Offshore Fishpond Wall

at

Kualoa Regional Park

Geographical Research and Mapping (G.R.A.M) Program

California State University Long Beach &

National Science Foundation – Research Experiences for Undergraduates

By

Scott Honda,  June 2013

Advisors  –  Dr. Carl Lipo,  Dr. Suzanne Wechsler,  Dr. Chris Lee,  Dr. Matt Becker

Off the shore of Kualoa Regional Park lies a prehistoric fishpond wall. The wall’s existence is of great importance, as it can give insight into the past practices and the history of Native Hawaiians. It can help in the planning and management of Kualoa Regional Park by the City and County of Honolulu and to aide in the general knowledge of the Kualoa area by academic and community members. The purpose of this project is to provide the location of the off-shore fishpond wall using technologies from both the remote sensing and geographical information science disciplines.

Kualoa Regional Park is run and operated by the City and State of Honolulu and there have been numerous archaeological studies of the area. Some of the past studies mentioned the existence of the remains of a fishpond wall off the shore of Kualoa Regional Park, but it seems none have done any extensive study or survey of the wall. According to A. F. Judd, the fishpond wall was destroyed by a huge storm, which can now be seen at low tide (Gunness 1987). The wall is quite possibly the remnants of a fishpond entirely separate from the near by Moli’i and Apua fishponds as one study wrote “a local resident of the area informed the staff that, as a child, he was told this pond was called ‘Pilihe’e’” (Gunness 1987, 47). The remnants of the wall is about 500 meters from the south-eastern tip of Kualoa Regional Park to the most southern point of the wall and is about 550 meters long.
In order to determine the location of the offshore fishpond wall both remote sensing and GIS disciplines were used. Methods from remote sensing included pansharpening, principal component analysis, and index ratios. As a result of the curse of dimensionality, all methods were then used in some combination to create an array of both useful and useless data pertaining to this project. With these methods combined with the various tools and concepts from geographical information systems, the still existing fishpond wall was located.

The fist initial step that was taken to determine the location of the offshore fishpond wall was to locate it using the World View 2 imagery from 2011, in ERDAS. The imagery was first pansharpened so that the multispectral bands could have the higher spatial resolution of 0.5 meters. Using the pansharpened imagery all terrestrial areas were masked out of the imagery in order to only process the offshore areas, which allows the digital numbers of offshore areas full range of the DN range. Here after two processing paths developed. The first was to run a Non-Homogenous Feature Difference, an index ratio which highlights areas that contrasts against the background that can be identified as man-made (Wolf 2010).

After the Index Ratio is executed the geometric shape of the off shore fishpond wall is highlighted as a curved brighter shape. It is uncertain if the wall actually extends to land as close inspection of the imagery may suggest. The second was to do a principle component analysis of the pansharpened image to help highlight features in the image. The results was promising. Once the principle component analysis was done different band combinations was used to create an image that highlighted the offshore fishpond wall. The process of finding the “right” band combination was done by trail and error. The resulting image highlights the same geometric shape of the fishpond wall as did the Non-Homogenous Feature Difference, only in an RGB image.

Out of curiosity, a index ratio was then done on the principle component analysis image. The formula used was as follows (Yellow – Coastal Blue)/(Yellow + Coastal Blue). The result highlights the same offshore fishpond wall.

UAV imagery of the offshore fishpond wall was made in an attempt to capture the wall in its present day conditions and for use in photogrammetry based processing software. The first challenge was judging the distance needed for the UAV (Phantom Quad-Copter) to fly for the sensor (Pentax Optio) to photograph the wall. Although more sophisticated UAV systems exist, the Phantom was the only tool available at the time. The second challenge was to figure out the altitude of the UAV needed. This is to ensure that the processing software (PhotoScan and Microsoft ICE) could be able to create a mosaic or orthophoto of the offshore fishpond wall. Previous attempts using low altitude (20 – 30 meters) flights provided imagery unusable in PhotoScan and Microsoft ICE. It was later found, through trial and error, that an altitude of 45 – 60 meters is sufficient in imaging the offshore fishpond wall. One mosaic was able to be derived from Microsoft ICE. A part of the offshore fishpond wall was photographed. On the bottom left corner of the mosaic the bend in the fishpond wall is imaged. PhotoScan was unable to create a orthorphoto or DSM with the same imagery.

A LiDAR point cloud of the area was used to create bathymetry of the area in ArcGIS. The point cloud was used in a Natural Neighbor interpolation to create the bathymetry. In the 0.25 m resolution bathymetry the shape of the offshore fishpond wall is also distinct.

GPS points were gathered using a recreational grade GPS (Garmin etrex) was gathered. Features of interest were coral heads and reef with high displacement. These points will be useful in an overall assessment of the area.

In any under water survey and mapping situation the use of sonar seems logical. For this project sidescan sonar was made available for one day. However, because the timing of refinement in focus of this research project, the sidescan sonar was of other areas and not of the area near the offshore fishpond wall.

Using the NHFD image a few relative measurements were made. The pond, of the offshore fishponds wall, would have incorporated more than 2,000 acres. The fishpond wall in it’s fully intact state may have been more than 1,300 meters in length and the remaining wall (as seen in the imagery) is about 500 meters in length. At it’s current state the remaining wall is about 20 meters wide. The original wall may have been around 20 meters wide or the wall may have been higher at one point.

From the imagery and data gathered, the offshore fishpond wall seems to be the remnants of a fishpond that is discrete from the Moli’i and Apua pond, perhaps it is the Pilihe’e fishpond referred to by the local resident Gunness mentioned. The use of remote sensing and geographical information systems has been a powerful tool in locating the offshore fishpond wall.

Based on the Non-Homogenous Feature Difference, the Principle Component Analysis imagery, and the LiDAR point cloud, the use of remote sensing (both active and passive) to locate the offshore fishpond wall were the most powerful tools. All three methods were the most definitive and capable in “finding” the offshore fishpond wall. It must also be mentioned that because only the contrast of the wall from the surrounding reef is being sought after, a spectral index ratio like the NHFD is useful as the Digital Numbers of the pixels can demonstrate contrast. If one were attempting to derive the health or biomass of vegetation the the use of reflectance would be more appropriate.

UAVs allowed for up-to-date imagery. Although more sophisticated UAV systems are available, the hobbiest grade Phantom Quad-Copter is capable (given enough time) of imaging the offshore fishpond wall. The post processing of UAV imagery in photogrammetry based software like PhotoScan, could be a useful and powerful method of finding and surveying under water features. However, in this case all images were not usable in PhotoScan this is probably due to distortions introduced by water. Perhaps if a multispectral camera, with bands that could penetrate water (i.e. Coastal blue) were used Photoscan would be capable of creating an orthophoto and a digital surface model of under water features.

Taking into consideration the GPS points of coral it seems that the offshore fishpond wall plays a role in protecting coral from ocean currents as most coral heads found during this project were landside of the wall. A more extensive survey of the area, both land-side and ocean-side of the wall, is needed in order to determine the extent of which the offshore fishpond wall plays in the coral’s survival and overall influence on the dynamics of the area.

The use of side scan sonar could also be a essential tool in surveying the area. However, the depth of the water is the limiting factor as the side scan sonar is capable of collecting the area under it at a 2 to 1 ratio. If the depth of the water is only 1 foot then a 2 feet area will be captured. Thus, the use of side scan sonar would gather the most data in the least amount of time during high-tide.
Although the offshore fishpond wall has been derived from the world view 2 imagery and the LiDAR based Bathymetry, more studies of the wall at a larger scale would be appropriate. A large (spatial) scale study of the wall could determine it’s condition, influence on offshore and on shore processes, and cultural significance.

Jo Lyn Gunness, “Archaeological Investigations at Kualoa Regional Park” (M.A. diss., University of Hawaii, 1987)

Antonio Wolf, “Using WorldView 2 Vis-NIR MSI Imagery to Support Land Mapping and Feature Extraction Using Normalized Difference Index Ratios,” Digital Globe (2010), http://www.exelisvis.com/portals/0/pdfs/envi/8_bands_Antonio_Wolf.pdf