Operational activities in the field of flood monitoring and prevention benefit from the availability of synthetic aperture radar (SAR) images. The main advantages of SAR data are synoptic views over wide areas, day and night acquisitions independent of weather conditions, as well as a reliable and high frequency data acquisition schedule. The Copernicus program, European Union's Earth observation (EO) programme, opens the door to disruptive innovation in the domain of floodwater monitoring and, more broadly, emergency management, due to its Sentinel-1 SAR mission’s capability to systematically, globally, and frequently acquire high quality EO data at 20 m spatial resolution with a revisit time of 2-3 days over Europe. In order to rapidly translate the large volume of SAR data into floodwater maps and value adding services, the European Commission’s Joint Research Centre (JRC) recently added Global Flood Monitoring (GFM) products based on Sentinel-1 as a new component to its Copernicus Emergency Management Service (CEMS). The GFM products are obtained by processing all incoming Sentinel-1 SAR images within 8 hours after data acquisition to systematically monitor flood conditions at global scope. While past analyses were limited to pre-identified flood images in the framework of CEMS, the current implementation processes all incoming images in a fully automatic way, thereby eliminating the time required for necessary human interventions. To reach this degree of automation, the system takes advantage of the constantly updated 20 m Sentinel-1 data cube made available by the Earth Observation Data Centre (EODC) facilities.

It is requisite that the Sentinel-1 based retrieval algorithm, as one of the core components of GFM, is both efficient and robust. Moreover, it is designed to balance two objectives: to detect water at high accuracy (i.e. permanent and seasonal water bodies, and floodwater), while minimizing the identification of false alarms due to water-look-alikes surfaces that can be confused with floodwater. To reach a high degree of robustness, an ensemble-based mapping algorithm is implemented, which combines three independent floodwater mapping algorithms driven by different approaches. 1) LIST’s algorithm that requires three main inputs: the most recent SAR scene to be processed, a previously recorded overlapping SAR scene acquired from the same orbit and the corresponding previously computed flood extent map. The change detection algorithm maps all increases and decreases of floodwater extent and makes use of this information to regularly update the flood extent maps. To do this, it uses a hierarchical split-based approach, region growing and an adaptive parametric thresholding. 2) DLR’s algorithm requires one scene as the main input and further exploits three ancillary raster datasets: i.e. a digital elevation model (DEM), areas not prone to flooding and a reference water map. To map flood extent, it makes use of non-parametric hierarchical tile-based thresholding, region growing and fuzzy logic. 3) TU Wien’s algorithm requires three input data sets: i.e. the SAR scene to be processed, a projected local incidence layer, and the corresponding parameters of a previously calibrated multitemporal harmonic model. Based on these inputs, the probability of a pixel belonging to the flood or non-flood class is defined.

The final floodwater map is obtained by integrating the results of the three independently developed algorithms. Pixelwise flood classifications are based on majority voting, such that at least two algorithms are in agreement. To contextualize the ensemble-based observed flood extent maps, the GFM system also provides a reference water mask derived from multi-temporal Sentinel-1 data. The combination of the reference water mask with the observed flood extent product results in the observed water extent.
The observed flood extent map is delivered with uncertainty values informing on the certitude of a pixel being classified as flooded. Moreover, an exclusion map identifies all areas where the detection of water using Sentinel-1 data is hampered by the presence of dense vegetation, urban areas, radar shadow regions, permanently low backscattering areas (e.g. sandy areas), and non-flood prone areas, i.e. those that have a Hight Above Nearest Drainage (HAND) value above 15 m. Finally, advisory flags are provided to make users aware of large-scale dryness and wet snow cover (both potential sources of over detection), or of wind (a major source of under detection). GFM is not only a system that systematically and fully automatically processes all images acquired by the Sentinel-1 mission in near real time, it also provides access to a global record of flood maps based on the processing of the entire Sentinel-1 collection since its start in 2015. This record provides valuable information to assess flood hazard and risk at 20 m resolution at a global scale. All these products are integrated in the Global Flood Awareness System (GloFAS), where end-users can visualize, analyze and download the data.

The algorithm is currently being extensively tested for different regions all over the world. A first quantitative evaluation shows encouraging results in relation to the accuracy for delineating the evolution of water bodies and further improvements to increase the accuracy of the GFM product is ongoing.

In this presentation the audience will be introduced the Worldwater project and the results of the Round Robin inter-comparison of inland surface water detection and monitoring algorithms using Sentinel-1, Sentinel-2 and Landsat 8 imagery.

Background
Water withdrawals globally have more than doubled since the 1960s due to growing demand – and showing no signs of slowing down. Population growth, socioeconomic development and urbanization are all contributing to increased water demand, while climate change induced impacts on precipitation patterns and temperature extremes further exacerbate water resource availability and predictability. In the future there will be an increasing need to balance the competing demands for water resources and have more efficient ways to manage water supply. The need for proper and timely information on water (non-) availability is probably the most important requirement for water management activities. In large, remote and inaccessible regions, in-situ monitoring of inland waters is sparse and hydrologic monitoring can benefit from information extracted from satellite earth observation (EO).

Earth Observation is an essential source of information, which can complement national data and support countries to collect regular information on the changes to their surface waters. Ever since the launch of the first earth observation satellites in the early 70ties the mapping and monitoring of surface water has been a subject attracting interest from researchers and practitioners in hydrology, environmental conservation, and water resource management. The field has gradually evolved and incentivized by the steady built up of long-term archives of global satellite data and the compute resources for analyzing those data. A significant breakthrough in the adoption of EO solutions has been the European Commission Joint Research Center’s Global Surface Water Explorer [JRC-GSWE] (Pekel et al., 2016) and the Global Land Analysis and Discovery group’s Global Surface Water Dynamics [GLAD-GSWD] (Pickens et al., 2020). Despite these developments and the large track record of related successful case studies on surface water mapping there is still a lack of clear, robust and efficient user-oriented methods and guidelines that allow for using Earth Observation data at scale and on an operational basis for surface water mapping and monitoring.

The WorldWater project
One of the main goals of the the WorldWater project is to contribute to the formulation of new best practices for mapping and monitoring surface water with Earth Observation (EO) data. A particular topic is to advance the monitoring of surface water extent dynamics by taken advantage of the new enhanced capabilities of the latest generation of open and free satellite data from the European Copernicus programme. For the first time in history, the Copernicus programme provides users with access to globally and systematically acquired Synthetic Aperture Radar (SAR) data. This is a major breakthrough which can contribute to more robust monitoring in environments challenged by frequent cloud cover and periods with limited light such as high latitudes. Yet, surface water mapping with SAR data is still complicated by a number of scientific challenges (e.g., topography, wind, low-backscatter surfaces other than water) and why the synergistic use of optical and SAR data emerges as an interesting alternative, with the potential to take advantage of the individual sensors’ strengths, while minimizing their weaknesses.

Lately there has been some promising studies showing the strength of such a sensor fused mapping approach, yet there has been no systematic evaluation against other leading approaches for surface water mapping. Therefore, the WorldWater project organised a Round Robin exercise aiming at the inter-comparison of EO algorithms for surface water detection, using the latest generation of free and open satellite data from Sentinel-1, Sentinel-2 and Landsat 8. In total the WorldWater Round Robin was joined by 15 organizations representing a mix of research institutions, private companies, government agencies and non-governmental organizations. All participants were asked to produce monthly maps of surface water presence at 10-meter spatial resolution for 2 consecutive years and over 5 challenging test sites. The outputs generated by the round robin participants across the test sites were evaluated individually and in cross-comparison using a harmonized independent reference data set.

The initial Round Robin results indicate that single sensor approaches can produce accurate and consistent water maps under ideal conditions, and yet across a range of challenging environments the synergistic usage of optical and SAR data delivers more accurate and consistent outputs. By comparing the robustness of the different algorithms, the Round Robin will help contribute to a better understanding of the pros and cons of EO approaches for mapping and monitoring the extent of inland open waters as well as shortfalls and areas of further research has been identified.

Dams are strategic tools for countries and their management of water resources. Within Space Climate Observatory (SCO) initiative and SWOT Downstream program, StockWater project aims to put in place a system for monitoring water volume in dams. It is on satellite data, and a specific processing system, thereby facilitating the work of the public authorities in this area.

Water resources monitoring, including surface and ground water, is a vital issue for governments and public institutions. Water resources are essential for society and economic activity (drinking water, irrigation, hydroelectricity, industry, flood control) and for natural and water ecosystems.

Generally, reservoirs stock information is collected and held by the local reservoir managers (public or private). Regional and national authorities might access this information with a certain latency, which depends on national water policies. Central authorities are then confronted to 2 issues: long latencies to retrieve water stock information and sparse or inexistent information about the small reservoirs.

The project proposes a global solution to monitor reservoirs stock volumes based on frequent satellite measurements. This solution is based on reservoirs area monitoring by imaging satellites (Sentinel 1&2) based in the Surfwater processing chain (Pena-Luque et al, 2021), which integrates a multitemporal approach to improve water masks. Furthermore, StockWater innovation relies on reservoir estimation of Area/Elevation/Volume relationships just from a DEM, even when acquired after the reservoir construction. Recent prototypes have been qualified on 29 reservoirs in France, ranging from 20 to 1600 hectares, yielding uncertainties below 15% on volume rates.

New versions are evaluated on Spain, France and India with different in situ datasets, and it will be extended to Burkina Faso, Laos and Tunisia with 250 reservoirs in next stages. This system will easily allow volume estimations from Elevation measurements (altimeters Jason, Sentinel3 with limited coverage or SWOT globally).

StockWater project , leaded by CNES and developed with CS-Group and SERTIT, holds a partnership initiative with CESBIO, GET, LISAH laboratories and their local partners in Tunisia, Laos, Burkina and India. StockWater is open to new countries willing to participate.

How to feed 10 billion people in 2050? Water is a critical input for agriculture and instrumental for achieving food security. But besides increasing food production, the agricultural sector also needs to use water resources more efficiently to respond to climate change and competition from other water use sectors. The challenge is thus not only to increase food production (in kilograms per hectare) but also to improve the water productivity (in kilograms per cubic meter) which is the production per unit of water consumed. For this reason the Dutch Development Cooperation (DGIS) identified the enhancing water productivity in the agricultural sector by 25% as a key policy priority.

In this context, and as part of its ‘Remote Sensing for Water Productivity’ programme, FAO developed and implemented the WaPOR portal with open-access satellite-based data on water productivity for Africa and the Near East. The FAO WaPOR database is updated every 10 days and supports national governments and other organisations to monitor and report on (agricultural) water productivity and to identify and mitigate water productivity gaps, which contributes to a sustainable increase of agricultural production.

Irrigation is key in achieving higher food production levels: only 20 percent of the agricultural land is irrigated, but irrigated agriculture contributes to 50 percent of the total food produced. At the same time is irrigation responsible for approximately 70% of all water abstractions. New methods and technologies to monitor irrigation performance and improve water productivity are required to sustainably manage irrigation water and decide on where to take action, and will also contribute to The Sustainable Development Goal 6.4 on improved water use efficiency.

The FAO WaPOR database is a cost-effective tool to support irrigation management and improvement. IHE Delft developed a diagnostic framework that applies WaPOR data to assess irrigation performance indicators (Chukalla et al., 2021; Safi et al, 2021). The irrigation performance framework helps to detect water productivity variations in irrigated agriculture and to set targets and actions for improvement. It is a standard procedure that can be used by practitioners to translate open-access remote sensing data to actionable information.

Under the WaterPIP project this framework has been incorporated in an open-access tool that allows users to assess irrigation performance for any area covered by the WaPOR database. The tool is built in Python code and will be available in a public Github repository by the end of 2021. The repository hosts tools to extract, interpret, analyse and visualize open-access geodata to improve water productivity.

The irrigation performance indicators incorporated build further on the work of IHE Delft and help to understand how agricultural systems are performing and their potential for improvement. The following irrigation performance indicators are currently included in the Github repository:
• Equity: The degree to which deliveries are considered fair by all.
• Adequacy: The ability of a system to reach targeted deliveries in terms of quantity (discharge and/or volume) service performance to the users.
• Reliability: The degree to which water delivery conforms to the prior expectations of users.
• Efficiency: System’s ability to minimize water losses due to oversupply.
• Productivity: Measures of the efficiency of production.

The open-access Github repository is the foundation for customized apps to be co-developed with local service providers in early 2022. These customized apps will address specific user needs. A potential customized app would address irrigation scheme managers that needs regular updates on irrigation performance to ensure all farmers have access to sufficient water throughout the growing season. Another customized app would support governments and donors to identify which areas are in need of irrigation improvement or irrigation rehabilitation and what would be the most critical problems to address.

The tool is currently applied for irrigation performance analytics for irrigation schemes in Sudan and Mali. Based on the area of interest provided, the tool automatically generated performance indicators on irrigation uniformity, equity, adequacy and land and water productivity. The results set a target for improvement and quantify the scope of improvement in terms of increase in yield and a simultaneous decrease in water consumption.

The increasing availability of open-access satellite data and derived data products and services makes satellite-based information accessible to a wider community. The Sentinel missions provide timely, continuous and independent satellite data on the land and how it is used. The FAO WaPOR database provides satellite-based products that are very suitable for irrigation performance assessment. The availability of open-access and automated tools for data interpretation such as the demonstrated irrigation performance indicator tool will further increase the uptake of satellite-based information outside the geospatial community.

[References
Chukalla, A. D., Mul, M. L., van der Zaag, P., van Halsema, G., Mubaya, E., Muchanga, E., den Besten, N., and Karimi, P.: A Framework for Irrigation Performance Assessment Using WaPOR data: The case of a Sugarcane Estate in Mozambique, Hydrol. Earth Syst. Sci. Discuss. [preprint], https://doi.org/10.5194/hess-2021-409, in review, 2021.
Safi, A.R., Karimi, P., Mul, M., Chukalla, A.D., de Fraiture, C., 2021. Translating open-source remote sensing data to crop water productivity improvement actions. Agric. Water Manag. (submitted)]

Consistent estimation of actual evapotranspiration (ET) from field to continental scales is critical for reliable monitoring and reporting of Sustainable Development Goal (SDG) indicator 6.4.1 (Change in Water Use Efficiency). With this in mind FAO is running an online portal called WaPOR, which provides access to ET estimates at different spatial scales derived from observations of Landsat, PROBA-V (replaced by Sentinel-2 since 2020) and Terra and Aqua satellites. The goal of ET4FAO project (https://et4fao.dhigroup.com/) was to demonstrate the use of Copernicus data, especially combined observations from Sentinel-2 (S2) and Sentinel-3 (S3) satellites, for consistent estimation of ET from 20 m to 300 m spatial resolutions. It showed that Copernicus data, together with advanced data fusion methods and physical models, greatly improves the accuracy of ET retrieval at scales in which Terra and Aqua (MODIS sensor) data was previously used.

The ET4FAO project also exposed an inherent limitation of the S2-S3 data fusion approach, which is used to derive high-resolution land surface temperature (LST) by enhancing spatial resolution of S3 LST observations (acquired at around 1 km resolution) using S2 optical observations (acquired at 20 m spatial resolution). This is the inability to reproduce very low LST values in heavily irrigated fields, thus leading to underestimation of ET from those fields. The reason for this is the difficulty in predicting values which are far beyond the LST range of the original S3 Sea and Land Surface Thermal Radiometer (SLSTR) LST. The most extreme LST values will not be present in the SLSTR data since LST from different landscape features aggregate within the 1 km pixels. The underestimation of ET in heavily irrigated fields could represent a critical issue when it comes to SDG indicator 6.4.1 reporting since irrigated agriculture is a focus area of this indicator.

In this follow-up study we aim to improve the ability of S2-S3 data fusion approach to reproduce low LST values present in irrigated fields by incorporating Landsat LST into the data fusion methodology. Based on comparisons performed in ET4FAO project, Landsat LST with a spatial resolution of around 100 m is sufficient to capture those low temperatures. It is however important to ensure that the inclusion of Landsat LST does not compromise the capability of the data fusion approach to produce high-resolution LST on all dates on which cloud-free S3 observations were acquired. In addition, energy must be conserved when re-aggregating sharpened LST back to S3 scale to ensure physical plausibility and consistency in ET retrieval across spatial scales. Preliminary investigation has shown that this is possible. The developed method should be directly applicable with future observations from S2, S3 and Copernicus high-priority candidate Land Surface Temperature Monitoring mission.

Additionally, in ET4FAO the validation of Copernicus-based ET was focused on sites in semi-arid Mediterranean climate. However, the FAO WaPOR portal is operational across Africa and Middle East. Therefore, we extended the validation and evaluation of Copernicus-based ET beyond Lebanon, Tunisia and southern Spain also into tropical and Sahelian regions. Those new sites were added for validation where field measurements are available and for evaluation where WaPOR Level 3 (30 m) data is being produced. In addition, field measurements of LST from southern Spain from 2018 and 2019 will be used for directly validating high-resolution LST produced through data fusion.

Agriculture, through growth-related evapotranspiration, is the largest water user on the Globe.
The Sustainable Development Goals demand both yield and agricultural water use efficiency (AWUE in kg fruit per m³ of evapotranspiration) to be maximized through sustainable farming practices. Monitoring systems have to be established, which document hot and cold spots of AWUE as well as the progress towards reaching the SDGs.

Remote sensing and specifically the Sentinel 1 and 2 Sensors create independent, homogeneous data sets on global agriculture but are not by themselves capable of determining AWUE and yield. Therefore we established a global high resolution monitoring system for AWUE and yield, which is based on coupling the high-resolution Sentinel-derived agricultural data streams with the sophisticated, dynamic agricultural crop growth model PROMET. PROMET is driven globally by regionalized meteorological inputs and produces, for each location on the Globe, a wide variety of crop growth trajectories based on different farm management, which range from extensive, low-fertilizer to intensive, mechanized high fertilizer practices including irrigation. In each location each selected farming practice creates a different course of simulated LAI for each crop that can grow there. Finally, the course of the LAI measured by Sentinel 2 in each location is compared with all simulated courses to determine the likely actual crop type and farming practice at the given location. The complete set of modelled results from PROMET of the selected farming practice further allows to infer actual AWUE and yield.

A sample of 120 Sentinel-2 tiles were randomly selected ensuring representativeness for all global agro-ecological zones. For these tiles the procedure described above was carried out to determine crop-specific courses of LAI from Sentinel-2 time series on the one side and to simulate corresponding LAI-courses together with AWUE and yield for a an ensemble of farming practices.
The results from the tiles were then upscaled to cover the global agriculture. Based on an example tile (33UUS) in Saxony, Germany, the process of acquiring the crop specific LAI time series as well as the calculation of AWUE and yield will be presented. Furthermore, results for globally distributed tiles will be shown.

The Sentinel-2 satellite imagery is processed using proprietary VISTA in-house software. This comprises sophisticated methods on cloud and cloud shadow detection, atmospheric correction and derivation of plant physiological parameters with the surface reflectance model SLC (Soil-Leaf-Canopy).

Since the LAI-courses to be determined from Sentinel-2 are crop specific, a crop type classification based on spectral characteristics and leaf area index (LAI) time series is carried out. Only pixels, which are classified with a high degree of accuracy as a considered crop type, are selected for further processing. The classified crop types are the main crops of the studied Sentinel-2 tile. Furthermore, the usage of the COPERNICUS High Resolution Layers (HRL) allows working only on agriculturally used lands.

A Sentinel-2 pixel has to fulfill a set of requirements to be selected. Beyond carrying a relevant crop type, its position should be sufficiently distant from the field boundary to guarantee homogeneity, and a continuous coverage of cloud free Sentinel-2 overpasses should be available to create a time series of the LAI. The amount of selected pixels is linked to the areal size, or, if available, to the crop specific agricultural area of regarded districts or counties.

Through a crop specific lookup-table inversion of the SLC model, exact plant physiological parameters (e.g. LAI) for the complete time series of satellite imagery are derived for the selected pixels. During processing sophisticated statistical methods are applied to identify and eliminate outliers. The result consists of harmonized, daily LAI values. The process of LAI derivation is schematically shown in Fig.1.

Furthermore extensive global simulations of AWUE and yield were carried out for < 200 combinations of crops and farming practices using the SuperMUC NG High Performance Computing (HPC) Facility of the Bavarian Academy of Science. A comparison of remote-sensing derived and simulated LAI-courses then enables to identify the most probable actual farming practice on the selected field. The results, which were exemplary achieved for tile Saxony (33UUS) for 2017 for maize, are shown in Fig.2. Fig.2 shows the selected fields and an exemplary comparison of simulated (grey, selected green) and observed (orange) LAI-courses for a pixel in the NE of the tile. The comparison of simulated farming practices and Sentinel-observations for maize revealed medium to high fertilization rates, an average maize yield across the tile of 9.61 t/ha (statistics 2017: 9,65 t/ha) and a high average AWUE of 2.27 kg/m³.

The global selection of Sentinel-2 tiles representative for regional agriculture allows to upscale our approach of determining yields and AWUE to the global level. Fig.3 shows, exemplary for maize and selected Sentinel-2 tiles, the graphs of the complete simulation results for all farming practices, showing yield (t/ha) as a function of water evaporated (m³) by the maize plants. Different combinations of rainfall, fertilizer application and irrigation result in different positions of each simulated scenario in the graph, which are represented by the black dots in each graph. The dots generally follow a saturation curve, which means that simulated maize yields in each tile approach region specific maxima at the expense of more water used. Since the saturation curves in general are not linear AWUE increase with increasing simulated yields. The LAI-curves, which were measured by Sentinel-2 correspond to the green dots in the graphs, which likely represent farming practices, which were realized by the farmers under the local environmental conditions. In graph USA one can see that the green dots represent high yields and relatively low water consumption (large actual AWUE) whereas in Ethiopia the green dots represent low yields and relatively high water consumption (small actual AWUE).

Altogether, the Sentinel-2 derived information on the dynamic LAI development during the growth of maize together with massive simulations of plant growth in the selected tiles (located at the small red in Fig.2) enables to, for the first time, create a very differentiated picture of the actual AWUE of maize around the Globe. This allows to identify global hot-spots of water waste in agricultural production e.g. in Ethiopia but also in Kenia and Zambia. Enabling to realize, in a sustainable manner, the existing potentials to improve both yield and AWUE by changing farming practices (e.g. introduce fertilizer and/or irrigation). The described procedure uses massive remote sensing information (app. 18000 Sentinel-2 scenes for the processed 120 tiles per year) and can easily be repeated on an annual basis within an operational, Sentinel-2-based AWUE monitoring system. The results will be uploaded and found on the Food Security TEP(https://foodsecurity-tep.net/).