EXECUTIVE SUMMARY INTRODUCTIONThe purpose of this report is to summarize a multi-participant effort which was initiated to demonstrate the potential benefits of using remote sensing data to enhance the State of Hawaii's ongoing efforts in developing spatial analyses capabilities. This pilot effort in processing and analyzing remotely sensed data was embraced by State agency participants as a step towards furthering the development of a Statewide geographic information system (GIS). When the pilot project was initiated, the implementation of a Statewide GIS program was already well underway with several data layers including base information and environmental data already digitized and being used on a daily basis by State agency personnel. The process of merging the raster-based remote sensing data with the existing as well as newly developed vector-based GIS data not only presented new challenges, but also offered exciting potential for enhancing planning and resource management activities in the State of Hawaii.
This executive summary supplements a detailed technical report prepared by the project consultants. Since it is unlikely that the technical report will be published, copies of it may be reviewed at the University of Hawaii's Office of Technology Transfer and Economic Development, the Office of State Planning or the Department of Land and Natural Resources.
In March of 1993, State of Hawaii officials were informed of a National Aeronautics and Space Administration (NASA) initiative to establish an Environmental Analysis and Applications Program within NASA's Mission to Planet Earth Office. A major component of this program would be to develop applications to monitor global change, which in part would include a renewed push to use LANDSAT data. A source of concern was that the U.S. Congress had been supporting the LANDSAT program for many years, but only a relatively small segment of the population in the United States was actually using the data being collected. While the LANDSAT program generated much interest in the use of remotely sensed satellite data in the 1970's, the decision in the early 1980's to privatize the distribution of LANDSAT data drove up the cost, thereby drastically decreasing the use of this data.
In response to NASA's intended global change program, a consortium of University of Hawaii and Hawaii State agency entities sought Federal funding support from NASA for a pilot project to demonstrate potential applications of satellite data and its integration with GIS information. The University of Hawaii's Office of Technology Transfer and Economic Development (OTTED), representing a number of State of Hawaii and University of Hawaii Departments, applied for a Federal grant from NASA to support such an effort. The grant proposal was submitted by OTTED, serving as the administrator of the grant, on behalf of the State of Hawaii Office of State Planning (OSP), Department of Land and Natural Resources (DLNR) and the University of Hawaii's Department of Planetary Geosciences and Department of Geography. In August 1993, the above consortium was awarded a $150,000 grant from NASA to in part, demonstrate the potential benefits of the Environmental Analysis and Applications Program being developed by NASA's Office of Mission to Planet Earth, while helping the State of Hawaii re-establish its in-State remote sensing analyses and image processing capabilities.
With limited usable land area coupled with an abundance of threatened and endangered natural and native resources, State of Hawaii lawmakers have over the years, been advocates for effective land use planning and resource management. As such, the State of Hawaii has historically been interested in effective land use planning and resource management and have traditionally been supportive of solutions which would facilitate these activities.
The State of Hawaii, for example, had previously been a participant in NASA Ames Research Center's Western Regional Application Program (WRAP) from the late 1970's until the program was discontinued in the early 1980's. While participating in the WRAP, State of Hawaii personnel were introduced to the use of digital data collected by satellite to discern various classes and patterns of land use and land cover. As an adjunct to the application of remote sensing technology, State personnel were also introduced to the use of the relatively new, but still not fully developed technology of geographic information systems to integrate mapped information with processed satellite images. When the WRAP was discontinued in the early 1980's due to Federal budget cuts, the State was unable to continue its work in this area since the cost of implementing remote sensing and GIS technology in-State was prohibitively high at that time.
Significant advances in computer technology and GIS software eventually enabled the State to initiate a GIS program in the late 1980's. Since the State's GIS was envisioned to be a multi-agency system accessible to all State agencies with an interest in spatial data, the Office of State Planning within the Office of the Governor, was designated as the lead agency for planning and developing the State's GIS program. The foundation of the system was the development of a centralized database which would be accessible to all user agencies to avoid duplication of effort in the costly area of database development, while maximizing efficiency in data management in State government.
Although the implementation of a Statewide GIS offered a wide range of applications, the initial focus of the database development efforts were concentrated on land use planning and resource management uses. As such, in addition to the development of the requisite base maps, most of the data layers which were initially developed included environmental and natural resources-related data layers.
When NASA representatives contacted State of Hawaii officials in March 1993 to discuss the potential benefits of incorporating remotely sensed data with the State's existing GIS to support resource planning and ecosystem management decision making, the State of Hawaii expressed its keen interest in such an initiative. The State believed that such an effort would not only help augment the existing GIS database, but would also provide an opportunity to evaluate the feasibility of using LANDSAT Thematic Mapper (TM) imagery to develop land use, land cover and vegetation data sets. In its previous experiences through the WRAP, the State's LANDSAT Demonstration Project utilized Multi-Spectral Scanner (MSS) data to develop land use/land cover classifications. Therefore, since the State had already worked with MSS data, this particular initiative offered the State an opportunity to evaluate the viability of using TM data which has a higher resolution and more spectral bands than the MSS.
By carrying out a pilot project involving the processing of LANDSAT TM data to develop a classified image, the State of Hawaii hopes to attain the following project goals.
A collaborative effort by and between the State and University of Hawaii entities was initiated in order to meet the stated project goals. OTTED, as the grantee, was in charge of overall project management and federal grant administration. OTTED also served as the State's point of contact with NASA's Earth Observing System Program. The University of Hawaii's Planetary Geosciences Department was in charge of receiving, processing, geo-rectifying and producing multiple sets of the acquired data onto CD-ROM. They also were the point of contact for distributing copies of the CD-ROM to interested users in the State of Hawaii. The University of Hawaii's Geography Department initiated a training curriculum in image processing techniques. This was to supplement their existing state-of-the-art cartography and GIS curriculum. The Department of Land and Natural Resources and Office of State Planning assumed the role of the application project team in order to develop in-house image processing expertise by actually carrying out an image analysis and classification pilot exercise for the Island of Hawaii. OSP served as the State agency liaison with OTTED and was also responsible for upgrading existing State GIS equipment and procuring image processing software for State agency use. DLNR was in charge of coordinating and leading the necessary field work efforts on the Island of Hawaii.
Since the State and the University of Hawaii did not have the necessary expertise to complete the pilot project, and in order to effectively carry out the technical requirements of the grant, the State opted to use a portion of the federal grant to hire a consultant to facilitate the technology transfer process. In this regard, OTTED solicited proposals to develop in-State expertise in analyzing and processing remotely sensed data and then applying such information to the State's GIS.
A proposal submitted by Geographic Decision Systems International (GDSI) along with subconsultant, Chris Hogan, who was involved with the State in the previously mentioned WRAP, was selected. The GDSI proposal offered image processing training, pilot project support and integration of the processed imagery to the State's Arc/Info GIS.
PROJECT SUMMARY Apart from the federal grant, NASA was committed to providing the State, up to ten LANDSAT TM scenes as part of their bulk purchase of data from EOSAT, Inc., the company responsible for distributing LANDSAT data. While the State requested ten cloud-free images of the major islands, NASA was only able to acquire seven scenes which are of marginal quality due to heavy cloud cover. For the Island of Hawaii, which was the designated pilot study area, four scenes were required to mosaic together a full island scene image.
The University of Hawaii's Department of Planetary Geosciences was in charge of receiving and pre-processing the raw digital TM data files. The data were delivered on 9-track tape. Once received by Planetary Geosciences, the data were processed, re-formatted, and transferred to CD-ROM. The pre-processed data were then provided to the State agencies for image processing analyses. The Planetary Geosciences Department used grant funds to help further establish their capabilities in pre-processing raw digital remote sensing data. The project helped demonstrate their ability to receive and prepare such data to facilitate the State's as well as other users' use of such data.
Once the data were pre-processed, the CD-ROM's were delivered to the State of Hawaii where they were loaded onto the State's GIS file server. The data were imported into ERDAS/Imagine for image processing. The scenes were then geo-rectified to UTM Zone 4 to comply with the State's GIS database.
Before actually beginning the process of classifying the image, a mosaic of the raw data for the four scenes was produced. This initial mosaic helped eliminate some of the problematic cloud cover and gave the project team a full island scene to serve as a visual source to refer to during the classification process. Figure 1 shows the resultant mosaicked raw image.
The unsupervised classification provides a "first look" statistical display of natural groupings of data based on spectral values in an image. Unsupervised classifications were performed on each of the scenes using ERDAS Imagine software. The resultant classifications were then evaluated for accuracy.
Evaluations included field verification of predicted vs. actual landcover for approximately 25 sites throughout all major landcover categories. Global Positioning System (GPS) technology was used to establish locations of ground cover types for the purpose of "training" the computer to recognize the spectral characteristics of these particular land cover types. A GIS point coverage was generated from the GPS points and field data were overlaid on the classified image. The results were evaluated visually by the project team. In addition to field verification, the classified map products were also distributed to environmental professionals around the State for comment. Feedback was generally good and the products were considered useful at the island scale, but were not considered useful without a quantitative assessment of accuracy. Figure 2 is the resultant unsupervised classification of one of the scenes.
Following evaluation of the unsupervised classifications, supervised classification was conducted on the four scenes. Using the "intelligence" gained by linking GPS sites with known coverages, a supervised classification was carried out. After this first supervised classification was completed, the project team took a second field trip to refine the supervised classification.
The classified images were then mosaicked. The mosaicked products were not aesthetically pleasing to the eye however, because of the differences in season and year for the four images. Differences in moisture, "greenness" and groundcover looked like fence lines or other artificial artifacts at the scene boundaries.
Simultaneously, the unsupervised geo-referenced image was mosaicked using a model developed by the technical consultant. The reason for this was that the mosaicked algorithm supplied with IMAGINE did not yield satisfactory results. Supervised classification was then run on the mosaicked image. This resulted in less conspicuous boundaries between scenes (Figure 3).
The final maps were again assessed for accuracy by field verification and evaluation by environmental professionals familiar with the Big Island. Assessments were qualitative and not quantitative. Those professionals consulted indicated that the supervised classification performed on the mosaicked imagery was the most accurate and useful of the final products.
RESULTS The imagery that was provided by EOSAT was marginal at best. Cloud cover, which is always a problem when collecting remotely sensed data of the Hawaiian Islands, masked significant areas of the land area. Furthermore, the fact that the data that were provided were collected in different years and different months, made the classification of the entire island that much more difficult.
As previously mentioned, ERDAS Imagine was used to carry out the image processing portion of this pilot project. ERDAS had a distinct advantage over other image processing software available because of its compatibility with the Arc/Info GIS software. Imagine had only just recently been ported to HP UNIX, the operating system for the State's GIS. As a result, the first release of Imagine for this OS was very "buggy" and support for peripheral devices including the State's CD-ROM, tape drives and printers was absent. Bugs and system incompatibilities initially slowed down the project team considerably until patch tapes and work around solutions were developed by ERDAS.
This project has demonstrated the value of remotely sensed data in the State of Hawaii GIS database. The raw imagery is used extensively as a background cover, and the classified products have already been used on several projects. Such projects include realignment of the saddle road on the Big Island through endangered species habitat, and for development of fire management plans on Mauna Kea. The State Department of Agriculture and the Department of Land and Natural Resources have expressed interest in using the classified images to monitor the distribution of weedy species such as gorse, fire tree and fountain grass. The State has indicated that it will continue to investigate the use of remotely sensed data in managing some of its more serious weed problems, including banana poka and miconia.
The State is continuously looking for new sources of data to incorporate into the GIS database. In 1995, for example, a Statewide purchasing consortium consisting of 26 participants was developed to purchase wall to wall coverage collected from the French satellite, SPOT. Full sets of panchromatic and multispectral imagery are expected to be delivered by SPOT Image Corporation by the end of 1996. Users are now evaluating a first partial delivery of this new imagery. Preliminary results indicate that users like the spatial resolution of the SPOT imagery, but prefer the spectral richness of the LANDSAT TM data.
Given the uncertain status of LANDSAT and any future launches, it is important for the State of Hawaii to explore the use of alternative sources of remotely sensed data. Other satellites launched by countries including India and Japan are circling the earth and are such potential sources for additional satellite imagery. It is anticipated that as State of Hawaii users become more comfortable with remotely sensed imagery, these different data types will become regular components of the State's GIS.
