Wetland Hydrology from Space

Wetlands are transition zones, where water flow, nutrient cycling, and the Sun’s energy meet to produce unique and productive ecosystems. They provide critical habitat for a wide variety of plant and animal species, including the larval stages of many ocean fish. Wetlands also filter nutrients and pollutants from fresh water used by humans, and provide aquatic habitats for outdoor recreation, tourism, and fishing. Globally, many such regions are under severe environmental stress, mainly from urban development, pollution, and rising sea level. However, there is increasing recognition of the importance of these habitats, and mitigation and restoration activities have begun in a few regions. A key element in wetlands conservation, management, and restoration involves monitoring its hydrologic system: the entire ecosystem depends on its water supply. In the past, hydrologic monitoring of wetlands was conducted almost exclusively by stage (water level) stations, which provide good temporal resolution, but suffer from poor spatial resolution, as stage station are typically distributed several, or even tens of kilometers, from each other.

Figure 1. Picturesque wetland in Big Cypress National Preserve, south Florida.







Wetland InSAR

InSAR provides the needed high spatial resolution hydrological observations, complementing the high temporal resolution of terrestrial observations. Although conventional wisdom suggests that interferometry should not work in vegetated areas, several studies have shown that L-, C- and X-band interferograms with short acquisition intervals (1-105 days) can maintain excellent coherence over wetlands. Our recent results [Wdowinski et al., 2004, 2008] indicates that wetlands InSAR provides high resolution water level change maps with 5 cm accuracy and 1-2 cm precision, providing direct observations of flow patterns and flow discontinuities, as well as excellent constraints for high resolution hydrologic flow models.

Figure 2. (a) RADARSAT-1 interferogram of central south Florid (2004/10/24-2004/11/17), overlying a Landsat ETM band8 and vectors maps showing the geographic location of the data. The interferogram shows backscatter phase changes between the two RADARSAT-1 Synthetic Aperture Radar (SAR) acquisitions. The observed phase changes measure cm-level changes in the wetland surface water level. (b) Enlarged section of the same interferogram showing phase discontinuities mainly along man-made structures (roads, levies) indicating uneven water level changes across the structures. Red circles mark the location of stage (water level) stations in the study area.
Interferometric Synthetic Aperture Radar (InSAR)

Space-based Synthetic Aperture Radar (SAR) is a very reliable technique for monitoring changes of both the solid and aquatic surfaces of the Earth. SAR measures two independent observables, backscatter amplitude and phase, over a wide swath (50-400 km) with pixel resolution of 10-100 m depending on the satellite acquisition parameters. Backscatter amplitude, which is often presented as gray-scale images of the surface (Figure 2a), is very sensitive to the surface dielectric properties, surface inclination towards the satellite, and wave direction in oceans. Amplitude images are widely used for studying surface classification, soil moisture content, ocean waves, and many other applications. The second observable, backscatter phase, measures the fraction of the radar wavelength that returns to the satellite’s antenna. It is mainly sensitive to the range between the surface and the satellite, but also to atmospheric conditions and changes in the surface dielectric properties. Phase data are mainly used in interferometric calculations (InSAR) for detecting cm-level displacements of the surface (Figure 1). The method compares pixel-by-pixel SAR phase observations of the same area acquired at different times from roughly the same location in space to produce high spatial-resolution displacement maps. Such maps, termed interferograms, are widely used in studies of earthquake induced crustal deformation, magmatic activity (volcanos), water-table fluctuations, and glacier movements.
Wetland InSAR is a relatively new application of the InSAR technique that detects water level changes in aquatic environments with emergent vegetation. Although conventional wisdom suggests that interferometry does not work in vegetated areas, several studies have shown that both L- and C-band interferograms with short acquisition intervals (1-105 days) can maintain excellent coherence over wetlands (Figure 1). Interferometric coherence is a measurement of how much the complex phase signal of two SAR images is coherent; it reflects a quality measure of an interferogram. In specific cases of wooded wetlands, coherence can be maintained over several years. The method works, because the radar pulse is backscattered twice (“double-bounce”), from the water surface and vegetation (Figure 2b). Interferometric phase is maintained over both woody and herbaceous vegetations, suggesting that double-bounce is the dominant backscatter mechanism in both wetland environments.

Figure 3. (a) RADARSAT-1 ScanSAR image of Florida showing location of study area (RADARSAT data © Canadian Space Agency / Agence spatiale canadienne 2002. Processed by CSTARS and distributed by RADARSAT International). (b) Cartoon illustrating the double-bounce radar signal return in vegetated aquatic environments. The red ray bounces twice and returns to the satellite, whereas the black ray bounces once and scattered away.





Wetland InSAR Applications
Wetland InSAR observations provide high special resolution maps of water level changes of the dynamic wetland environment. Because InSAR observations are relative in both space and time, it is important to tie the space-based observations with ground observations of water level (stage monitoring). In the Everglades, there is a dense stage monitoring network, which allows us to calibrate and validate the InSAR observations and tie them to an absolute reference frame.
Our calibration studies suggest an accuracy level of 5-10 cm. The man-made structures in the Everglades create many flow discontinuities, which require such dense network for accurate and reliable monitoring. However, in natural undisruptive wetland area, such as in the Everglades National Park in the southern part of the Everglades eco-system, the flow is continuous and can be monitored by a less dense network. Thus, sparsely distributed stage stations in natural wetland areas may be sufficient for calibrating the InSAR observations. In remote wetland areas, where no stage monitoring exists, one can use altimetry data for InSAR calibration. However, altimeter observations are also characterized by low temporal resolution, which may acquire at a different time than the SAR acquisitions. Nevertheless, altimeter observations will be useful for calibration, if no other accurate measurement can be obtained.

One very useful and important observation can be derived directly from the raw interferogram, without the need of stage data for calibration. The high-resolution wetland interferograms provide direct observations of flow patterns and flow discontinuities, as shown in the figure 2b. As water level and water level changes tend to be different across barriers, these differences will be shown in the interferogram as phase discontinuities. This observation is important for wetland restoration efforts that aim to restore a managed or degraded wetland, such as the Everglades, to its natural undisturbed condition.

Another very important application of wetland InSAR is constraining high-resolution flow models, which are important tools for wetland management and restoration. This application also does not require stage calibration, as the model results are converted into the interferogram phase domain, as shown in Figure 5. The stage data are typically used as boundary conditions of the flow model. We conducted such preliminary study by comparing the InSAR observations with the TIME model (Tides and Inflows in the Marshes of the Everglades), which was developed by US Geological Survey and University of Miami. Our study indicates that the models predict well longer wavelength water levels, which are constrained by the stage data, but miss many of the shorter wavelength features (figure 3).


Figure 4. Comparison between synthetic (TIME model) and observed (InSAR) interferograms describing water level changes occurring between the two RADARSAT-1 acquisition dates on 1997/07/13 and 1997/08/06. The comparison shows similarities in the orientation and shape of the longer wavelength fringes, but many differences in the shorter wavelength features.

Figure 5. Remote sensing products of the Everglades wetland, South Florida. Lowest level (base): L-band SAR image showing the backscatter amplitude of the study area. Second level: SAR interferogram illustrating lateral phase changes between two SAR acquisitions (1994/08/09 and 1994/12/19). Third level: map of water level changes occurring between the two acquisition times (the map is computed by multiplying each phase cycle by 15.1 cm). Highest level (top): high spatial-resolution 3-D map of “absolute” water levels calculated by integrating the space- and ground-based observations. The 3-D map shows dynamic water topography (up to 1 m elevation difference across a 20 km distance) caused predominantly by flood gate operation.








Figure 6. RADARSAT-2 interferogram of western south Florida (2008/09/23-2008/10/17) showing tide-induce water level changes along the transition between the saltwater mangrove marsh in the southwest and freshwater swamp in the northeast. Interferometric SAR analysis was produced by S.-H. Hong and S. Wdowinski.




More details at Wdowinski et al. (2004), Wdowinski et al. (2008), Hong et al. (2010a), Hong et al. (2010b), and Gondwe et al. (2010).