Floods

In this Story Map we will show how Synthetic Aperture Radar (SAR) images can be used to map floods that occur as part of ecosystem dynamics or disaster events. The capabilities presented here are focused on L-band radars such as NISAR (NASA-ISRO SAR Mission).

In populated areas, flooding can result from heavy rainfall and storm surge associated with hurricanes and typhoons. Flood extent maps provide critical guidance to first responders following extreme weather events. In some ecosystems, flood is an integral part of seasonal dynamics that animals and plants must adapt to. In both cases, researchers need to accurately map temporal changes in surface water. This includes water under low vegetation, water in urban areas, water under tree canopies, and open water not associated with vegetation (See  NISAR White Paper ).

SAR sensors send microwave pulses and record the strength of the signal reflected off the ground and back to the sensor, also referred to as the backscatter. By convention, SAR images are displayed in grayscale with bright areas indicating stronger detected signal or backscatter. The examples below show dry and flooded sites observed on a single date, as well as changes that can be observed when comparing images before and after a flood event. In general, radar brightness is highest over flooded forests and tall vegetation, while short vegetation and open water will appear dark.

Hurricane Florence was a category 4 hurricane (August 31 - September 18, 2018) that caused significant freshwater and storm surge flooding along the coast of North and South Carolina. From September 17 through September 23, UAVSAR (NASA JPL's L-Band airborne radar and NISAR testbed instrument) flew five times over the Pee Dee River in South Carolina to collect of flooding from the hurricane. 

Swipe below to see SAR acquisitions from during and after Hurricane Florence along the Pee Dee River in South Carolina. In the left image captured on September 17th, there is already significant inundation along the Pee Dee River, but the right image from September 23rd captures further rapid expansion of inundated vegetation (bright) and open water (dark).

Below is a subset of the Pee Dee River in an area of rapidly expanding flooding extent. The first image is an optical acquisition from Google Earth. A portion of the Pee Dee River is in the lower left corner. Moving left, this subset is composed of forested areas and agricultural fields with shorter vegetation. Notice the flood advancing between the September 17th to 23rd SAR images. The ability to detect changes in flooding extent is relevant for rapid response in disaster scenarios, particularly near inhabited areas.

In the next image, see if you can find inundation based on brightness. Swipe through to compare SAR (left) and Optical (right) images over the Napo River, a tributary to the Amazon River in Ecuador and Peru. Notice the bright features in the SAR image. These are areas of inundated forest below the canopy not visible in optical acquisitions.

Synthetic Aperture Radars (SARs) transmit microwave pulses and receive the echoes that backscatter from the Earth's surface. The amount of signal that is backscattered depends upon the properties of the objects on the surface. This backscatter information can be transformed into high-resolution radar remote-sensing products with detailed information about surface characteristics. 

Radar can penetrate cloud cover and operate day or night, which can prove particularly useful during flooding events from storms when there is persistent cloud cover. With all these capabilities, radar can be leveraged for flood mapping and detection in both vegetated and non-vegetated areas. 

Radar is highly sensitive to the dielectric properties of water. As an active technology, long radar wavelengths can penetrate the vegetation canopy and detect water on the ground underneath the vegetation. Radar can also detect areas of open water because most of the energy is scattered in the forward direction from the water's smoother surface, so radar returns from open water are ‘radar dark’.  However, when there is water below vegetation, the pulses can be reflected back towards the receiver by the vegetation, bouncing twice, i.e., from water and vegetation.  In fact, backscatter from water below trees can be very bright.

NISAR (NASA-ISRO SAR) is an upcoming radar instrument collecting data in the L-band wavelength (23.8 cm). The longer L-band wavelength can penetrate further into vegetation compared to other common SAR sensors measuring in C-band. NISAR will provide high-resolution, near-global, and continual mapping of Earth's natural resources and hazards, such as flooding.

Explore the NISAR Sample Data Product Suite, or planned NISAR data products and their product specifications, here:  https://nisar.jpl.nasa.gov/data/sample-data/ 

Radar data is collected at different polarizations, or combinations of horizontal and vertical radar waves (i.e. HH, HV, etc.). For example, the images shown in this Story Map are HH, which denotes a horizontally emitted, horizontally collected wave. Each polarization is a separate data product generated by the radar, and the different polarizations have unique interactions with the ground cover.

The expected scattering mechanisms in a flooded area are described below: 

• Specular / Surface Scattering: occurs when there is a smooth surface, such as open water. In specular scattering, the signal scatters away from the radar, and as a result, open water appears dark in the image. 
• Rough Scattering: can occur where some, but not all, of the signal scatters away from the radar. This could be due to wind roughening on open water or areas of shorter vegetation.
• Double Bounce Scattering: occurs when most of the signal is returned to the sensor, and these areas appear very bright in the image. This can occur in areas of flooded vegetation due to the interaction pulse scattering from both the smooth water surface and vegetation that sticks out of the water.
• Volume Scattering: is most common in vegetated areas as the signal interacts with all components (leaves, branches, trunk, etc.) of the canopy before returning to the sensor.

For flood applications, the HH channel from NISAR will be the most relevant due to stronger backscatter returns from open water, compared to HV, which is more suitable to detect vegetation. HH has higher sensitivity to surface and double bounce scattering, while HV is more suitable for volume scattering detection.

A model is usually employed to infer inundation classes from a backscatter image. The simplest approach is to identify a range of backscatter values associated with different classes. We expect the lowest backscatter values to be associated with open water, but the threshold will vary depending on the study site being considered. The example below from Martinez and Le Toan (2007) shows a classification for the Amazon forest.

The map below made by North Carolina researchers (Salem and Hashemi-Beni 2022), shows how multi-temporal images can be used to classify permanent water (in blue) and flood water (in yellow) in areas impacted by Hurricane Florence.

(2022). ARSET - Disaster Assessment Using Synthetic Aperture Radar. NASA Applied Remote Sensing Training Program (ARSET).  http://appliedsciences.nasa.gov/join-mission/training/english/arset-disaster-assessment-using-synthetic-aperture-radar 

Chapman, B., McDonald, K., Shimada, M., Rosenqvist, A., Schroeder, R., & Hess, L. (2015). Mapping regional inundation with Spaceborne L-band SAR. Remote Sensing, 7(5), 5440–5470. https://doi.org/10.3390/rs70505440 

Kundu, S., Lakshmi, V., & Torres, R. (2022). Flood depth estimation during Hurricane Harvey using Sentinel-1 and UAVSAR data. Remote Sensing, 14(6), 1450. https://doi.org/10.3390/rs14061450 

Martinez, J., & Le Toan, T. (2007). Mapping of flood dynamics and spatial distribution of vegetation in the Amazon floodplain using multitemporal SAR Data. Remote Sensing of Environment, 108(3), 209–223. https://doi.org/10.1016/j.rse.2006.11.012 

Melancon, A. M., Molthan, A. L., Griffin, R. E., Mecikalski, J. R., Schultz, L. A., & Bell, J. R. (2021). Random Forest Classification of inundation following Hurricane Florence (2018) via L-band Synthetic Aperture Radar and ancillary datasets. Remote Sensing, 13(24), 5098. https://doi.org/10.3390/rs13245098 

NISAR: The NASA-ISRO SAR Mission. (2017). Timely Maps of Flooding [White paper]. NASA.  https://nisar.jpl.nasa.gov/documents/5/NISAR_Applications_Floods.pdf 

Flores-Anderson, A. I., Herndon, K. E., Thapa, R. B., & and Cherrington, E. (2019) The SAR handbook: comprehensive methodologies for forest monitoring and biomass estimation. No. MSFC-E-DAA-TN67454. https://doi.org/10.25966/nr2c-s697

Salem, A., & Hashemi-Beni, L. (2022). Inundated vegetation mapping using SAR data: A comparison of polarization configurations of UAVSAR L-Band and Sentinel C-Band. Remote Sensing, 14(24), 6374. https://doi.org/10.3390/rs14246374 

Wang, C., Pavelsky, T. M., Yao, F., Yang, X., Zhang, S., Chapman, B., Song, C., Sebastian, A., Frizzelle, B., Frankenberg, E., & Clinton, N. (2022). Flood extent mapping during Hurricane Florence with repeat‐pass l‐band uavsar images. Water Resources Research, 58(3). https://doi.org/10.1029/2021wr030606 

UAVSAR data courtesy NASA/JPL-Caltech