Deployment of Sentinel-1 (S1) satellite constellation carrying a $C$-band synthetic aperture radar (SAR) enables regular and timely monitoring of floods from their onset until returning to nonflooded (NF) conditions. The major constraint on using SAR for near-real-time (NRT) flood mapping has been the inability to rapidly process the obtained imagery into reliable flood maps. This study evaluates the efficacy of S1 time series for quantifying and characterizing inundation extents in vegetated environments. A novel algorithm based on statistical time-series modeling of flooded (F) and NF pixels is proposed for NRT flood monitoring. For each new available S1 image, the probability of temporarily F conditions is tested against that of NF conditions by means of likelihood ratio tests. The likelihoods for the two conditions are derived from early acquisitions in the time series. The algorithm calibration consists of adjusting two likelihood ratio thresholds to match the reference F area extent during a single flood season. The proposed algorithm is applied to the Caprivi region, the resulting maps were compared to cloud-free Landsat-8 (LS8) derived maps captured during two flood events. A good spatial agreement (85-87%) between LS8 and S1 flood maps was observed during the flood peak in both 2017 and 2018 seasons. Significant discrepancies were noted during the flood expansion and recession phases, mainly due to different sensitivities of the data sources to the emerging vegetation. Overall, the analysis shows that S1 can stand as an effective standalone or gap-filling alternative to optical imagery during a flood event.

Small reservoirs play an important role in mining, industries, and agriculture, but storage levels or stage changes are very dynamic. Accurate and up-to-date maps of surface water storage and distribution are invaluable for informing decisions relating to water security, flood monitoring, and water resources management. Satellite remote sensing is an effective way of monitoring the dynamics of surface waterbodies over large areas. The European Space Agency (ESA) has recently launched constellations of Sentinel-1 (S1) and Sentinel-2 (S2) satellites carrying C-band synthetic aperture radar (SAR) and a multispectral imaging radiometer, respectively. The constellations improve global coverage of remotely sensed imagery and enable the development of near real-time operational products. This unprecedented data availability leads to an urgent need for the application of fully automatic, feasible, and accurate retrieval methods for mapping and monitoring waterbodies. The mapping of waterbodies can take advantage of the synthesis of SAR and multispectral remote sensing data in order to increase classification accuracy. This study compares automatic thresholding to machine learning, when applied to delineate waterbodies with diverse spectral and spatial characteristics. Automatic thresholding was applied to near-concurrent normalized difference water index (NDWI) (generated from S2 optical imagery) and VH backscatter features (generated from S1 SAR data). Machine learning was applied to a comprehensive set of features derived from S1 and S2 data. During our field surveys, we observed that the waterbodies visited had different sizes and varying levels of turbidity, sedimentation, and eutrophication. Five machine learning algorithms (MLAs), namely decision tree (DT), k-nearest neighbour (k-NN), random forest (RF), and two implementations of the support vector machine (SVM) were considered. Several experiments were carried out to better understand the complexities involved in mapping spectrally and spatially complex waterbodies. It was found that the combination of multispectral indices with SAR data is highly beneficial for classifying complex waterbodies and that the proposed thresholding approach classified waterbodies with an overall classification accuracy of 89.3%. However, the varying concentrations of suspended sediments (turbidity), dissolved particles, and aquatic plants negatively affected the classification accuracies of the proposed method, whereas the MLAs (SVM in particular) were less sensitive to such variations. The main disadvantage of using MLAs for operational waterbody mapping is the requirement for suitable training samples, representing both water and non-water land covers. The dynamic nature of reservoirs (many reservoirs are depleted at least once a year) makes the re-use of training data unfeasible. The study found that aggregating (combining) the thresholding results of two SAR and multispectral features, namely the S1 VH polarisation and the S2 NDWI, respectively, provided better overall accuracies than when thresholding was applied to any of the individual features considered. The accuracies of this dual thresholding technique were comparable to those of machine learning and may thus offer a viable solution for automatic mapping of waterbodies.

The Caprivi basin in Namibia has been affected by severe flooding in recent years resulting in deaths, displacements and destruction of infrastructure. The negative consequences of these floods have emphasized the need for timely, accurate and objective information about the extent and location of affected areas. Due to the high temporal variability of flood events, Earth Observation (EO) data at high revisit frequency is preferred for accurate flood monitoring. Currently, EO data has either high temporal or coarse spatial resolution. Accurate methodologies for the estimation and monitoring of flooding extent using coarse spatial resolution optical image data are needed in order to capture spatial details in heterogeneous areas such as Caprivi. The objective of this work was the retrieval of the fractional abundance of water (γ_w) by applying a new spectral indices-based unmixing algorithm to Medium Resolution Imaging Spectrometer Full Resolution (MERIS FR) data using a minimum number of spectral bands. These images are technically similar to the OLCI image data acquired by the Sentinel-3 satellite, which are to be systematically provided in the near future. The normalized difference wetness index (NDWI) was applied to delineate the water surface and combined with normalized difference vegetation index (NDVI) to account for emergent vegetation within the water bodies. The challenge to map flooded areas by applying spectral unmixing is the estimation of spectral endmembers, i.e., pure spectra of land cover features. In our study, we developed and applied a new unmixing method based on the use of an ensemble of spectral endmembers to capture and take into account spectral variability within each endmember. In our case study, forty realizations of the spectral endmembers gave a stable frequency distribution of γ_w. Quality of the flood map derived from the Envisat MERIS (MERIS) data was assessed against high (30 m) spatial resolution Landsat Thematic Mapper (TM) images on two different dates (17 April 2008 and 22 May 2009) during which floods occurred. The findings show that both the spatial and the frequency distribution of the γ_w extracted from the MERIS data were in good agreement with the high-resolution TM retrievals. The use of conventional linear unmixing, instead, applied using the entire available spectra for each image, resulted in relatively large differences between TM and MERIS retrievals.