In view of recent dramatic floods and drought events, detection of trends in the frequency and magnitude of long time series of flood data is of scientific interest and practical importance. It is essential in many fields, from climate change impact assessment to water resources management, from flood forecasting to drought monitoring, for the planning of future water resources and flood protection systems. In these fields, the system design is traditionally based on the assumption of stationarity in hydrological processes such as river water level or discharge, usually obtained from traditional monitoring network measuring water level, river flow velocity and cross section geometry. The lack of long continuous observed time series, due to actual decreasing of the number of sensors across the world and to the cost of installation and maintenance of the monitoring network, prevent to identify trends over large areas. The most direct method of deriving such information on a global scale involves satellite earth observation. Over the last two decades, the growing availability of satellite sensors, and the results so far obtained in the estimation of river discharge from the monitoring of the water level through satellite radar altimetry has fostered the interest on this subject.
In this study, satellite altimetry water level data are used to set-up a consistent, continuous and up-to-date daily discharge dataset for different sites across the world. Satellite-derived water levels are related to river discharge data available over overlapping time spans to estimate empirical rating curves. Once validated, the rating curves are used to fill and extrapolate discharge data over the whole period of altimetry water level observations. The long continuous discharge time series so obtained have allowed to perform a trend analysis on extreme flood and drought events. Specifically, annual maximum discharge and peak-over threshold values are extracted from the simulated daily discharge time series, as proxy variables of independent flood events. Moreover, the threshold level approach is considered to discriminate between low/moderate flow regimes and to identify hydrological drought events.
For flood and drought events, a trend analysis is also carried out to identify changes in the frequency and magnitude of extreme events through the Mann-Kendall (M-K) test and a linear regression model between time and the flood magnitude. The analysis has permitted to identify areas of the world prone to floods and drought, so that appropriate actions for disaster risk mitigation, and continuous improvement in disaster preparedness, response, and recovery practices can be adopted.

Area A - Board 142

In our changing world, floods are a threat of increasing importance causing major fatalities and economic losses. Within this perspective, flood extent mapping is of great importance for emergency management as well as damage assessment and flood forecast improvement. SAR sensors are particularly suited for this purpose due to their independence of daylight and their ability to penetrate clouds.

Following the rapidly growing data archive resulting from recently launched SAR-sensors such as the Sentinel-1 and COSMO-SkyMed constellations, which provide unequalled revisit times, a large number of flood mapping algorithms has been developed in the past years. However, most of these algorithms focus on the delineation of open flood surfaces in rural areas, while the detection of floods in both urban and vegetated areas remains challenging and thus requires further investigation. Additionally, the majority of these algorithms consider all pixels independently. As a consequence, context information cannot be taken into account and a potentially valuable source of information is neglected. In contrast to the prevailing per-pixel mapping approaches, object-based image analysis (OBIA) groups pixels into objects before further analysis. This allows taking into account spatial context, as well as some other relevant properties such as object shape, proximity and homogeneity. Moreover, it is possible to include additional information sources such as different polarisation bands, elevation, land cover and optical imagery.

This study aims to improve flood mapping in diverse landscapes comprising flooded vegetation, including both partially submerged vegetation and flooded forests, by means of an object-based mapping approach. This approach is applied on Sentinel-1 GRD imagery and exploits both the VV and VH polarised bands. Two additional freely available information sources, i.e. CORINE land cover and SRTM topography information, are also included. The resulting classification comprises four instead of two classes, i.e. open flood, dry land, low flooded vegetation and high flooded vegetation. The performance of this object-based approach is tested on two test sites, situated in Belgium and Croatia, with a selection of established pixel-based approaches serving as benchmark.

Area A - Board 144

Flooding is a rapidly growing concern in West Africa, projected to increase with climate change. It is both a threat and a blessing for human lives, agriculture and infrastructure. There is a great need for reliable access to operational flood forecasts and alerts produced by a robust information and communication technology (ICT) system adapted to regional conditions and operated by capable West African institutions.

The Hydrology Thematic Exploitation Platform (H-TEP) provides cloud capabilities to systematically generate flood forecast production in the Niger River basin. Local discharge observation and satellite-based water level observations of the Niger River basin are sequentially assimilated into the Niger-HYPE hydrological model produced by the Swedish Meteorological and Hydrological Institute (SMHI). The hydrological modelling service integrated in the H-TEP provides data on river discharge, lake outflow, water level in lakes and rivers, as well as land surface water balance components (precipitation, evapotranspiration, soil moisture) within the area covered by the model domain. Assimilation of earth observation data from the H-TEP processing services is a core feature of the flood forecast application which can also integrate local observations from virtual stations as well as in-situ measurements when available, however, the later are scarce. This results in a 10-day flood forecast generating an early-warning system when river discharge thresholds are met.
This case study will show how the flood forecasting model used in the H-TEP is improved when calibrated with local discharge observations or derived data from virtual stations to replace the simulated discharge in specific areas along the Niger River. Ten-day flood forecasts at different points across the Niger River basin during a past flood event will be compared to test the accuracy of the model.

Area A - Board 143

Hydrodynamic models are in widespread use and vary widely in their complexity, resolution and data inputs. Therefore, validation is vital to understand which models and input data are more reliable. Without validation, the inevitable inaccuracies in model outputs may go unchecked and lead to misinformed decisions about response, mitigation and adaption. Before remotely sensed data is assimilated into flood models, better understanding of model accuracy is also vital to enable an assessment of the degree of improvement that assimilation may provide. However, calculating the accuracy of simulated flood extents is a universal and ongoing challenge. Gathering the required data either from ground or air can be expensive and effective surveys during major flood events are rare. Often, model outputs are therefore compared with results from other models due to the lack of available observations. This model-model comparison is highly problematic as it can easily paint an overly optimistic picture of accuracy, especially when the same or similar data sources are used in both setups. The SAR on board Sentinel 1 provides a possible solution to these issues, with a source of real observations. A major benefit of SAR is that data collection is not hindered by the presence of cloud cover, often present during the large storm events which tend to cause flooding. The high frequency of passes also means the sensors are more likely to be overhead at or near to the time of maximum flood extent and the resolution is high enough for use in model validation. Satellites also provide the capability to scan much larger areas in less time and more cheaply per image than using aeroplanes or conducting ground-based surveys. An example of this application of satellite data is presented and some of the limitations are discussed.

In 2015, Storm Desmond produced widespread flooding across Northern England. To characterise the event, a hydrodynamic simulation has been developed at 90m resolution using hourly rainfall from the Met Office and the HydroSHEDS variant of SRTM. To create a satellite-observed extent outline, observations from Sentinel 1 were processed using a change detection algorithm by Clement et al (2017). The available time steps were accumulated over the period of the event to create an outline of maximum extent. This outline was then used to calculate metrics including hit rate, false alarm ratio and critical success index. Results indicate a poor match between the modelled and observed extents for this event, however a number of limitations in the SAR imagery and classification algorithm used mean that this still does not provide convincing evidence of model accuracy.

Three major limitations of using remotely sensed extent for validation in this way have been identified. Firstly, the disparity in temporal resolution between the 30 minute flood model intervals and much less frequent satellite passes mean that even though it is more likely that observations will be collected at the time of maximum flood extent, it is still not certain. In the specific event discussed here, the closest overpass occurred approximately a day before the maximum discharge peak was seen at river discharge gauges in the area. This means the accumulated satellite observations are likely to underestimate the maximum flooded area during the event and therefore produce an excessively high FAR when used for validation. Secondly, inherent problems with change detection methodologies mean floods on flat surfaces such as roads and runways may not be classified correctly as they appear similar in the SAR data whether inundated or not. Finally, SAR can be unreliable in built up urban areas where signals are interrupted before reaching the ground surface. Despite these problems, remote sensing and specifically SAR data from Sentinel 1 currently provides one of the most consistent if not entirely accurate methods to verify the accuracy of large scale simulations of flood events. Therefore further development of sensors and algorithms can only benefit their usability and effectiveness for hydrodynamic model validation.

Figure 1 - Extent comparison NE of Carlisle, UK, between maximum output depths (m) of a hydrodynamic flood model (CityCAT) using MIDAS rainfall and the HydroSHEDS DEM to an accumulated remotely sensed flood extent from change detection results using SAR on board Sentinel 1.

Figure 2 - Times at which SAR snapshots were recorded by Sentinel 1 (vertical black lines) over Northern England during Storm Desmond in 2015, relative to hydrographs (m3/s) from 12 flow gauges in the region.

Clement, M. A., Kilsby, C. G. and Moore, P. (2017) ‘Multi-temporal synthetic aperture radar flood mapping using change detection’, Journal of Flood Risk Management, pp. 1–17. doi: 10.1111/jfr3.12303.

Area A - Board 145

Change analysis in combined time series of multilook polarimetric Sentinel 1 (S1) SAR data with Sentinel 2 (S2) optical data is a challenging and promising topic. This is one interest for the French CNES , that is financing this R&T study, for the definition of new products for automatic change detection for flood monitoring compatible with the incoming SWOT mission. We propose here a new mathematical framework based on an omnibus likelihood ratio test statistic (LRT) for the equality of several variance-covariance matrices following the Wishart distribution. We introduce two approaches; i/ the extension to a combined S1+S2 covariance matrix, including both log-based normalisation and Nataf transformation of the covariances for the two sensors; ii/ measurements of T-norme or T-conorme (Martin, 2005) to take a posteriori decisions considering the two independant change detection performed using each sensor.

In a second step we compare our results for the flood event monitoring of the bay of Osaka , that occurred in Japan in July 2018, with the results obtained with the Copernicus emergency monitoring service. We show how both the resolution of 10 meters and the use of the combined S1+S2 images to monitor the flood event enhance the monitoring of this disaster. Finally we discuss on the requirements for a further and possible high resolution product for flood monitoring with the incoming SWOT monitoring product.




Figure 1 : Flood monitoring using 3 images S1A GRD VV&VH 10m résolution, & 3 images S2A 10m resolution (detection threshold at 99%), between the 28 of June and the 10 July 2018 in the bay of Osaka, Japan. Automatic analysis performed for the french « Disaster Charter » plateform.
Nielsen, A. A., Conradsen, K., & Skriver, H. (2016, August). Change detection in a short time sequence of polarimetric C-band SAR data. In ESA Living Planet Symposium 2016. European Space Agency, ESA.
SWOT Science Requirements Document, 2016
Martin, A. (2005). La fusion d’informations. Polycopié de cours ENSIETA-Réf, 1484, 117.
https://emergency.copernicus.eu/

Area A - Board 146

The paper presents the results of 3-year Sentinel-1 observations over the Biebrza floodplain (NE Poland)and their exploatation in the context of seasonal inundations monitoring. The extent of the open water as well as partially submerged vegetation has been mapped repeatedly using Sentinel-1 SLC and GRD data. This mapping was based on backscattered amplitude and Shannon Entropy (SE)measure of the backscattered SAR complex signal. The most difficult task in open water mapping was the treshold settlement permitting to divide the area into two categories: water/non-water. Very smooth land surfaces (fine sands) and wind-induced roughness of the water surface created ambiguous situations in image interpretation and proper thresholding. The partially submerged herbaceous vegetation could be mapped only when the double bounce effect had translated to low SE value. Otherwise the volume scattering of vegetation masked the water hidden beneath the dense vegetation cover, especially sedge. In order to verify the real water extents on the ground as well as beneath the vegetation the network of piezometers has been used.
In the relatively flat wetland areas, composed of organic soils, covered by mixte vegetation where the digital terrain model (DTM) has inhomogeneous quality the hydrological modeling encounters several problems. The use of remotely sensed data , especially satellite multispectral images, can help to estimate vegetation cover density and its influence on water transfer rates. This is a very important advantage but the problem of imperfect DTM still persits. In this context the estimates of water extent at given moment, done with Sentinel-1, are the key information validating the results on the hydrological modeling. In this paper the results of modeling confronted with Sentinel-1-based inundation monitoring and settled on piezometers (divers) measuring precise level of water are presented and discussed. The main advantage of Sentinel-1 is its very high frequency of observations, wide swath an freely available data permitting checking the results of the modeling.

Area A - Board 147

River runoff is the single largest contributor to the Arctic Ocean's freshwater budget. The Lena River in Eastern Siberia is one of the major rivers flowing into the Arctic Ocean. More than 50% of the total runoff from the mainland originates from the four large rivers Ob, Yenisei, Lena and Mackenzie. Global warming is expected to significantly influence the amount and temporal dynamic of river. The average annual discharge of fresh water from the six largest Eurasian rivers into the Arctic Ocean has already increased by 7% over the last century.

It has been demonstrated, that the timing and magnitude of discharge of Arctic rivers is strongly linked to snow mass storage during the winter and the subsequent melt. The melting of the accumulated snow during spring leads to extreme flooding that represents the major hydrologic event of the year. About 40% of the Lena River's annual discharge is released during this few weeks lasting spring flood. Studies of the spring flood process mainly rely on hydrological modelling and sparsely available in situ gauge measurements. Also remote sensing techniques are applied to map and monitor the spatial and temporal evolution of the event.

We present a new approach to delineate water levels during spring flood events that solely relies on the evaluation of high resolution SAR data provided by the TerraSAR-X and TanDEM-X satellites. The German Earth Observation satellite TerraSAR-X was launched in 2007. It provides SAR data in different modes and resolutions. Operationally single and dual polarization is possible. Since 2010 it is flying together with its twin satellite TanDEM-X in close formation enabling single pass interferometry. Here, the primary mission goal is the generation of a global digital elevation model (GDEM) in outstanding quality and resolution that allows classifying the shoreline and land’s topography at an unprecedented level of detail. The spatial resolution is 12.5 meters with a vertical accuracy of 2.5 meters.

Our test site covers the central part of the Lena River Delta where the main channel diverges into its major distributaries, the Trofimowskaya, Bykowskaya and Olenekskaya channels. The spatial extent of the flooding is derived from 22 TerraSAR-X images acquired between end of May and mid of July 2013. The satellite's footprint pattern allows observations on three consecutive days, followed by a three days gap. The GDEM product of the TanDEM-X mission is utilized to refine the water mask and to derive the bathymetric 3d information of the flooding. The high spatial resolution of three meters and the dual polarization of the TerraSAR-X data acquired in the StripMap mode, allow for a fine delineation of the water line and an improved separation of open water, ice cover and land, respectively.

Area A - Board 148

Currently, ice situation monitoring in Russia is carried out primarily on the network of permanent and temporary stream gauge of Federal Service for Hydrometeorology and Environmental Monitoring of Russia (Roshydromet). However, the large the country area and the relatively small number of existing stream gauge lead to a lack of information about the current conditions on the rivers. Therefore, an important source of information are space images, whose large spatial coverage and high spatial resolution allow for almost same time monitoring of the situation on the rivers. The main source of information is radar images (SAR imagery), which can be obtained in cloudy conditions and do not depend on the sun illumination of the surface.
The study area is selected from the gauge Poldarsa (the Sukhona river) to the lower course of the river Yug and river Malaya Severnaya Dvina near the Velikiy Ustyug town. On the site below the city are all conditions for the formation of large congestion. Morphological features is a sharp meander of the river bed near the Veliky Ustyug and the long, erratic bar river Malaya Severnaya Dvina (Hristovska, Bobrovnikova and Golodaevsky thresholds), which are contribute its formation almost yearly.
This paper presents the results of the ice situation monitoring based on radar images Sentinel-1 for the season of 2017-2018. Also, we used optical imagery (OLI/Landsat-8 and MSI/Sentinel 2) to control the SAR-imagery interpretation results, the data from stream gauges, results of field works in February 2018 and meteorological dataset.
The spatial resolution of available SAR-images from Sentinel-1A and B is 10 m. The period of acquisition is 12 days. Radar data was processed by the software SNAP. The pre-processing steps included radiometric calibration, speckle filtering, geometric correction. Processing parameters were selected from the series of experiments and are the same for multitemporal SAR-images. It allows to lead to one conditions all images for correct multitemporal analysis of the reflected signal from the ice and snow cover for the season 2017-2018.
Figure 1 shows a graph of changes in the values of the backscattering coefficient (σ) between December 2017 and April 2018. It’s necessary to involve additional information about weather conditions during imagery acquisition and the day before for correct interpretation of radar imagery. In autumn 2017, the first transition to negative average daily air temperatures was marked on October 20. In November, the air temperature fluctuated around 0°C, liquid precipitation was observed. First ice formation (sludge and ice fat) appeared only in the 3rd week of November. Radar images from December 3 displayed ice formation on the Sukhona river ( σ= -25 dB), the Yug river is free of ice. The ice formation on the Sukhona river was accompanied by the rippling of the ice cover. Areas of high ice roughness on radar image from January 20, 2018, have a bright white color (σ= -11 dB).
From December 28 to January 4 there was a thaw with temperatures around 0...+1°C. According to results of field works at the end of January 2018, the height of hummocks on the Sukhona river near the Porog and Kalikino villages was up to 1 m; from the site the Podsosenie village to the Velikiy Ustyug road bridge – up ti 1,5 m, near the Velikiy Ustyug – up to 0,5m. Snowfalls in late January and early February contributed to leveling the ice cover surface, mainly in the urban area, where the height of hummocks was the lowest. At the same time, there were no changes in the snow and ice conditions in the visible and near infrared bands of winter optical images. However, the attraction of the Landsat-8 thermal infrared band also allowed to identify areas of snow and ice cover with the different moisture content.
Throughout river ice period ice-holes were observed in the area of the convergence of the Sukhona river and the Yug river and along the Malay Severnay Dvina river. In the first month of freeze-up, areas of open water were clearly displayed on radar imagery with a black color. The future ice-hole development was analyzed on the basis of optical imagery because of ice-cover moisture increasing leads to the fact that such ice cover areas can look the same as open water.
In March and April 2018, clear and low-cloud conditions were prevailed. It h made possible to analyze a series of optical images OLI/Landsat-8 and MSI/Sentinel-2 and monitor the start of the spring ice drift.
Thus, the use of radar and optical imagery with results of field works and meteorological observation allow to obtain a complete picture of the ice-cover development along the river and provide information for hydrological forecasts of the spring ice drift.
The reported study was funded by RFBR according to the research project № 18-05-60021-Arctic.

Area A - Board 149

1.INTRODUCTION
Flooding maps with high accuracy are important because they allow: a) the identification of those areas at risk of flooding and therefore to plan a more effective emergency response; b) flood risk mapping that allows the identification and quantification of the elements subject to flooding (e.g. houses, businesses, etc.); c) to promote a greater awareness of the risk of floods for the inhabitants of the area concerned; e) identifying high-risk "hazardous buildings"; f) in general, in the long term, support planning and development by identifying high-risk positions and directing development away from these areas.

Currently, flooding maps are produced using satellite-based flood detection algorithms or through simulations by hydraulic modeling.

Most exploited satellite sensors for flood detection are SAR (including InSAR approach) and optical. Each technology corresponds to a different way of recognizing water bodies. For SAR, Change Detection (CD) technique (Auynirundronkool et al., 2012) and InSAR (Pulvirenti et al. al., 2016; D'addabbo, 2016) are used, while for optical through a combination with high resolution DEM (Irwin et al., 2017; Liu X et al., 2017).

These technologies, however, if taken individually, have significant limitations in identifying flooded areas. Concerning SAR, through the CD, it isn’t possible to detect floods in urban contexts, in dense vegetation or in presence of shadows and in areas characterized by a high roughness of the ground (Giustarini, 2013; Brivio et al., 2002). For InSAR approach, multitemporal and interferometric archive images aren’t always available and often the analyzes are affected by "false positives and negatives errors" (Pierdicca et al., 2016; Ferretti et al., 2001; Refice et al., 2014). These errors occur especially in urban areas or where there are small thicknesses of water under the vegetation.

Hydraulic modeling flooding maps are based on the use of 2-D flood simulations, (see, for example, Proust et al., 2010; Costabile & Macchione, 2012).
Two-dimensional approach is preferred when the flood must be simulated on a large scale (more than 10000 km2) and where there are urban contexts (Hunter et al., 2008; Mignot et al., 2006). The information provided from this numerical method refers to the depth, velocity and the extension of the flooding. Another approach that is widely found in the literature refers to 1D + 2D, in which the river is discretized in one dimension while the surrounding environment in two-dimensional (Mazzoleni et al., 2013; Leandro et al., 2009).
Flood hazard assessments are affected by several sources of uncertainties which have significant consequences on the flooding maps accuracy by hydraulic modeling. In particular, uncertainty concerns the hydrological data, the hydraulic parameters, calibration and validation data, the spatial and temporal discretization of the governing equations describing the physical processes, the way to take into account man-made structures interacting with the flow and so on (among the most recent ones see: Di Baldassarre & Montanari, 2009; Bales & Wagner, 2009; Di Baldassarre et al., 2010; Grimaldi et al., 2013; Domeneghetti et al., 2013; Dottori et al., 2013; Jung and Merwade, 2015).
The uncertainties above underlined and related to the two approaches - remote sensing and hydraulic modelling- used separately, suggest to explore integrated approaches which in dependence of the availability of data (rainfall or streamflow measurements, satellite imagines, markers for social media) might allow a more accurate reconstruction of the flooding map.
In this context it is relevant to explore: a) how data from the hydraulic simulations , satellite maps and the markers deriving from social media might be integrated, in order to carry out a more accurate reconstruction of the flooding map; b) how the accuracy of the satellite instrument could be improved in relation to the context in which flooding occurs (urban, suburban, dense vegetation, etc.); c) if remote sensing can be useful or not, in the case of non-instrumented areas, in the retrieval of hydrological data as input of hydraulic models (e.g. precipitation); d) if the markers of social media allow to visualize or not the elements to date blind to the satellites and to evaluate, therefore with greater resolution, the damages produced by the flood event.
In order to provide a first preliminary answer to the above listed issues, for the Houston areas (USA) and for the extreme rainfall events due to Harvey hurricane, a comparation of flooding map obtained by satellite data with that obtained by hydraulic model simulations validated by comparison with ground-based measurements and social media markers is carried out. This has allowed to quantify the limitation of flooding mapping by satellite images, and furthermore to explore the potentiality of integrated approaches which use satellite data, social media markers and hydraulic models.

2. METHODS
The area considered for the analysis is Houston (TX) and the hydrological extreme event the category 4 Harvey hurricane occurred between 27 and 31 August 2017. The choice of this area is justified by the fact that, being a dense monitored area, it is possible to have a large amount of ground-based measurements available for the hydraulic model and for the validation of the results.
Satellite SAR images, were downloaded from the Copernicus website (http://emergency.copernicus.eu) and are the result of the Change Detection (CD) technique applied on high-resolution X-band images (from 1 to 4 m) of the Italian COSMO-SkyMed constellation (Chini et al., 2012).
The social media markers used were extracted from the Evolution of Emergency Copernicus (E2mC) database owned by Copernicus and developed by the Polytechnic of Milano (Scalia et al., 2017).
In the present study, USACE Hydrologic Engineering Center’s River Analysis System (HEC-RAS 5.0.3) Model is used for the analysis of the flood. This software turns out to be a very robust software for the analysis of large-scale flooding in urban areas, with results that are not significantly affected by numerical diffusivity and allows the use of high time-steps in order to reduce computational times. Furthermore that software allows the direct application of precipitation on the integration domain. Rainfall temporal and spatial distribution during the event was obtained both from ground based measurement in a large number of raingauges and from TRMM satellite. This has allowed also to assess the accuracy of the latter data in comparison with ground based measurements.

3. RESULTS
The comparison of the satellite map with that obtained from model simulation is based on: a) correspondence of the position of the floods and on the ratio between flood surfaces detected by satellite and obtained by hydraulic model which has been calibrated, varying the Manning coefficients, and comparing simulations with observations by ground-based hydrological measurements; b) on the temporal progress of the floods of the hydraulic network; c) characterization of the flood through the photographs and comments found by social media markers thought as further elements of model validation; d) using the rain data derived from satellite (TRMM satellite) instead of using ground-based precipitation measurements to explore the possibility to use rainfall data from satellite for not instrumented regions.
From the first comparison of the results obtained from the hydraulic modeling, it has emerged how the flooding surfaces in the considered area, have the same positions as those "seen" by the satellite (Figure 1 a). The second comparison demonstrates that the hydraulic model followed the real-time trend of the floods and that the accuracy, in terms of the hydrometric level, between the calculated one and the real one is 91% (Figure 1b). The third comparison allowed, through photographs and comments found in social media markers, to validate with greater precision the positioning and depth of flood emerged from hydraulic model (Figure 1c). The comparison between rainfall temporal and spatial distribution during the event was obtained both from ground based measurement in a large number of raingauges and from TRMM satellite has shown that the latter underestimates significantly (49%) the actual precipitation rate (Figure 1d).

4. DISCUSSIONS AND CONCLUSIONS
The results obtained have clearly shown that it is possible to create very precise flooding maps by combining: a) hydraulic model simulations, b) social media marker; c) satellite data; d) ground based measurements.
However, it should be noted that maps obtained from satellite, used for the validation of the results of the hydraulic model, were very effective only for the verification of flood in correspondence of suburban areas. In fact, it has been shown that in presence of buildings, roads, shadows, rough terrain, dense vegetation and all those problematic elements for SAR, this tool is useless for our purpose. The analysis showed how the use of the social media marker allowed to overcome this important limit, in fact it allowed to obtain information for validation even where the satellite was found to be "blind".
The integrated approach followed is reliable mainly for areas of the world equipped with ground-based measurement stations. The issue remains open how to improve flood detection in non-instrumented areas, since that the precipitation obtained by remote sensing are affected by significant inaccuracies. This suggests to explore the possibility to combine by suitable algorithms the recorded from different sensors, for example SAR, InSAR, MODIS, Landsat in order to improve the flooding detection or precipitation field by satellite data.

Area A - Board 150

Nowadays the exploitation of satellite data in hydrological-hydraulic field is strongly consolidated. Nevertheless, even if the approaches to measuring the water surface elevation, slope and river width are established, deducting the river discharge directly from satellite is still a current challenge. Among the procedures proposed in the literature, Data Assimilation (DA) techniques combine available information from observations (remotely sensed or in-situ) and mathematical models simulations to provide an optimal estimation of unknown variables (i.e., river discharge) and parameters (i.e., roughness coefficient, bed elevation). Focusing on hydraulic modeling, DA methods are based on optimization algorithms that require prior knowledge of the model inputs, such as roughness and fluvial bathymetry. However, the absence of information on the submerged river portion limits the implementation of hydraulic models in remote or poorly surveyed areas, where topographic data are usually retrieved from remotely sensed and freely available Digital Elevation Models (DEMs), such as the Shuttle Radar Topography Mission (SRTM), which do not provide information on the wetted river portion.
The aim of this study is to evaluate the improvement in river discharge estimation using spaceborne digital elevation model (DEM) integrated with river bathymetry information. The study refers to a 132 km-stretch of the Po River (Italy) whose geometry information derives from a freely available satellite DEM SRTM (Shuttle Radar Topography Mission) with a resolution of 30 meters. River discharge is estimated using SIC4DVAR algorithm that combines the 1.5 full Saint-Venant hydraulic model SIC² and a variant of the 4D-Var method, while prior bathymetry information is retrieved from two different approaches: i) Channel Bankfull Approach (CB) that estimates river bottom from an empirical relationship identified by a set of hydromorphological variables available for a limited number of river sections; ii) Slope Break Approach (SB), which is based on the relationship between water surface width and elevation retrieved exclusively from spaceborne data. Simulation results further prove that fundamental role of the bathymetry in river discharge estimation and highlight the potential of the proposed methodologies to integrate remotely-sensed DEMs with bathymetric data.

Area A - Board 151

With the increasing demography, inland water is a more and more pressured resource for the population needs as well as a societal risk for local populations. It is also a fundamental element for industry and agriculture, therefore becoming an economic and political stake. The monitoring of inland water level, proxy to freshwater stocks, conditions of navigability on inland waterways, discharge, flood prevention, is thus an important challenge.
With the decreasing number of publicly available in situ water level records, the altimetry constellation brings a powerful and complementary alternative. It allows a massive continuous and homogeneous access to a wide network of data over lakes, reservoirs and rivers where the satellite ground track intersects them, defining so called virtual stations (hereafter VS).
Due to the complexity of the terrain at the VS positions, cautious measurement selection must be performed on the raw altimetry data to provide a valuable water level estimate. In order to ensure the quality of the observation, the Hydroweb/HYSOPE water level production line accounts for the natural hydrological cycle that is responsible for water level variations, reaching up to several meters. This allows to discriminate the “true” water echoes from off-nadir land echoes. To that respect, introduction of new virtual stations in HYSOPE operational products usually requires to wait for the processing of almost 1 year of data to provide hydrological constraints.
This talk aims at presenting a new approach and its automatization for massive systematic processing at continental scale. The objective is to build such hydrological constraints from other previously existing and causally linked VS on the same hydrological basin. The implementation of this solution requires an already existing dense network of VS but also the a priori knowledge of the hydrographic basin and the causal links between VS. Thanks to this approach, it is possible to integrate rapidly new repetitive mission (e.g. Sentinel-3B), but also drifting (e.g. Cryosat2) missions in operational water surface height products.

Area A - Board 152

Monitoring of arctic rivers is important in the sense that
they are a significant freshwater source to the Arctic ocean, which is sensitive to the overall freshwater budget.

River levels typically show large seasonal variations. In Arctic rivers, the water level rises
rather quickly in the spring when the snow and ice are melting. As an example, the water level in Lena river rises up to 20-25 meter within a few weeks. This abrupt rise is challenging
to capture for satellite altimetry missions such as Sentinel-3 and SARAL/AltiKa that has a repeat period of approximately one month. However, by combining levels from several missions the temporal resolution is significantly improved.

Here we use data from CryoSat-2, SARAL/AltiKa, and Sentinel-3A for a river segment to make a joint solution of the river level. We set up a model that consists of a state-space model and two spline functions. The state-space model is composed of a process part consisting of an AR1 process and an observational part, where the error follows a mixture between a Gaussian and a Cauchy distribution. The spline functions account for the change in topography and a potential variation in the water level amplitude along the river. The model approach has the advantage that the river level time series can be reconstructed at any location along the considered river segment. Such a framework further allows for an easy calculation of other interesting parameters such as the longitudinal water slope.

Here We demonstrate the model for the Ob, Lena Rivers, and other selected Arctic rivers.
We find a clear improvement of the agreement with in situ data
when all missions are used compared to only using single mission data. The amplitude of the spring flood is better reconstructed and the noise in the modeled is reduced.
The presented work is part of the ESA funded project ArcFlux.

Area A - Board 153

Radar altimetry has significantly progressed over the past decades. Nevertheless, altimetry-derived water surface elevation (WSE) suffers from land contamination near the water-land transition zones due to the large altimeter footprint and topography. The launch of CryoSat-2 opened a new era of Delay-Doppler or SAR (Synthetic Aperture Radar) altimetry. The SAR altimeter provides an improved along-track resolution and a higher precision, which offers the capability to monitor near coastal regions and smaller inland water bodies. The technical heritage from CryoSat-2 has been integrated into Sentinel-3 mission. Many studies have approved the benefits of SAR altimeter over coastal regions, However, the benefits of this novel Delay-Doppler technology have not fully exploited over inland water.
With multiple altimetry data sets (i.e. CryoSat-2 SARIn mode, Sentinel-3 SAR mode, and SARAL LRM mode), we evaluate the performance over a mountainous lake, Namco in Tibetan Plateau. Sentinel-3 is less affected by surrounding topography, and can deliver observations even the ground track is very close to the lake shore. SARAL is prone to be contaminated by off-nadir features due to its large footprint. However, CryoSat-2 has problem when it flies over mountain towards the lake.
We find that the performance of CryoSat-2 is closely related to topography. The CryoSat-2 waveforms are noisy when the satellite is flying over mountains towards the lake, where only a few looks capture the water surface shown in L1B stacks. This is most likely because the tracking window is still adjusted to the higher topography for most of the bursts that is used to form the multi-looked waveform. However, the waveforms close to lakeshore are still good when the satellite is approaching the mountains, because the tracking window is still positioned referencing to the lake elevation instead of the higher mountain elevation. Moreover, CryoSat-2 multi-looks from different bursts are probably misaligned in range, which leads to unwanted smearing of multi-looked waveform after the multi-looking processing. Therefore, this problem is very likely attributed to the fact that CryoSat-2 operates in closed-loop combined with how fast the tracker can adjust based on the observations.

Area A - Board 154

Satellite altimeters have been used to monitor rivers and reservoirs water level from which water storage can be derived. Inland water altimetry can play an important role in continental water resource management. Traditionally satellite altimeters have been designed to monitor homogeneous surfaces such as oceans or ice sheets, which results in poor performance over small inland water bodies because of the land contamination contribution in the returned waveforms. The advent of SAR altimetry (with its improved along-track spatial resolution) has allowed measuring inland water level with better accuracy and increased spatial resolution. Sentinel-3 is the first altimeter with SAR mode covering the whole globe and there is a need to evaluate the retrackers implementation and their accuracy. One of our study objectives is to test the performances of Sentinel-3 over middle and small size water bodies that could not be monitored previously with LRM altimeters. The application of water level in human impact river basin is also our goal to make a step further on the practicality of the research. Water levels retrieved from the satellite are integrated into a river flow simulation model that includes the buffering effects of dams to improve the accuracy of the model.

This work focused on three retrackers to retrieve water levels from SAR altimeter data of Sentinel-3 over inland water bodies dedicated to restricting the land contamination of the signal. We modeled three different retrackers including the threshold retracker, OCOG (Offset Center of Gravity) retracker and 2-step physical-based retracker applied directly over Level-1 data of Sentinel-3 and adjusted the parameters to fit the inland water bodies. Also, we make an effort to limit the land contamination for the received waveform and improve the accuracy for small and middle size water body monitoring by including the DEM information at 30m resolution to select the waveform portion from nadir. The waveforms with heavy land contaminations are filtered out, and after that, the time series of water level results are retrieved with water masks over water bodies in the Ebro river basin (Spain).

We studied eight water bodies with width varies from 130 meters to 4.5 km. All performances of the three retrackers are compared with the ESA’s ocean retracker in Sentinel-3 level-2 product as well. The results show good agreement with the in-situ measurements over lake Ebro (width about 1.8 km) and Ribarroja reservoir (width about 400m) with unbiased RMSE about 0.28 m and 0.16 m respectively. Accuracy over water bodies located in mountainous areas still need to be improved, but results show the possibility of retrieving water levels over a very small water body with the width about 130m.

With the water level products, we turn them into volumes and integrate with SAFRAN-SURFEX-RAPID (SASER) model over the same river basin and forcing dams in the model to see the impact on the outflow. The preliminary results with historically observed volumes already show that forcing dams improves the results. The work with the integration of satellite retrieved water levels is undergoing.

Area B - Board 155

Rivers and lakes are a key component of the continental hydrological cycle. The storage and amount and dynamic of water stored in lakes, reservoir and wetlands globally is known only crudely. Over rivers, surface water stage and slope are a key measurement for derivation of streamflow. Although river discharge and lake water storage are critical elements of land surface hydrology, they are poorly observed globally. Freely available stream gauge networks cover only a few river basins. They record water surface elevation at fixed points along river channels. Since the beginning of the 1990s, numerous in-situ networks have declined or stopped working, because of political and economic factors
Since the mid-nineties, satellite radar altimetry has been a successful technique for monitoring height variations of continental surface water, such as lakes and rivers. Hydroweb was developed at Legos (France) in 2003 in order to produce water level variations on large lakes and some virtual stations on the biggest rivers on the Earth. In 2016 a new database (http://hydroweb.theia-land.fr) has been developped jointly with Legos, CNES and CLS aiming at providing the time series of water level and volume calculated from the historical altimetry data and at continuously updating them with the current (Jason-3 & Sentinel-3A) and future (Sentinel-3B, Jaspon-CS, and ultimately SWOT) altimetry mission. It now produces and delivers river level data from satellite altimetry on 2200 virtual stations spread over river basins worldwide. It also provides lakes level variations for 160 lakes. A goal of the development of the new Hydroweb database is to make the availability of the products operationnal and in near real time as much as possible. A tool, named Hysope, running in CLS, is now calculating the water level more than 1000 virtual stations, using Jason-3 and Sentinel-3A, and on 64 lakes in an automatic mode and in NRT (the data are processed 1.5 days after their acquisition). In paralel, for almost 80 lakes, uing satellite images, the ara and the volume changes are also produced in Hydroweb.

Area B - Board 158

Recent developments in river modeling combined with global hydrographic information on river networks and their associated watersheds have allowed for efficient simulations of both river discharge and water levels on a global scale and at an increasingly fine spatial and temporal resolution. Coincidentally, the amount of available hydrologic remotely-sensed data from space have also unprecedently grown for the last two decades. Remotely-sensed river data is currently mainly provided by satellite nadir altimetry missions such as TOPEX-Poseidon (1992-2005), Jason series (2001-present)), EnviSat (2002-2012), or more recently Sentinel 3 (2018-present), and will soon be complemented by the upcoming wide-swath altimetry SWOT mission (2021). The merging of hydrologic models with data (satellite and/or in situ), for example through data assimilation, can be used for either long-term reanalysis, to improve our understanding of underlying physical processes, as well as for operational applications such as flood damage prevention. However, the global coverage of these satellite missions results in limited spatiotemporal observing capabilities, and space missions inevitably have to choose between fine spatial sampling or fine temporal sampling. As an example, the TOPEX-Poseidon and Jason missions series observe the same spot every 10 days but their characteristic inter-track distance at the equator is 315 km, resulting in sparser spatial coverage. In contrast, the EnviSat mission has a longer repeatitivity period of 35 days, but covers more land due to a narrower inter-track distance, which is typically only 80 km at the equator. Therefore, individual altimetry missions may not be able to capture all single hydrological events, and more generally, they may not be able to accurately observe the main characteristics of river flow. In an attempt to better understand the observational requirements of satellite missions, we assess the relative capability of altimetry missions, both individually or combined, to capture the main characteristics of river flow such as minimum, maximum, and mean flow.

Area B - Board 159

Satellite radar altimetry has been used to measure river and lake water levels with large successes for the last 20 years. However, it is well known that the altimeters, being designed to observe the oceans, can have some difficulties to track the surface of the water bodies especially in cases where the surrounding topography is rugged. To overcome this issue, recent radar altimeters (on Jason3 and Sentinel3 satellites) can be operated in two tracking modes: the classic Close Loop (CL) mode and the Open Loop (OL) mode. In OL mode the position of the tracking window is no more defined autonomously by the altimeter but is imposed by internal tables that are built from a database of hydrology targets. This database provides position and approximate width and elevation of the water bodies, the so called potential virtual stations (PVS). This mode improves the number of available measurements for hydrology [1] and their accuracy. Recently, the methods used for the definition of PVS have been refined and extended, so that a large number of new PVS have now been defined [2]: 33 000 for Sentinel-3A and 32 500 for Sentinel-3B on its operational orbit. They will be put in operation by the end of November 2018.
This database has been fully validated before delivery to CNES and ESA using as much as possible of external database (DEM, watermasks) and altimetry measurements when they were available. However, taking into account that the height of the PVS must be known with an exactitude of plus or minus ten meters typically, it is expected that it will be necessary to correct a small fraction of them after validation in orbit (for example because a dam has been built after the measurement of the global DEM used in building the database). Therefore, a gradual improvement process has been defined with ESA and CNES [3] in order to obtain in a couple of years targets of very high quality in the OLTC (the first version of which is already a very big step in quantity of measurements of good quality compared to what is already available with the CL mode).
So, taking into account the large size of this database and the plans to extend it further in the next future, it is crucial to have automatic tools adapted to the validation. Given the raw data produced by the altimeters (waveforms, agc, …) and derived data available in the Geophysical Data Records (GDR), a set of criteria have been defined to classify the overpasses for which the altimeter tracks a waterbody and those where it fails to do that. These criteria have been implemented in automatic tools.
Obviously any automatic classification fails for some cases. So we have also validated these automatic criteria and characterized their classification accuracy (confusion matrix). This has been done manually over a significant number of randomly sampled PVS using a visualization tool developed in Qgis which is very complementary to the automatic tools. This tool allows us to visualize some graphical measurements, temporal evolution of measurement with animations and geophysical environment of the station.
Today, we have a first result on 400 stations in Open Loop mode that are already in the Sentinel-3A OLTC version 4.2. On this stations, more than 90 % are classified as water. Only 10 % of stations requires a manual expertise. Their height or location will probably have to be adjusted for the next update of the DEM onboard Sentinel-3A. This result show that the great majority of stations are validated and this is very encouraging for the future validation.
The poster will show the results of the validation of the 65000 new PVS for Sentinel-3A and 3B and the methods used for this validation.
[1] Biancamaria et al., Validation of Jason-3 tracking modes over French rivers, Rem. Sens. Env. 209, 2018, 77-89
[2] Blumstein et al., Extending the database of hydrology targets for DEM onboard altimeters, OSTST 2018
[3] Le Gac et al., Benefits of the Open Loop Tracking Command (OLTC): Extending Conventional Nadir Altimetry to Inland Water Monitoring, OSTST 2018

Area B - Board 160

The Sentinel-3A satellite, successfully launched on 16 February 2016, carries the SAR Altimeter (SRAL). SRAL is the main topographic instrument and is expected to provide accurate topography measurements over sea ice, ice sheets, rivers and lakes. It operates in dual frequency mode (Ku and C bands) and is supported by a microwave radiometer for wet tropospheric correction over ocean and a DORIS receiver for orbit positioning.

Lake Bracciano is of volcanic origin, located 32 km northwest of Rome. It is one of the major lakes of Italy and has a circular perimeter of approximately 32 km and a diameter of 7 km. Its inflow is from precipitation only as there are no inflowing rivers. As the lake serves as a drinking water reservoir for the city of Rome, it has been under control since 1986 to avoid the pollution of its waters. For this reason, Bracciano is among the cleanest lakes of Italy. The absence of motorized navigation favors sailing, canoeing, swimming and scuba-diving.

Because of a drought that has hit Italy in 2017 and for the intensive water drainage by the Regional Water Supply Agency (ACEA), Lake Bracciano has suffered a significant drop of the water level between March and December 2017, as registered as well by in-situ measurements (Bracciano Smartlake [1]).

Since Sentinel-3A Orbit Number 213 overflies Lake Bracciano and considering the satellite’s 27-day repeat period, water level variations detected by the SRAL altimeter can be regularly compared to in-situ measurements to measure the performance of the instrument and of the data processing algorithms estimating the radar range over such a very small lake. The purpose of the experiment is to design and verify novel processing algorithms that are specialized for “difficult” (ie., small) targets. Furthermore, an important aspect of this development is to produce a realistic error characterisation both from intrinsic statistics and from auxiliary data such as precipitation or soil moisture informing on the variation of the backscatter from land. Once matured and validated, these specialized algorithm can be run systematically in the Sentinel-3 ground segment and monitor the water level and volume variation of many small ungauged or transboundary reservoirs from space.

In this study, Sentinel-3 products (level 1A/FBR) made available by the ESA GPOD SARvatore for Sentinel-3 online and on-demand processing service (https://gpod.eo.esa.int/services/SENTINEL3_SAR) have been processed with specialised retrackers and analyzed. The SARvatore service exploits the computational power provided by ESA Grid Processing on Demand system (GPOD) that is a generic GRID-based operational computing environment where specific data-handling Earth-Observation services can be seamlessly plugged-in. One of the goals of GPOD is to provide users with a fast computing facility without the need to handle bulky data.

SARvatore for S3 Products have been processed in SAR mode at 20 Hz (330 m along-track resolution) and 80 Hz (83 m) and retracked with the SAMOSA+ retracker [2] and the novel enhanced SAMOSA++ retracker. SAMOSA++ is an advancement of the SAMOSA+ that exploits the RIP (Range-Integrated Power) to carry out the retracking. Results obtained have been compared to the ground segment Sentinel-3 products, not specifically designed for inland water, but providing estimates from the resident physical and empirical retrackers used for ocean, ice-sheets and sea-ice.

Furthermore, the novel specialized retrackers will be validated also on a small river with accessible in situ data, in order to demonstrate the capacity provided by SAR altimetry to accurately monitor the seasonal cycle of such small rivers.

Considering the expected signal degradation recorded in L1b waveforms (multiple-peaks), well known criteria in post-processing have been adopted to correctly screen out outliers improving the quality of the estimates.

Future developments will aim at improving the SAR processing by properly selecting the waveforms composing the stack before multi-looking (e.g. as successfully done in the ESA-funded CRUCIAL Project [3]).

[1] https://doi.org/10.1016/j.asr.2017.12.018
[2] https://braccianosmartlake.com
[3] research.ncl.ac.uk/crucial

Area B - Board 161

The capacity of radar altimetry for monitoring rivers and lakes water level has been constantly improving over the years: the expansion of the constellation allows better spatial and temporal coverage; new technologies combined with innovative processing improve the quality of the estimations. Particularly, recent studies have demonstrated the improvement of performance offered by the new generation of altimeters featuring SAR acquisition modes, the improvement being mostly attributed to the reduced footprint brought by SAR processing.

Recently introduced for high-resolution applications, the fully-focused SAR processing (FF-SAR) allows to fully take advantage of the altimeter capacity in terms of resolution gain, reaching resolutions around 1 meter. Consequently, the traditional 20 Hz posting rate can be drastically increased leading to a great number of height estimations on the same water target. While being already applied to current SAR mode data (Cryosat-2, Sentinel-3 A-B with global coverage) with some limitations, it will be fully efficient with the upcoming Sentinel-6 mission. Preliminary work has already provided insights into the potential of the FF-SAR to precisely and accurately estimate water levels over a wide range of water bodies, including very small targets. [1], [2]

The work that will be presented takes one step further the evaluation of the FF-SAR benefits using a large dataset. Additionally, new methodologies to optimally exploit the high number of WSH estimates, including editing/averaging strategies and classification methods, will be shown. We will also present the first results of a long reprocessing campaign of S3-A data over French rivers and lakes that are referenced in the S3 OLTC tables. Systematic comparisons with unfocused SAR and LRM data will be shown, as well as comparisons with available in-situ dataset.

[1] Ejido A., Smith W., “Fully Focused SAR Altimetry: Theory and Applications”, IEEE Trans. Geosci. Remote Sens., vol 55, n°1, January 2017.
[2] Rieu et al., “From unfocused to fully-focused SAR processing: benefits for different surfaces”, OSTST 2018

Area B - Board 162

For many years, satellite altimetry is providing valuable information on water level variations of inland water bodies. The “Database of Hydrological Time Series of Inland Waters” (DAHITI, http://dahiti.dgfi.tum.de) developed by the “Deutsches Geodätisches Forschungsinstitut der Technischen Universität München” (DGFI-TUM) already provides more than 800 globally distributed water level time series of lakes, rivers, reservoirs and wetlands.
In this contribution, we demonstrate the potential to improve the DAHITI water level time series by combining multi-mission satellite altimetry with monthly land-water masks derived from Landsat and Sentinel-2. Hereby, optical images are used two times: On the one hand, the masks are used to select the best altimeter measurements at each overflight, on the other hand, derived surface area time series are used to detect outliers in the water level time series. Finally, surface area time series are used to increase the length and temporal resolution of the water level time series by using the derived height-area relationship. Moreover, a combination of both time series allows for the computation of water storage changes.
For validation, the resulting water level time series are compared with in-situ data as well as with the original altimetry time series in order to demonstrate the advantage of using time-variable land-water masks.

Area B - Board 164

The Copernicus Sentinel-3A mission was successfully launched in February 2016. It is a multi-instrument mission to measure surface topography, sea- and land-surface temperature, ocean colour and land colour with high-end accuracy and reliability. To reach the data coverage required by the Copernicus service, its twin Sentinel-3B was launched on 25th of April 2018. After 5 months in a tandem flight configuration (Sentinel-3B 30 seconds ahead of Sentinel-3A), Sentinel-3B was moved to its final orbit, the same as Sentinel-3A but shifted by 140°.

As for ocean, the main objective of the Sentinel-3 Surface Topography Mission (STM) constellation over inland waters is to provide accurate measurements of water surface height to support the monitoring of freshwater stocks for drinking water and agriculture, reduce risks for navigation, floods, droughts…etc. To ensure the missions requirements are met, ESA and EUMETSAT have defined the Mission Performance Centre (MPC), a consortium in charge of the qualification and the monitoring of the instrument performances.

In this paper we propose to present the results obtained in the frame of the MPC project focusing on the assessment of Sentinel-3 A and B altimeters performances over Inland Water. The SRAL (Sentinel Radar Altimeter) sensors on board Sentinel-3 missions differ from previous conventional pulse limited altimeters by providing observations with the Synthetic Aperture Radar mode (SARM). The onboard enhanced Digital Model Elevation, tuned for inland water monitoring, and allowing a precise Open Loop Tracking Command (OLTC) is also an essential asset to improve the precision of the measurements. The assessment of the SARM measurements performances will be assessed over a large set of lakes and rivers worldwide. It will include statistical analyses as well as cross comparisons between the two satellites and in situ gauges. The assessment of the OLTC mode compared to the conventional Close Loop mode will performed thanks to the tandem phase between Sentinel-3a and 3b where the altimeters recorded water surface height measurements with different acquisition modes.

Area B - Board 165

More than 25 years of observations have been amassed to date from a series of satellite altimeters with different designs and tracking systems, generating a wealth of information over the earth’s land surfaces including inland water. However, satellite radar altimetry has a substantially greater potential contribution to global river, wetland and soil moisture monitoring than is currently realised. From mapping historical flooding and paleo-hydrology to generating accurate height profiles along river courses and measuring surface soil moisture, these instruments provide a wealth of information. The objective of this global study is to examine river systems in their terrain context, to assess how a wider analysis of altimeter data can inform river system studies.

This work quantifies the extent to which the earth’s river systems can be monitored using current and past altimeters, drawing on a global database of over 86,000 quality graded river and lake time series from Ku band altimeters, including ERS2, Envisat, Jason-1, Jason-2 and TOPEX/Poseidon. All data were processed through an augmented system based on the ESA “River and Lake” system [1] and incorporating a variety of detailed regional river masks. These results are supplemented by the outcomes of CryoSat-2 target analyses over multiple river basins [2] and augmented by a first assessment of Sentinel-3 river monitoring capability. The outcome of this work is a series of river system “accessibility maps”, quantifying the extent to which river systems can be monitored by current and past altimeters. These maps are then evaluated in geographic context, using geodetic mission land height datasets [e.g. 3], study outcomes from Envisat 1800Hz data [e.g.4] and multi-mission backscatter analysis using detailed Dry Earth ModelS (DREAMS) [5]. Initially constructed over desert regions to derive surface soil moisture, DREAMS are now being extended to model river basins and surrounding areas.

Diverse river systems are used to illustrate the challenges of these complex targets for conventional and SAR mode altimeters, and assess the extent to which additional data from satellite altimetry can inform river and wetland studies. Presented case studies include the Congo, Nile, Okavango, Niger, Zambezi, Amazon and Orinoco basins.
The study results demonstrate that prior and current satellite altimeters already successfully monitor substantive parts of river systems. Waveform analysis is found to be the single most important factor in consistent height retrieval both over rivers and the surrounding terrain; over inland water, sub-waveform editing is beneficial and outcomes demonstrate that the increased along-track sampling offered by the new generation of SAR altimeters is very significant.

However, the unique and valuable contributions to river and wetland monitoring offered by satellite altimetry are not limited to measuring inland water heights. This study shows that these instruments collect a wealth of additional information over river basins. Waveform shape analysis offers additional insights into distribution of surface water, greatly enhanced with higher Pulse Repetition Frequency; backscatter data allows derivation of soil moisture estimates and surface texture and composition information, and geodetic missions provide spatially dense height profiles.

The wealth of data amassed by current and past altimeter missions and the varied capabilities of the next generation of satellite radar altimeters promise a bright future for river, wetland and soil moisture monitoring.


1 Salloway, M., Berry, P.A.M., Smith, R.G., Benveniste, J., 2012. Global Inland Water Monitoring in Near Real Time: Current Capabilities. Proceedings of the 20 Years of Progress in Radar Altimetry Symposium, Venice, 24–29 September 2012, ESA SP-710. http://dx.doi.org/10.5270/esa.sp-710.altimetry2012.
2 Berry, P.A.M. & Balmbra, R., 2016. Inland water height retrieval – from Cryosat2 to Sentinel3. Proc. ‘Living Planet Symposium 2016’, Prague, Czech Republic, 9–13 May 2016, ESA SP-740.
3 Berry, P.A.M., Smith, R., Salloway, M., Benveniste, J, 2012. Global Analysis of Envisat Burst Echoes over Inland Water, IEEE Trans. Geoscience & Remote Sensing V50 issue 5.
4 Smith, R.G. & Berry, P.A.M., 2011. Evaluation of the differences between the SRTM and satellite radar altimetry height measurements and the approach taken for the ACE2 GDEM in areas of large

Area B - Board 166

The past GRACE (Gravity Recovery And Climate Experiment) mission has clearly demonstrated over its 15 years’ mission lifetime its capability to monitor global mass redistribution. It has allowed us a deeper understanding of the interactions between several components of the complex Earth system. In particular, GRACE has quantified the mass variations due to the hydrological cycle at planetary scale. Due to its gravimetric nature, the GRACE-observable is even able to access deep ground water. Nonetheless, GRACE products are limited by their relatively low spatial (~300 km) and temporal resolutions (30 days) as well as by the precision and accuracy of the background models (e.g., atmospheric and oceanic de-aliasing products) used in the processing.

An on-going ESA-funded study (ADDCON: Additional Constellation & Scientific Analysis Studies of the Next Generation Gravity Mission Concept) investigates multi-pair GRACE-like satellite constellations and their potential applications in the Earth sciences. Among several other goals the project aims to make the step towards near-real time products, with a special focus on hydrological services and applications. In particular, it aims to contribute to gravimetric drought monitoring and to the characterization of flood potential.

We review the current ADDCON concept for the Next Generation Gravity Mission architecture and its gravity-based products. We discuss the potential for hydrological services, in particular for drought monitoring. It is explicitly understood that gravimetric drought is fundamentally different from conventional drought indices. Because drought is inherently defined as an anomaly relative to a normal situation, we critically assess how to define an appropriate climatology in the context of gravimetric drought.

We review the current ADDCON concept for the Next Generation Gravity Mission architecture and its gravity-based products. We discuss the potential for hydrological services, in particular for drought monitoring. It is explicitly understood that gravimetric drought is fundamentally different from conventional drought indices. Because drought is inherently defined as an anomaly relative to a normal situation, we critically assess how to define an appropriate climatology in the context of gravimetric drought.

Area B - Board 167

Terrestrial water storage (TWS) changes corresponding continuously changing distribution of ice, snow, soil moisture, groundwater, and surface water have an important role in the dynamic system of the Earth. In order to study the changes in TWSs over large basins, the Gravity Recovery and Climate Experiment (GRACE) satellite mission data as well as hydrological models can be used. In this study, we focus the results obtained from a hydrology model (Water – Global Assessment and Prognosis – WaterGAP 2.2b) (Müller Schmied, H. 2017) and the GRACE Level-2 gravity field product of German Research Centre for Geosciences – GFZ (Release 05) (Swenson, S.C. 2012) to estimate the TWS changes in the basins of Turkey over 2002–2014. The GRACE monthly gravity data in terms of Equivalent Water Thickness (EWT) and the WaterGAP monthly total water storage data are analyzed by a least-squares fit of the following equation:
TWS(t)=TWS₀+vt+∑Ak*cos(ωk*t+Φk)+ε(t)
where TWS(t) = the TWS time series, t = time, TWS₀ = the value of TWS when t = 0 (at the start time), v = the rate of TWS change (linear trend), k = 1 annual signal, k = 2 semi-annual signal, A = amplitude, ω = angular frequency, Φ = phase, and ε(t) = unmodelled residual term. For this, the 13 missing solutions in the GFZ Release 05 monthly time series from April 2002 to December 2014 have been interpolated linearly. Three variants of the WaterGAP model (CRU TS3.23, CRU TS3.23 + GPCCv7, and WFDEI+GPCCv7 until 2013 and WFDEI+CRU for 2014) based on different meteorological input data and processes have been assessed by considering the temperature and precipitation data in the study region, provided from Willmott and Matsuura (2001). The highest correlation has found with third meteorological forcing containing WFDEI, and thus this variant has been preferred in the study. For outlier detection, 3σ-rule has been used. The correlation between both data is 0.83 (Figure 1). The linear trends, annual amplitudes and phases of the TWS changes from both time series are mapped and compared. Analysis of the spatial patterns of TWS anomaly trends for both results indicates that a declining trend is dominant over a large part of Turkey. Figure 2 shows the spatial distribution of trends derived from the GRACE data. The mean trend is -3.8 ± 0.8 mm/year from the GFZ Release 05, and -2.9 ± 0.9 mm/year from the WaterGAP 2.2b model, which have a good agreement (Figure 3). The maximum annual amplitude of the TWS changes from GRACE occurs in March with almost one month earlier than that of the hydrology model. Generally speaking, the largest annual variation (about 9 cm) is seen in the southeastern part of Turkey. Our results show that the TWS change from 2002 to 2014 in the parts of Turkey along the Tigris-Euphrates river basins has the fastest rate in our study region. So, the results of this study support the findings of the NASA about on the decreasing of water storage in these river basins.

Area B - Board 168