Application of airborne Hyperspectral imaging
in Wetland delineation
Chung-hsin Juan1, Jonathan D. Jordan1, and Chih-Hung Tan2
1Agricultural and Biological Engineering Department, University of Florida
PO Box 110570, Gainesville, Florida 32611-0570,USA
Tel: +1 (352) 3928030
E-mail: crs@ufl.edu
2 Agricultural Engineering Research Center
196-1, ChungYuan Rd., ChungLi
Tel: (886)-3-4521314
E-mail: chtan@aerc.org.tw
TAIWAN
Keywords:Hyperspectral image, Wetland delineation, Wetland Vegetation Mapping, Remote sensing
Abstract
The study site is located in Fort Drum Marsh, Indian River county, Florida, where the main vegetation types are sawgrass (Cladium Jamaicense) and cattail (Typha domigensis) with some floating vegetation and scattered clusters of bushes and trees. A complete hyperspetral imaging mission for hyperspectral imaging was performed in May, 2000. The mission included the preparation of ground targets and calibration panels, the aerial imaging flight, and the ground truth trip. The mounted hyperspectral imaging sensor has 64 wavebands from 399.2 nm to 920.5 nm. The ground truth work included taking spectral reflectance measurements by a hand-held radiometer, multi-spectral images by a multi-spectral imager, measurements of leaf area indices (LAI) by a LAI meter in the locations where only one interested plant species grew in the surrounding area. After the hyperspectral image was rectified and spectrally calibrated, the pixels of the ground truthing points were extracted from the images and then their spectral reflectances were compared with the ground spectral reflectances acquired by the spectral radiometer. The results demonstrated noticeable differences in the spectral reflectances of different wetland plant species. Therefore, wetland could be delineated into species level using aerial hyperspectral imaging.
1. Introduction
Wetland vegetation species are very sensitive to the environmental changes, and vice versa wetland vegetation species may be a good indicator of the environmental changes. As the concerns of wetland protection grow, wetland delineation is also getting more important. In freshwater marsh, both cattail (Thyha, spp.) and sawgrass (Cladium jamincese) can grow in similar geomorphological and geographical locations and may mix together, but sawgrass tends to be found in lower nutrient level conditions (Kadlec and Knight, 1996). Therefore, the spatial distribution of the cattail and sawgrass in freshwater marshes are monitored by the government agencies routinely. However, both of the species looked very similar from a distance and hardly identified without close observations, and thus wetland delineation to the specie level usually needs more intensive tasks in field trips. In order to save the time and labor consuming filed work in identifying different species, remote sensing was employed as an ideal and convenient tool to serve this purpose (Doren, 1999).
The most frequently used conventional remote sensed data were either satellite images or airborne color infrared (CIR) photos and images (Madden, 1999; Welch, 1996). The satellite data mostly have not enough high spatial resolution to differentiate detailed ground information. Although the airborne CIR photos and images may have enough spatial resolution, they may have too broad spectral wavebands to identify two similar color looking species. Therefore, as to properly delineate wetlands of the study purpose, airborne hyperspectral images would fit the needs in terms of both spatial and spectral resolutions.
The main goal of this study is to use ground-based and airborne remote sensing tools to study the feasibility of delineate wetlands at species level. The mission of the aerial hyperspectral imaging have three stages: 1) Preparations before aerial hyperspectral imaging, 2) Setup during aerial imaging, and 3) Ground truthing.
2. Materials and Methods
2.1 Study Area
The study site is located in the Fort Drum marsh, upper St. John's River basin, in Indian River county, Florida (27E35'12"N, 80E41'17"W, Figure1). It is the headwater of the St. Johns River. The site is a freshwater marsh with cattail and sawgrass as the dominant vegetation species. The desired imaging area with the illustration of the mission plan was also show in Figure 1.
Figure 1. The location of the study site and the illustration of the mission plan.
2.2 Mission Plan for Aerial Hyperspectral Imaging
The employed hyperspectral imager in this study is band-adjustable and can have 128 wavebands in maximum. Due to the limitation of the transfer rate and the storage capacity, to get more wavebands will be traded off by less available imaging area and larger pixel size. In this study, 64 wavebands with the pixel size of 1 meter with a range from 399.2 nm to 920.5 nm were selected.
2.2.1 Preparations of Ground Targets and Calibration Panels Before Aerial Imaging
To utilize the aerial hyperspectral images usually has two main problems of concerns. First, due to the current data transferring speed and storage capacity of storage devices, many hyperspectral imagers are pushbroom imaging systems. Because the instability of the airplane, to keep the imager scanning ground areas exactly along the designated route is difficult and usually results in more possible distortion than fixed-frame imagers (Lee and et al, 2000). Second, the hyperspectral imager measures reflected radiances of ground objects, which may change with respect to incoming radiances and may be hardly compared with other images of the same objects but different light conditions. More useful measuring unit to represent the signatures of measured objects is the ratio of the reflected radiance of an object to the incoming radiance. Therefore, during the aerial imaging, a set of white bright ground targets and a set of calibration panels with different gray levels all the way from visible wavebands and infrared wavebands were necessary to be properly placed in the desired imaging area.
Because the imaging area was a marsh, the ground targets needed to be water-proof, floatable, and anchorable in the marsh. One meter by one meter white foam boards coated with plastic films were used as ground targets, because they were water-proof, floatable and more resistant to bending stress with plastic films. Two wood sticks cross each other were affixed on the back of each white foam board to increase its resistance to winds and waves. A fishing line with 100 lbs resistance was tied up onto the wood frame and linked to a concrete block served as an anchor. An airboat trip was taken in the Ft. Drum Marsh before the imaging day to place the ground targets and record their coordinates using a code differential GPS unit (Model Pathfinder Pro XR, Trimble Navigation, California).
