Synthetic aperture radar (SAR) acquisitions are mainly deemed suitable for mapping dynamic land-cover and land-use scenarios due to their timeliness and reliability. This particularly applies to Sentinel-1 imagery. Nevertheless, the accurate mapping of regions characterized by a mixture of crops and grasses can still represent a challenge. Radar time-series have to date mainly been exploited through backscatter intensities, whereas only fewer contributions have focused on analyzing the potential of interferometric information, intuitively enhanced by the short revisit. In this paper, we evaluate, as primary objective, the added value of short-temporal baseline coherences over a complex agricultural area in the São Paulo state, cultivated with heterogeneously (asynchronously) managed annual crops, grasses for pasture and sugarcane plantations. We also investigated the sensitivity of the radar information to the classification methods as well as to the data preparation and sampling practices. Two supervised machine learning methods—namely support vector machine (SVM) and random forest (RF)—were applied to the Sentinel-1 time-series at the pixel and field levels. The results highlight that an improvement of 10 percentage points (p.p.) in the classification accuracy can be achieved by using the coherence in addition to the backscatter intensity and by combining co-polarized (VV) and cross-polarized (VH) information. It is shown that the largest contribution in class discrimination is brought during winter, when dry vegetation and bare soils can be expected. One of the added values of coherence was indeed identified in the enhanced sensitivity to harvest events in a small but significant number of cases.

@en

In this study, research on productivity and land cover monitoring is presented, with a focus on sugarcane, based on space-based remote sensing observations that were collected by Synthetic Aperture Radar (SAR) and optical sensors. The study aims to provide new insights into techniques and methodologies that allow for cost-efficient monitoring of sugarcane productivity and the wide-scale expansion of sugarcane over long time series. It is part of a wider framework of research projects, sharing the overarching goal to contribute to a biobased society, where the resources for the production of chemicals, materials and energy are based on biomass, produced in a competitive and sustainable manner. São Paulo state, Brazil, was selected as study area, primarily because it is one of the most prominent regions worldwide that hosts sugarcane, which is one of the most prominent crops for bio-energy production. The study includes the following research. First of all, results are presented that are based on a wide range of biophysical measurements that were collected during a year-long ground measurement campaign in several sugarcane fields. These results are accompanied by detailed quality assessments, illustrating their reliability when collecting such measurements through such campaigns. In addition, the methodology for setting up and carrying out the ground campaign is explained, which was designed to minimize biomass alterations in the field in light of the use of the measurements for validation of space-based SAR and optical remote sensing signals. Secondly, remote sensing signals from various satellites are compared to the ground reference measurements in order to develop space-based sugarcane productivity monitoring techniques. It includes an analysis on the sensitivity of C-band and L-band SAR and optical observations to sugarcane biomass growth, to precipitation events and to SAR sensor configurations. In addition, the spatial features in satellite imagery from the various sensors are analyzed for their temporal consistencies in order to deduce time windows during which the satellite observations are most effective for productivity monitoring. It was found that especially saturation, precipitation and sensor configurations dictate the effectiveness, particularly for SAR. Furthermore, the highest spatial resolution optical imagery proved to perform best for mapping intra-field productivity differences that were measured in the field. In addition to this study, two related but smaller studies are presented. The first focuses on a specific remote sensing technique to identify patterns in a sugarcane field that occur persistently in time. The second demonstrates how plant gaps in a densely ground-measured sugarcane field affect signals from various SAR and how this effect is influenced by spatial averaging windows, precipitation events, sugarcane height and SAR sensor type. Thirdly, the performance of a specific Bayesian land cover monitoring model that combines SAR and optical observations is demonstrated. The model is an adaptation of the Hidden Markov Model, which allows for the temporally-consistent tracking of vegetation states regardless of gaps in satellite observations. Attention is paid to the effect of precipitation during SAR observations on the model's performance and to certain vegetation conditions that cause classification confusion between land cover types. The research finally provides detailed insights into when SAR-only observations outperform optical-only observations and vice versa, in addition to the advantages when combining them. Finally, a technique is introduced that exploits SAR signal fluctuations caused by varying (ground and plant) surface wetness conditions in order to improve the characterization of vegetation. Three scenarios that define the selection of SAR observations were investigated for their effect on the classification performance: (i) no distinction between wetness conditions, (ii) distinction between wetness conditions at the time of the SAR acquisitions and (iii) distinction between wetness conditions between consecutive SAR acquisitions. Particularly when the wetness conditions differ under the last scenario, it was found that performances improve. When combining this information with a-priori knowledge on soil types, the accuracy of the classification further increases. For this, maps are used that are a result from applying the previously introduced Hidden Markov Model over the entire state of São Paulo. The datasets that are used in these studies were mainly acquired by the SAR satellites Sentinel-1, Radarsat-2 and ALOS-2, and by the optical satellites Landsat-8 and Worldview-2. For the studies that are related to land cover monitoring and vegetation characterization, high performance computing was required due to the vast amount of observation data and the complexity of the applied techniques. These facilities were mainly provided by the Dutch national supercomputer of SURF and by Google Earth Engine.@en

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

@en

Agriculture is of huge economic significance in The Netherlands where the provision of real-time, reliable information on crop development is essential to support the transition towards precision agriculture. Optical imagery can provide invaluable insights into crop growth and development but is severely hampered by cloud cover. This case study in the Flevopolder illustrates the potential value of Sentinel-1 for monitoring five key crops in The Netherlands, namely sugar beet, potato, maize, wheat and English rye grass. Time series of radar backscatter from the European Space Agency's Sentinel-1 Mission are analyzed and compared to ground measurements including phenological stage and height. Temporal variations in backscatter data reflect changes in water content and structure associated with phenological development. Emergence and closure dates are estimated from the backscatter time series and validated against a photo archive. Coherence data are compared to Normalized Difference Vegetation Index (NDVI) and ground data, illustrating that the sudden increase in coherence is a useful indicator of harvest. The results presented here demonstrate that Sentinel-1 data have significant potential value to monitor growth and development of key Dutch crops. Furthermore, the guaranteed availability of Sentinel-1 imagery in clouded conditions ensures the reliability of data to meet the monitoring needs of farmers, food producers and regulatory bodies.@en

Following publication, it was noticed that the horizontal brackets labelling the two groups of precisions present in Equation 7 are incorrectly rendered in the PDF version of this Data Descriptor. The correct Equation 7 is as follows: (Formula presented.) (Formula presented.) In addition, in the Biomass subsection of the Methods section in both the HTML and PDF versions, the term “ESUs” is incorrectly rendered as “ESU’s” and the term ESUBs is incorrectly rendered as “ESUB’s” Finally, throughout the manuscript, references to sections and subsections include the prefixes “sec:” and “subsec:”, respectively. These prefixes and any hyphen between the reference words that follow the prefixes can be ignored.

@en

Space-based remote sensing imagery can provide a valuable and cost-effective set of observations for mapping crop-productivity differences. The effectiveness of such signals is dependent on several conditions that are related to crop and sensor characteristics. In this paper, we present the dynamic behavior of signals from five Synthetic Aperture Radar (SAR) sensors and optical sensors with growing sugarcane, focusing on saturation effects and the influence of precipitation events. In addition, we analyzed the level of agreement within and between these spaceborne datasets over space and time. As a result, we produced a list of conditions during which the acquisition of satellite imagery is most effective for sugarcane productivity monitoring. For this, we analyzed remote sensing data from two C-band SAR (Sentinel-1 and Radarsat-2), one L-band SAR (ALOS-2), and two optical sensors (Landsat-8 and WorldView-2), in conjunction with detailed ground-reference data acquired over several sugarcane fields in the state of São Paulo, Brazil. We conclude that satellite imagery from L-band SAR and optical sensors is preferred for monitoring sugarcane biomass growth in time and space. Additionally, C-band SAR imagery offers the potential for mapping spatial variations during specific time windows and may be further exploited for its precipitation sensitivity.@en

In order to make effective decisions on sustainable development, it is essential for sugarcane-producing countries to take into account sugarcane acreage and sugarcane production dynamics. The availability of sugarcane biophysical data along the growth season is key to an effective mapping of such dynamics, especially to tune agronomic models and to cross-validate indirect satellite measurements. Here, we introduce a dataset comprising 3,500 sugarcane observations collected from October 2014 until October 2015 at four fields in the São Paulo state (Brazil). The campaign included both non-destructive measurements of plant biometrics and destructive biomass weighing procedures. The acquisition plan was designed to maximize cost-effectiveness and minimize field-invasiveness, hence the non-destructive measurements outnumber the destructive ones. To compensate for such imbalance, a method to convert the measured biometrics into biomass estimates, based on the empirical adjustment of allometric models, is proposed. In addition, the paper addresses the precisions associated to the ground measurements and derived metrics. The presented growth dynamics and associated precisions can be adopted when designing new sugarcane measurement campaigns.@en

SAR as an active remote sensing technique is capable of providing insights into the physical features of agricultural vegetation. However, the noisy nature of SAR signals makes the direct conversion to effective productivity metrics challenging. This study sheds light on the effect of gaps present in a sugarcane field on L-band and C-band SAR signals and demonstrates the variability of this effect with changing spatial averaging windows, changing precipitation conditions and changing vegetation height.@en

Current space-based SAR offers unique opportunities to classify vegetation types and to monitor vegetation growth due to its frequent acquisitions and its sensitivity to vegetation geometry. However, SAR signals also experience frequent temporal fluctuations caused by precipitation events, complicating the mapping and monitoring of vegetation. In this paper, we show that the influence of a priori known precipitation events on the signals can be used advantageously for the classification of vegetation conditions. For this, we exploit the change in Sentinel-1 backscatter response between consecutive acquisitions under varying wetness conditions, which we show is dependent on the state of vegetation. The performance further improves when a priori information on the soil type is taken into account.@en

Reliable and timely mapping of crop growth conditions and of their water resources is considered a prioritary application in light of the abrupt climate changes. With this view, the paper presents a novel approach that makes use of dense C-Band time-series for the timely estimation of crop growth stages and for the detection of changes in crop water conditions, especially related to precipitation and irrigation events. Aided by vegetation indexes extracted from Landsat and Sentinel-2 imagery, the proposed Sentinel-l centered method exploits both temporal patterns of crop growth and spatial patterns of water anomalies to enhance its classification robustness.@en

Multi-temporal and multi-sensor solutions are essential to increase timeliness and reliability of land monitoring systems. This paper advocates the exploitation of the temporal contextual information provided by temporally dense SAR and optical data series series through the use of a Hidden Markov model (HMM)-based approach. An efficient strategy to incorporate the C-Band SAR data into the HMM framework, relying so far on Landsat, will be debated and assessed over a dynamic agricultural scenario, i.e. characterized by high temporal and spatial diversity in cropping practices. The site is located in the state of São Paulo (Brazil), where recent ground surveying activities has been conducted.

@en

The paper debates a novel approach for land cover (LC) mapping based on the Hidden Markov Model. The proposed methodology is aimed to address both the urgent demand of off-line (or historic) LC information retrieval and of near-real time LC monitoring. The discrete-time model employs short steps of 16 days, that conveniently fits the Landsat revisit time while providing a continuous and temporally dense representation of the land cover dynamics. Two temporal pattern typologies were identified and modeled within the proposed Markov chain architecture: a seasonal and synchrounous behavior which can be associated to the observables of LC classes such as forest and grasses, and a highly asynchronous behaviour, which characterizes the crop observables. The first typology is addressed by introducing time-dependency in state output probabilities, whereas the latter is rendered through a sequence of (sub-class) states interlinked by means of a 'left-right' based model. Such model inherently incorporates crop growth tracking functionalities as an added value. In this paper the methodology has been tailored to Sao Paulo state (Brazil) scenario, showing overall accuracy above 80% on the test sample. A particular emphasis is attributed to the identification of sugarcane plantations, that are indeed responsible for major land use changes.

@en

During the 2014-2015 sugarcane growth season in São Paulo, Brazil, a considerable dataset was acquired consisting of space-based remote sensing images from radar and optical sensors, together with intensive ground measurements. In this work, images from the Sentinel-1, Radarsat-2 and Landsat-8 satellites are used to test the effectiveness of satellite-based indicators in sugarcane growth monitoring. A two-fold hypothesis testing is applied, in order to find statistically significant emerging hot spots and cold spots, both in space and time. Especially the comparison of results from the radar and optical sensors gives an insight into the difference in capability of these sensors to detect spatial and temporal patterns and trends.

@en

Land cover reference data containing original point layer, relocated point layer to the fields and field polygons. Information includes date of acquisition, geolocation, photograph web address, crop type, phenological stage, crop height, additional notes of field. The objective was to collect data for training and validation of optical and radar remote sensing-based land cover monitoring models. The ground reference data was acquired through the freely available mobile app EpiCollect, developed at Imperial College London. The work has been carried out within the framework of the joint BE-Basic FAPESP project: 2013/50943-9@en

Intensive measurements of sugarcane taken at four fields close to Capivari, São Paulo state, Brazil, from October 2014 to October 2015. The objective was to collect data for training and validating optical and radar remote sensing-based sugarcane growth models. The measurements include LAI, stalk thickness, stalk height, leaf length, cane density, plant density, biomass, soil moisture, plant and leaf water content, and photos. Devices used include LAI (Leaf Area Index) LICOR-2000, measurement tape and pole, caliper rule, GPS device, Falcker HidroFarm HFM2010 device, weight scale, drying ovens. The work has been partially carried out within the framework of the joint BE-Basic FAPESP project: 2013/50942-2.@en

In this study, space remote sensing data and crop specific information from the ESA-led AgriSAR 2009 campaign are used for studying the profiles of C-band SAR backscatter signals and multispectral-based leaf area index (LAI) over the growth period of canola, pea and wheat. In addition, the correlations between radar backscatter parameters and the crop yields were analyzed, based on extracted statistics of temporal profiles. The results show that the HV backscatter and LAI are correlated differently before and after LAI peak. In addition, the coefficient of determination between peakrelated statistics from polarimetric indicator profiles and yield for pea fields can reach up to 0.68, and for canola and wheat up to 0.47 and 0.5, respectively. HV backscatter and coherence between HH and VV are most.@en