For agronomic, environmental, and economic reasons, the need for spatialized information about agricultural practices is expected to rapidly increase. In this context, we reviewed the literature on remote sensing for mapping cropping practices. The reviewed studies were grouped into three categories of practices: crop succession (crop rotation and fallowing), cropping pattern (single tree crop planting pattern, sequential cropping, and intercropping/agroforestry), and cropping techniques (irrigation, soil tillage, harvest and post-harvest practices, crop varieties, and agro-ecological infrastructures). We observed that the majority of the studies were exploratory investigations, tested on a local scale with a high dependence on ground data, and used only one type of remote sensing sensor. Furthermore, to be correctly implemented, most of the methods relied heavily on local knowledge on the management practices, the environment, and the biological material. These limitations point to future research directions, such as the use of land stratification, multi-sensor data combination, and expert knowledge-driven methods. Finally, the new spatial technologies, and particularly the Sentinel constellation, are expected to improve the monitoring of cropping practices in the challenging context of food security and better management of agro-environmental issues.

Developing better agricultural monitoring capabilities based on Earth Observation data is critical for strengthening food production information and market transparency. The Sentinel-2 mission has the optimal capacity for regional to global agriculture monitoring in terms of resolution (10–20 meter), revisit frequency (five days) and coverage (global). In this context, the European Space Agency launched in 2014 the “Sentinel­2 for Agriculture” project, which aims to prepare the exploitation of Sentinel-2 data for agriculture monitoring through the development of open source processing chains for relevant products. The project generated an unprecedented data set, made of “Sentinel-2 like” time series and in situ data acquired in 2013 over 12 globally distributed sites. Earth Observation time series were mostly built on the SPOT4 (Take 5) data set, which was specifically designed to simulate Sentinel-2. They also included Landsat 8 and RapidEye imagery as complementary data sources. Images were pre-processed to Level 2A and the quality of the resulting time series was assessed. In situ data about cropland, crop type and biophysical variables were shared by site managers, most of them belonging to the “Joint Experiment for Crop Assessment and Monitoring” network. This data set allowed testing and comparing across sites the methodologies that will be at the core of the future “Sentinel­2 for Agriculture” system.

The water cloud model (WCM) can be inverted to estimate leaf area index (LAI) using the intensity of backscatter from synthetic aperture radar (SAR) sensors. Published studies have demonstrated that the WCM can accurately estimate LAI if the model is effectively calibrated. However, calibration of this model requires access to field measures of LAI as well as soil moisture. In contrast, machine learning (ML) algorithms can be trained to estimate LAI from satellite data, even if field moisture measures are not available. In this study, a support vector machine (SVM) was trained to estimate the LAI for corn, soybeans, rice, and wheat crops. These results were compared to LAI estimates from the WCM. To complete this comparison, in situ and satellite data were collected from seven Joint Experiment for Crop Assessment and Monitoring (JECAM) sites located in Argentina, Canada, Germany, India, Poland, Ukraine and the United States of America (U.S.A.). The models used C-Band backscatter intensity for two polarizations (like-polarization (VV) and cross-polarization (VH)) acquired by the RADARSAT-2 and Sentinel-1 SAR satellites. Both the WCM and SVM models performed well in estimating the LAI of corn. For the SVM, the correlation (R) between estimated LAI for corn and LAI measured in situ was reported as 0.93, with a root mean square error (RMSE) of 0.64 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$ and mean absolute error (MAE) of 0.51 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$. The WCM produced an R-value of 0.89, with only slightly higher errors (RMSE of 0.75 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$ and MAE of 0.61 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$) when estimating corn LAI. For rice, only the SVM model was tested, given the lack of soil moisture measures for this crop. In this case, both high correlations and low errors were observed in estimating the LAI of rice using SVM (R of 0.96, RMSE of 0.41 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$ and MAE of 0.30 ${\mathrm{m}}^2{\mathrm{m}}^{-2}$). However, the results demonstrated that when the calibration points were limited (in this case for soybeans), the WCM outperformed the SVM model. This study demonstrates the importance of testing different modeling approaches over diverse agro-ecosystems to increase confidence in model performance.

The impact of land cover change across the planet continues to necessitate accurate methods to detect and monitor evolving processes from satellite imagery. In this context, regional and global land cover mapping over time has largely treated time as independent and addressed temporal map consistency as a post-classification endeavor. However, we argue that time can be better modeled as codependent during the model classification stage to produce more consistent land cover estimates over long time periods and gradual change events. To produce temporally-dependent land cover estimates—meaning land cover is predicted over time in connected sequences as opposed to predictions made for a given time period without consideration of past land cover—we use structured learning with conditional random fields (CRFs), coupled with a land cover augmentation method to produce time series training data and bi-weekly Landsat imagery over 20 years (1999–2018) across the Southern Cone region of South America. A CRF accounts for the natural dependencies of land change processes. As a result, it is able to produce land cover estimates over time that better reflect real change and stability by reducing pixel-level annual noise. Using CRF, we produced a twenty-year dataset of land cover over the region, depicting key change processes such as cropland expansion and tree cover loss at the Landsat scale. The augmentation and CRF approach introduced here provides a more temporally consistent land cover product over traditional mapping methods.

The availability of spatially explicit information about agricultural crops for large regions in Argentina is scarce. In particular, due to temporal dynamics of agricultural production (i.e. changes in planted crops from year to year) and spectral similarities among herbaceous crops it is difficult to generate crop type maps from remote sensing. Large regions with marked climatic variations, like the main agricultural areas of Argentina, represent an additional challenge. Here we generated a map based on supervised classifications using field samples along 14 agricultural zones. Best classification accuracies were obtained by combining seasonal indices (year, summer and winter), with indices that describe the temporal dynamics of vegetation. Accuracy was increased at regions with high and balanced number of samples and with longer growing seasons. The map allows to identify areas with clusters of one, two or three crops and to characterize areas with different spatial distribution between cropland and no cropland areas.

Land transformation is one of the most significant human changes on the Earth’s surface processes. Therefore, land use land cover time series are a key input for environmental monitoring, natural resources management, territorial planning enforcement at national scale. We here capitalize from the MapBiomas initiative to characterize land use land cover (LULC) change in the Gran Chaco between 2010 and 2017. Specifically we sought to a) quantify annual changes in the main LULC classes; b) identify the main LULC transitions and c) relate these transitions to current land use policies. Within the MapBiomas project, Landsat based annual maps depicting natural woody vegetation, natural herbaceous vegetation, dispersed natural vegetation, cropland, pastures, bare areas and water. We used Random Forest machine learning algorithms trained by samples produced by visual interpretation of high resolution images. Annual overall accuracy ranged from 0,73 to 0,74. Our results showed that, between 2010 and 2017, agriculture and pasture lands increased ca. 3.7 Mha while natural forestry decreased by 2.3 Mha. Transitions from forests to agriculture accounted for 1.14% of the overall deforestation while 86% was associated to pastures and natural herbaceous vegetation. In Argentina, forest loss occurred primarily (39%) on areas non considered by the territorial planning Law, followed by medium (33%), high (19%) and low (9%) conservation priority classes. These results illustrate the potential contribution of remote sensing to characterize complex human environmental interactions occurring over extended areas and timeframes.