LiDAR has become an increasingly appealing technology for remote sensing of tidal wetlands because the high cost of conducting ground-based elevation surveys and because of LiDAR’s high vertical accuracy with higher point densities. However, small errors in a tidal wetland DEM can create large errors when the tidal inundation models are created. Another challenge is tidal wetland vegetation, which tends to be too dense for the LiDAR to penetrate to the wetland surface (in studies seeking to create bare earth models). The following annotated bibliography represents major work that has been done that tests LiDAR and other remote sensing techniques in tidal wetland settings.

Gilmore, M., E. Wilson, N. Barrett, D. Civco, S. Prisloe, J. Hurd and C. Chadwick. 2008. Integrating multi-temporal spectral and structural information to map wetland vegetation in a lower Connecticut River tidal marsh. Remote Sensing of Environment, 112(11):4048-4060.

This study used multi-temporal satellite imagery, field spectral data (due to plant pigments, structure and health) and LiDAR data for top of vegetation canopy to classify and map the principle plant communities in Ragged Rock Creek Marsh, Connecticut. Visible to near infrared reflectance spectra were measured in order to obtain the phenological variability of the dominant marsh species (Spartina patens, Phragmites australis and Typha spp.), which are spectrally distinguishable at certain times of the year. The field spectra and canopy height data were used to classify the three plant species in multi-temporal QuickBird multispectral imagery. The maximum fuzzy accuracies, using the method of Gopal and Woodcock (1994), were 80 percent for Spartina patens, 97 percent for Phragmites australis and 63 percent for Typha spp. It would be interesting to test the effectiveness of this method in tidal marshes with diverse vegetation structures and attempting to identify more than three marsh plant species.

Sadro, S., M. Gastil-Buhl and J. Melack. 2007. Characterizing patterns of plant distribution in a southern California salt marsh using remotely sensed topographic and hyperspectral data and local tidal fluctuations. Remote Sensing of the Environment, 110: 226–239.

The authors (researchers at the University of California – Santa Barbara) tested the hypothesis that plant distribution within Carpinteria Salt Marsh (open and closed canopy) dependent on tidal hydrology, tied to elevation. A tidal inundation-elevation regression model was combined with a LiDAR-based DEM to characterize spatial patterns of inundation. An overlay of Airborne Visible and Infrared Imaging Spectrometer (AVRIS) plant classification data was applied to link the distribution of plants to elevation and inundation regime. One of the main points researchers found was although the commercial LiDAR vendor reported uncorrected accuracy of 9 cm based on paved surface measurements, LiDAR was frequently unable to penetrate the marsh vegetation (3 percent incidence of vegetation penetration to ground elevation) and registered elevated ground heights between plant canopy height and actual ground elevation. However, after tying-in the LiDAR data to ground-based surveys, the root mean square error was reduced to 6.3 cm in vegetated areas. In tidal marshes, accurate topographical data is imperative for researchers trying to correlate ecosystem variables that respond to subtle changes in elevation (i.e.: differences in only a few inches can change inundation regime). AVRIS overall accuracy was 58.8%, likely due to large pixel size (3.4 meters) and degree of plant heterogeneity within the marsh, frequently creating mixed pixel situations.

Belluco, B., M. Camuffo, S. Ferrari, L. Modenese, S. Silvestri, A. Marani and M. Marani. 2006. Mapping salt-marsh vegetation by multispectral and hyperspectral remote sensing. Remote Sensing of Environment, 105:54–67.

At the San Felice marsh of the Lagoon of Venice, Italy, authors tested  a variety of remote sensing tools, including a Reflective Optics System Imaging Spectrometer (ROSIS), Compact Airborne Spectrographic Imager (CASI), Multispectral Infrared and Visible Imaging Spectrometer (MIVIS), IKONOS and Quickbird data, coupled with “extensive” sets of field observations. Authors found optimal conditions when the tide level was low when it was unlikely that there was standing water on the marsh surface and minimizes the soil moisture variability. Authors emphasize the importance of spectral and spatial resolution for identification of marsh plant species. Higher spatial resolution reduces within-pixel heterogeneity and increases spectral separability. While some species are consistently identified correctly by sensors, plants like Juncus spp. that form small patches (a few square meters) were often problematic if the sensor resolution was large. Spartina maritima presented occasional problems for the sensors as well, due to its density and structure leave a large amount of soil visible. The spectral signature is therefore a mix of the species and that of bare soil, which can be confused with soils containing microphytobenthos that have significant amounts of chlorophyll.

Montané, J. and R. Torres. 2006. Accuracy Assessment of Lidar Saltmarsh Topographic Data Using RTK GPS. Photogrammetric Engineering & Remote Sensing, 72(8):961–967.

This study was done to evaluate the accuracy of LiDAR against a representative array of Real-Time Kinematic (RTK) GPS data from a salt marsh island of the North Inlet-Winyah Bay NOAA National Estuarine Research Reserve in South Carolina. The authors state that RTK GPS data took 120 fieldwork hours to collect (2-person team) for a minimum sample size of 265 LiDAR target points for comparison. The LiDAR data overestimated the topographic RTK data by a 7.2 cm overall average and 8.3 cm overall precision. Results showed that LiDAR worked well on the marsh platform, but more care and ground site data should be employed around the creek networks (higher vegetation) and levees.

Nayegandhi, A., J. Brock, C. Wright, and M. O’Connell. 2006. Evaluating a Small Footprint, Waveform-resolving Lidar over Coastal Vegetation Communities. Photogrammetric Engineering & Remote Sensing, 72(12):1407–1417.

This study tested the mapping capabilities of NASA’s Experimental Advanced Airborne Research LiDAR (EAARL; green wavelength), which is designed to simultaneously map near-shore bathymetry, topography and vegetation structure at laser pulses within a 20 cm diameter (“small footprint”). This technology enables characterization of vegetation canopy structure and bare earth topography for a variety of vegetation types. Multiple “small footprints” pulses are synthesized into one “large-footprint” waveform in order to view the vertical structure of the vegetation canopy. This instrument’s green-wavelength laser can penetrate water to about one Secchi-disk depth, allowing researchers to map submerged and unsubmerged topography simultaneously. Comparisons between the waveforms and field surveys had strong correlation and absolute errors (RMSE 0.93 meters at a North Atlantic barrier island site and RMSE 0.8 meters at a marsh site near Tampa Bay, Florida) within the accuracy of field measurements.

Prisloe, S., E. Wilson, D. Civco, J. Hurd and M. Gilmore. Use of lidar data to aid in discriminating and mapping plant communities in tidal marshes of the lower Connecticut River: Preliminary results. ASPRS 2006 Annual Conference (Reno, Nevada), May 1-5, 2006.

The authors report their preliminary LiDAR research findings along a 300 km² stretch of coastal Connecticut. Researchers want to learn how LiDAR reacts to coastal plant communities in order to characterize mean species elevations and transition zones between species and characterize mixed vegetation communities. Mean species heights based on LiDAR were 3.01 meters for Phragmites australis, 1.37 for Typha spp. and 0.64 meters for Spartina patens. It appears to authors that no LiDAR points penetrated Phragmites australis or Typha spp. to the ground surface, due to species’ vegetation density. Though MHHW is not reported, it is assumed that the future publication will include this parameter.

Rosso, P., S. Ustin and A. Hastings. 2006. Use of lidar to study changes associated with Spartina invasion in San Francisco Bay marshes. Remote Sensing of Environment, 100:295–306.

This study tests LiDAR data quality in mapping tidal marshes in San Francisco Bay, with specific attention to the hybrid Spartina alternaflora x foliosa expansion. Comparing LiDAR datasets with ground elevation measurements, authors found that the LiDAR was unable to penetrate the vegetation to the marsh surface with 2.3 points/m² density (though no absolute accuracy was tested, only relative accuracy between LiDAR components). On the mud flats, change detection from surface models showed the effects that the hybrid Spartina has had on accretion and erosion patterns along the shoreline. Water drainage patterns on the mudflat were also clearly evident. LiDAR also proved effective at discriminating Spartina species.

Morris, J., D. Porter, M. Neet, P. Noble, L. Schmidt, L. Lapine and J. Jensen. 2005. Integrating LIDAR elevation data, multi-spectral imagery and neural network modeling for marsh characterization.  International Journal of Remote Sensing, 26(23):5221-5234.

In this paper, authors used a vegetation map derived from aerial photographs and LiDAR elevations to compute frequency distribution of marsh vegetation (Spartina alternaflora and Juncus roemerianus) and elevation relative to tidal elevations at North Inlet estuary in South Carolina. Salt marsh dominated by Spartina alternaflora had median elevation 0.349 meters relative to NAVD88 and mean high water was 0.618 m with mean tidal range 1.39 meters. Salt marsh dominated by Juncus roemerianus had a median elevation of 0.519 meters with a broader and skewed distribution between 0.296 and 0.981. The mean difference between the LiDAR data and ground surveyed elevations was 13 cm, with RMSE 6.5 cm. This study used a 5 meter by 5 meter LiDAR grid, which authors admit is a low density grid. Other problems contributing to elevation discrepancies were vegetation interference (incomplete LiDAR penetration), benchmark proximity to nearby creeks and waterlogged sediments.

Paine, J., W. White, R. Smythe, J. Andrews and J. Gibeaut. 2005. Combining EM and LiDAR to map coastal wetlands: an example from Mustang Island, Texas. In Proceedings, Symposium on the Application of Geophysics to Engineering and Environmental Problems: Environmental and Engineering Geophysical Society, p. 745-756 (CD-ROM).

Authors Paine et al. combined LiDAR and frequency-domain electromagnetic induction (EM) to examine the relationships between elevation, soil and water salinity, electrical conductivity and coastal habitat along transects at Mustang Island, Texas. The LiDAR point data that were used to produce the digital elevation models (DEM’s) had stated vertical accuracy of 15 cm and spaced 0.5 m² apart. At one third of the 38 locations tested, the vegetation was sufficiently dense that the authors question the derived elevations representing the ground surface or incomplete penetration of the vegetation mass. In these dense areas, the avg. vegetation height was 0.5 meters. The authors found geomorphology correlated well with measured conductivity, with highest conductivities in beaches, salt marshes (high and low) and wind tidal flat environments (low and high); lowest conductivities were found in dunes, vegetated-barrier flats (VBF) and freshwater marshes (high and low). The authors state that both EM and LiDAR achieved similar levels of detail that were superior to that of the U.S. Fish and Wildlife Service in its National Wetland Inventory (NWI) aerial photo maps. Authors identified misclassifications on NWI maps based on conductivity data. They state that further investigation should be done to evaluate the effect of vegetation density on LiDAR beam penetration.

Töyrä, J. and A. Pietroniro. 2005. Towards operational monitoring of a northern wetland using geomatics-based techniques. Remote Sensing of the Environment, 97: 174–191.

The authors (professionals from the National Water Research Institute, Canada) combined vegetation and flood duration maps to see how the two correlated in the large Peace–Athabasca Delta (PAD) in northeastern Alberta, Canada. Their objective was to illustrate how multisensory and multi-platform remote sensing could be used in the operational spatio-temporal monitoring of the PAD. A combination of Radarsat Synthetic Aperture Radar (SAR) and optical satellite images (Landsat or SPOT multispectral) were used to generate time-series flood maps for the six-year period from 1996 to 2001. Airborne scanning LiDAR data was used to generate a Digital Elevation Model (DEM) of non-flooded areas (root mean square error 0.24 m). General vegetation patterns were also mapped using multi-temporal SPOT-4 images with an accuracy of 86 percent. Results showed that wetland basins with long inundation periods (3–5 years) were dominated by relatively more productive graminoid vegetation and areas flooded under two years were characterized by less productive shrub vegetation. Authors found that although other parameters were likely also important, the results indicate that flood duration and elevation were important factors influencing vegetation distribution in the PAD.

Yang, X. 2005a. Remote sensing and GIS applications for estuarine ecosystem analysis: an overview. International Journal of Remote Sensing, 26:5347-5356.

This paper covers advances and limitations in four main areas of estuarine remote sensing (estuarine water quality, submergent aquatic vegetation, coastal wetlands and landscape structure). As relates to coastal wetland research, hyperspectral remote sensing has proven very useful in distinguishing between tidal wetland species. Both spectral mixture analysis (SMA) and multiple endmember spectral mixture analysis (MESMA) were applied on AVRIS, though MESMA was preferred because it is able to incorporate more than one endmember per class. The author also discusses areas of continuing research, including the lack of ground data that impedes wetland plant species mapping from hyperspectral imagery. Another area of need is building comprehensive spectral libraries for different wetland plant species and efficient methods for determining the best endmembers for hyperspectral data processing.

Yang, X. 2005b. Use of LIDAR elevation data to construct a high-resolution digital terrain model for an estuarine marsh area. International Journal of Remote Sensing, 26:5163-5166.

The author begins by explaining the inadequacies of the currently available digital elevation datasets for estuarine marsh areas (USGS DEMs, NASA’s SRTM) because the terrain often has limited contrast. This study looks at the effectiveness of LiDAR data to form a DEM of a 6 km² portion of the North Inlet estuary, South Carolina that contains primarily salt marshes behind barrier islands. The dataset was processes using vegetation removal algorithms to create a bare-earth model georeferenced to the UTM projection with WGS84 horizontal datum and WGS84 ellipsoid. The model had a RMSE of 0.164 foot. This study also used a high-resolution multi-spectral image acquired by an ADAR 5500 digital camera to validate the DEM’s terrain features reproduced by the LiDAR data.

Töyrä, J., A. Pietroniro, C. Hopkinson and W. Kalbfleisch. 2003. Assessment of airborne scanning laser altimetry (lidar) in a deltaic wetland environment. Canadian Journal of Remote Sensing, 29(6):718–728.

This was one of the earlier studies evaluating the effectiveness of LiDAR to create DEM’s of wetland settings with dense vegetation. The authors tested LiDAR technology to see if they could create an elevation model in the Peace–Athabasca Delta (PAD; 3900 km²) with sufficient detail and accuracy needed for flood forecasting because traditional in situ elevation surveys would have been difficult over large areas with difficult terrain and airborne stereophotography would have been time consuming and costly. High vertical accuracy was needed because the PAD has a very low relief, so small water level changes can result in large increases in water surface area. The LiDAR elevation data were evaluated by a comparison with in situ survey data. The authors also switched from Canada’s official vertical datum (CGVD28, not based on gravimetric measurements) to the WGS84 ellipsoid and CGG2000 geoid model in order to calculate the orthometric heights for their datasets. The RMSE for the adjusted lidar point elevations was 0.15 m in graminoid vegetation, 0.26 m in willow-covered areas and 0.07 m for bare ground (overall RMSE of the DEM was 0.22 m). Authors recommend that future studies with LiDAR conduct their surveys during leaf-off periods to maximize vegetation penetration. The authors used automated vegetation-removal algorithms to compensate for the inability to penetrate thick vegetation. However, this resulted in some levee heights, which are crucial for determining the point and extent of flooding, being either overestimated (lack of vegetation removal) or underestimated (lack of ground points on the levee crest).

Valentine, J. 2002. Scrub-Shrub/emergent wetland ecotone migration along Delaware tidal rivers in response to relative sea-level change, natural impacts, and human modifications. Thesis (Ph. D.), University of Delaware. 126 pp.

This paper evaluates freshwater tidal wetland plant migration in response to local relative sea-level changes n 5 tidal rivers and creeks of Delaware Bay. The author evaluated plant migration through a series of historical aerial photo overlays from the past several decades. The overlays were digitized and converted from raster to vector format (using polygons to outline dominant marsh plant groups) using ArcInfo. The polygon centroids were generated to use in calculating migration distance. For the five sites sampled, migration rates from main channel sites ranged from 13 to 90 meters per year; while the range for tributary sites ranged from 3.0 to 4.6 meters migration per year. This finding supports the author’s hypothesis that sea-level changes can force tidal wetlands across large geographic areas.

Comments or questions to: custerj@onid.orst.edu

Back