Estimation of Water Level Changes of Large-Scale Amazon ... - MDPI

remote sensing Article

Estimation of Water Level Changes of Large-Scale Amazon Wetlands Using ALOS2 ScanSAR Differential Interferometry Ning Cao 1,2 1 2 3

*

ID

, Hyongki Lee 1,2, *, Hahn Chul Jung 3 and Hanwen Yu 1,2

Department of Civil & Environmental Engineering, University of Houston, Houston, TX 77004, USA; [email protected] (N.C.); [email protected] (H.Y.) National Center for Airborne Laser Mapping, Houston, TX 77204 USA Hydrological Sciences Laboratory, NASA Goddard Space Flight Center, Science Systems and Applications, Inc., Lanham, MD 20706, USA; [email protected] Correspondence: [email protected]





Received: 23 May 2018; Accepted: 15 June 2018; Published: 17 June 2018

Abstract: Differential synthetic aperture radar (SAR) interferometry (DInSAR) has been successfully used to estimate water level changes (∂h/∂t) over wetlands and floodplains. Specifically, amongst ALOS PALSAR datasets, the fine-beam stripmap mode has been mostly implemented to estimate ∂h/∂t due to its availability of multitemporal images. However, the fine-beam observation mode provides limited swath coverage to study large floodplains and wetlands, such as the Amazon floodplains. Therefore, for the first time, this paper demonstrates that ALOS2 ScanSAR data can be used to estimate the large-scale ∂h/∂t in Amazon floodplains. The basic procedures and challenges of DInSAR processing with ALOS2 ScanSAR data are addressed and final ∂h/∂t maps are generated based on the Satellite with ARgos and ALtiKa (SARAL) altimetry’s reference data. This study reveals that the local ∂h/∂t patterns of Amazon floodplains are spatially complex with highly interconnected floodplain channels, but the large-scale (with 350 km swath) ∂h/∂t patterns are simply characterized by river water flow directions. Keywords: SAR; InSAR; DInSAR; water level change; floodplain; wetland

1. Introduction The wetlands and floodplains of the Amazon River are massive in size and in volumetric fluxes. The Amazon Basin, with an area of ~6 million km2 and containing the largest tropical rainforest in the world, contributes 15%–20% of the global river discharge to the oceans (annual averaged discharge of ~200,000 m3 s−1 ) [1,2]. Basin-wide hydrologic and hydrodynamic modeling suggests that surface water accounts for most of the Amazon’s total storage (56%), followed by soil water (27%) and ground water (8%) [3]. Therefore, the amount of water on the Amazon floodplains and wetlands and the amount exchanged with river channels become essential information for understanding large-scale flood propagation, sediment delivery and nutrient exchanges, and the emission of methane and carbon dioxide [4–12]. However, measuring storage and its flux using in situ stage recording devices is not effective because its water flow is complex and not bounded, as in typical channel flow. Consequently, the data scarcity and complexity of the hydrological process still remain a challenge to fully understand the hydrologic dynamics of seasonally flooded wetlands and floodplains [13–15]. Given the vast size and remote location of this large tropical basin, satellite-borne observations (such as radar imaging and satellite altimetry) provide the only viable approach to understanding the spatial and temporal distributions of its water balances and to constraining and validating

Remote Sens. 2018, 10, 966; doi:10.3390/rs10060966

www.mdpi.com/journal/remotesensing

Remote Sens. 2018, 10, 966

2 of 20

basin-scale hydrologic and hydrodynamic models. Among the various remote sensing technologies, differential synthetic aperture radar (SAR) interferometry (DInSAR) has been proven to be a useful technique for measuring ground displacement [16,17]. It has also been used to estimate water level changes (∂h/∂t) in floodplains and wetlands [18–25]. The rationale of using DInSAR to measure ∂h/∂t is based on the fact that the vegetation and water surface create a double-bounce scattering mechanism [26]. The strength of the double-bounce scatterings depends on factors such as aboveground biomass levels, canopy structure, and water levels [27]. A detailed analysis of the connections between the biomass levels and double-bounce scattering can be found in [28–30]. A theoretical analysis of the canopy backscatter modeling of the floodplain can also be found in [31]. When the water surface rises or falls, the range from the radar to the double-bounce scatterers changes, which can be captured by DInSAR. In other words, vegetation is needed to measure ∂h/∂t. For still open water, the backscattering signal would reflect away from the radar and no measurements can be obtained. For open water with waves, correlation is generally lost for DInSAR because of the temporal randomness of the waves. The SAR data from the Space Shuttle Imaging Radar mission (SIR-C) was first used to estimate ∂h/∂t in Amazon floodplains [18]. In recent years, L-band JERS data and ALOS PALSAR data have also been used to estimate ∂h/∂t in both Congo and Amazon floodplains [21–25]. The unwrapped differential interferograms can provide only a spatial gradient of ∂h/∂t. Therefore, a vertical reference (or an offset) is needed to convert it into absolute ∂h/∂t. Recently, satellite altimetry has been used as the vertical reference in order to generate “absolute” ∂h/∂t maps [21,24–32]. The Satellite with ARgos and ALtiKa (SARAL) altimetry measurements are used as a reference in this study. The SARAL altimeter is a Ka-band system with repetivity of 35 days [33]. All of the above previous studies only focus on a small portion of the floodplains using fine-beam data. For PALSAR fine-beam mode, the observation swath is around 70 km, which is too small to study large floodplains and wetlands, such as the Amazon floodplains. Figure 1 shows the study area of the Amazon floodplains with PALSAR wrapped fine-beam differential interferograms. The master–slave acquisition dates (yyyymmdd) for the six swaths from left to right are 20100430–20100615, 20100529–20100714, 20100512–20100627, 20100425–20100610, 20100524–20100709, and 20100507–20100622, respectively. It can be seen that at least six swaths (each with five to six frames) are needed to cover the entire study area. However, the repeat time of the PALSAR system is 46 days and the basic observation scenarios are designed to be provided three times per year for global coverage for fine-beam mode in order to meet the requirements of a multitude of users [34]. It has been found that there are approximately only three to four acquisitions for the same area during the flood season of Amazon wetlands between February and July. In other words, only two to three ∂h/∂t maps can be generated for every flood season, which is not efficient since ∂h/∂t changes rapidly in Amazon floodplains. Besides the temporal variations, the large-scale spatial ∂h/∂t patterns also cannot be directly analyzed using PALSAR fine-beam datasets. For example, six swaths of the interferograms generated from different time spans can be mosaicked in Figure 1 to cover the entire study area. However, the large-scale ∂h/∂t patterns cannot be directly analyzed since the six segments are from different temporal spans. Therefore, the small coverage and large repeat time constrain the applications of PALSAR fine-beam data to study the large-scale ∂h/∂t of Amazon floodplains. Besides PALSAR fine-beam data, ScanSAR data has also been used to study the hydrologic dynamics of large scale wetlands, such as hydroperiod estimation and flood extent monitoring [35,36]. Moreover, PALSAR ScanSAR data can also be used to estimate ∂h/∂t in Amazon wetlands as shown in Figure 2a. It can be seen that the pattern of ∂h/∂t for most of the study area can be obtained. The ALOS ScanSAR mode has a swath of 350 km with basic observation scenario of eight times per year for wetland focus areas. Therefore, the routine use of ScanSAR data for DInSAR processing can significantly improve our understanding of ∂h/∂t in Amazon floodplains and wetlands. The most critical challenge in processing ALOS ScanSAR for DInSAR is the synchronization of the signal bursts [37]. The basic idea of synchronization is that the scene should be viewed from the same

Remote Sens. 2018, 10, 966

3 of 20

azimuth angle by corresponding pulses to acquire enough coherence. Since the synchronization of Remote Sens. 2018, 10, x FOR PEER REVIEW 3 of 20 the bursts is not well programmed in ALOS PALSAR mission, the burst overlap is not always large Remote Sens. 2018, 10, x FOR PEER REVIEW 3 of 20 enough for interferometric analysis, which limits the DInSAR processing. We have found that only a interferometric analysis, which limits the DInSAR processing. We have found that only a few ALOS few ALOS ScanSAR pairs which can acquire reliable ∂h/∂t for the study area. interferometric analysis, limits thefor DInSAR processing. ScanSAR pairs can acquire reliable ∂h/∂t the study area. We have found that only a few ALOS ScanSAR pairs can acquire reliable ∂h/∂t for the study area.

Figure 1. Study areas of Amazon floodplains with ALOS PALSAR fine-beam wrapped differential Figure 1. Study areas of Amazon floodplains with ALOS PALSAR fine-beam wrapped differential Figure 1. Studyacquired areas of between Amazon April floodplains with ALOS PALSAR fine-beam about wrapped differential interferograms and July 2010. Each fringe represents 12 cm of water interferograms acquired between April and July 2010. Each fringe represents about 12 cm of water interferograms between and July Each fringe represents 12 cm of ALOS water level changes inacquired line-of-sight (LOS)April direction. Black2010. and white rectangles are the about coverage of the level changes in line-of-sight (LOS) direction. Black and white rectangles are the coverage of the ALOS level changes in line-of-sight (LOS) direction. Black and whiteused rectangles the coverage of the ALOS PALSAR ScanSAR and ALOS2 PALSAR2 ScanSAR datasets in thisare study. PALSAR ScanSAR and ALOS2 PALSAR2 ScanSAR datasets used in this study. PALSAR ScanSAR and ALOS2 PALSAR2 ScanSAR datasets used in this study.

Figure 2. ScanSAR wrapped differential interferograms. (a) ALOS PALSAR ScanSAR with acquisition Figure ScanSAR wrapped differential interferograms. (a) ALOS PALSAR ScanSAR with acquisition dates 2. of2.ScanSAR 20090522–20090707. (b) ALOS2 interferograms. ScanSAR with acquisition dates of 20150628–20150726. Figure wrapped differential (a) ALOS PALSAR ScanSAR with acquisition dates of 20090522–20090707. (b) ALOS2 ScanSAR with acquisition dates of 20150628–20150726. dates of 20090522–20090707; (b) ALOS2 ScanSAR with acquisition dates of 20150628–20150726.

Since the launch of the ALOS2 PALSAR2 system, ALOS2 ScanSAR data has become available to Since the launch of thefloodplains. ALOS2 PALSAR2 system,system ALOS2isScanSAR dataupgrade has become available to estimate ∂h/∂t in Amazon The PALSAR2 a significant of the PALSAR Since the launch of thefloodplains. ALOS2 PALSAR2 system, ALOS2 data has become available estimate ∂h/∂t in Amazon Thehas PALSAR2 system is aScanSAR significant upgrade of the PALSAR radar. The synchronization of the bursts been well planned for repeat-pass interferometry. An toradar. estimate in Amazon The system isALOS2 a significant upgrade ofAn the The synchronization of floodplains. the burstsishas beenPALSAR2 well planned for interferometry. example of∂h/∂t a repeat-pass interferogram demonstrated in Figure 2b.repeat-pass ScanSAR has a swath example a repeat-pass interferogram is demonstrated in Figure 2b. ALOS2 a swath of 350 kmofwith basic observation scenarios of around nine times per year forScanSAR wetland has monitoring, of 350 means km with basic observation scenarios offor around ninefloodplains times per year forthe wetland monitoring, which four to five ∂h/∂t measurements Amazon during flood season. which means four to five ∂h/∂t measurements for Amazon floodplains during the flood season.

Remote Sens. 2018, 10, 966

4 of 20

PALSAR radar. The synchronization of the bursts has been well planned for repeat-pass interferometry. An example of a repeat-pass interferogram is demonstrated in Figure 2b. ALOS2 ScanSAR has a swath of 350 km with basic observation scenarios of around nine times per year for wetland monitoring, which means four to five ∂h/∂t measurements for Amazon floodplains during the flood season. Because of the dense canopy in Amazon floodplains, only L-band (or P-band) signals can sufficiently penetrate the canopy and reach the water surface. Among the available space-borne SAR missions, L-band SAR systems are preferred to measure ∂h/∂t in Amazon floodplains. Besides the historical JERS, ALOS, and currently operating ALOS2 systems, several L-band SAR systems are planned to be launched in the future, such as the Tandem-L and NISAR systems [38,39]. Hitherto, only ALOS2 ScanSAR data can be used currently to estimate ∂h/∂t in large-scale Amazon floodplains. None of the previous studies have been conducted to estimate ∂h/∂t of Amazon wetlands with ALOS2 ScanSAR DInSAR. Therefore, for the first time, this paper demonstrates that ALOS2 ScanSAR data can be used to map and analyze ∂h/∂t in Amazon floodplains. This paper is organized as follows. The basic procedures and challenges of DInSAR processing with ALOS2 ScanSAR data are given and discussed in Section 2. Section 3 provides the results and discussions on ∂h/∂t patterns and the final ∂h/∂t maps. We conclude with Section 4 which summarizes our findings. 2. Data Processing Strategy In this section, the DInSAR approach using ALOS2 ScanSAR data is introduced to estimate ∂h/∂t, where the SARAL 40 Hz Sensor Geophysical Data Record (SGDR) dataset of year 2015 are used to provide references. 2.1. SAR Dataset Used The ALOS2 ScanSAR datasets with HH polarization acquired during the flood season (i.e., between February to July) of 2015 were used to estimate the large-scale ∂h/∂t in the Amazon floodplains. The ALOS2 datasets used in this study are summarized in Table 1. The perpendicular baselines of the four pairs are relatively small, which means the interferometry is less affected by the geometrical decorrelation. In order to analyze ∂h/∂t patterns for different periods, adjacent scanning pairs were combined to perform DInSAR. Such a configuration can also reduce the effect of temporal decorrelation. The 30-m resolution Shuttle Radar Topography Mission (SRTM) DEM was used in the DInSAR procedure. Table 1. ALOS2 ScanSAR datasets used in this study. Master-Slave Dates (yyyymmdd)

Perpendicular Baseline (m)

Incidence Angle (◦ )

20150222–20150405 20150405–20150517 20150517–20150628 20150628–20150726

111 21 209 159

39 39 39 39

Note: The perpendicular baselines and incidence angles are obtained from the third swaths of the ScanSAR datasets.

2.2. Basic Procedures of ScanSAR DInSAR Processing A flowchart of ALOS2 ScanSAR DInSAR processing is shown in Figure 3. The input single-look-complex (SLC) images were level 1.1 full-aperture images, which means the ScanSAR bursts were processed by a stripmap processor by filling burst gaps with zero echoes before focusing. The major advantage of using the full-aperture image is that existing stripmap InSAR processors can still be used. Every SLC dataset contains five subswath images, each of which were processed individually before mosaicking.

Remote Sens. 2018, 10, 966

5 of 20

Remote Sens. 2018, 10, x FOR PEER REVIEW

5 of 20

Figure 3. Flowchart of ALOS2 ScanSAR DInSAR processing. Figure 3. Flowchart of ALOS2 ScanSAR DInSAR processing.

Compared with conventional DInSAR processing with stripmap mode data, DInSAR processing Compared with conventional processingand with stripmap mode DInSAR processing with ALOS2 ScanSAR mode data isDInSAR more complicated special cautions aredata, needed [40]. The major with ALOS2 ScanSAR mode data is more complicated and special cautions are needed [40]. difference is the removal of the nonoverlap spectrum caused by burst misalignment. First, the crossThe major difference removal the nonoverlap spectrum caused by burstand misalignment. correlation was usedistothe estimate the of offsets between master and slave SLC images, slave SLC First, cross-correlation was used estimate the offsets master slave SLC wasthe coregistered to the master gridtoby resampling. After between coregistration, theand common rangeimages, and and slave SLC was coregistered to the master grid by resampling. After coregistration, the common azimuth spectrums of master and slave SLC images were estimated and extracted by bandpass filters range and azimuth spectrums and slave images were estimated and extracted separately. Once we obtained of twomaster SLC images withSLC overlap spectrums, an interferogram was andfilters the topographic contribution was removed a reference digital spectrums, elevation byobtained bandpass separately.phase Once we obtained two SLCusing images with overlap (DEM) towas generate a differential interferogram. many cases, residual phases by anmodel interferogram obtained and the topographic phaseIn contribution was removed usingcaused a reference baseline error and ionospheric error existed and needed to be removed. Then, the five differential digital elevation model (DEM) to generate a differential interferogram. In many cases, residual phases interferograms were geocoded and mosaicked. Finally,and a deformation map was generated after caused by baseline error and ionospheric error existed needed to be removed. Then, the five filtering and unwrapping. differential interferograms were geocoded and mosaicked. Finally, a deformation map was generated after filtering and unwrapping. 2.3. Coregistration 2.3. Coregistration It has been found that the cross-correlation coregistration method used in stripmap mode InSAR canItalso be used to coregister full-aperture ScanSAR imagesmethod [41]. For an in InSAR pair, mode the offsets has been found that the cross-correlation coregistration used stripmap InSAR between master and slave images are estimated using cross-correlation. Then, the offsets are can also be used to coregister full-aperture ScanSAR images [41]. For an InSAR pair, the offsets parameterized by a low-order polynomial function of range and azimuth using least-squares curve

Remote Sens. 2018, 10, 966

6 of 20

between master and slave images are estimated using cross-correlation. Then, the offsets are Remote Sens. 2018, by 10, xaFOR PEER REVIEW of 20 parameterized low-order polynomial function of range and azimuth using least-squares6 curve fitting. Finally, the slave image is resampled to the master coordinate system using the polynomial fitting. Finally, the slave image is resampled to the master coordinate system using the polynomial function. Compared with the stripmap mode image, the azimuth pixel number of ScanSAR image function. Compared with the stripmap mode image, the azimuth pixel number of ScanSAR image is is much larger than range pixel number. For example, in this study, the azimuth pixel number was much larger than range pixel number. For example, in this study, the azimuth pixel number was usually 20-times larger than the range pixel number. Moreover, ScanSAR InSAR requires very high usually 20-times larger than the range pixel number. Moreover, ScanSAR InSAR requires very high azimuth coregistration precision. Therefore, it is necessary to acquire more offset measurements in azimuth coregistration precision. Therefore, it is necessary to acquire more offset measurements in azimuth direction than range direction. In this study, the number of offset measurements in azimuth azimuth direction than range direction. In this study, the number of offset measurements in azimuth and range directions were set to be 64 and 512, respectively. and range directions were set to be 64 and 512, respectively. 2.4. Baseline Error Removal 2.4. Baseline Error Removal As shown in Figure 3, the topographic phases are estimated from baselines and external As shown in Figure 3, the topographic phases are estimated from baselines and external DEM. DEM. In order to avoid discontinuity between subswaths in ScanSAR DInSAR, only the orbit In order to avoid discontinuity between subswaths in ScanSAR DInSAR, only the orbit information information was used to estimate the topographic phases. Because of the inaccurate orbit measurement, was used to estimate the topographic phases. Because of the inaccurate orbit measurement, residual residual phases caused by baseline error could exist in the final differential interferograms. phases caused by baseline error could exist in the final differential interferograms. The baseline error The baseline error can be estimated and removed by fitting the residual phase interferogram can be estimated and removed by fitting the residual phase interferogram by predefined polynomial by predefined polynomial models. First, the differential interferogram should be unwrapped. models. First, the differential interferogram should be unwrapped. The unwrapped results are fitted The unwrapped results are fitted by a low-order polynomial function of range and azimuth using the by a low-order polynomial function of range and azimuth using the least-squares method. After the least-squares method. After the estimation of the optimal parameters, the estimated residual phase is estimation of the optimal parameters, the estimated residual phase is subtracted from the differential subtracted fromFor the example, differential For example, the wrapped differential interferogram. theinterferogram. wrapped differential interferograms before and after interferograms baseline error before and after baseline error removal are shown in Figure 4. It can be that thehas baseline removal are shown in Figure 4. It can be seen that the baseline phase errorseen component been phase error component has been successfully removed after the baseline error removal. It should be successfully removed after the baseline error removal. It should be noted that the floodplains with noted floodplains with level should be masked out in thethe above procedure. waterthat levelthe changes should be water masked outchanges in the above procedure. Moreover, baseline error Moreover, the baseline error removal can also be implemented after geocoding and mosaicking the removal can also be implemented after geocoding and mosaicking the five subswaths to avoid steps five subswaths to avoid steps between subswaths. between subswaths.

Figure 4. Wrapped differential interferograms of 20150517–20150628 pair before (a) and after (b) Figure 4. Wrapped differential interferograms of 20150517–20150628 pair before (a) and after (b) baseline error removal. baseline error removal.

2.5. Ionospheric Error Removal 2.5. Ionospheric Error Removal It is known that the phase of the radar pulse would be advanced when a radar pulse propagates It is known that the phase of the radar pulsedistortion would beinadvanced when radar pulseto propagates through the ionosphere. The ionospheric phase radar images is aproportional the total through the ionosphere. The ionospheric phase distortion in radar images is proportional to the total electron content (TEC) and the inverse of the radar frequency [42,43]. The L-band SAR systems are electron contentto(TEC) and the inverse of the radar frequency [42,43]. SAR systems more sensitive the ionospheric phase distortions than the X- and C-bandThe SARL-band systems. Moreover, are sensitive to the phase distortions than main the X-regions and C-band SAR systems. themore ionospheric phase errorionospheric can be particularly severe in three including equatorial (such as the Amazon floodplains), middle, and high latitude regions [44,45]. In recent years, various ionospheric correction methods have been proposed for DInSAR processing [45–50]. These methods can generally be divided into two main categories: (1) the range split-spectrum methods are based on

Remote Sens. 2018, 10, 966

7 of 20

Moreover, the ionospheric phase error can be particularly severe in three main regions including equatorial (such as the Amazon floodplains), middle, and high latitude regions [44,45]. In recent years, various ionospheric correction methods have been proposed for DInSAR processing [45–50]. Remote Sens. 2018, 10,generally x FOR PEERbe REVIEW of 20 These methods can divided into two main categories: (1) the range split-spectrum 7methods are based on the dispersive propagation of the ionosphere to estimate the difference of the phase delay the dispersive propagation of the ionosphere to estimate the difference of the phase delay from from different wavelengths [45–51]; (2) the azimuth shift methods are based on the proportional relation different wavelengths [45–51]; (2) the azimuth shift methods are based on the proportional relation between azimuth displacement and and azimuth derivative of theofionospheric phasesphases [52–54]. The azimuth between azimuth displacement azimuth derivative the ionospheric [52–54]. The displacement can be estimated accurate coregistration or multiple-aperture interferometry azimuth displacement can beby estimated by accurate coregistration or multiple-aperture (MAI) [53,54]. The(MAI) ionospheric screen canphase thenscreen be obtained bybe integrating the scaled azimuth interferometry [53,54]. phase The ionospheric can then obtained by integrating the shiftscaled in the azimuth direction. In this study, the MAI-based azimuth shift method was used azimuth shift in the azimuth direction. In this study, the MAI-based azimuth shift method was to estimate the ionospheric errors. errors. As shown in Figure 5, a5, segment fifthsubswath subswath of pair used to estimate the ionospheric As shown in Figure a segmentof of the fifth of pair 20150222–20150405 was used demonstrate theionospheric ionosphericcorrection. correction. It It can can be seen from 20150222–20150405 was used to to demonstrate the from Figure Figure 5a that the upper-right shows clear ionosphericdistortions. distortions. Using band that5athe upper-right area area shows clear ionospheric Using an anazimuth azimuthcommon common band filter, the full aperture can be split into two looks (forwardand backward-looks) for both master and filter, the full aperture can be split into two looks (forward- and backward-looks) for both master andand corresponding interferograms can be can generated. As shownAs in Figure MAI 5b, andslave slavedatasets datasets corresponding interferograms be generated. shown5b, inthe Figure interferogram can then be obtained from the conjugate multiplication of the forwardand backwardthe MAI interferogram can then be obtained from the conjugate multiplication of the forward- and looking interferograms. The streaks shown in Figure 5b are mainly caused by the ionospheric TEC backward-looking interferograms. The streaks shown in Figure 5b are mainly caused by the ionospheric variation. As shown in Figure 5c, the ionospheric phase screen can be generated by integrating the TEC variation. As shown in Figure 5c, the ionospheric phase screen can be generated by integrating the scaled MAI interferogram in azimuth direction. Finally, the corrected interferogram can be obtained scaled MAI interferogram in azimuth direction. Finally, the corrected interferogram can be obtained by by subtracting the ionospheric phase screen from the original interferogram. It can be seen from subtracting thethat ionospheric phase distortions screen fromhave the original interferogram. It can be seeninfrom Figure 5d Figure 5d the ionospheric been effectively removed, especially the upperthatright the ionospheric distortions have been effectively removed, especially in the upper-right area. area.

Figure 5. Ionospheric correction results. (a) Original differential interferogram. (b) MAI interferogram.

Figure 5. Ionospheric correction results. (a) Original differential interferogram; (b) MAI interferogram; (c) Ionospheric phase screen. (d) Corrected interferogram. (c) Ionospheric phase screen; (d) Corrected interferogram.

Besides ionospheric TEC variation, ground deformation can also cause residual phases in the Besides ionosphericInTEC deformation can level also cause residual theto MAI MAI interferogram. thisvariation, study, theground floodplains with water changes were phases maskedinout estimate theIn ionospheric screen. Then, the ionospheric phases were for the masked-out were the interferogram. this study,phase the floodplains with water level changes masked out toarea estimate interpolated. However, if the the ionospheric distortions significantly in area the floodplains, the ionospheric phase screen. Then, ionospheric phases vary for the masked-out were interpolated. interpolation of the ionospheric phases becomes unreliable. Moreover, the numerical integration can However, if the ionospheric distortions vary significantly in the floodplains, the interpolation of the introducephases a numerical error that grows as azimuthal distance increases, which could cause phase ionospheric becomes unreliable. Moreover, the numerical integration can introduce a numerical ramps in the final result. Therefore, the ionospheric correction may not be correctly implemented, error that grows as azimuthal distance increases, which could cause phase ramps in the final result. especially for noisy data with low coherence. In this study, it was a challenge to entirely remove the Therefore, the ionospheric correction may not be correctly implemented, especially for noisy data with ionospheric error, something which requires more work in the future.

low coherence. In this study, it was a challenge to entirely remove the ionospheric error, something which more work in the future. 2.6.requires Phase Unwrapping For DInSAR processing, phase unwrapping is a critical step, especially for areas with strong phase variations. For single-baseline phase unwrapping, the phase jumps between adjacent pixels are

Remote Sens. 2018, 10, 966

8 of 20

2.6. Phase Unwrapping For DInSAR processing, phase unwrapping is a critical step, especially for areas with8 strong Remote Sens. 2018, 10, x FOR PEER REVIEW of 20 phase variations. For single-baseline phase unwrapping, the phase jumps between adjacent pixels areassumed assumedtotobebesmaller smallerthan thanππand andthe thephase phasecan canbebeunwrapped unwrapped integrating the phase gradient byby integrating the phase gradient among adjacent pixels. encountersdifficulties difficultieswhen whenthe the phase jumps among adjacent pixels.However, However,phase phaseunwrapping unwrapping encounters phase jumps larger than thisstudy, study,phase phasediscontinuities discontinuities mainly existed border between areare larger than π.π. InInthis existedin intwo twocases: cases:the the border between floodplain and nonfloodplainareas, areas,and andthe the border border between separated floodplain and nonfloodplain betweendifferent differentfloodplain floodplainsegments segments separated channels. byby channels. example, segmentofofthe thewrapped wrapped differential differential interferogram is is ForFor example, a asegment interferogramofofpair pair20150628–20150726 20150628–20150726 shown in Figure 6a. It can be seen that the fringes generally change along the river flow direction and shown in Figure 6a. It can be seen that the fringes generally change along the river flow direction thethe phase jumps are are clearclear along the borders of theof floodplains. In this study, branch-cut region and phase jumps along the borders the floodplains. In thisthe study, the branch-cut growing algorithm was used to perform phase unwrapping [55,56]. Directly performing phase region growing algorithm was used to perform phase unwrapping [55,56]. Directly performing unwrapping would cause discontinuities in the final results, as shown in Figure 6b. The phase phase unwrapping would cause discontinuities in the final results, as shown in Figure 6b. The phase discontinuities are caused by the inappropriate formations of the branch-cuts that join the residues. discontinuities are caused by the inappropriate formations of the branch-cuts that join the residues. In other words, the phase unwrapping integration path should climb the “hill” (i.e., along the In other words, the phase unwrapping integration path should climb the “hill” (i.e., along the direction direction of the river) in order to avoid phase jumps. However, the default strategy of the current of the river) in order to avoid phase jumps. However, the default strategy of the current branch-cut branch-cut algorithm would not automatically guarantee such an integration path. Moreover, the algorithm would not automatically guarantee such an integration path. Moreover, the floodplains are floodplains are separated into different segments by the channels, which also causes phase separated into different segments by the channels, which also causes phase discontinuities. discontinuities.

Figure 6. (a) A segment of wrapped differential interferogram of pair 20150628–20150726. (b) Figure 6. (a) A segment of wrapped differential interferogram of pair 20150628–20150726; Unwrapped result without masking. (c) Unwrapped result with masking. The black dots are the (b) Unwrapped result without masking; (c) Unwrapped result with masking. The black dots are reference point (with phase set to be zero) for phase unwrapping. the reference point (with phase set to be zero) for phase unwrapping.

Therefore, we extracted the floodplain area by masking out upland areas from the original Therefore, we the floodplain area by masking uplandisareas frominto theseveral original interferogram andextracted then performed phase unwrapping. Since the out floodplain separated interferogram then by performed unwrapping. Since the for floodplain is separated disconnectedand regions the river,phase we performed unwrapping individual segments into and several then connected them to get the river, final results, as shown in Figure 6c. connection can beand easily disconnected regions by the we performed unwrapping for The individual segments then implemented byget adding an offset value if assuming the6c. phase between is connected them to the final results, as shown in Figure The jump connection canadjacent be easilysegments implemented less than π. The connection of disconnected areas is straightforward in Figure 6, since the pattern of the fringes is clear and simple (i.e., the phase jump between adjacent segments is less than π). However, there exist cases when the phase jumps are larger than π and the connection of different

Remote Sens. 2018, 10, 966

9 of 20

by adding an offset value if assuming the phase jump between adjacent segments is less than π. The connection of disconnected areas is straightforward in Figure 6, since the pattern of the fringes is clear and simple (i.e., the phase jump between adjacent segments is less than π). However, there exist cases when the phase jumps are larger than π and the connection of different segments can be incorrect. Moreover, a reference is PEER needed to obtain the absolute ∂h/∂t measurements. In this study,9we used Remote Sens. 2018, 10, x FOR REVIEW of 20 SARAL altimetry measurements as references to correct the phase jumps between disconnected areas. segments can beway incorrect. Moreover, reference neededistoby obtain absolute ∂h/∂tphase measurements. Another potential to deal with the alarge phaseisjumps usingthe multibaseline unwrapping In this study, we used SARAL altimetry measurements as references to correct the phase jumps methods [57,58], which will be tested in the future. between disconnected areas. Another potential way to deal with the large phase jumps is by using

multibaseline phase unwrapping methods [57,58], which will be tested in the future. 3. Results and Discussion Results Discussion 3.1. 3.Results of and SARAL Altimetry

The SRTM DEM over the study area is shown in Figure 7 along with 16 segments of tracks of 3.1. Results of SARAL Altimetry the SARAL altimetry used in this study. It is easy to see that the topography generally decreases The SRTM DEM over the study area is shown in Figure 7 along with 16 segments of tracks of the from upstream to downstream. Besides the Amazon mainstem, the floodplains of the three tributaries SARAL altimetry used in this study. It is easy to see that the topography generally decreases from (i.e.,upstream Japura River, Jutai River, and Jurua River) are also included in this study area. From the SARAL to downstream. Besides the Amazon mainstem, the floodplains of the three tributaries (i.e., altimetry the water levelsRiver) of theare floodplains of the rivers area. can be obtained as shown Japurameasurement, River, Jutai River, and Jurua also included in four this study From the SARAL in Figure 8. The water levels are obtained by spatially averaging the track segments. From Figure altimetry measurement, the water levels of the floodplains of the four rivers can be obtained as shown 8a, we can see that the water levels are decrease from to (8), which is expected since theFigure topography in Figure 8. The water levels obtained bysegment spatially(1) averaging the track segments. From 8a, decreases segment (1) tolevels (8). We can also that the floodplains generally reach we canfrom see that the water decrease fromsee segment (1)mainstem to (8), which is expected since the topography decreases from (1) Compared to (8). We with can also see that thethe mainstem floodplains peak water levels between Maysegment and June. the mainstem, three tributaries reach generally reach peak water different levels between and June. mainstem, peak water levels at slightly times,May especially in Compared the case ofwith the the Japura River. the Wethree can see tributaries reach peak levels at slightly different times, especially in the the case Japura River. from Figure 8b that thewater Japura River reaches its peak water level at end ofofthe June. The offsets We can see lines from in Figure 8b that Japura River reaches peak water differences level at the end of June. The between two Figure 8c,dthe can be caused by the its topographic between different offsets betweenspeaking, two lines the in Figure be caused by the topographic tracks. Generally water 8c,d levelcan increases before June and starts differences to fall afterbetween June over different tracks. Generally speaking, the water level increases before June and starts to fall after June the study area. Even though satellite altimetry can successfully measure the water level changes, over the study area. Even though satellite altimetry can successfully measure the water level changes, the measurements are limited to the 1-D track and it is difficult to analyze the spatial pattern of ∂h/∂t. the measurements are limited to the 1-D track and it is difficult to analyze the spatial pattern of ∂h/∂t. Therefore, it is necessary to use DInSAR to acquire the 2-D ∂h/∂t maps. Therefore, it is necessary to use DInSAR to acquire the 2-D ∂h/∂t maps.

Figure 7. SRTM DEM of the study area. Blue lines indicate segments of the SARAL altimetry tracks

Figure 7. SRTM DEM of the study area. Blue lines indicate segments of the SARAL altimetry tracks used in this study. used in this study.

Remote Sens. 2018, 10, 966

10 of 20

Remote Sens. 2018, 10, x FOR PEER REVIEW

10 of 20

Figure 8. 8. Water Water levels levels measured measured by by SARAL SARALaltimetry altimetryfor for year year2015. 2015. (a–d) (a–d) indicates indicates the the floodplains floodplains of of Figure the four four rivers. rivers. Labels Labels (1)–(16) (1)–(16) indicate indicate the the 16 16 segments segments of of the altimetry tracks the the altimetry tracks as as shown shown in in Figure Figure 7. 7. For every segment, the altimetry water levels are averaged. For every segment, the altimetry water levels are averaged.

3.2. Interpretations of ∂h/∂t Patterns on a Small Scale 3.2. Interpretations of ∂h/∂t Patterns on a Small Scale The final ALOS2 ScanSAR differential interferograms are obtained as shown in Figure 9. Three The final ALOS2 ScanSAR differential interferograms are obtained as shown in Figure 9. segments of the study area are zoomed in to interpret the ∂h/∂t patterns on a small scale, as shown in Three segments of the study area are zoomed in to interpret the ∂h/∂t patterns on a small scale, Figure 10. It should be noted that there still exist some residual phases caused by ionospheric errors, as shown in Figure 10. It should be noted that there still exist some residual phases caused by especially for Figure 9a,b. However, such residual phases generally do not affect the analysis of the ionospheric errors, especially for Figure 9a,b. However, such residual phases generally do not affect spatial patterns of ∂h/∂t maps. From the altimetry measurements in Figure 8, it is known that water the analysis of the spatial patterns of ∂h/∂t maps. From the altimetry measurements in Figure 8, it is rises from February to mid-May (i.e., Figure 9a,b) and begins to fall at the end of June (i.e., Figure known that water rises from February to mid-May (i.e., Figure 9a,b) and begins to fall at the end of June 9c,d). Several key interpretations of ∂h/∂t patterns can be obtained from Figures 9 and 10. (i.e., Figure 9c,d). Several key interpretations of ∂h/∂t patterns can be obtained from Figures 9 and 10. First, the patterns of ∂h/∂t vary significantly as the water level changes. At the mid-rising water First, the patterns of ∂h/∂t vary significantly as the water level changes. At the mid-rising water period (i.e., Figure 9a,b), ∂h/∂t tends to be localized and related to the topography of scroll bars, levees, period (i.e., Figure 9a,b), ∂h/∂t tends to be localized and related to the topography of scroll bars, channels, and depressions, which is consistent with the previous studies [22]. At the high water levees, channels, and depressions, which is consistent with the previous studies [22]. At the high period (i.e., Figure 9c), the mainstem floodplain has smooth ∂h/∂t patterns extended uniformly across water period (i.e., Figure 9c), the mainstem floodplain has smooth ∂h/∂t patterns extended uniformly the floodplain, which means the water is flowing in a diffuse pattern. At the mid-falling water period across the floodplain, which means the water is flowing in a diffuse pattern. At the mid-falling water (i.e., Figure 9d), ∂h/∂t patterns of the mainstem floodplain become complex and localized again. In period (i.e., Figure 9d), ∂h/∂t patterns of the mainstem floodplain become complex and localized again. Figure 9d, the ∂h/∂t pattern of the Japura floodplain becomes uniformly extended since Japura River In Figure 9d, the ∂h/∂t pattern of the Japura floodplain becomes uniformly extended since Japura reaches its peak water level at the end of June. River reaches its peak water level at the end of June.

Remote Sens. 2018, 10, 966 Remote Sens. 2018, 10, x FOR PEER REVIEW

11 of 20 11 of 20

Figure 9. Final wrapped differential interferograms. Three white rectangles in (a) are zoomed in as Figure 9. Final wrapped differential interferograms. Three white rectangles in (a) are zoomed in as shown in Figure 10. The time span between master and slave dates for the four interferograms are 42, shown in Figure 10. The time span between master and slave dates for the four interferograms are 42, 42, 42, and 28 days, respectively. 42, 42, and 28 days, respectively.

The change of ∂h/∂t patterns with water levels can be easily seen in the zoomed-in areas, as The of ∂h/∂t patternsthe with water geomorphology levels can be easily in the zoomed-in as shown in change Figure 10. For example, complex can seen be seen in Figure 10(i.a)areas, around shown Figure For example, the complex geomorphology can be in Figure 10(i.a) the areainwith high10. elevation (i.e., scroll bars), where the ∂h/∂t pattern is seen complex, as shown inaround Figure the area with high elevation (i.e., scroll bars), where the ∂h/∂t pattern is complex, as shown in 10(i.b). The water level continues to rise and fills the depressions between April and May. Compared Figure 10(i.b). The water level continues to rise and fills the depressions between April and May. with Figure 10(i.b), the ∂h/∂t pattern of Figure 10(i.c) becomes less complex, especially for the scroll Compared with Figure the pattern ∂h/∂t pattern of Figure becomes less distributed. complex, especially for bars. During high water,10(i.b), the ∂h/∂t of Figure 10(i.d) 10(i.c) becomes uniformly Moreover, the scroll bars. During high water, the ∂h/∂t pattern of Figure 10(i.d) becomes uniformly distributed. the color pattern of the interferometric fringes is inversed since the water level is falling. As the water Moreover,to thefall, color of the becomes interferometric fringes is inversed the water level as is shown falling. continues thepattern ∂h/∂t pattern complex and related to thesince topography again, AsFigure the water continues to fall, the ∂h/∂t pattern becomes complex and related to the topography again, in 10(i.e). as shown in Figure 10(i.e). The ∂h/∂t patterns of the floodplains are also shown by arrows in Figure 10. Black arrows The positive ∂h/∂t patterns the floodplains also shown in ∂h/∂t Figure 10. Black arrows indicates ∂h/∂t ofof rising water while are white arrows areby forarrows negative of falling water. The indicates positive ∂h/∂t of rising water while white arrows are for negative ∂h/∂t of falling water. sign of ∂h/∂t is determined by the absolute ∂h/∂t results in Section 3.3. The arrows point toward The sign ofwith ∂h/∂t is determined the absolute ∂h/∂tblack results in Section 3.3. Thedirections arrows point directions larger water levelby changes. Therefore, arrows point toward that toward directions with larger water level changes. Therefore, black arrows point toward directions accumulate more water and white arrows point toward directions that depart more water. From that accumulate water and white directions that depart water. Figure 10i, we canmore see that both black andarrows white point arrowstoward point toward upstream, whichmore means the From Figure 10i, we can see that both black and white arrows point toward upstream, which means upstream receives more water in the rising water season while departing more water in the falling the upstream more water in directions the rising of water seasonarrows while change departing water in the water season. receives For Figure 10(ii.b,ii.d), the black formore different periods, falling water season. For Figure 10(ii.b,ii.d), directions of the black arrows change for different periods, which can be caused by the sources of water flows being different when water fills the floodplains from rivers. From Figure 10(iii.b), we can see the spatial patterns of the overbank flow from the river

Remote Sens. 2018, 10, 966

12 of 20

which can be caused by the sources of water flows being different when water fills the floodplains Remote Sens. 2018, 10, Figure x FOR PEER REVIEW 12 river of 20 from rivers. From 10(iii.b), we can see the spatial patterns of the overbank flow from the to the floodplains. For Figure 10(iii.b,iii.e), arrows uniformly point toward upstream, which means the to the floodplains. For Figure 10(iii.b,iii.e), arrows uniformly point toward upstream, which means upstream departs more water than the downstream. the upstream departs more water than the downstream. It is known that ∂h/∂t equals to spatial changes in discharge (i.e., ∂h/∂t ≈ −∇Q, where Q is It is known that ∂h/∂t equals to spatial changes in discharge (i.e., ∂h/∂t ≈ −∇Q, where Q is the the water discharge) [22]. For rising water with positive ∂h/∂t, water is assumed to be flowing to water discharge) [22]. For rising water with positive ∂h/∂t, water is assumed to be flowing to floodplain pockets with greater ∂h/∂t [22]. Based on this assumption, the black arrows in Figure 10 floodplain pockets with greater ∂h/∂t [22]. Based on this assumption, the black arrows in Figure 10 can also indicate the directions of water flow. However, such an assumption may not be accurate can also indicate the directions of water flow. However, such an assumption may not be accurate because of the inner channels in the floodplains. How to infer the directions of water flow from spatial because of the inner channels in the floodplains. How to infer the directions of water flow from spatial variations in ∂h/∂t is beyond the scope of this paper. variations in ∂h/∂t is beyond the scope of this paper.

Figure 10. Differential interferograms for three zoomed-in areas (i), (ii), and (iii) labeled in Figure 9. Figure 10. Differential interferograms for three zoomed-in areas (i), (ii), and (iii) labeled in Figure 9. SRTM DEMs are also included in (a). SRTM DEMs are also included in (a).

3.3. Interpretations of ∂h/∂t Patterns on a Large Scale 3.3. Interpretations of ∂h/∂t Patterns on a Large Scale The previous section investigated three zoomed segments of the study area in order to interpret The previous investigated threeshowed zoomed segments ofpatterns the study in order to interpret ∂h/∂t patterns on section a small scale. The results that the ∂h/∂t of area Amazon floodplains are ∂h/∂t patterns on a small scale. The results showed that the ∂h/∂t patterns of Amazon floodplains are spatially complex and localized for mid-rising and mid-falling water seasons. Similar conclusions spatially complex and localized for mid-rising and mid-falling water seasons. Similar conclusions have have also been introduced in the previous studies using DInSAR with fine-beam datasets [21,22]. In also introduced the previous studies DInSAR fine-beam datasets In this thisbeen section, the ∂h/∂tin patterns on a large scaleusing observed fromwith the ScanSAR datasets are[21,22]. presented. section, the ∂h/∂t patterns on a large scale observed from the ScanSAR datasets are presented. The absolute ∂h/∂t maps can be obtained as shown in Figure 11. It should be noted that ∂h/∂t Theinabsolute ∂h/∂t can be obtained as shown in Figure 11. should be noted ∂h/∂t maps Figure 11 use maps the altimetry measurements as references. TheIttime series of thethat altimetry maps inheight Figureare 11 use the altimetry measurements as references. time series ofcan thebe altimetry water water interpolated for the SAR acquisition dates andThe ∂h/∂t references obtained for height are interpolated for the SAR acquisition dates and ∂h/∂t references can be obtained for different different DInSAR pairs. Since there are few in situ measurements available in the study area, the DInSAR Since therebe aredirectly few in situ measurements the study area,the theperformance DInSAR results DInSARpairs. results cannot validated by true available data. In in order to verify of cannot be directly validated by true data. In order to verify the performance of DInSAR measurements, DInSAR measurements, the comparisons of the DInSAR ∂h/∂t with altimetry ∂h/∂t along the SARAL the comparisons of the ∂h/∂t altimetry along SARAL tracks are altimetry shown in tracks are shown in DInSAR Figure 12. Thewith DInSAR ∂h/∂t∂h/∂t values are the interpolated for the Figure 12. The DInSAR ∂h/∂t values are interpolated for the altimetry measurement We can measurement points. We can see that the four DInSAR ∂h/∂t maps agree well points. with SARAL see that the four DInSAR ∂h/∂t maps agree well(rmse) with SARAL with measurements, with root-mean-square errors of 0.2 m,measurements, 0.1 m, 0.1 m, and 0.1root-mean-square m, respectively. errors (rmse) of 0.2 m, 0.1 m, can 0.1 m, 0.1 m, The bias of the comparison canofbe caused The bias of the comparison be and caused by respectively. the inconsistencies of the acquisition dates DInSAR by thealtimetry. inconsistencies of the acquisition of DInSAR and altimetry. It should also be noted that and It should also be noted dates that such “cross validation” is only for DInSAR–altimetry agreement evaluation and is not a strict way to validate the DInSAR measurements.

Remote Sens. 2018, 10, 966

13 of 20

such “cross validation” is only for DInSAR–altimetry agreement evaluation and is not a strict way to validate the DInSAR measurements. Remote Sens. 2018, 10, x FOR PEER REVIEW 13 of 20

Figure 11. Final DInSAR ∂h/∂t maps. The black lines in (a) are profiles along the floodplains which Figure 11. Final DInSAR ∂h/∂t maps. The black lines in (a) are profiles along the floodplains which will be examined for further analysis. will be examined for further analysis.

During the water increasing period (i.e., Figure 11a,b), the water level increases up to 3 m. During the water increasing period (i.e., Figure the level waterdecreases level increases 3 During the water decreasing period (i.e., Figure 11c,d),11a,b), the water up to 4up m.toFor m. During the water decreasing period (i.e., Figure 11c,d), the water level decreases up to 4 m. different parts of the floodplains, the patterns of absolute ∂h/∂t are different. The water levels of the For parts of the floodplains, the patterns of absolute ∂h/∂t different. The water levels of Jutaidifferent and Jurua floodplains mainly increase between February and are April and decrease significantly the Jutai and Jurua mainly increase between February April in and decrease significantly after mid-May. Forfloodplains the Japura floodplain, the water level mainly and increases Figure 11b and starts to after mid-May. For the Japura floodplain, the water level mainly increases in Figure 11b and starts to decrease in Figure 11d. For the mainstem floodplain, the water level increases dramatically between decrease in Figure 11d. For the mainstem floodplain, the water level increases dramatically between February and mid-May. Compared with the upstream floodplain, the water level in the downstream February mid-May. Compared with thedecreases upstream less, floodplain, the water level11c,d, in thewhich downstream floodplainand along the mainstem floodplain as shown in Figure can be floodplain along the mainstem floodplain decreases less, as shown in Figure 11c,d, which can be caused by the late arrival of the peak stages of the Japura River, as shown in Figure 9b. caused by the late arrival of the peak stages of the Japura River, as shown in Figure 9b.

Remote Sens. 2018, 10, 966 Remote Sens. 2018, 10, x FOR PEER REVIEW

14 of 20 14 of 20

Figure 12. Comparison of DInSAR ∂h/∂t with altimetry ∂h/∂t. The parameters p and r2 represent the Figure 12. Comparison of DInSAR ∂h/∂t with altimetry ∂h/∂t. The parameters p and r2 represent the slopes of the least-square fitted lines and R2 values, respectively. slopes of the least-square fitted lines and R2 values, respectively.

In order to further analyze the ∂h/∂t patterns on a large scale, the ∂h/∂t along the four arbitrarily In order to further the ∂h/∂tJapura, patterns on aand large scale, the ∂h/∂tare along theinfour arbitrarily chosen center profilesanalyze of the mainstem, Jutai, Jurua floodplains shown Figures 13– chosen center profiles of the mainstem, Japura, Jutai, andand Jurua floodplains in Figures 13–16, 16, respectively. Figure 13a shows that the upstream downstream ofare theshown mainstem floodplain respectively. Figure 13aincreases shows that themupstream downstream of the mainstem floodplain have have dramatic water of ~2 between and February and April. As shown in Figure 13b, the dramatic water increases of ~2 m between February and April. As shown in Figure 13b, the water water level continuously increases about 1 m and there exist abrupt ∂h/∂t changes around the longitude of −67° increases where theabout profile1 crosses Amazon River. Such changes have been level continuously m and the there exist abrupt ∂h/∂tabrupt changes around thealso longitude observed in previous studies andthe areAmazon spatiallyRiver. coincident floodplain channels [22]. We observed can see of − 67◦ where the profile crosses Suchwith abrupt changes have also been Figure 13c that values graduallywith increase from upstream downstream, which in from previous studies andthe are∂h/∂t spatially coincident floodplain channelsto[22]. We can see from verifies that the mainstem floodplain has smooth ∂h/∂t patterns extending uniformly across the Figure 13c that the ∂h/∂t values gradually increase from upstream to downstream, which verifies floodplain at high water. In has Figure 13d, the abrupt ∂h/∂t extending change reoccurs around longitude −67° as that the mainstem floodplain smooth ∂h/∂t patterns uniformly across the floodplain ◦ the water level decreases. In spite of the abrupt ∂h/∂t changes around the river, the ∂h/∂t patterns are at high water. In Figure 13d, the abrupt ∂h/∂t change reoccurs around longitude −67 as the water smoothly distributed from upstream to downstream. From Figure 13c,d, we can observe larger level decreases. In spite of the abrupt ∂h/∂t changes around the river, the ∂h/∂t patterns are smoothly upstream water level changes than downstream during water falling period. distributed from upstream to downstream. From Figure 13c,d, we can observe larger upstream water We can see from Figure 14a–c that the water level of the Jupura floodplain continuously level changes than downstream during water falling period. increases from February to the end of June. However, the ∂h/∂t patterns are interestingly different for We can see from Figure 14a–c that the water level of the Jupura floodplain continuously increases different periods. Figure 14a shows that the downstream of the Jupura floodplain has larger ∂h/∂t from February to the end of June. However, the ∂h/∂t patterns are interestingly different for different than the upstream from February to April. On the contrary, Figure 14b shows that the ∂h/∂t of the periods. Figure 14a shows that the downstream of the Jupura floodplain has larger ∂h/∂t than the upstream becomes larger than the downstream from April to mid-May. Figure 14c indicates that ∂h/∂t upstream from April. On thedownstream contrary, Figure 14b shows the ∂h/∂t ofFrom the upstream is around 0.5 February m for bothtoupstream and from mid-May tothat the end of June. Figure becomes larger than the downstream from April to mid-May. Figure 14c indicates that ∂h/∂t is 14d, we can see that the water level of the downstream starts to fall while the upstream still shows around 0.5 m for both upstream and downstream from mid-May to the end of June. From Figure 14d, slightly increasing water. we can see that the water level of the downstream starts to fall while the upstream still shows slightly increasing water.

Remote Sens. 2018, 10, 966 Remote Sens. 2018, 10, x FOR PEER REVIEW Remote Sens. 2018, 10, x FOR PEER REVIEW

15 of 20 15 of 20 15 of 20

Figure 13. DInSAR ∂h/∂t along the center profile of the mainstem floodplain shown in Figure 11a. Figure 13.13.DInSAR mainstem floodplain floodplainshown shownininFigure Figure11a. 11a. Figure DInSAR∂h/∂t ∂h/∂talong along the the center center profile of the mainstem

Figure 14. DInSAR ∂h/∂t along the center profile of the Japura floodplain shown in Figure 11a. Figure 14. DInSAR ∂h/∂t along the center profile of the Japura floodplain shown in Figure 11a. Figure 14. DInSAR ∂h/∂t along the center profile of the Japura floodplain shown in Figure 11a.

Remote Sens. 2018, 10, 966 Remote Sens. 2018, 10, x FOR PEER REVIEW

16 of 20 16 of 20

Figure 15. DInSAR ∂h/∂t along the center profile of the Jutai floodplain shown in Figure 11a. Figure 15. DInSAR ∂h/∂t along the center profile of the Jutai floodplain shown in Figure 11a.

Comparing Figures 15 and 16, we can see that the ∂h/∂t patterns of the Jutai and Jurua Comparing 15 andthe 16,center we can see thatBoth the ∂h/∂t patterns theshow Jutai that and the Jurua floodplains floodplains are Figures similar along profiles. Figures 15a andof16a downstream are similar along the center profiles. Both Figures 15a and 16a show that the downstream has a slightly has a slightly larger water level increase than the upstream. Even though the ∂h/∂t of the Jurua larger water level increase the16a, upstream. Even pattern though the ∂h/∂t of the Jurua floodplain fluctuates floodplain fluctuates in than Figure the general is that ∂h/∂t increases from upstream to indownstream. Figure 16a, the patternthe is that increases from16a upstream to downstream. In other words, Ingeneral other words, ∂h/∂t∂h/∂t patterns of Figure are complex in local scale and have a the ∂h/∂tincreasing patterns of Figure arescale. complex in local scaleFigures and have simple trend simple trend on a16a large We can see from 15baand 16b increasing that the water of on thea large scale. falls We can see fromwhile Figures and 16b that the water of the upstream fallsApril dramatically upstream dramatically the 15b downstream still slightly accumulates water from to midMay.the From Figures 15c and 16c, weaccumulates can see that water the water falls between mid-May while downstream still slightly fromlevel April to dramatically mid-May. From Figures 15c and andwe thecan end ofthat Junethe andwater the upstream more between than the mid-May downstream. and 16d 16c, see level fallsdecreases dramatically andFigures the end15d of June and show that thedecreases water level continuously decreases from the end15d of June to July. In general, canlevel see the upstream more than the downstream. Figures and 16d show that the we water that the downstream a larger ∂h/∂t during the water increasing period andthe the upstream has continuously decreaseshas from the end of June to July. In general, we can see that downstream hasaa larger∂h/∂t absolute ∂h/∂t during fallingperiod period.and Moreover, compared the absolute mainstem∂h/∂t floodplain, larger during the waterwater increasing the upstream has with a larger during the three tributaries have much simpler ∂h/∂t patterns. water falling period. Moreover, compared with the mainstem floodplain, the three tributaries have In summary, the above analysis, we can see that the large-scale ∂h/∂t pattern can be much much simpler ∂h/∂tfrom patterns. simpler even though the patterns are complex, as large-scale shown in Section 3.2. Therefore, it is In summary, from the local above∂h/∂t analysis, we can see that the ∂h/∂t pattern can be much beneficial to analyze both the large-scale and local ∂h/∂t pattern of Amazon floodplains using simpler even though the local ∂h/∂t patterns are complex, as shown in Section 3.2. Therefore, it is ScanSARtodatasets. It should be noted that the quantitative accuracy analysis of the above beneficial analyze both the large-scale and local ∂h/∂t pattern of Amazon floodplains using DInSAR ScanSAR results cannot be provided due to the limited in situ measurements. datasets. It should be noted that the quantitative accuracy analysis of the above DInSAR results cannot

be provided due to the limited in situ measurements.

Remote Sens. 2018, 10, 966 Remote Sens. 2018, 10, x FOR PEER REVIEW

17 of 20 17 of 20

Figure 16. DInSAR ∂h/∂t along the center profile of the Jurua floodplain shown in Figure 11a. Figure 16. DInSAR ∂h/∂t along the center profile of the Jurua floodplain shown in Figure 11a.

4. Conclusions 4. Conclusions The spatial resolution and coherence of fine-beam DInSAR are generally better than ScanSAR The spatial coherence of fine-beam DInSAR are generally better than ScanSAR DInSAR, whichresolution means the and fine-beam DInSAR can capture more accurate spatial variation of the ∂h/∂t DInSAR, which means the fine-beam DInSAR can capture more accurate spatial variation of be the patterns and get better ∂h/∂t estimate accuracy. However, the large-scale ∂h/∂t patterns cannot ∂h/∂t patterns and get better ∂h/∂t estimate accuracy. However, the large-scale ∂h/∂t in patterns cannot directly obtained from mosaicking fine-beam images since ∂h/∂t changes rapidly floodplains. beTherefore, directly obtained from mosaicking fine-beam images since ∂h/∂t changes rapidly in floodplains. even though the accuracy of the ScanSAR estimate is not as good as the fine-beam data, it Therefore, even though the accuracy of the ScanSAR estimate is not as∂h/∂t good as the is still necessary for processing ScanSAR datasets to obtain large-scale maps. Infine-beam this paper,data, for it the is still processinghow ScanSAR to obtain data large-scale ∂h/∂t maps. Inwater this paper, firstnecessary time, we for demonstrated to usedatasets ALOS2 ScanSAR to acquire large-scale level forchanges the first(∂h/∂t) time, in weAmazon demonstrated how to basic use ALOS2 ScanSAR data toprocessing acquire large-scale wetlands. The procedures of DInSAR are given.water The level changes (∂h/∂t) in Amazon wetlands. The basic procedures of DInSAR processing are given. major difficulties caused by orbit error, ionospheric error, and phase unwrapping in ScanSAR DInSAR also analyzed. Finally, four large-scale kmunwrapping swath) two-dimensional The majorprocessing difficultiesare caused by orbit error, ionospheric error,(with and 350 phase in ScanSAR ∂h/∂t maps are generated year 2015. The obtained ∂h/∂t maps can350 provide reliable information to DInSAR processing are alsofor analyzed. Finally, four large-scale (with km swath) two-dimensional study the wetland storages fluxes Amazon wetlands, are critical in understanding ∂h/∂t maps are generated forand year 2015.over Thethe obtained ∂h/∂t maps which can provide reliable information to theirthe flow processes and biogeochemical cycles. The proposed procedures cancritical also beinused over other study wetland storages and fluxes over the Amazon wetlands, which are understanding large-scale wetlands, such as the Congo wetland, which will be studied in the future. their flow processes and biogeochemical cycles. The proposed procedures can also be used over other large-scale wetlands, such as the Congo wetland, which will be studied in the future.

Author Contributions: Ning Cao: Conceptualization; Data curation; Investigation; Methodology; Validation; Visualization; Writing Ning – original Hyongki Lee: Conceptualization; Supervision; Investigation; Author Contributions: Cao:draft. Conceptualization; Data curation; Investigation; Methodology;Validation. Validation; Hahn Chul Jung: Conceptualization; Hanwen Yu: Methodology; Validation.Validation. Visualization; Writing – original draft.Investigation; Hyongki Lee:Validation. Conceptualization; Supervision; Investigation; Hahn Chul Jung: Conceptualization; Investigation; Validation. Hanwen Yu: Methodology; Validation. Funding: This paper is supported by NASA’s New Investigator Program (NNX14AI01G), SWOT Science Team Funding: This paper is supported NASA’s(80NSSC18K0423). New Investigator Program (NNX14AI01G), SWOT Science Team Project (NNX16AQ33G), and GEOby Program Project (NNX16AQ33G), and GEO Program (80NSSC18K0423). Acknowledgments: The ALOS2 data are provided by Japan Aerospace Exploration Agency (JAXA) (PI No. 1976 Acknowledgments: The ALOS2 data are provided by Japan Aerospace Exploration Agency (JAXA) (PI No. 1976 3069). && 3069). Conflicts Interest:The Theauthors authorsdeclare declareno noconflict conflictofofinterest. interest. Conflicts ofof Interest:

References

Remote Sens. 2018, 10, 966

18 of 20

References 1.

2.

3.

4. 5. 6. 7.

8. 9. 10.

11.

12.

13. 14. 15.

16. 17. 18.

19.

20.

Molinier, M.; Guyot, J.L.; Oliveira, E.; Guimaraes, V.; Chaves, A. Hydrologie du bassin de l’Amazone. In Grands bassins fluviaux périatlantiques: Congo, Niger; Actes du Colloque PEGI/INSU/CNRS, 22-24/11/1993; Amazone, J.B., Olivry, J.C., Eds.; ORSTOM: Paris, France, 1995; pp. 335–344. Ronchail, J.; Guyot, J.L.; Villar, J.C.E.; Fraizy, P.; Cochonneau, G.; Oliveira, E.; Filizola, N.; Ordenez, J.J. Impact of the Amazon tributaries on major floods at Óbidos. In Climate Variability and Change: Hydrological Impacts; IAHS Publication Red book; Demuth, S., Gustard, A., Planos, E., Scatena, F., Servat, E., Eds.; IAHS: London, UK, 2006; pp. 220–225. Paiva, R.C.D.; Buarque, D.C.; Collischonn, W.; Bonnet, M.-P.; Frappart, F.; Calmant, S.; Mendes, C.A.B. Large-scale hydrologic and hydrodynamic modeling of the Amazon River basin. Water Resour. Res. 2013, 49, 1226–1243. [CrossRef] Paiva, R.C.D.; Collischonn, W.; Tucci, C.E.M. Large scale hydrologic and hydrodynamic modeling using limited data and a GIS based approach. J. Hydrol. 2011, 406, 170–181. [CrossRef] Yamazaki, D.; Kanae, S.; Kim, H.; Oki, T. A physically based description of floodplain inundation dynamics in a global river routing model. Water Resour. Res. 2011, 47, W04501. [CrossRef] Dunne, T.; Mertes, L.A.K.; Meade, R.H.; Richey, J.E.; Forsberg, B.R. Exchanges of sediment between the flood plain and channel of the Amazon River in Brail. Geol. Soc. Am. Bull. 1998, 110, 450–467. [CrossRef] Melack, J.M.; Forsberg, B. Biogeochemistry of Amazon floodplain lakes and associated wetlands. In The Biogeochemistry of the Amazon Basin and its Role in a Changing World; McClain, M.E., Victoria, R.L., Richey, J.E., Eds.; Oxford University Press: Oxford, UK, 2001; pp. 235–276. Hamilton, S.K.; Sippel, S.J.; Melack, J.M. Comparison of Inundation Patterns among Major South American Floodplains. J. Geophys. Res. 2002, 107, 8038. [CrossRef] Melack, J.M.; Engle, D. An organic carbon budget for an Amazon floodplain lake. Verh. Internat. Verein. Limnol. 2009, 30, 1179–1182. [CrossRef] Richey, J.E.; Melack, J.M.; Aufdenkampe, A.K.; Ballester, V.M.; Hess, L. Outgassing from Amazonian rivers and wetlands as a large tropical source of atmospheric carbon dioxide. Nature 2002, 416, 617–620. [CrossRef] [PubMed] Melack, J.M.; Hess, L.L.; Gastil, M.; Forsberg, B.R.; Hamilton, S.K.; Lima, I.B.T.; Novo, E.M.L.M. Regionalization of methane emissions in the Amazon basin with microwave remote sensing. Glob. Chang. Boil. 2004, 10, 530–544. [CrossRef] Melack, J.M. Aquatic ecosystems. In Interactions between Biosphere, Atmosphere and Human Land Use in the Amazon Basin. Ecological Studies 227; Nagy, L., Forsberg, B., Artaxo, P., Eds.; Springer: Berlin, Germany, 2016; pp. 117–145. Alsdorf, D.E.; Lettenmaier, D.P. Tracking fresh water from space. Science 2003, 301, 1485–1488. [CrossRef] [PubMed] Yamazaki, D.; Sato, T.; Kanae, S.; Hirabayashi, Y.; Bates, D.P. Regional flood dynamics in a bifurcating mega delta simulated in a global river model. Geophys. Res. Lett. 2014, 41, 3127–3135. [CrossRef] Pontes, P.R.M.; Fan, F.M.; Fleischmann, A.S.; Paiva, R.C.D.; Buarque, D.C.; Siqueira, V.A.; Jardim, P.F.; Sorribas, M.V.; Collischonn, W. MGB-IPH model for hydrological and hydraulic simulation of large floodplain river systems coupled with open source GIS. Environ. Model. Softw. 2017, 94, 1–20. [CrossRef] Rosen, P.A.; Hensley, S.; Joughin, I.R.; Li, F.K.; Madsen, S.N.; Rodriguez, E.; Goldstein, M.R. Synthetic aperture radar interferometry. Proc. IEEE 2000, 88, 333–382. [CrossRef] Bamler, R.; Hartl, P. Synthetic aperture radar interferometry. Inv. Probl. 1998, 14, R1–R54. [CrossRef] Alsdorf, D.E.; Melack, J.M.; Dunne, T.; Mertes, L.A.K.; Hess, L.L.; Smith, L.C. Interferometric radar measurements of water level changes on the Amzon floodplain. Nature 2000, 404, 174–177. [CrossRef] [PubMed] Alsdorf, D.E.; Birkett, C.; Dunne, T.; Melack, J.M.; Hess, L. Water level changes in a large Amazon lake measured with spaceborne radar interferometry and altimetry. Geophys. Res. Lett. 2001, 28, 2671–2674. [CrossRef] Alsdorf, D.E.; Smith, L.C.; Melack, J.M. Amazon Floodplain Water Level Changes Measured with Interferometric SIR-C Radar. IEEE Trans. Geosci. Remote Sens. 2001, 39, 423–431. [CrossRef]

Remote Sens. 2018, 10, 966

21.

22. 23.

24. 25.

26. 27.

28.

29. 30.

31. 32.

33. 34. 35.

36.

37. 38.

39.

40.

19 of 20

Jung, H.C.; Hamski, J.; Durand, M.; Alsdorf, D.; Hossain, F.; Lee, H.; Hossain, A.K.M.A.; Hasan, K.; Khan, A.S.; Hoque, A.K.M.Z. Characterization of complex fluvial systems via remote sensing of spatial and temporal water level variations in the Amazon, Congo, and Brahmaputra Rivers. Earth Surf. Process. Landforms 2010, 35, 294–304. [CrossRef] Alsdorf, D.E.; Bates, P.; Melack, J.; Wilson, M.; Dunne, T.B. Spatial and temporal complexity of the Amazon flood measured from space. Geophys. Res. Lett. 2007, 34. [CrossRef] Alsdorf, D.E.; Beighley, E.; Laraque, A.; Lee, H.; Tshimanga, R.; O’Loughlin, F.; Mahé, G.; Dinga, B.; Moukandi, G.; Spencer, R.G.M. Opportunities for hydrologic research in the Congo Basin. Rev. Geophys. 2016, 54, 378–409. [CrossRef] Yuan, T.; Lee, H.; Jung, H.C. Congo floodplain hydraulics using PALSAR InSAR and Envisat altimetry data. In Remote Sensing of Hydrological Extremes; Springer: Berlin, Germany, 2017; pp. 65–81. Yuan, T.; Lee, H.; Jung, H.C.; Aierken, A.; Beighley, E.; Alsdorf, D.; Tshimanga, R.; Kim, D. Absolute water storages in the Congo River floodplains from integration of InSAR and satellite radar altimetry. Remote Sens. Environ. 2017, 201, 57–72. [CrossRef] Jung, H.C.; Alsdorf, D. Repeat-pass multi-temporal interferometric SAR coherence variations with Amazon floodplain and lake habitats. Int. J. Remote Sens. 2010, 31, 881–901. [CrossRef] Salvia, M.; Grings, F.; Ferrazzoli, P.; Barraza, V.; Douna, V.; Perna, P.; Karszenbaum, H. Estimating flooded area and mean water level using active and passive microwaves: The example of Paran River delta floodplain. Hydrol. Earth Syst. Sci. Discuss. 2011, 8, 2895–2928. [CrossRef] Grings, F.M.; Ferrazzoli, P.; Jacobo-Berlles, J.C.; Karszenbaum, H.; Tiffenberg, J.; Pratolongo, P.; Kandus, P. Monitoring flood condition in marshes using EM models and Envisat ASAR observations. IEEE Trans. Geosci. Remote Sens. 2006, 44, 936–942. [CrossRef] Kim, J.W.; Lu, Z.; Gutenberg, L.; Zhu, Z. Characterizing hydrologic changes of the Great Dismal Swamp using SAR/InSAR. Remote Sens. Environ. 2017, 198, 187–202. [CrossRef] Neumann, M.; Saatchi, S.S.; Ulander, L.M.H.; Fransson, R.E.S. Assessing performance of L- and P-band polarimetric interferometric SAR data in estimating boreal forest above-ground biomass. IEEE Trans. Geosci. Remote Sens 2012, 50, 714–726. [CrossRef] Kasischke, E.S.; Tanase, M.A.; Bourgeau-Chavez, L.L.; Borr, M. Soil moisture limitations on monitoring boreal forest regrowth using spaceborne L-band SAR data. Remote Sens. Environ. 2011, 115, 227–232. [CrossRef] Kim, J.-W.; Lu, Z.; Lee, H.; Shum, C.K.; Swarzenski, C.M.; Doyle, T.W.; Baek, S.-H. Integrated Analysis of PALSAR/Radarsat-1 InSAR and ENVISAT Altimeter Data for Mapping of Absolute Water Level Changes in Louisiana Wetlands. Remote Sens. Environ. 2009, 113, 2356–2365. [CrossRef] Bronner, E.; Gulliot, A.; Picot, N.; Noubel, J. SARAL/AltiKa Products Handbook; SALP-MU-M-OP-15984-CN; Centre National d’Etudes Spatiales: Toulouse, France, 2013. Japan Aerospace Exploration Agency. ALOS Data Users Handbook; Revision C; Japan Aerospace Exploration Agency: Tokyo, Japan, 2008. Lee, H.; Yuan, T.; Jung, H.C.; Beighley, E. Mapping wetland water depths over the central Congo Basin using PALSAR ScanSAR, envisat altimetry, and MODIS VCF data. In Remote Sensing of Environment; Elsevier: New York City, NY, USA, 2014. Arnesen, A.S.; Silva, T.S.F.; Hess, L.L.; Novo, E.M.L.M.; Rudorff, C.M.; Chapman, B.D.; McDonald, K.C. Monitoring flood extent in the Lower Amazon River floodplain using ALOS/PALSAR ScanSAR images. Remote. Sens. Environ. 2013, 130, 51–61. [CrossRef] Bamler, R.; Eineder, M. ScanSAR processing using standard high precision SAR algorithms. IEEE Trans. Geosci. Remote Sens. 1996, 34, 212–218. [CrossRef] Krieger, G.; Hajnsek, I.; Papathanassiou, K.; Eineder, M.; Younis, M.; de Zan, F.; Huber, S.; Lopez-Dekker, P.; Prats, P.; Werner, M. Tandem-L: An innovative interferometric and polarimetric SAR mission to monitor Earth system dynamics with high resolution. In Proceedings of the IEEE Geoscience and Remote Sensing Symposium (IGARSS), Honolulu, HI, USA, 25–30 July 2010; pp. 253–256. Rosen, P.; Hensley, S.; Shaffer, S.; Edelstein, W.; Kim, Y.; Kumar, R.; Misra, T.; Bhan, R.; Satish, R.; Sagi, R. An update on the NASA-ISRO dual-frequency DBF SAR (NISAR) mission. In Proceedings of the 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Beijing, China, 10–15 July 2016. Liang, C.; Fielding, E.J. Interferometry with ALOS-2 fullaperture ScanSAR data. IEEE Trans. Geosci. Remote Sens. 2017, 55, 2739–2750. [CrossRef]

Remote Sens. 2018, 10, 966

41. 42. 43. 44. 45.

46.

47.

48.

49. 50. 51. 52.

53. 54. 55. 56.

57. 58.

20 of 20

Liang, C.; Fielding, E.J. Interferometric processing of ScanSAR data using stripmap processor: New insights from coregistration. IEEE Trans. Geosci. Remote Sens. 2016, 54, 4343–4354. [CrossRef] Meyer, F.; Bamler, R.; Jakowski, N.; Fritz, T. The potential of low-frequency SAR systems for mapping ionospheric TEC distributions. IEEE Geosci. Remote Sens. Lett. 2006, 3, 560–564. [CrossRef] Hanssen, R.F. Radar Interferometry: Data Interpretation and Error Analysis; Kluwer: Norwell, MA, USA, 2002. Shim, J.S. Analysis of Total Electron Content (TEC) Variations in the Low- and Middle-Latitude Ionosphere. Ph.D. Thesis, Utah State University, Logan, UT, USA, 2009. Fattahi, H.; Simons, M.; Agram, P. InSAR Time-Series Estimation of the Ionospheric Phase Delay: An Extension of the Split Range-Spectrum Technique. IEEE Trans. Geosci. Remote Sens. 2017, 55, 5984–5996. [CrossRef] Brcic, R.; Parizzi, A.; Eineder, M.; Bamler, R.; Meyer, F. Estimation and compensation of ionospheric delay for SAR interferometry. In Proceedings of the 2010 IEEE International Geoscience and Remote Sensing Symposium, Honolulu, HI, USA, 25–30 July 2010; pp. 2908–2911. Rosen, P.A.; Hensley, S.; Chen, C. Measurement and mitigation of the ionospheric in L-band interferometric SAR data. In Proceedings of the 2010 IEEE Radar Conference, Washington, DC, USA, 10–14 May 2010; pp. 1459–1463. Gomba, G.; Parizzi, A.; Zan, F.D.; Eineder, M.; Bamler, R. Toward operational compensation of ionospheric effects in SAR interferograms: The split-spectrum method. IEEE Trans. Geosci. Remote Sens 2016, 54, 1446–1461. [CrossRef] Gomba, G.; Rodriguez, F.; de Zan, F. Ionospheric Phase Screen Compensation for the Sentinel-1 TOPS and ALOS-2 ScanSAR Modes. IEEE Trans. Geosci. Remote Sens. 2017, 55, 223–235. [CrossRef] Gomba, G.; de Zan, F. Bayesian Data Combination for the Estimation of Ionospheric Effects in SAR Interferograms. IEEE Trans. Geosci. Remote Sens. 2017, 55, 6582–6593. [CrossRef] Liang, C.; Fielding, E.J. Measuring azimuth deformation with L-band ALOS-2 ScanSAR interferometry. IEEE Trans. Geosci. Remote Sens. 2017, 55, 2725–2738. [CrossRef] Raucoules, D.; de Michele, M. Assessing ionospheric influence on L-band SAR data: Implications on coseismic displacement measurements of the 2008 Sichuan earthquake. IEEE Geosci. Remote Sens. Lett. 2010, 7, 286–290. [CrossRef] Jung, H.S.; Lee, D.T.; Lu, Z.; Won, J.S. Ionospheric Correction of SAR Interferograms by Multiple-Aperture Interferometry. IEEE Trans. Geosci. Remote Sens. 2013, 51, 3191–3199. [CrossRef] Chen, A.C.; Zebker, H.A. Reducing Ionospheric Effects in InSAR Data Using Accurate Coregistration. IEEE Trans. Geosci. Remote Sens. 2014, 52, 60–70. [CrossRef] Goldstein, R.M.; Zebker, H.A.; Werner, C.L. Satellite radar interferometry: Two-dimensional phase unwrapping. Radio Sci. 1988, 23, 713–720. [CrossRef] Rosen, P.A.; Werner, C.W.; Hiramatsu, A. Two-dimensional phase unwrapping of SAR interferograms by charge connection through neutral trees. In Proceedings of the IGARSS’94, Pasadena, CA, USA, 8–12 August 1994. Yu, H.; Lee, H.; Yuan, T.; Cao, N. A novel method for deformation estimation based on multi-baseline InSAR phase unwrapping. IEEE Trans. Geosci. Remote Sens. 2018. to be published. [CrossRef] Cao, N.; Yu, H.; Lee, H. A Multi-baseline InSAR Phase Unwrapping Method using Designed Optimal Baselines obtained by Motion Compensation Algorithm. IEEE Geosci. Remote Sens. Lett. 2018. to be published. [CrossRef] © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).