Reliable crop monitoring is paramount to achieve the objectives of the Common Agricultural Policy (CAP) and Food and Agriculture Organization. Synthetic Aperture Radar (SAR) provides high-resolution imaging and all-weather data acquisition capabilities for crop monitoring. This study investigates the sensitivity of parcel-level Sentinel-1 interferometric coherence to farming activities (e.g. planting, emergence, harvest and tillage) and weather events. A methodology to detect activities was developed and validated using ground-truth data from four crop types, collected over four years. The proposed approach was able to detect over 60% of all nine different farming activities. The results show that interferometric coherence is a reliable indicator for farming activities that can be considered as events resulting in a clear structural change (e.g. tillage 100%), but less reliable for gradual changes (e.g. Emergence 40%).

In this article, our aim is to estimate synthetic aperture radar (SAR) observables, such as backscatter in VV and VH polarizations, as well as the VH/VV ratio, cross ratio, and interferometric coherence in VV, from agricultural fields. In this study, we use the decision support system for agrotechnology transfer (DSSAT) crop-growth simulation model to simulate parcel-level phenological and growth parameters for over 1500 parcels of silage maize in the Netherlands. The crop model was calibrated using field data, including silage maize phenological phases, leaf area index, and above-ground dry biomass (AGB). The simulations incorporate fine-resolution gridded precipitation data and soil parameters to model the interaction between soil-plant-atmosphere and genotype in DSSAT. The crop variables produced by DSSAT are then used as inputs to a support vector regression model. This model is trained to simulate SAR observables in 2017, 2018, and 2019, and its performance is evaluated using independent fields in each of these years. The results show a close fit between modeled and observed SAR C-band observables. The importance of vegetation variables in the estimation of SAR observables is assessed. The AGB showed significant importance in the estimation of backscatter. This study demonstrates the potential value of combining crop-growth simulation models and machine learning to simulate SAR observables. For example, the SVR model developed here could be used as an observation operator in an assimilation context to constrain vegetation and soil water dynamics in a crop-growth model.

Biophysical parameters are descriptors of crop growth and production estimates. Retrieval of these biophysical parameters from synthetic aperture radar sensors at operational scales is highly interesting given the increase in access to data from radar missions. Vegetation backscattering can be simulated using the water cloud model (WCM). Crop biophysical parameters are obtained by inverting this model. However, the inversion problem is ill-posed, and existing methods, which include the lookup table (LUT) and iterative search algorithms, are often computationally intensive and lack good generalization capacity. This might make retrieval of the biophysical parameters computationally intensive for large study areas. In addition, the new generation of operational missions, which are often associated with a large volume of data, poses a challenge for estimating crop parameters. In this work, we use the cloud computing potentials of the Google Earth Engine (GEE) to demonstrate a unified processing pipeline for WCM inversion. The processing pipeline (GEE4Bio) uses Sentinel-1 radar measurements for WCM inversion and subsequently produces crop biophysical maps. Inversion is achieved by employing Random Forest regression, which is trained with radar backscatter measurements at Vertical transmit and vertical receive (VV) and Vertical transmit and horizontal receive (VH) channels. The model is trained and validated with independent calibration and validation datasets consisting of ground measurements for five major crops over the Joint Experiment for Crop Assessment and Monitoring–Carman test site in Canada. The inversion accuracies indicate strong correlation coefficients (r) of 0.83 and 0.87, with the estimated and in situ measured plant area index and wet biomass, respectively, with low root mean square error values. The GEE4Bio processing chain produced crop inventory maps with a reasonable time and apprehended the variability in plant growth across the test site.

For a good interpretation of radar backscatter sensitivity to vegetation water dynamics, we need to know which parts of the vegetation layer control that backscatter. However, backscatter sensitivity to different depths in the canopy is poorly understood. This is partly caused by a lack of observational data to describe the vertical moisture distribution. In this study, we aimed to understand the sensitivity of L-band backscatter to water at different heights in a corn canopy. We studied changes in the contribution of different vertical layers to total backscatter throughout the season and during the day. Using detailed field measurements, we first determined the vertical distribution of moisture in the plants, and its seasonal and sub-daily variation. Then, these measurements were used to define different sublayers in a multi-layer water cloud model (WCM). To calibrate and validate the WCM, we used hyper-temporal tower-based polarimetric L-band scatterometer data. WCM simulations showed a shift in dominant scattering from the lowest 50 cm to 50–100 cm during the season in all polarizations, mainly due to leaf and ear growth and corresponding scattering and attenuation. Dew and rainfall interception raised sensitivity to upper parts of the canopy and lowered sensitivity to lower parts. The methodology and results presented in this study demonstrate the importance of the vertical moisture distribution on scattering from vegetation. These insights are essential to avoid misinterpretation and spurious artefacts during retrieval of soil moisture and vegetation parameters.

This chapter investigates multi-frequency (C-, L-, and P-bands) single-date AIRSAR data using Random Forest (RF) based polarimetric parameter selection for crop separation and classification. The RF classifier has an inherent parameter ranking and partial probability plot ability which gives not only the important parameters but also their optimal dynamic range. Crop separation was assessed among crop types by identifying polarimetric parameters having highest difference of Mean Decrease Accuracy (MDA) scores as measured by RF. Earlier studies primarily focused on polarimetric backscattering coefficients for crop analysis. In this study in addition to these parameters, the scattering decomposition powers along with the backscattering ratio parameters were also analyzed and found vital for multi-frequency crop classification. The Yamaguchi model-based decomposition, the Cloude-Pottier and the Touzi decomposition parameters provided complimentary information which were further used for critical analysis of crops in this study. In this study, the classification accuracy using RF was obtained as: C-band (71.9%); L-band (80.7%); P-band (75.8%). The long-stem crops: barley and rapeseed had the best accuracy in L-band (91.7%) and C-band (91.4%), respectively, while for the short-stem broad-leaf crops: sugarbeet (86.2%) in L-band and potatoes (95.4%) in L-band and (94.5%) in P-band, respectively.

Synthetic Aperture Radar (SAR) data handling, processing, and interpretation are barriers preventing a rapid uptake of SAR data by application specialists and non-expert domain users in the field of agricultural monitoring. To improve the accessibility of Sentinel-1 data, we have generated a reduced-volume, multi-year Sentinel-1 SAR database. It includes mean and standard deviation of VV, VH and VH/VV backscatter, pixel counts, geometry, crop type, local incidence angle and azimuth angle at parcel-level. The database uses around 3100 Sentinel-1 images (5 TB) to produce a 12 GB time series database for approximately 770,000 crop parcels over the Netherlands for a period of three years. The database can be queried by Sentinel-1 system parameters (e.g. relative orbit) or user application-specific parameters (e.g. crop type, spatial extent, time period) for parcel level assessment. The database can be used to accelerate the development of new tools, applications and methodologies for agricultural and water related applications, such as parcel-level crop bio-geophysical parameter estimation, inter-annual variability analysis, drought monitoring, grassland monitoring and agricultural management decision-support.

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 m²m^-2 and mean absolute error (MAE) of 0.51 m²m^-2. The WCM produced an R-value of 0.89, with only slightly higher errors (RMSE of 0.75 m²m^-2 and MAE of 0.61 m²m^-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 m²m^-2 and MAE of 0.30 m²m^-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.

Using the cross-validation approach, strategies for estimating biophysical parameters are still pre-operational with synthetic aperture radar (SAR) data. In this regard, the Joint Experiment for Crop Assessment and Monitoring (JECAM) SAR inter-comparison experiments provide an opportunity for the potential implementation of cross-validation strategies for biophysical parameters retrieval utilizing the next-generation compact polarimetric (CP) modes available from the RADARSAT Constellation Mission (RCM). This work first uses the conventional semi-empirical Water Cloud Model (WCM) modified by exploiting the scattering power decompositions of CP measurements to estimate the Plant Area Index (PAI) for rice. The modified WCM (MWCM) is then inverted using the scattering power components from the (Formula presented.) decomposition. We compare the PAI estimates using MWCM- (Formula presented.) between the estimates obtained from (1) the conventional WCM using the RH and RV backscatter intensities and (2) MWCM- (Formula presented.) decomposition scattering powers. We exploit a time series of simulated compact-pol SAR data over the JECAM test site in Vijayawada, India, throughout 2018 and 2019. We use the C-band RADARSAT-2 full-pol data to simulate the RADARSAT Constellation Mission (RCM) compact-pol mode data. Utilizing the advantage of systematically collected multi-year SAR data and in-situ measurements, the present research also assesses the calibrated model transferability performances to another data set and cross-validation of a model in a multi-year experiment setting. The comparative analysis indicates potential improvements in PAI estimation with MWCM- (Formula presented.) scattering powers. A high range of correlation coefficient ((Formula presented.)) between the estimated and observed PAI is observed with good Root Mean Square Error (RMSE) of (Formula presented.) m² m⁻², and Mean Absolute Error (MAE) of (Formula presented.) m² m⁻².

Sentinel-1 Synthetic Aperture Radar (SAR) data have provided an unprecedented opportunity for crop monitoring due to its high revisit frequency and wide spatial coverage. The dual-pol (VV-VH) Sentinel-1 SAR data are being utilized for the European Common Agricultural Policy (CAP) as well as for other national projects, which are providing Sentinel derived information to support crop monitoring networks. Among the Earth observation products identified for agriculture monitoring, indicators of vegetation status are deemed critical by end-user communities. In literature, several experiments usually utilize the backscatter intensities to characterize crops. In this study, we have jointly utilized the scattering information in terms of the degree of polarization and the eigenvalue spectrum to derive a new vegetation index from dual-pol (DpRVI) SAR data. We assess the utility of this index as an indicator of plant growth dynamics for canola, soybean, and wheat, over a test site in Canada. A temporal analysis of DpRVI with crop biophysical variables (viz., Plant Area Index (PAI), Vegetation Water Content (VWC), and dry biomass (DB)) at different phenological stages confirms its trend with plant growth dynamics. For each crop type, the DpRVI is compared with the cross and co-pol ratio (σ_VH⁰/σ_VV⁰) and dual-pol Radar Vegetation Index (RVI = 4σ_VH⁰/(σ_VV⁰ + σ_VH⁰)), Polarimetric Radar Vegetation Index (PRVI), and the Dual Polarization SAR Vegetation Index (DPSVI). Statistical analysis with biophysical variables shows that the DpRVI outperformed the other four vegetation indices, yielding significant correlations for all three crops. Correlations between DpRVI and biophysical variables are highest for canola, with coefficients of determination (R²) of 0.79 (PAI), 0.82 (VWC), and 0.75 (DB). DpRVI had a moderate correlation (R²≳ 0.6) with the biophysical parameters of wheat and soybean. Good retrieval accuracies of crop biophysical parameters are also observed for all three crops.

Crop discrimination with synthetic aperture radar (SAR) data primarily depends on the characterization of crop geometry using radar backscatter response. Differences in phenological development of crops lead to dissimilar temporal signatures of backscatter intensities, which may influence the separability of the crop classes. This principle leads to multi-date classification approach. In this work, kernel principal component (KPCA) is adopted for feature selection from multi-date datasets, and the selected features are used for classification using support vector machine (SVM) classifier. The classification is investigated for both the KPCA-based SVM and only SVM approaches using quad-pol C-band RADARSAT-2 data acquired over the test site in Vijayawada, India. KPCA-based SVM classification shows an overall accuracy of 89%, which is better than 82% obtained using the SVM-based classification. The proposed methodology effectively incorporates the temporal crop information during classification.

Estimation of bio-and geophysical parameters from Earth observation (EO) data is essential for developing applications on crop growth monitoring. High spatio-temporal resolution and wide spatial coverage provided by EO satellite data are key inputs for operational crop monitoring. In Synthetic Aperture Radar (SAR) applications, a semi-empirical model (viz., Water Cloud Model (WCM)) is often used to estimate vegetation descriptors individually. However, a simultaneous estimation of these vegetation descriptors would be logical given their inherent correlation, which is seldom preserved in the estimation of individual descriptors by separate inversion models. This functional relationship between biophysical parameters is essential for crop yield models, given that their variations often follow different distribution throughout crop development stages. However, estimating individual parameters with independent inversion models presume a simple relationship (potentially linear) between the biophysical parameters. Alternatively, a multi-target inversion approach would be more effective for this aspect of model inversion compared to an individual estimation approach. In the present research, the multi-output support vector regression (MSVR) technique is used for inversion of the WCM from C-band dual-pol Sentinel-1 SAR data. Plant Area Index (PAI, m² m⁻²) and wet biomass (W, kg m⁻²) are used as the vegetation descriptors in the WCM. The performance of the inversion approach is evaluated with in-situ measurements collected over the test site in Manitoba (Canada), which is a super-site in the Joint Experiment for Crop Assessment and Monitoring (JECAM) SAR inter-comparison experiment network. The validation results indicate a good correlation with acceptable error estimates (normalized root mean square error–nRMSE and mean absolute error–MAE) for both PAI and wet biomass for the MSVR approach and a better estimation with MSVR than single-target models (support vector regression–SVR). Furthermore, the correlation between PAI and wet biomass is assessed using the MSVR and SVR model. Contrary to the single output SVR, the correlation between biophysical parameters is adequately taken into account in MSVR based simultaneous inversion technique. Finally, the spatio-temporal maps for PAI and W at different growth stages indicate their variability with crop development over the test site.