Data Application of the Month: River Flooding

Figure 2: scattering mechanisms of grass, vegetation and built-up areas under flood and non-flooded conditions  (Schumann et al., 2018).

When flooded, the reflection becomes specular (assuming the water surface is smooth). In forested areas volume scattering dominates short wavelengths and signals only reach the canopy. Longer wavelengths experience both volume scattering (from the branches of the trees) and double-bounce (penetration through the canopy causes signals to reach the surface and bounce on tree trunks). However, when inundated, double-bounce backscatter becomes more dominant. In urban landscapes, under both flooded and non-flooded conditions, double-bounce backscatter mechanisms dominate due to the presence of many perpendicular structures. However, if sensed with short wavelengths, the backscatter is stronger when there is an inundation than under non-flooded conditions suggesting that short wavelengths can better detect flooded areas in urban areas. Longer wavelengths return the same backscatter under both flooded and non-flooded conditions (Schumann et al., 2018). 

HH polarisation is the preferred polarization for flood detection and extent mapping because it is less sensitive to minor vertical differences on the water surface caused by waves. VH is also widely suggested for flood mapping, since it is more sensitive to changes on the land surface, while VV is often the least advised because it is rather susceptible to vertical structures(Rahman & Thakur, 2018).

River flood mapping combining optical and radar imagery

In general, three combinations of satellite imagery are used for flood inundation detection, respectively applied to the periods before and during the flood: (i) optical imagery / optical imagery, (ii) optical imagery / SAR imagery and (iii) SAR imagery / SAR imagery. The second combination is used as a multi-temporal and multi-source data approach for flood progression and flood detection analysis. The objective is to use radar images to accurately identify large-scale water area during floods and utilise optical imagery for mapping permanent water areas. Comparing individual results differentiates between permanent water bodies and flooded areas (Tong et al., 2018). 

Table 1: List of satellite sensors that are typically used for river flood mapping and monitoring.

Web-based tools and data sources available for mapping and monitoring river floods

1. The Global Flood Detection System

The Global Flood Detection System (GFDS, http://www.gdacs.org/flooddetection) uses passive microwave remote sensing to monitor and observe floods worldwide daily. Unlike conventional in situ gauging stations located in remote areas, it acts as an effective substitute and provides systematic detection of riverine flooding by using microwave radiation to differentiate between dry pixels (non-flooded areas) and wet pixels (flooded areas). GDFS uses the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E), and deviations in surface water over large extents (10 × 10 km) can be observed as it affects the emitted microwave signal, which allows the system to flag flooded areas when surface water significantly surges. This system also provides precise time series of flood surface variations, near-real-time flood maps and animations in over 10,000 monitoring areas (Schumann et al., 2018). Because of its importance, GFDS is being integrated within operations of flood forecasting centres in some countries (Brazil, Namibia, and Haiti) and used by some international aid organizations. 

2. The Global Flood Awareness System

GloFAS is a web-based global hydrological forecasting and monitoring system jointly owned by the European Commission and the European Center for Medium-Range Weather Forecasts (ECMWF). The system incorporates contemporary weather forecasts with a hydrological model and with its continental scale set-up provides downstream countries with information on upstream river conditions as well as continental and global overviews. GloFAS comprises two complementary systems; i) GloFAS 30-day - a daily hydrological forecast which provides a quick overview of upcoming flood events for the next 30 days, and ii) GloFAS Seasonal - a monthly hydrological forecast which provides river flow outlooks highlighting unusually high or low river flow up to 16 weeks ahead. GloFAS forecasts of possible river flood episodes and unusually high/low flows for all major rivers of the world are displayed through its web platform, the GloFAS web viewer.

3. The Dartmouth Flood Observatory (DFO)

The Dartmouth Flood Observatory (DFO, http://floodobservatory.colorado.edu) seeks the acquisition and preservation of digital map records of the ever-changing Earth’s characteristics especially related to floods, droughts, wetlands, shorelines, lakes and reservoirs. DFO has tracked, monitored and archived flood events on a global scale since 1985, offering long historic data of flood disasters. Data is acquired through remote sensing, analysed and made open source in different formats (GIS maps, spreadsheets, graphics). In developing flood maps, presently all sensors available and useful are utilized. DFO primarily utilizes MODIS, however ESA's Sentinel-1A and -1B, Landsat, Sentinel-2 are also used. Radarsat, Cosmo SkyMed, EO-1 products have also been provided for aggregation that assist flood response teams through situational awareness across large scale coverages. In addition to these provided services, DFO, in collaboration with NASA, has ensured the provision and distribution of near-real-time (NRT) flood maps whose role in disaster response decisions cannot be overemphasized (Schumann et al., 2018). The DFO “River Watch” represents a fully automated platform where river discharges are measured and have a period-of-record from 1998 to the present (http://floodobservatory.colorado.edu/CriticalAreas/DischargeAccess.html).

4. The Global Flood Monitoring System

The GFMS is a NASA-funded experimental system at the University of Maryland (UMD’s GFMS http://flood.umd.edu) that uses real-time TRMM Multi-satellite Precipitation Analysis (TMPA) and Global Precipitation Measurement (GPM) Integrated Multi-Satellite Retrievals for GPM (IMERG) precipitation information as input for real-time quasi-global (50°N - 50°S) hydrological calculations at 0.125 degrees for river networks and 1km resolution for streamflow, surface water storage and inundation variables. While the GFMS has some shortcomings (low resolution, inaccuracy of data), the timely delivery of system outputs and its lead-time capability makes it invaluable to flood relief services and flood disaster response organizations (Schumann et al., 2018). 

5. Automated SAR Flood Mapping with ESA’s G-POD

ESA’s G-POD (https://gpod.eo.esa.int/) provides a free registration-based service where SAR data (flood images) can be analyzed by registered users to generate flood maps. The system uses an algorithm that delineates flooded areas and non-flood areas based on the backscatter intensity of the image (open surface water has lower backscatter values as opposed to other land characteristics) (Schumann et al., 2018). The system uses a thresholding method (all pixels of an intensity image whose backscattering coefficient (σ0) is smaller than a given threshold value are classified as flooded) which is computationally inexpensive and suitable for rapid mapping purposes (Pulvirenti et al., 2011).

6. Rapid Flood Mapping from NOAA’s VIIRS Sensor

The Visible Infrared Imaging Radiometer Suite (VIIRS) instrument is one of the five major EO instruments onboard NOAA’s Suomi-NPP and JPSS satellites, a continuation of NOAA’s AVHRR legacy sensors. With a very large swath width of 3060 km, it provides full daily coverage of the Earth both during the day and at night. Consequently, it can be utilized for large scale river floods. Optionally, the low-resolution flood map derived from VIIRS can be downscaled to a higher spatial resolution using a DEM. 

7. Aqueduct Global Flood Products

The Aqueduct flood products is a collection of products that analyse past, present and future extreme flood events and their impacts. The Aqueduct Global Flood Analyzer is a web-based platform which measures river flood impacts by urban damage, affected GDP and affected population at the country, state, and river basin scale across the globe. The Aqueduct Floods Hazard Maps measures river and coastal floods risks under both current baseline conditions and future projections in 2030, 2050 and 2080. River flood risk is estimated with the Aqueduct Global Flood Risk Maps. This online platform provides current and future river flood risk estimates in urban damage, affected GDP and affected population by country, state and river basin. Finally, the Aqueduct Global Flood Risk Country Ranking ranks 163 countries by their current annual average population affected by river floods. 

8. SERVIR-Mekong’s Surface Water Mapping Tool

This web-based mapping tool was developed by SERVIR-Mekong, Technical University Delft, the Asian Disaster Preparedness Center (ADPC), eartH2Obsrve, Google, Stockholm Environmental Institute and Spatial Informatics Group covering the lower Mekong geographical region. This region comprises Cambodia, Laos, Myanmar, Thailand and Vietnam. The tool is used to map spatio-temporal changes in surface water distribution and provides insight into the ecological structure and function of rivers, patterns of flooding, and flood risks. Initially developed to document the historical dynamics of seasonal flooding cycles on the Mekong River, other uses include flood risk assessment for disaster preparedness, identifying areas of permanent water etc.

9. Flood Simulator for African Basins

This was developed by SERVIR-Eastern and Southern Africa in collaboration with the Famine Early Warning Systems Network (FEWS NET) and is used to estimate the extent of river flooding in Kenya, Uganda, Namibia and Rwanda. the tools is available both as a desktop console tool and as an online tool. The application combines CREST (Coupled Routing and Excess Storage) hydrologic model with digital elevation maps and sends users an email alert with real-time and short-term forecast flood inundation maps of select stream-gauge locations.

10. Historical Flood Analysis Tools for the Lower Mekong Region

The historical flood analysis tool is designed to provide the information regarding flood prone areas (e.g. frequency of seasonal flooding cycles) in the Lower Mekong Region. This tool provides essential information for river flood disaster preparedness and flood disaster management, especially in the context of advanced relief resource provision. With this tool, variable risk for floods can be found by identifying areas prone to such disasters.

11. HYDrologic Remote Sensing Analysis for Floods Viewer (HYDRAFloodsViewer) 

HYDRAFloods is an open source Python application for downloading, processing, and delivering flood maps derived from remote sensing data.  This tool leverages data acquired by multiple satellite platforms to automate the creation of daily flood maps. The HYDRAFloods tool is the backend behind the web based HYDRAViewer Geospatial tool which provides daily (sometimes twice daily) flood maps.

12. High-Impact Weather Assessment Toolkit (HIWAT) Streamflow Prediction Tool 

This web-based tool was developed for steamflow forecasting, flood mapping, and data sharing that is openly accessed by a variety of stakeholders in the Hindu-Kush-Himalaya (HKH) region. The project uses advanced modeling, mapping, and visualization to make results intuitive and accessible for decision support. The following water resources applications are included:(i) Streamflow Prediction Tool - A medium-range, 15-day stream flow forecasting and flood awareness web app for the entire SERVIR-Himalaya region based on global ECMWF forecasts. (ii) Flood Map Viewer – A tool supporting a higher resolution forecast that includes inundation maps (for watersheds of interest) derived from the 15-day forecasts. (iii) Earth Observation Data Explorer – An Earth observation data explorer web app that leverages and is patterned after and can extend SERVIR ClimateSERV Viewer functionality. (iv) Water Observations Data Integrator – A customized version of the CUAHSI (Consortium of Universities for the Advancement of Hydrologic Science, Inc.) HydroServer web app to support managing, hosting, and discovering observational or modelled time series data with accepted World Meteorological Organization (WMO) and Open Geospatial Consortium (OGC) standards.

13. ESA Hydrology Thematic Exploitation Platform (TEP)

This is a web-based platform used to access, process, upload, visualise, manipulate and compare hydrological data. It offers a number of hydrological service products for flood monitoring, hydrological modelling, water quality, water level, small water body mapping, arctic inland water monitoring and forecasting, FANFAR (water-related applications for West Africa), and Water Observation and Information System (WOIS). Although some service features are open access, it also includes a paid self-managed processing option for expert users excluding FANFAR which is entirely free.

Limitations

• Seasonal changes account for large-scale changes in river extent in many areas of the Earth leading to complexities in correctly isolating unusual changes from seasonal events.
• The spectral signature of river flooding is complex. Floods affect tropical riverine forests, agricultural lands, desert, steppe landscapes as well as urban areas. The spectral variability of these landscapes under flooded and non-flooded conditions poses difficult challenges to flood detection, mapping and monitoring. 
• The unavailability of open access VHR satellite imagery hinders the use of EO data for mapping and monitoring flooded conditions of small rivers because river reaches tend to be monitored at a much smaller scale than that typically acquired with wide-swath EO imagery. Imagery from coarse and medium resolution satellites is highly generalised and more suitable for detecting and monitoring large scale flood events. 
• Although SAR data are more suitable for river flood, optical sensor data are more cost-effective, visually interpretable and have an extensive historical archive. In developing countries, revisit frequency for radar images is between 6-12 days. Hence, images often provide fragmented moments and not the full evolution of flood events. 
• In cases where urban areas are affected by river floods, it can be difficult to determine flooded areas using SAR. First, urban landscapes consist of strong scatterers (concrete, steel and building concerns) which impedes the detection of floods. Also, radar shadow, smooth surfaces and steep incidence angles are often confused with floods because shadow areas yield no backscatter as do smooth water surfaces (Rubinato et al., 2019). 
• Adequate utilization of the quantity of available EO data requires new and more powerful computational tools that are often not available due to financial and capacity constraints, especially, in the developing world. There is limited time and personnel available during disasters and emergencies to understand and process geospatial data into meaningful products. Other issues include cumbersome production and data distribution due to limited near-real-time data accessibility, bandwidth and sharing capacity. 
• Many of the web-based service interfaces are not intuitive and require personnel with geospatial skill sets to run specific queries for complex datasets from multiple data sources which can be time-consuming for emergency responders during flood disaster response. 
• Stakeholders often require at least daily status updates on affected regions and at spatial resolutions detailed enough to assess local infrastructure assets at risk. However, nontrivial issues, such as inadequate satellite revisit times, adverse weather conditions (e.g., cloud cover, fog, or heavy precipitation) and densely vegetated and built-up areas, typically hinder analysts to meet end-user expectations.