Rice is the most important food crop in Asia, and the timely mapping and monitoring of paddy rice fields subsequently emerged as an important task in the context of food security and modelling of greenhouse gas emissions. Rice growth has a distinct influence on Synthetic Aperture Radar (SAR) backscatter images, and time-series analysis of C-band images has been successfully employed to map rice fields. The poor data availability on regional scales is a major drawback of this method. We devised an approach to classify paddy rice with the use of all available Envisat ASAR WSM (Advanced Synthetic Aperture Radar Wide Swath Mode) data for our study area, the Mekong Delta in Vietnam. We used regression-based incidence angle normalization and temporal averaging to combine acquisitions from multiple tracks and years. A crop phenology-based classifier has been applied to this time series to detect single-, double- and triple-cropped rice areas (one to three harvests per year), as well as dates and lengths of growing seasons. Our classification has an overall accuracy of 85.3% and a kappa coefficient of 0.74 compared to a reference dataset and correlates highly with official rice area statistics at the provincial level (R² of 0.98). SAR-based time-series analysis allows accurate mapping and monitoring of rice areas even under adverse atmospheric conditions.

The Sentinel-1 GRD (ground range detected) Level-1 product generated by the Instrument Processing Facility of the European Space Agency has noise artifacts at the image borders, which are quite consistent at both left and right sides of the satellite’s cross track and at the start and end of the data take along track. The Sentinel-1 border noise troubles the creation of clean and consistence time series of backscatter. Data quality control and management become very challenging tasks, when it comes to the large-scale data processing, both in terms of spatial coverage and data volume. In this paper, we evaluate three techniques for removing the Sentinel-1 border noise and compare the results with the existing “Sentinel-1 GRD Border Noise Removal” algorithm implemented in the Sentinel-1 toolbox of the Sentinel application platform. Validation and evaluation of the newly proposed algorithms was done using random samples containing 1500 Sentinel-1 scenes selected from a complete Sentinel-1 archive. The newly proposed approach has successfully achieved the required level of accuracy and solved the issue of time-series anomalies due to the border noise.

Recent studies have shown the potential of Synthetic Aperture Radars (SAR) for mapping of rice fields and some other vegetation types. For rice field classification, conventional classification techniques have been mostly used including manual threshold-based and supervised classification approaches. The challenge of the threshold-based approach is to find acceptable thresholds to be used for each individual SAR scene. Furthermore, the influence of local incidence angle on backscatter hinders using a single threshold for the entire scene. Similarly, the supervised classification approach requires different training samples for different output classes. In case of rice crop, supervised classification using temporal data requires different training datasets to perform classification procedure which might lead to inconsistent mapping results. In this study we present an automatic method to identify rice crop areas by extracting phonological parameters after performing an empirical regression-based normalization of the backscatter to a reference incidence angle. The method is evaluated in the Red River Delta (RRD), Vietnam using the time series of ENVISAT Advanced SAR (ASAR) Wide Swath (WS) mode data. The results of rice mapping algorithm compared to the reference data indicate the Completeness (User accuracy), Correctness (Producer accuracy) and Quality (Overall accuracies) of 88.8%, 92.5 % and 83.9 % respectively. The total area of the classified rice fields corresponds to the total rice cultivation areas given by the official statistics in Vietnam (R²  0.96). The results indicates that applying a phenology-based classification approach using backscatter time series in optimal incidence angle normalization can achieve high classification accuracies. In addition, the method is not only useful for large scale early mapping of rice fields in the Red River Delta using the current and future C-band Sentinal-1A&B backscatter data but also might be applied for other rice cultivation areas.