, in Chile, a large proportion of the SOC data have been collected in areas of intensive agricultural or forestry use; however, vast areas beyond these forms of land use have few or no soil data available. Here we present a new SOC database for the country, which is the result of an unprecedented national effort under the framework of the Global Soil Partnership. This partnership has helped build the largest database of SOC to date in Chile, named the Chilean Soil Organic Carbon database (CHLSOC), comprising 13 612 data points compiled from numerous sources, including unpublished and difficult-to-access data. The database will allow users to fill spatial gaps where no SOC estimates were publicly available previously. Presented values of SOC range from $\mathrm{6}\times {\mathrm{10}}^{-\mathrm{5}}$ % to 83.3 %, reflecting the variety of ecosystems that exist in Chile. The database has the potential to inform and test current models that predict SOC stocks and dynamics at larger spatial scales, thus enabling benefits from the richness of geochemical, topographic and climatic variability in Chile. The database is freely available to registered users at https://doi.org/10.17605/OSF.IO/NMYS3 (Pfeiffer et al., 2019b) under the Creative Commons Attribution 4.0 International Public License.