Mapping of Eelgrass and Other SAV Using Remote Sensing and GIS Chris Mueller NRS 509 November 30, 2004

Transcription

1 Mapping of Eelgrass and Other SAV Using Remote Sensing and GIS Chris Mueller NRS 509 November 30, 2004 Of the 58 species of seagrass that grow worldwide, Zostera marina, commonly called eelgrass, is by far the most common along the eastern coast of the US. Its range once extended almost continuously from Nova Scotia down to South Carolina on the East coast and from Alaska to the Gulf of California on the West coast. Its ability to grow is usually limited by the availability of light, and it is therefore confined to water less than approximately 2 meters deep. Eelgrass is a highly productive marine subaquatic vegetation (SAV) that provides important benefits to ecosystems in which it grows. These benefits include improving sediment stability, serving as a food source for various organisms, reduction of shoreline erosion due to lessening of wave energy, and providing refuge habitat for juveniles of some commercially important finfish species. In the 1930 s, North Atlantic populations were nearly decimated by a virulent outbreak of a marine slime mold. In the decades following World War II, public and scientific awareness of the importance of this habitat allowed eelgrass to recover much of its former range. However, anthropogenic impacts such as increased fertilizer use and a steady increase in coastal development have degraded near shore water quality by increasing turbidity and eutrophication. In turn, these factors slowed and eventually reversed the recovery process of eelgrass. Studies have estimated that since the 1970 s, between 45 and 70% of the eelgrass habitat that recovered after the wasting disease outbreak has been lost again. In the past, field surveys have been the predominant method used to determine the existence and extent of eelgrass habitats. This method can be very accurate, however it is also very labor intensive and time consuming and the quality and accuracy of the data can vary depending on the survey methods employed. Aerial photography has also been used to map eelgrass, however the photos are often very expensive to obtain, which could prove problematic for an application such as this, where the extent and shape of a seagrass bed can change from year to year. This literature review will give an overview of the technologies used to map eelgrass and other subaquatic vegetation as well as elucidating some of the difficulties associated with this endeavor. To do this comprehensively, it is necessary to look beyond eelgrass in particular, and focus attention on the mapping of seagrass habitats in general. The environment plays a critical role in the ability of remote sensing technology to accurately describe SAV habitats. Factors such as bottom type, water column clarity, water depth, and the presence or absence of habitat with similar reflective characteristics to that of the SAV can all result in the misclassification of pixels in an image. This is true for the majority of the commonly employed technologies, including Landsat and SPOT imagery and multispectral images taken from aircraft (such as the digital images taken by a Compact Airborne Spectrographic Imager). These issues are in addition to the normal difficulties encountered when using these remote sensors such as atmospheric distortions and cloud cover. Aerial photography has been the most widely used method of mapping SAV habitats for many years. The photographs can be manually interpreted and can result in an incredibly accurate delineation of the habitat, usually with more than 90% accuracy. The largest drawbacks to this method are the labor intensive process of interpreting and digitizing the images, the difficulty of correcting the image for distortion (this can be very

2 difficult if the image does not contain any land), and the relatively high cost of obtaining the photographs in the first place. Despite these difficulties, the images produced by the aforementioned technologies have many important advantages over the labor and time intensive process of field surveying. They can cover relatively large areas in one image, they can be manipulated in various ways to illuminate different aspects of the environment, and perhaps most importantly, they can be easily integrated into a GIS and combined with other data to create a remarkably accurate picture of the location, size and even the density of SAV habitat. One quite new technology that avoids many of the problems associated with remote sensing has recently come out of Japan (Komatsu et. al, 2003). A team used multi-beam sonar to map an SAV bed, producing some startling results and showing a great deal of promise for the technique. The scientists were able to map an area of over 17,000m 2 in just under an hour, and the resulting data was easily processed to create a three dimensional visualization of the seagrass bed as well as providing estimates of both the biomass and coverage area of the bed. This technique does not have the misclassification problems associated with other remote sensing technologies, however it can image only a small fraction of the area covered by one aerial photograph or satellite image and the costs of equipment, labor and boat time could prove to be substantial for a large area survey. The particular characteristics of the seagrass habitat can also affect the effectiveness of remote sensing. Studies repeatedly find that satellite and aerial images are most accurate when the bed is in shallow water (<3m) and when the bottom type surrounding the bed is sandy. Deep water and certain bottom types tend to result in spectral characteristics that are very similar to those produced by SAV. The issue of water depth can be corrected for to some degree by incorporating accurate bathymetric data to the image prior to analysis, thereby allowing the differences in depth to be adjusted for during classification of the reflective signals. Additionally, the proportion of misclassification increases when the bottom habitat is composed of a variety of substrates, such as when a seagrass bed is in close proximity to large areas of macroalgae, mussel beds, or coral reef. The magnitude of this error can be lessened by segmenting the image into the various habitat types using visual interpretation, and then classifying each segment independently. Overall, small scale satellite images, such as those taken using SPOT and Landsat, are most accurate when attempting to get a big picture of an SAV habitat, especially when the habitat is in shallow, clear water that does not have large bathymetric relief. On the other hand, aerial photography and airborne multispectral imagers such as CASI can provide remarkably detailed information on the size, shape and relative density of an SAV habitat in addition to being able to handle a larger range of depths and water qualities. Once an image is incorporated into a GIS, the value of these technologies can be truly appreciated. The data can be combined with other datasets such as bathymetry, sediment type, coastal land cover and population (just to name a few) to result in novel interpretations of how SAV habitat is affected by these factors. With repeated images of the same location, it is also possible to gain an understanding of how an SAV habitat changes temporally or how it responds to storms and various anthropogenic influences such as dredging or point source pollution. It can even be used to determine the degree of fractionation within a seagrass bed by calculating perimeter to volume ratios for the various seagrass polygons. Conversely, areas where seagrass beds were actively

3 growing could also be determined. This information can in turn be utilized by managers to determine the best places for restoration or conservation efforts.

4 Annotated Bibliography Ferguson, R. L. and K. Korfmacher (1997). Remote sensing and GIS analysis of Seagrass meadows in North Carolina, USA. Aquatic Botany 58: This paper by Ferguson and Korfmacher presents the results of a study in which they used three Landsat TM images to determine the extent of seagrass meadows along the coast of North Carolina as compared to aerial photographs of the same location. The authors describe the costs associated with the use of Landsat TM images as being less expensive than aerial photography. A comparison of the two techniques is given, highlighting the advantages and disadvantages of each. The most interesting portion of this article is the conclusion that the major limitation of using Landsat imagery is the necessity that numerous environmental factors be favorable to detection of seagrass coincidently. In addition to the atmospheric and seasonal constraints faced when determining dry land cover, hydrologic constraints such as water depth and turbidity make the accurate detection of seagrasses difficult. Mumby, P.J., E.P. Green, A.J. Edwards, and C.D. Clark (1997). Measurement of seagrass standing stock using satellite and airborne remote sensing. Marine Ecology Progressive Series. 159: This study presents a comparison of seagrass habitat delineation by Landsat Thematic Mapper, SPOT XS multispectral imagery, and CASI digital airborne imagery (also multispectral). Limitations and benefits of each technique are discussed especially with respect to resolution and study area size. The remote sensor techniques were compared to field measurements to show that standing crop estimates by the three methods were highly accurate. In addition, the cost associated with each method is briefly discussed. The most useful portion of this paper is the suggestion that smaller scale changes to individual beds is most accurately and easily determined using aerial images, while satellite imagery is more efficient for changes on a system-wide scale. Komatsu, T., C. Igarashi, K. Tatsukawa, S. Sultana, Y. Matsouka, and S. Harada (2003). Use of multi-beam sonar to map seagrass beds in Otsuchi Bay on the Sanriku Coast of Japan. Aquatic Living Resources. 16: In this paper the authors discuss a case study in which they employed multi-beam sonar to map the distribution, biomass and canopy height of a seagrass bed off the coast of Japan. The sonar signals were processed to remove the bottom signal and the result was a clear, 3-D image of the seagrass bed. Previous studies using side-scan sonar or echo sounding to map beds resulted in either horizontal or vertical images respectively, but not a three dimensional view. The authors also collected field data to verify the accuracy of the technique. A gyrocompass and a DGPS unit were used to ensure spatial accuracy during the collection of the sonar data. The most interesting aspect of this paper is the suggestion that this method allowed almost instantaneous view of the seagrass bed, and that with only minimal processing they were able to determine biomass and volume of the plants. Additionally, the speed at which the data was collected (40 minutes for a 115m X 156m area) is very fast for the level of accuracy attained in the experiment.

5 Ackelson, S.G., and V. Klemas (1987). Remote Sensing of Submerged Aquatic Vegetation in Lower Chesapeake Bay: A Comparison of Landsat MSS to TM Imagery. Remote Sensing of Environment. 22: This paper compares the ability of Landsat Multispectral Scanner imagery (80m resolution) and Landsat Thematic Mapper imagery (30m resolution) to detect submerged aquatic vegetation (SAV). Images from the two systems were compared to known seagrass beds delineated from color aerial photographs on a pixel-by-pixel basis to determine their relative accuracy. While the authors found that there was no significant advantage of Landsat TM images over the older MSS imagery, they do caution that differences in sediment reflectance, canopy height and the reflective properties of the water itself could result in some added benefit from the TM images. One very interesting conclusion that the authors draw is that the accuracy of classification can be greatly increased by merging the classified image with depth information, as deep water was often misclassified as SAV cover. Pasqualini, V., C. Pergent-Martini, G. Pergent, M. Agreil, G. Skoufas, L. Sourbes, and A. Tsirika (2005). Use of SPOT 5 for Mapping Seagrasses: An Application to Posidonia oceanica. Remote Sensing of Environment. 94: In Press. The authors of this paper used imagery from the SPOT 5 satellite to determine seagrass distribution in Laganas Bay, Greece. Images were taken at two resolutions (10m pixels and 2.5m pixels) and analyzed to determine seagrass coverage. Image pixels were categorized into 3 classifications and the accuracy of classification was checked against field measurements. The study found that the 10m pixel image provided much more accurate classification overall (~90%) than the 2.5m pixel image (73%). The authors also found that, while the 2.5m images provided a better picture of patchiness in the seagrass bed, due to the fact that the 10m pixels were simply too large to elucidate such small spatial features. Most interestingly, both images had great potential for fine-level habitat mapping and that the accuracies produced by the SPOT system were comparable to those produced by IKONOS imagery. Pasqualini, V., C. Pergent-Martini, P. Clabaut, H. Marteel, and G. Pergent (2001). Integration of Aerial Remote Sensing, Photogrammetry, and GIS Technologies in Seagrass Mapping. Photogrammetric Engineering & Remote Sensing. 67(1): This paper highlights the use of color, infrared, and black and white, aerial photographs to evaluate the spatial distribution of a seagrass bed. Aerial photographs with a 1m pixel resolution were georeferenced using photogrammetry and GIS was then used to combine the aerial images with bathymetric and cartographic datasets, thereby allowing accurate spatial distributions to be determined. Once again, the authors found that depth and water clarity were the biggest problems with this technique. This paper is interesting because it utilizes aerial photography from various dates to observe temporal change of the seagrass bed. The authors were also able to use the bathymetric data to create a three-dimensional model of the bed, which could also be observed to change over time. Additionally, the authors conclude that while black and white or color aerial photographs can provide a detailed environmental map, the infrared photography was able to optimize the detection of marine plant formations.

6 Additional References Lehmann, A., and J.B. Lachavanne (1997). Geographic Information Systems and Remote Sensing in Aquatic Botany. Aquatic Botany. 58: Green, E.P., P.J. Mumby, A.J. Edwards, and C.D. Clark (1996). A Review of Remote Sensing for the Assessment and Management of Tropical Coastal Resources. Coastal Management. 24: Long, B., T.D. Skewes, and I.R. Pointer (1994). An Efficient Method for Estimating Seagrass Biomass. Aquatic Botany. 47: