GAP - Bulletin Number 6 - A Preliminary Analysis of GAP Land Cover Mapping Procedures

A Preliminary Analysis of GAP Land Cover Mapping Procedures

Marlen D. Eve, James W. Merchant, and K.C. Kroll
Center for Advanced Land Management Information Technologies (CALMIT), University of Nebraska-Lincoln

Accurate maps of land cover are clearly critical to the GAP Program. In order to facilitate regional and national-level Gap Analysis and to allow for studies of change over time, individual states must be mapped so as to ensure state-to-state consistency. Each state project, however, has a unique blend of technical expertise and capabilities. Each is dealing with a unique mix of land cover types. And, each has a unique set of statewide cooperators and project goals to satisfy. In order to balance all of the needs and create the most useful products possible, the GAP community has shown a great deal of innovation in mapping land cover. A wide variety of methods have been used by individual states to map their land cover. At the Center for Advanced Land Management Information Technologies (CALMIT) at the University of Nebraska, we have been conducting a study of land cover mapping protocols and methods used to develop land cover maps in support of the Gap Analysis Program. This paper is a preliminary report on the findings of this project.

Our working premise has been that land cover mapping is, in many ways, as much an art as a science. As such, the project has been a learning experience and is providing the opportunity to document and synthesize the current development of land cover mapping within the Gap program. The principal objectives of this project are to:

Inventory, compare, and (to the extent possible) evaluate land cover mapping protocols and methods used by the states to develop land cover maps for Gap Analysis; and
Provide information and recommendations to the National GAP office that will enhance future GAP-related land cover mapping efforts. Furthermore, we have been trying to identify the "common threads" that seem to be woven throughout each of the individual projects. We have used personal communication, published literature, written summaries and reports, and Internet home pages to gather information on logistical issues (such as hardware, software, and staffing); data utilized; land cover classification and delineation, labeling, and accuracy assessment; and data handling, archiving, and dissemination. Following is a sampling of what we have learned thus far.

Cooperators and collaborators have seen the usefulness and importance of the land cover information being generated under the Gap program. Almost every state project reported gaining supplemental funding in some form from other agencies. About 44 percent of the people working on land cover mapping are graduate students, showing that GAP is contributing to education and the development of technological skills. While numerous hardware and software configurations are being utilized, most of the mapping is being conducted in a UNIX environment using ERDAS Imagine and ARC/INFO software. Early GAP projects had to acquire their own satellite imagery, but later projects have had the benefit of the MRLC national Landsat TM purchase. Most states are attempting to utilize multiple dates of TM imagery where they are readily available. Seasonal differences in the vegetation have helped in some cases and added confusion in others.

The actual delineation of vegetation classes has been quite variable. Approaches have included any or all of: photointerpretive techniques, supervised and unsupervised clustering, photointerpretive or machine labeling of the MRLC-generated hyperclusters, and modeling using ancillary data sets. Numerous types of ancillary data have been used to aid the land cover mapping. The most commonly used data are aerial photographs, field data, airborne video, and existing land cover maps. The mapping is being done in most states on a scene-by-scene basis, commonly with each scene being stratified using ecoregions or other physiographic or edaphic data. Each project has had to develop mapping techniques that fit the type of land cover in their area and that matches their level of expertise and specialization. Accuracy assessment has been especially variable. Some projects have not had the time or funding to conduct formal accuracy assessment, while others have spent considerable time and effort collecting and analyzing thousands of field or video samples. No matter what approach or level of accuracy assessment is undertaken, the common thread seems to be the need for some field-based data and collaboration with local vegetation experts.

Overall, the development of the GAP land cover mapping effort has resulted in several very positive things. It has provided seed money for further advancements in land cover mapping, stimulated cooperation and collaboration in mapping, increased the acceptance and adoption of remote sensing and GIS as mapping tools, and assisted in the development of numerous new techniques for land cover mapping.

This success has not come easily, though. Trouble spots have included: getting access to adequate and appropriate seasons of imagery, dealing with shadowing effects and clouds, collecting adequate field and ancillary data, edge-matching with adjoining states, and achieving consistency of vegetation map legends.

At CALMIT, we are currently working on gathering additional project information and synthesizing and analyzing all of the information. We will be finished with our analysis and submit our report to the National GAP office by early 1998. Currently, an Internet home page is under construction for the project. This site is where the project report will be assembled and, as such, will contain all of our findings and analysis, as well as any conclusion or recommendations. The site will also feature a link for feedback, corrections, or other information. The URL for the home page is http://www.calmit.unl.edu/gapmap.