Plain Language Summary
Hurricane induced flooding causes damages to property and the
environment. It is important to assess the intensity of damage by
estimating flood extent and depth. It is difficult to estimate the flood
extent during the hazard in the forested and remote areas using in-situ
observations, which makes it imperative to assess real time flooding
using satellite data. We assessed the flood characteristics during
Hurricane Florence in September 2018. Flood extent and depth was
observed to be maximum on September 17 and 18, and gradually decreased
till September 24. Flood depths were compared with the in-situ
observations from gauges and the study emphasized the importance of
using satellite data that can help to improve flood management and
recovery efforts.
1 Introduction
Extensive flooding causes significant damage to infrastructure and
limits access to natural resources (Kuenzer et al. 2013). Precise
information on the flood extent can help property owners and government
agencies to cope with the economic losses and for developing mitigation
measures (Smith 1994). In order to provide essential information, it is
necessary to assess the full extent of inundation, however, it is
difficult to visualize a large flooded area due to several constraints
(Martinis et al. 2009). Remote sensing using Synthetic Aperture Radar
(SAR) is an effective method to monitor inundated areas. The major
advantages are that it is independent of atmospheric conditions and can
carry out mapping even under vegetated conditions (Woodhouse, 2017).
Recently, natural disasters worldwide have shown that remote sensing
technology can be used in different phases of disaster management such
as preparedness, prevention, relief, and reconstruction (van Westen,
2000).
The return period of floods is estimated to decrease (Pall et al., 2011;
Arnell and Gosling, 2016) and studies have already indicated an increase
in the extreme precipitation (Min et al., 2011; Westra et al., 2013).
However, there is limited global evidence regarding the trend in the
magnitude of annual maximum floods and their prediction (Kundzewicz et
al., 2014).
The majority of literature on the use of SAR has focused on flood
inundation in forested areas and wetlands (Schlaffer et al., 2015; Twele
et al., 2016; Pradhan et al., 2017; Chini et al., 2017; Amitrano et al.,
2018). In addition, most of these studies focus solely on the areal
extent of inundation (Huang and Jin, 2020; Zeng et al., 2020) and only a
handful address changes in water level and dynamics of flood depth.
Alsdorf et al., (2001; 2000) examined water level dynamics along parts
of the Amazon floodplain using Shuttle Imaging Radar (SIR-C) L-band
observations. They found that water level changes in the inundated
vegetation can be measured using the L-band HH polarization
observations. Brown et al. (2016) estimated the flood boundary from the
SAR data and flood surface elevation using the digital terrain model
derived from LiDAR data to obtain flood depth. Zhang et al. (2018)
mapped the flood extent and change in the water level during Hurricane
Irma using Sentinel-1 data and concluded that the majority of their
study area in South Florida was impacted by flooding. However, some
recent studies have used only SAR or integrated SAR and hydrological
modeling for estimating flood inundation. Dasgupta et al. (2020) have
compared SAR derived flood extent to the hydrodynamic modeling in
Mahanadi. Psomiadis et al. (2019) estimated flood depth in Greece using
Sentinel-1 and Digital Elevation Model (DEM). Grimaldi et al., 2020 have
used SAR to map flood under vegetated cover while urban flood inundation
and mapping was done by Bhatt et al. (2019) in Jammu and Kashmir of
India using SAR observations. Urban flood such as the one in Houston in
2017 and Joso of Japan were studied by Li et al. (2019) using Sentinel-1
and ALOS-2 data and Houston flood caused by Hurricane Harvey was studied
by integrated SAR and hydraulic modeling by Scotti et al. (2020).
Hydraulic model was used here to estimate the flood depth and velocity
and compared with the SAR derived flood inundation. However, for large
area flooding, hydraulic models may not be appropriate to estimate flood
extent and depths. Recent studies on flood detection using SAR data are
carried out by Wan et al. (2019); Wu et al. (2019), Sharifi (2020).
Hultquist and Cervone (2020) have used SAR product for estimation,
Wdowinski et al (2004) worked on wetland inundation, and Liang and Liu
(2020) proposed a method to estimate daily inundation based on flood map
information from multiple sources. However, in our study difficulties in
estimating the hurricane flood was experienced even while using high
quality data of UAVSAR and Sentinel-1. In this paper both Sentinel-1 and
UAVSAR were used and results from both were presented so that it can be
compared and best method with better accuracy can be indicated. All
these studies have observed flood using SAR and integrating other
methods or have used products, but very few have directly estimated and
compared depth from different methods using DEM and interferometry for a
major hurricane.
Most of the remote sensing studies use either L (1-2 GHz) or C (4-8 GHz)
frequency to obtain changes in the water level. The L-band is better
suited for identifying water in forested areas as compared to the
C-band, however, for sparsely vegetated areas and in leaf-off
conditions, C- or X-band (10 GHz) can also be used to map flooded
vegetation (Lang et al., 2008; Voormansik et al. 2013; Martinis and
Rieke 2015; Plank et al., 2017). Lu and Kwoun (2008) used European
Remote Sensing (ERS)-1/ERS-2 with C-band and VV polarization and found
that the Interferometric Synthetic Aperture Radar (InSAR) maintained
sound coherence and could measure the changes in phase over the wet
forest areas in Southeast Louisiana.
The main aim of the present study is to investigate and assess depths
obtained by two different methods (DEM and InSAR) to provide the best
result during hurricanes with a short duration, where water recedes
quickly but results in extensive damages. Results and analysis of the
study has the potential that can be followed to estimate extensive flood
in case of other hurricanes or disaster induced floods, with
comprehensive comparison against the observed USGS ground data. This
analysis provides a significant direction to the techniques that can be
best suited in hurricane flood depth estimates. To accomplish this, we
estimated changes in the flood extent and water level in South Carolina
during the passage of Hurricane Florence (September 2018) using two high
resolution data - the C-band (5.405 GHz) Sentinel-1 and L-band (1.25
GHz) Unmanned Aerial Vehicle Synthetic Aperture Radar (UAVSAR) and
compared the results with the USGS gauge data.
2 Materials and Methods
2.1 Study area
The Pee Dee River Basin covers about 18,702 km2 in
North and South Carolina, United States, and drains into the Atlantic
Ocean via the Winyah Bay in South Carolina. The Black River, Waccamaw
River, and Pee Dee River are considered as the largest rivers within the
Yadkin Pee Dee River Basin in South Carolina (Majidzadeh et al., 2017).
Our study site is the Pee Dee River basin in South Carolina with a
flooded area of 5,607 km2 (Figure 1) on September 18,
22, and 23, 3,486 km2 for the September 17 and 4,622
km2 for September 24. The study sites were selected
based on the availability of the UAVSAR flight lines and Sentinel-1 data
during the passage of Hurricane Florence between September 17 to
September 24, 2018. The Light Detection and Ranging (LiDAR) digital
elevation model (DEM) was used in the Pee Dee River Basin area to obtain
the flood depth. The elevation of the study site is around 132.85 m in
the north. The extreme southeast of our study area has a low elevation
near the coastline and it increases towards the north along both sides
of the Pee Dee River. Observed data from the United States Geological
Survey (USGS) gauge stations in the Pee Dee River Basin were used to
compare with the computed flood depth from the DEM and the differences
in the water level inferred from the interferogram obtained from SAR
during the passage of Hurricane Florence.
2.2 Data
The UAVSAR and Sentinel-1 data were used in the study for estimation of
flood extent, depth and changes in the water level. Observed data from
the 9 USGS gauges located within the study area were used that recorded
the gauge heights during Hurricane Florence (Table S1). The LiDAR DEM of
3 m spatial resolution of vertical datum of North American Vertical
Datum (NAVD88) was used in the analyses. All the USGS gauges used here
have NAVD88 datum (Table S1, S2, and S3).
Our flood estimates rely on two sets of SAR data - UAVSAR and
Sentinel-1. UAVSAR is a Jet Propulsion Laboratory (JPL) based airborne
pod-mounted polarimetric instrument that provides a repeat-pass
interferometric observation system with a 16 km swath and operates at
1.26 GHz frequency and HH polarization. The UAVSAR used in this work has
a fine spatial resolution (Table S2). The horizontal transmit and
horizontal receive (HH) polarization of InSAR pair of Ground Range
Detected (GRD) data of UAVSAR was used to identify the flooded area as
it can penetrate the vegetation canopy and displays a good contrast
between land and water (Brisco et al., 2008; 2009).
Sentinel-1 is a C-band system developed by European Space Agency with a
250 km Interferometric Wide Swath (IW). Sentinel-1 data were obtained
from the Alaska SAR Facility (ASF) EarthData site. The Sentinel-1 data
constellation consists of two satellites in the same orbital plane,
Sentinel-1A, and Sentinel-1B
(https://Sentinel.esa.int/web/Sentinel/missions/Sentinel-1/overview).
The Sentinel-1 products are useful for flood mapping and monitoring due
to its frequent revisit time (6 days for the constellation). The level-1
GRD and Single Look Complex (SLC) data from Sentinel-1A/B were used for
September 18 and 24 of 2018 to estimate the difference in the level of
flood water between the two dates. VV polarization (vertical transmit
and vertical receive of waves by the antenna) was selected in this study
since it is considered to have better accuracy than VH (Twele et al.
2016). Co-polarization such as VV has the ability to detect partially
submerged features, which is beneficial in flood damage assessments
(Manjusree et al. 2012). However, HH polarization is preferred for
estimating flood inundation because it is considered to be less
sensitive to minor vertical differences due to waves on the water
surface (Martinis et al., 2009, Gan et al., 2012). Therefore, VV and HH
polarization were selected for this study.
2.3 Methodology
The remotely-sensed (UAVSAR and Sentinel) amplitude data were corrected
and geo-referenced, and converted to backscatter values to identify the
flooded and non-flooded areas (Manjusree et al., 2012; Liang and Liu,
2020). The permanent water bodies were identified from the Landsat 8
image during a non-flooded period (May 2018) using the Normalized
Difference Water Index (NDWI) method (Chen et al., 2006). NDWI is
related to the vegetation water content based on physical principles
(Gao, 1996). Month of May was the dry month and hence water available
during this time remains throughout the year as permanent water. The
permanent water body extracted by this method was eliminated from the
inundated areas to obtain the flooded areas. The threshold method was
used to differentiate between flooded and non-flooded areas (Zhang et
al., 2018) and permanent water features were used as a base reference
for selecting flood thresholds. At the C-band VV polarization, the
backscatter coefficient varies between -6 and -15 dB for water while in
VH polarization it varies between -15 and -24 dB (Manjusree et al.,
2012). In our study also, the threshold values in Sentinel-1 with VV
polarization was within this given limit. Flooded area extraction using
backscatter observations relies on threshold-based methodology, which
requires to carefully identify the proper threshold backscatter value
for water body detection (Manjusree et al., 2012; O’Grady et al., 2013).
Interferometric processing was carried out using SLC data from
Sentinel-1 (Alsdorf et al., 2000; Jung et al., 2010) to obtain a change
in the water level (∂h/∂t) (Zhang et al., 2018). Preprocessing of the
Sentinel-1 data was carried out in VV polarization with IW mode and
along with the UAVSAR data was used to estimate the extent of floodwater
using the threshold method. Co-polarization like VV and HH are preferred
for smooth surfaces like for flood over cross-polarization (Gan et al.,
2012). Use of interferometry processing of L-band SAR data was
demonstrated by Alsdorf et al. (2000) for the first time to detect
changes in the wetland water level. Change in the water level is
recorded between two acquisition dates that shows a change in phase of
interferogram. We have followed the pre-processing instructions of the
SAR data provided by the Sentinel Application Platform (SNAP) of the
European Space Agency including speckle and terrain correction.
Additional filtering was not done to retain the fine pixel resolution of
the image. The ground range detected (GRD) product comprises information
for converting digital pixel values into backscatter intensity. UAVSAR
data are high precision aerial data processed and provided by the JPL
lab. To retain the fineness of this data, no speckle correction was
carried out on these images.
The LiDAR DEM was used to determine the flood depth on the different
dates using the extent of the flooded area. LiDAR data were obtained
from the South Carolina Department of Natural Resources (SCDNR). The
corresponding LiDAR DEM-based water depths were then compared against
the USGS gauge data. Estimation of water depth on different flooded days
was carried out by extracting the boundary cells of the flooded area and
assigning their DEM elevations in the surrounding area by iteration
(Cohen et al., 2018).
Our analyses use the LiDAR DEM and radar interferometry to estimate the
depth and the difference in the water level between two dates. The water
elevation changes measured from the interferogram along the direction of
the radar line-of-sight (LOS) was converted to the vertical displacement
with the wavelength and incidence angle (Jung et al., 2012).
3 Results
3.1 Water level during Florence and flood depth estimation
Variation in the gauge heights obtained from the 9 USGS gauges during
Hurricane Florence is depicted in Figure S1. The X-axis indicates dates
from September 14 to September 24, 2018. Few gauges have data starting
from September 17 and gauge 2110802 (A) and 2130561 (F) have missing
data. The gauge heights reach the highest level on September 17 and in
some gauges the peak level was reached on September 20-21, following
which there is a gradual decrease in heights to September 24. 3 gauges -
gauge 2130980 (TA), gauge 2130561 (F), and gauge 02130000 (G) showed
that the maximum height is reached on September 17 and one on September
18 (gauge 2130930 i.e. TB); 3 gauges - gauge 2131010 (C), gauge 2131000
(D), and gauge 2130810 (E) showed gradual increase to a maximum on
September 20-21 and the remaining 2 gauges - gauge 2110802 (A) and gauge
2135200 (B) showed continuous increase up to September 24.
The spatial distribution of flood depths is shown in Figure 2, which was
estimated using the flood extent and the LiDAR DEM. Depth estimation by
this method indicated a range of depths from less than 0.5 m to greater
than 4.5 m over the study area. About 7% of flooded area on September
17, 9% and 8.8% on September 18 in Sentinel-1 and UAVSAR respectively,
13.5% on September 22 in UAVSAR, 14% on September 23 in UAVSAR, and
15% in Sentinel-1 on September 24, 2018, showed the flooded depth of
less than 1 m. Depth of floodwater was highest on September 17 and 18
and decreased gradually to September 24. However, errors in the
estimation process were observed as negative values of depth, which
imply no flood depth in the area. Negative values are observed in about
5% of the area on September 17, 2018, 7% of the area on September 18,
2018, decreasing to about 3% on September 24, 2018. Areas of high
values of more than 15 m are estimated to vary from 0.09 to 0.65% of
the flooded area. Underestimation of flood depth values is observed in
most of the study area, which might be due to error caused by the
location of the flood boundary cells on big water bodies (Cohen et al.,
2018).
Table S4 shows the distribution of flood depth with respect to the total
flooded area for each day during Hurricane Florence computed from the
flood extent and DEM. As observed from the data, September 17 and 18 of
2018 experienced maximum area (>4.5 m) under high flood
water depth (4.64% on September 17, and 2.53% and 5.83% on September
18 for Sentinel-1 and UAVSAR respectively), and September 22 and 23
showed lowest percentage areas of maximum flood depth (0.14 and 0.12%
respectively). From September 17 to 24, areas of low flood depth
(<1 m) increased while areas of maximum flood depth
(>4.5 m) decreased. The peak flood was observed on
September 18 of 2018.
The water depths are obtained by differencing water surface elevation
and land surface elevation (i.e., DEM), which were then compared with
the observed data from USGS gauges (Figure 3A). The Pearson correlation
(R2) for all dates for this comparison varies from
0.79 on September 22 to the highest R2 of 0.96 on
September 18, 2018 (UAVSAR). For other dates we observe that
R2 varied between 0.86 to 0.95. The root mean square
error (RMSE) varies from 1.69 m on September 17 to 13.59 m on September
24. The depths estimated from UAVSAR and Sentinel underestimated the
observed depth from the USGS gauge data but for a few exceptions. We
observe that the USGS gauge F (Pee Dee River near Bennettsville)
displays overestimation for September 18, 22, and 23 of 2018 (UAVSAR)
and slight overestimation was also observed in the USGS gauge of A
(Waccamaw River at Bucksport) for September 17 and 18, 2018.
3.2 Estimation of the difference in water level using interferogram
The temporal variation of the water heights (∂h/∂t) between September 18
and 24 of 2018 obtained from the USGS gauge data were compared with that
obtained from the interferogram using the Sentinel-1 data (Figure 3B).
The two dates of September 18, which corresponded to the peak flood, and
September 24 corresponding to the lowest floodwater were used to show
the changes in elevation of the water level. A correlation of 0.9 was
observed between the USGS gauges and the ∂h/∂t of the interferogram.
Sentinel-1 images do not cover the entire study area on September 24,
therefore only the 6 USGS gauge stations within the available study area
were used.
Comparison of errors of two methods in change of water level between two
dates is illustrated in Figure 3C. 6 gauges are indicated here, where
change of flood water level is negative in gauges D and E. Similar
results are observed in the water level change obtained from the
interferogram, but the change in water level obtained from DEM are
positive. In gauge F, the change in water level is underestimated in
DEM, while it is overestimated in interferogram in respect to the
observed data. In gauge G, the interferogram underestimated the change
more than the result obtained from DEM. In gauges TA and TB, change in
level obtained from both the methods are overestimated, however
overestimation is more in gauge TB by DEM than interferogram. Analysis
of daily change in depth during hurricanes is very crucial. UAVSAR and
Sentinel-1 data indicated change in the flood level with gradual
recession in the entire hurricane duration.
However, the peak of the flood occurs on September 18 after the
precipitation declines - as observed from gauge heights of different
gauge stations, precipitation was high between September 14 to 16, 2018
in the study area. The gauge height increases between September 17 and
thereafter declines from September 21, 2018. A similar observation in
Figure 3 implies a rise in the flood water to peak on September 18 and
the study area remained submerged till September 22-24, 2018.
4 Conclusions
This study includes the combined approach of determination of flood
depth and the difference in water level (∂h/∂t) using Sentinel-1,
UAVSAR, and LiDAR DEM. The analyses provide a noteworthy direction to
the type of data and techniques used that would give precise result in
estimating hurricane flood depth, and also for storms where water
recedes quickly. Previous studies of inundation by radar remote sensing
typically focused on flood mapping and do not include depth analysis
that determines the intensity and damage from a major flood event
induced by the hurricane. Use of modeling is time exhausting and is not
always suitable for large areas. This study provides a breakthrough,
firstly, by estimating daily changes in the flood depth, which is
difficult to obtain due to absence of high quality daily SAR data, and
secondly, we endeavored at highlighting the method that would be precise
in hurricane flood depth estimation with short duration. We observed the
maximum inundation with a higher range of depth (>2m) due
to flooding on September 18 using Sentinel-1 and UAVSAR data. We also
found good correlation in the water level variation - ∂h/∂t from USGS
gauge height and that generated by interferometry. Interferometry
indicated a comparatively better accuracy when compared with the gauge
data. Therefore, short duration hurricane floods, which are difficult to
assess can follow this using high quality data. This approach is
particularly useful in locations with little or no ground observations
as this can help to schedule relief operations and help in land use
management.
Acknowledgments and Data
The financial support for this work was received from NASA (grant no.
NASA SC EPSCoR award NNX16AR02A). The authors are thankful to the JPL,
NASA for the UAVSAR data (https://uavsar.jpl.nasa.gov/cgi-bin/data.pl),
and the European Space Agency (ESA) for the Sentinel-1 data
(https://vertex.daac.asf.alaska.edu/#). The South Carolina Department
of Natural Resources (SCDNR)
(http://www.dnr.sc.gov/GIS/lidarstatus.html) provided the LiDAR DEM, and
the United States Geological Survey (USGS) provided gauge data
(https://waterdata.usgs.gov/sc/nwis/rt/) and Landsat data obtained from
the Earthexplorer (https://earthexplorer.usgs.gov/).
References
Alsdorf, D. E., J. M. Melack, T. Dunne, L. A. Mertes, L. L. Hess, and L.
C. Smith (2000), Interferometric radar measurements of water level
changes on the Amazon flood plain, Nature, 404(6774), 174-177.
Alsdorf, D. E., L. C. Smith, and J. M. Melack (2001), Amazon floodplain
water level changes measured with interferometric SIR-C radar, IEEE
Transactions on Geoscience and Remote Sensing, 39(2), 423-431.
Amitrano, D., G. Di Martino, A. Iodice, D. Riccio, and G. Ruello (2018),
Unsupervised rapid flood mapping using Sentinel-1 GRD SAR images, IEEE
Transactions on Geoscience and Remote Sensing, 56(6), 3290-3299.
Arnell, N. W., and S. N. Gosling (2016), The impacts of climate change
on river flood risk at the global scale, Climatic Change, 134(3),
387-401.
Bhatt, C. M., G. S., Rao, and S. Jangam (2020), Detection of urban flood
inundation using RISAT-1 SAR images: a case study of Srinagar, Jammu and
Kashmir (North India) floods of September 2014, Modeling Earth Systems
and Environment, 6(1), 429-438.
Brisco, B., N. Short, J. v. d. Sanden, R. Landry, and D. Raymond (2009),
A semi-automated tool for surface water mapping with RADARSAT-1,
Canadian Journal of Remote Sensing, 35(4), 336-344.
Brisco, B., R. Touzi, J. J. van der Sanden, F. Charbonneau, T. Pultz,
and M. D’Iorio (2008), Water resource applications with RADARSAT-2–a
preview, International Journal of Digital Earth, 1(1), 130-147.
Brown, K. M., C. H. Hambidge, and J. M. Brownett (2016), Progress in
operational flood mapping using satellite synthetic aperture radar (SAR)
and airborne light detection and ranging (LiDAR) data, Progress in
Physical Geography, 40(2), 196-214.
Chen, X.-L., H.-M. Zhao, P.-X. Li, and Z.-Y. Yin (2006), Remote sensing
image-based analysis of the relationship between urban heat island and
land use/cover changes, Remote sensing of environment, 104(2), 133-146.
Chini, M., R. Hostache, L. Giustarini, and P. Matgen (2017), A
hierarchical split-based approach for parametric thresholding of SAR
images: Flood inundation as a test case, IEEE Transactions on Geoscience
and Remote Sensing, 55(12), 6975-6988.
Cohen, S., G. R. Brakenridge, A. Kettner, B. Bates, J. Nelson, R.
McDonald, Y. F. Huang, D. Munasinghe, and J. Zhang (2018), Estimating
floodwater depths from flood inundation maps and topography, JAWRA
Journal of the American Water Resources Association, 54(4), 847-858.
Dasgupta, A., P. K. Thakur, and P. K. Gupta (2020), Potential of
SAR-Derived Flood Maps for Hydrodynamic Model Calibration in Data Scarce
Regions, Journal of Hydrologic Engineering, 25(9), 05020028.
Gan, T.Y., Zunic, F., Kuo, C.C. Strobl, T. (2012). Flood mapping of
Danube river at Romania using single and multi-date ERS2SAR images. Int.
J. Appl. Earth Obs. Geoinf., 18, 69-81
Gao, B.-C. (1996), NDWI—A normalized difference water index for remote
sensing of vegetation liquid water from space, Remote sensing of
environment, 58(3), 257-266.
Grimaldi, S., J. Xu, Y. Li, V. R. Pauwels and J. P. Walker (2020), Flood
mapping under vegetation using single SAR acquisitions. Remote Sensing
of Environment, 237, 111582.
Huang, M., and S. Jin (2020), Rapid flood mapping and evaluation with a
supervised classifier and change detection in shouguang using sentinel-1
SAR and sentinel-2 optical data, Remote Sensing, 12(13), 2073.
Hultquist, C., and G. Cervone (2020), Integration of Crowdsourced
Images, USGS Networks, Remote Sensing, and a Model to Assess Flood Depth
during Hurricane Florence, Remote Sensing, 12(5), 834.
Jung, H. C., J. Hamski, M. Durand, D. Alsdorf, F. Hossain, H. Lee, A. A.
Hossain, K. Hasan, A. S. Khan, and A. Z. Hoque (2010), Characterization
of complex fluvial systems using remote sensing of spatial and temporal
water level variations in the Amazon, Congo, and Brahmaputra Rivers,
Earth Surface Processes and Landforms: The Journal of the British
Geomorphological Research Group, 35(3), 294-304.
Jung, H. C., M. Jasinski, J. W. Kim, C. Shum, P. Bates, J. Neal, H. Lee,
and D. Alsdorf (2012), Calibration of two‐dimensional floodplain
modeling in the central Atchafalaya Basin Floodway System using SAR
interferometry, Water Resources Research, 48(7).
Kuenzer, C., H. Guo, J. Huth, P. Leinenkugel, X. Li, and S. Dech (2013),
Flood mapping and flood dynamics of the Mekong Delta: ENVISAT-ASAR-WSM
based time series analyses, Remote Sensing, 5(2), 687-715.
Kundzewicz, Z. W., S. Kanae, S. I. Seneviratne, J. Handmer, N. Nicholls,
P. Peduzzi, R. Mechler, L. M. Bouwer, N. Arnell, and K. Mach (2014),
Flood risk and climate change: global and regional perspectives,
Hydrological Sciences Journal, 59(1), 1-28.
Lang, M. W., P. A. Townsend, and E. S. Kasischke (2008), Influence of
incidence angle on detecting flooded forests using C-HH synthetic
aperture radar data, Remote Sensing of Environment, 112(10), 3898-3907.
Li, Y., S. Martinis, M. Wieland, S. Schlaffer and R. Natsuaki (2019),
Urban flood mapping using SAR intensity and interferometric coherence
via Bayesian network fusion, Remote Sensing, 11(19), 2231.
Liang, J., and D. Liu (2020), Estimating daily inundation probability
using remote sensing, riverine flood, and storm surge models: A case of
hurricane Harvey, Remote Sensing, 12(9), 1495.
Lu, Z., and O.-i. Kwoun (2008), Radarsat-1 and ERS InSAR analysis over
southeastern coastal Louisiana: Implications for mapping water-level
changes beneath swamp forests, IEEE Transactions on Geoscience and
Remote Sensing, 46(8), 2167-2184.
Majidzadeh, H., H. Uzun, A. Ruecker, D. Miller, J. Vernon, H. Zhang, S.
Bao, M. T. Tsui, T. Karanfil, and A. T. Chow (2017), Extreme flooding
mobilized dissolved organic matter from coastal forested wetlands,
Biogeochemistry, 136(3), 293-309.
Manjusree P, Kumar L.P., Bhatt C.M, Rao G.S, Bhanumurthy V. (2012).
Optimization of threshold ranges for rapid flood inundation mapping by
evaluating backscatter profiles of high incidence angle SAR images.
International Journal of Disaster Risk Science. Jun 1;3(2):113-22.
Martinis, S., A. Twele, and S. Voigt (2009), Towards operational near
real-time flood detection using a split-based automatic thresholding
procedure on high resolution TerraSAR-X data, Natural Hazards & Earth
System Sciences, 9(2) 303-314.
Martinis, S., and C. Rieke (2015), Backscatter analysis using
multi-temporal and multi-frequency SAR data in the context of flood
mapping at River Saale, Germany, Remote Sensing, 7(6), 7732-7752.
Min, S.-K., X. Zhang, F. W. Zwiers, and G. C. Hegerl (2011), Human
contribution to more-intense precipitation extremes, Nature, 470(7334),
378-381.
O’Grady, D., M. Leblanc, and D. Gillieson (2013), Relationship of local
incidence angle with satellite radar backscatter for different surface
conditions, International journal of applied earth observation and
geoinformation, 24, 42-53.
Pall, P., T. Aina, D. A. Stone, P. A. Stott, T. Nozawa, A. G. Hilberts,
D. Lohmann, and M. R. Allen (2011), Anthropogenic greenhouse gas
contribution to flood risk in England and Wales in autumn 2000, Nature,
470(7334), 382-385.
Plank, S., M. Jüssi, S. Martinis, and A. Twele (2017), Mapping of
flooded vegetation by means of polarimetric Sentinel-1 and
ALOS-2/PALSAR-2 imagery, International Journal of Remote Sensing,
38(13), 3831-3850.
Pradhan, B., M. I. Sameen, and B. Kalantar (2017), Optimized rule-based
flood mapping technique using multitemporal RADARSAT-2 images in the
tropical region, IEEE Journal of Selected Topics in Applied Earth
Observations and Remote Sensing, 10(7), 3190-3199.
Psomiadis, E., K. X. Soulis, M. Zoka and N. Dercas (2019), Synergistic
approach of remote sensing and GIS techniques for flash-flood monitoring
and damage assessment in Thessaly Plain Area, Greece. Water, 11(3), 448.
Schlaffer, S., P. Matgen, M. Hollaus, and W. Wagner (2015), Flood
detection from multi-temporal SAR data using harmonic analysis and
change detection, International Journal of Applied Earth Observation and
Geoinformation, 38, 15-24.
Scotti, V., M. Giannini and F. Cioffi (2020), Enhanced flood mapping
using synthetic aperture radar (SAR) images, hydraulic modelling, and
social media: A case study of Hurricane Harvey (Houston, TX), Journal of
Flood Risk Management, 13(4), e12647.
Sharifi, A. (2020), Flood Mapping Using Relevance Vector Machine and SAR
Data: A Case Study from Aqqala, Iran, Journal of the Indian Society of
Remote Sensing, 48(9), 1289-1296.
Singh, A. (1989), 1989: Digital change detection techniques using
remotely-sensed data. International Journal of Remote Sensing 10,
989-1003.
Smith, D. I. (1994), Flood damage estimation-A review of urban
stage-damage curves and loss functions, Water Sa, 20(3), 231-238.
Twele, A., W. Cao, S. Plank, and S. Martinis (2016), Sentinel-1-based
flood mapping: a fully automated processing chain, International Journal
of Remote Sensing, 37(13), 2990-3004.
Van Westen, C. (2000), Remote sensing for natural disaster management,
International archives of photogrammetry and remote sensing, 33(B7/4;
PART 7), 1609-1617.
Voormansik, K., J. Praks, O. Antropov, J. Jagomägi, and K. Zalite
(2013), Flood mapping with TerraSAR-X in forested regions in Estonia,
IEEE Journal of Selected Topics in Applied Earth Observations and Remote
Sensing, 7(2), 562-577.
Wan, L., M. Liu, F. Wang, T. Zhang and H. J. You (2019), Automatic
extraction of flood inundation areas from SAR images: a case study of
Jilin, China during the 2017 flood disaster. International journal of
remote sensing, 40(13), 5050-5077.
Wdowinski, S., F. Amelung, F. Miralles‐Wilhelm, T.H. Dixon, and R.
Carande (2004), Space‐based measurements of sheet‐flow characteristics
in the Everglades wetland, Florida, Geophysical Research Letters,
31(15).
Westra, S., L. V. Alexander, and F. W. Zwiers (2013), Global increasing
trends in annual maximum daily precipitation, Journal of climate,
26(11), 3904-3918.
Woodhouse, I. H. (2017), Introduction to microwave remote sensing, CRC
press.
Wu, C., X. Yang and J. Wang (2019), November, Flood Detection in Sar
Images Based on Multi-Depth Flood Detection Convolutional Neural
Network. In 2019 6th Asia-Pacific Conference on Synthetic Aperture Radar
(APSAR) (pp. 1-6). IEEE.
Zeng, Z., Y. Gan, A. J. Kettner, Q. Yang, C. Zeng, G. R. Brakenridge and
Y. Hong (2020), Towards high resolution flood monitoring: An integrated
methodology using passive microwave brightness temperatures and Sentinel
synthetic aperture radar imagery, Journal of Hydrology, 582, 124377.
Zhang, B., S. Wdowinski, T. Oliver-Cabrera, R. Koirala, M. Jo, and B.
Osmanoglu (2018), Mapping the extent and magnitude of severe flooding
induced by Hurricane IRMA with multi-temporal SENTINEL-1 SAR and Insar
observations, Int. Arch. e Photogramm. Remote Sens. Spatial Inf. Sci,
42(43), 2237-2244.
Figure 1. The SAR image in the left plot shows the geographic
extent of the study domain. The red boundary in the left plot indicates
Pee Dee River basin and the shaded portion inside the basin boundary
indicates the study area (aircraft flight domain). Hydrograph in the
left plot shows the discharge at gauge B and red dots in it indicate the
dates used in the analysis. The right plot displays a LiDAR-based DEM of
the study area along with locations of USGS gauge points and approximate
distance from the river mouth.
Gauge IDs: A=02110802; B=02135200; C=02131010; D=02131000; E=02130810;
F=02130561; G=02130000 TA=02130980; and TB=02130930 (T=tributary)
Figure 2. Spatial plots of the estimated flood depth with LiDAR
DEM and flood extent using Sentinel-1 and UAVSAR data
Figure3. A) Daily scatter plots showing the agreement
between the estimated flood depth using LiDAR DEM and flood extent and
the USGS gauge data. The Correlation coefficient R2between calculated and observed depth varies from 0.79 and 0.96 for the
different dates and RMSE varies from 1.69 to 13.59 B) Scatter
plots showing the change in water surface depth between September 18 and
September 24, 2018 for the calculated using Sentinel-1 versus that
observed using USGS gages. The correlation between the two is 0.9.(C) The (right) figure indicates change in water level between
18 and 24 September obtained from two different methods and their
differences from the observed gauge data. The black column indicates
difference in observed gauge data between two dates, light blue column
indicates difference in depth obtained from DEM, and dark blue column
indicates difference obtained from InSAR.