A Near-Real-Time Flood Mapping Chain using Synthetic Aperture Radar - - PowerPoint PPT Presentation

A Near-Real-Time Flood Mapping Chain using Synthetic Aperture Radar Imageries

Qing Yang, Guangxi University, University of Connecticut, Xinyi Shen and Emmanouil Anagnostou, University of Connecticut Xi Chen, Peking University Albert Kettner, and Robert Brakenridge, University of Colorado, Boulder Jack Eggleston, Hydrological Remote Sensing Branch, USGS

SAR Data Frequency During Recent Events

• Sentinel-1 （2014）

– 8 days interval – 2-3 days potential – 10 m spatial resolution

• NISAR (2020)

– 4 times per month – 5-10 m

Sentinel Revisits

• Aug. 27-Sept. 10,2017

Sentinel Revisits

• Sept. 10-Nov. 2, 2017

Shen, et al., (2019). Remote Sensing

The RAdar Produced Inundation Diary (RAPID)-System Overview

Flood- Retrieval Trigger SAR data Query Retrieval Algorithm Execution

The Flood Trigger

• Daily Flood status of

~4000 stations

– USGS Water Watch

• Drainage regions

– Watershed algorithm-

• Potential Flooded Area

– Flooded-unflooded

A Binary classification B Morphologic processing

C

Multi-threshold compensation D Machine learning-based refinement

Overview of the RAdar Produced Inundation Diary(RAPID)

Shen, et al., (2019). Remote Sensing of Environment

Theoretical PDF

20 40 60 80 100 10 20 30 40 50 60 70 80 90 100

Real PDF

20 40 60 80 100 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 x 10

𝑞 𝐽1, 𝐽2 → ℎ 𝐽1, 𝐽2 ∆𝐽1∆𝐽2

• Water source tracing

(WST)

• Improved change

detection (ICD)

Input Datasets

Dataset Name Source/Type Producer Time span Coverage Spatial Res. Revisiting intervals Sentinel-1 SAR ESA Since 2014 Global 3.5/10 m 12 days (after 2013 Oct) 6 days (after 2015 Oct) NLCD Landsat/LCC USGS 1992-2011 US 30 m 5 years GLR-FROM Landsat/LCC Tsinghua Unvi. 2010, 2015 Global 30 m 5 years Water Occurance Landsat/water probability Pekel et al. 1984-2015 Global 30 m static Hydrography NHD-HR USGS N/A US 30 m static DEM STRM USGS N/A Global 10 m US 30 m global static NARWidth TM/River Width George Allen et al. N/A North America 30 m static GWD-LR STRM/River Width & Hydrograph Dai Yamazaki et al. N/A Global 90 m static HydroLakes STRM WWF N/A Global 90 m static US-Detailed Stream Body Survey USGS/USEPA/E SRI N/A US 10 m static

Case study-Typhoon Nepartak, July 17, 2016 Yangtze River, China

Case study-Oct. 10, 2017, Vietnam

Case Study- Hurricane Harvey, Aug. 27-Sept. 10, 2017

Confusion Matrix Reference User accuracy Wet Dry Retrieval Wet 11.09% 3.29% 77% Dry 3.73% 81.90% Producer accuracy 75% 93% RAPID Map DFO Comprehensive Map

https://floodobservatory.colorado.edu/E vents/2017USA4510/2017USA4510.html

• Aug. 30, 2017
• Aug. 27-Sept. 10, 2017

Case Study- Hurricane Florence

• Sept. 18, 2018

Case Study- Hurricane Florence

• Sept. 19, 2018

2019 Midwestern U.S. floods

Summary

• An automated open-flood

mapping chain has been mapped

• Quantitative validation shows

satisfactory

• It has been tested over many

events

Next

• Synergizing active/passive

microwave data

• Wetlands/Vegetation flood

mapping by SAR

• Base flow estimation

Combining the Active and Passive Inundation Mapping

Flood Scan M/C

April 16, 2008

Inundation mapping beneath vegetation

• Backscatter enhancement-4a & 4b

𝑇𝐼𝐼 2 𝑇𝐼𝐼𝑇𝑊𝑊

∗

2 𝑇𝐼𝑊 2 𝑇𝑊𝑊𝑇𝐼𝐼

∗

𝑇𝑊𝑊 2 = 𝑔

𝑇 𝛾 2 + 𝑔 𝐸 𝛽 2 + 3

8 𝑔

𝑊

𝑔

𝑇𝛾 + 𝑔 𝐸𝛽 + 𝑔 𝑊

8 2 8 𝑔

𝑊

𝑔

𝑇𝛾∗ + 𝑔 𝐸𝛽∗ + 𝑔 𝑊

8 𝑔

𝑇 + 𝑔 𝐸 + 3

8 𝑔

𝑊

• Freeman-Durden 3-component Model

Open Water Dry Lands Wetlands

σSD

c

σHV

c

σSD

c

σHV

c

σSD

c

σHV

c

𝜏𝑇𝐸

𝑑 = σ𝐼𝐼 𝑑

− 3σ𝐼𝑊

𝑑

𝜏𝑟𝑞

𝑑

𝑢2 = 𝑔

𝑟𝑞 21𝜏𝑟𝑞

𝑢2

• Temporal Calibration
• Incomplete polarimetric

decomposition

• Wetland Signature

Inundation beneath Vegetation Detection

Acknowledgement

• John W. Jones, Hydrologic Remote Sensing Branch, USGS
• David M. Bjerklie, New England Water Science Center, USGS
• John F. Galantowicz, Atmospheric and Environmental Research, Boston