Water Bodies

Since more than 25 years EO data are exploited in the water domain to recover at least two major aspects, the water resource monitoring and the devastating floods extents and impacts retrieval. The topic of this paper is focused on water resource. Water resources access and management is become a more and more critical issue all around the world, the future will be darker in the context of global chance and would involve tension in numerous regions. Earth observation would play an important role to assess and monitor this resource. 2014, with the launch of the first satellite of the Sentinel constellation, Sentinel 1A and in October the online availability, represents a real break in Earth Observation world, with massive data acquired at high resolution, high revisit rate, ie. about every 5 days.

Based on this new data, a monitoring of water bodies was carried out over the Yangtze flood plain (PR China). These water bodies, ie lakes and associated wetlands within the middle and lower reaches of the Yangtze, are representing 25% of Eastern Asia water resources, and provide service to 330 million of inhabitants such as regulating services and flood storage; provisioning services such as fishing, and being biodiversity holders. These water bodies are fluctuating, influenced by the river runoff, controlled as well as the Eastern Asian monsoon climate and anthropogenic activities. Within this context Poyang Lake, first fresh water resource in China, is particularly interesting as greatly reacting to seasonal climatic variations, involving water height variations of 8 to 10 meters, and surfaces variations from 700 to 3500 km².

Since the year 2000, a monitoring of water surface is carried out in order to get a better knowledge of hydro-system dynamics as well of its driven forces. It was based mostly on the middle resolution (MR) optical and SAR data. The latest years, more and more high resolution data have been integrated to the monitoring process (ie. ScanSAR data from TerraSAR-X and Cosmo-SkyMed as well as optical such as Deimos and HJ1) but on a bi or a monthly or bi- monthly basis. After a first Sentinel-1 acquisition, on the 12 May 2014, at a moment when the satellite was not yet at its nominal orbit, since fall 2014 there are a major chance of paradigm with the arrival of the Sentinel-1 data through the ESA Hub. Now with Sentinel-1A acquiring, thanks to the pre-programmed, conflict-free data acquisition using the Interferometric Wide Swath (IW) , it is an image every 5 and then 7 days later that is done available, ie. 5 image by month, and with Sentinel-1B launch in spring 2016 the frequency should be again doubled.

The presented results illustrate the gain moving from Envisat ASAR data towards Sentinel-1 ones for water detection, as well as the systematic monitoring of the Poyang Lake water surface over near 18 months period covering by the way near two hydrological years. This impressive data flow highlights the importance of accessing to high revisiting data, as surface variations are occurring in a very short time, ie. less than 5 days during the infilling or lake’s draw off. In addition, the high radiometric quality, associated with a pixel spacing of 10m, allows to monitor not only the large water bodies but also all the small ones with a size of a few hundred hectares which have been excluded up to now of most of the systematic monitoring. The full development of the Sentinel constellation will allow the monitoring of enables the build-up of long data time series at equidistant and short time intervals.

Further work, the beginning of the operational of Sentinel-2 will be carried out on the coming synergy between the two Sentinels constellation, presenting the advantages and complementarities of the Sentinel “dual” optical SAR systems.

SAR polarimetry and polarimetric decomposition techniques have proven to be useful tools for wetland mapping. In this study we monitor natural lakes in northeastern Germany using dual-polarimetric (HH/VV) TerraSAR-X time series. The time series has 19 images, acquired between August 2014 and May 2015. We calculated different polarimetric indices using the HH and VV intensities, the dual-polarimetric coherency matrix including dominant and mean alpha scattering angles, entropy and anisotropy (normalized eigen-value difference) as well as combinations of entropy and anisotropy for the analysis of the scattering scenarios. The image classifications were performed with randomForest and validated with high-resolution aerial photos and in situ measured data. This study shows that dual-polarimetric (HH/VV) TerraSAR-X data allows the classification of reed at shallow areas of the lakes and the classification of water surface areas covered by vegetation (reed). However, seasonal changes of the reed and the surrounding vegetation influence the mapping accuracy. Images acquired in early springtime are most suitable for the mapping of reed.

The Sentinel-1 mission ensures continuity of C-band SAR data following the ERS and Envisat missions thus allowing repeated mapping over long time periods in several thematic applications developed in recent times. One application is the detection and delineation of water bodies. Envisat Advanced Synthetic Aperture Radar (ASAR) observations of the backscattered intensity (referred to as SAR backscatter) were used to generate a global record of open water bodies (Santoro et al., submitted)  in support of the Climate Change Initiative (CCI) Water Body Product (WBP) generated by the CCI Land Cover (CCI-LC) project. The temporal variability of the SAR backscatter and the minimum backscatter from observations acquired between 2005 and 2012 were input to a rather straightforward set of thresholding schemes to produce an indicator of water bodies, which was then consolidated with auxiliary datasets to provide a highly accurate portrait of open and permanent water bodies at moderate resolution (150 m) (Lamarche et al., in preparation). The dataset can be considered representative for a 2010 epoch. The major limitation on the use of ASAR data to detect water bodies was the commission of sandy deserts, glaciers and very arid surfaces to water because the multi-temporal metrics behaved similarly to what was observed in the case of open water bodies.

The capability of Sentinel-1A and -1B (S1) to provide global acquisitions with a 6-day repeat-pass interval suggests that a similar data product of open and permanent water bodies could be obtained for an epoch around 2015. Such S1-based data product is likely to improve the thematic accuracy of the ASAR-based data product because of the higher spatial resolution (20-30 m) and the availability of large scale acquisitions of cross-polarized images and interferometric coherence, which are supposed to reduce the confusions observed with ASAR data. It is noteworthy reminding that the ASAR dataset assembled to generate the CCI-WBP consisted of co-polarized observations of the SAR backscatter only. A preliminary investigation on a one-year dataset of dual-polarized (VV-VH) S1A observations of the SAR backscatter at two locations in Sweden indicates that the combination of cross-polarized average VH-backscatter and minimum VH-backscatter performs best among several combinations of co- and cross-polarized multi-temporal metrics (Santoro et al., 2015). Currently, we are consolidating our results by encompassing more areas, including interferometric coherence in the analysis and developing a classification scheme in support of global mapping of water bodies in the frame of the CCI-LC project. Results of our work will be presented at the Symposium and compared with the ASAR-based data product. Furthermore, we will review the usefulness of the S1 acquisitions towards a global mapping endeavor of water bodies with respect to the requirements imposed by the classification algorithms here developed.

References

Lamarche, C., Santoro, M., Bontemps, S., d'Andrimont, R., Giustarini, L., Brockmann, C., Defourny, P., Arino, O., "Climate Change Initiative Water Body Products (CCI-WB) : Generation, evolution and validation of global SAR-only and synergy datasets of open permanent water bodies.," in preparation.

Santoro, M., Wegmuller, U., Lamarche, C., Bontemps, S., Defourny, P., Arino, O., "Strengths and weaknesses of multi-year ENVISAT ASAR backscatter measurements to map open water bodies globally," Remote Sensing of Environment, submitted.

Santoro, M., Wegmüller, U., Wiesmann, A., Lamarche, C., Bontemps, S., Defourny, P., Arino, O., "Assessing Envisat ASAR and Sentinel-1 multi-temporal observations to map open water bodies," Proceedings of 5th Asia-Pacific Conference on Synthetic Aperture Radar (APSAR 2015), Singapore, 1-4 September, 2015.

In the era of Earth Observation big data, with the huge amount of information coming from the recent Sentinels constellation, we are entering in a new paradigm for disaster monitoring and EO data exploitation. For example,  the Sentinel-1 A (S1A) radar satellite is able to map entire Earth every 6 days (with interferometric capabilities) giving an unprecedented opportunity to access a huge number of archived scenes, which is of key importance to detect changes and assess economic impacts in case of disasters.

Furthermore, in the context of climate change with a foreseen increase in the number of extreme events such as extreme precipitation and consequent flash and riverine floods, it becomes even more important to monitor and map these phenomena with increased accuracy and rapidity. An improved capacity of characterizing risks is vital especially for managing urban areas and planning economic activities, not only to save lives, but also to reduce losses and build more resilient livelihood.

Based on the statistical analysis of SAR multi-temporal series, an innovative flood index is being developed for identifying non-permanent water bodies that can be suitably applied even by non-experts. Specifically, by taking into consideration a large amount of reference pre/post event scenes, in addition to those acquired during the investigated flood; the index easily allows to categorize as “flooded”, those areas solely temporarily covered by water with respect to permanent water bodies and non-water land cover classes.

Moreover, the index has two different versions. A simple version only takes into account the backscattering σ₀ of each scene and then performs a statistical analysis of its value throughout the whole temporal series covering the area of interest.

In addition, a more complex version is derived following the same approach described above, but in lieu of considering σ_{0 ,} multi-temporal coherence maps are used. This is foreseen to be more effective in identifying floods occurring in vegetated and urban areas, which cannot be effectively delineated with current state-of-the-art methods.

Three case studies are used to test the index: i) the flood event that occurred in the Veneto region (city of Vicenza and its surroundings) in the North-East of Italy for which multi-temporal COSMO-SkyMed stripmap data have been used; ii) the flood of January 2015 that occurred in south Malawi, well covered by S1A GRDH data ; and iii) the flood that occurred in central Japan at the beginning of September 2015, covered by about twenty S1A SLC images.

The index exhibits very good performances and improved simplicity in flood mapping compared to state-of-art methods, which are usually more complex and strongly user-dependent (results might sensibly vary based on the user experience). The accurate maps obtained have been validated with products delivered during the emergency management period by qualified providers.

The precise mapping provided by the application of the index also permits the estimation of the flood depth in case a high resolution digital elevation model is available for the study area. Validation and results are presented for the Veneto case study.

Global, accurate and up-to-date mapping of inland surface water is of paramount importance for a variety of fields ranging from governance and mitigation to satellite data processing and climate modelling. A series of maps of open water bodies are currently available at high and moderate spatial resolutions. Yet, taken separately, they do not fully meet these requirements. The scope of this work was therefore to identify complementarities between existing water body data productsand auxiliary datasets,and harmonize them in a truly global,accurate and consolidated product.

The combined data product was obtained from a synergy of the Shuttle Radar Topography Mission (SRTM) Water Body Dataset (SWBD)  for year 2000 (Farr et al., 2007), the ASAR Water Bodies Indicator at 150 m spatial resolution built in the framework of the ESA Climate Change Initiative Land Cover and based on multi-temporal metrics of Envisat Advanced Synthetic Aperture Radar between 2005 and 2012  (Santoro et al., submitted), the Landsat-based Global Forest Change 2000 - 2012 sub-dataset named “datamask” (Hansen et al., 2013) and the Global Inland Water v1.0 product for year 2000 (Feng et al., 2015). Additionally, post-editions and support from auxiliary datasets helped to improve the spatial characterization of water bodies and completeness of the consolidated product. An inland water/ocean repartition is also provided as a separate layer.

The Synergy map of open water bodies was qualitatively and quantitatively assessed against a reference validation database of 2342 samples of 150 m x 150 m spread throughout the globe and visually interpreted against Google Earth Imagery. To increase the confidence in the synergy data product, the assessment was undertaken for each of the input water body datasets and the MODIS global raster water mask at 250 m resolution (MOD44W) (Caroll et al., 2009). The validation sampling strategy was intentionally biased towards areas that are considered challenging for water body mapping. This was done to increase the sensitivity to the relative improvements with respect to the different water body products. This biased validation yielded more informative accuracy figures in terms of relative improvement assessments.

The overall accuracy of the Synergy map of open water bodies was then calculated on a stratified random sampling and compared to the input water body products considered separately and the MOD44W. Clear improvements in spatial details and water representation with respect to each of the water body products here considered were confirmed by the visual qualitative assessment. The Synergy map of open water bodies could serve as a baseline to be updated on a regular basis using both SAR and optical high resolution imagery, such as Sentinel-1, Landsat-8, and Sentinel-2. The Synergy map of open water bodies at 150 m spatial is publically available from the ESA Climate Change Initiative Land Cover viewer:  http://maps.elie.ucl.ac.be/CCI/viewer.

Carroll, M. L., Townshend, J. R., DiMiceli, C. M., Noojipady, P., & Sohlberg, R. a. (2009). A new global raster water mask at 250 m resolution. International Journal of Digital Earth, 2(4), 291–308. doi:10.1080/17538940902951401

Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., Duren, R., Hensley, S., … Alsdorf, D. (2007). The Shuttle Radar Topography Mission. Reviews of Geophysics, 45(2), RG2004. doi:10.1029/2005RG000183

Feng, M., Sexton, J. O., Channan, S., & Townshend, J. R. (2015). A global, high-resolution (30-m) inland water body dataset for 2000: first results of a topographic–spectral classification algorithm. International Journal of Digital Earth, (July 2015), 1–21. doi:10.1080/17538947.2015.1026420

Hansen, M. C., Potapov, P. V, Moore, R., Hancher, M., Turubanova, S. a, Tyukavina, a, … Townshend, J. R. G. (2013). High-resolution global maps of 21st-century forest cover change. Science (New York, N.Y.), 342(6160), 850–3. doi:10.1126/science.1244693

Santoro, M., Wegmüller, U., Lamarche, C., Bontemps, S., Defourny, P., & Arino, O. (n.d.). Strengths and weaknesses of multi-year ENVISAT ASAR backscatter measurements to map permanent open water bodies at global scale. Submitted to Remote Sensing of Environment.