[an error occurred while processing this directive][an error occurred while processing this directive]
Using GIS for Image Analysis and Land Cover Classification
Preserving Biodiversity in Russia's Meshora Lowland
The Russian Federation is an enormously large country with an equally large and extensive natural resource base. Its biological diversity is impressive. For the most part, the federation's 145 million inhabitants are confined to the Russian Plain west of the Ural Mountains. Therefore, the human impact on the environment tends to be somewhat less severe in those vast Siberian expanses east of the Urals but much more pronounced in the west. While a number of nationally protected environments are in jeopardy in the Russian Federation, one area of great concern to Russian and non-Russian scientists is the Meshora Lowland.
The Meshora Lowland, although largely unprotected, contains two national parks (Meshora and Meshchorskiy) and a strictly protected reserve (Okskiy Zapovednik), all three of which were created in an attempt to protect and preserve hundreds of species of rare and endangered flora and fauna.
The problem is that the lowland, including the parks and the reserve, is located adjacent to nearly 15 million people, the majority of whom live in the cities of Moscow, Vladimir, and Ryazan'. Thousands of other persons, however, live in hundreds of small agricultural villages and a few larger towns. All of these people, their juxtaposition to the lowland, and their ability to access and utilize the resources of the lowland potentially will have a long-term, damaging impact on it unless comprehensive, rather than merely local, policies regarding use are developed. Additionally, the lowland is located within the political jurisdiction of three separate political entities.
It is vital, therefore, for scientists to carefully study the Meshora Lowland so its unique biodiversity may be sufficiently managed, protected, and preserved for the enjoyment of future generations. It is also vital that all political entities work together in that process.
The Pra River Watershed
Since 1999, Dr. Brooks Green, associate professor of geography at the University of Central Arkansas, and his Russian colleagues (Dr. Elena V. Biryukova and Dr. Alexander Pribylov) at Ryazan' State Pedagogical University have been studying the Pra River Watershed, a significant drainage basin in the central portion of the lowland. The Pra River Watershed is a 180- by 48-kilometer, north-to-south trending system of small streams, shallow rivers and lakes, extensive marshes, vast meadows, and a dense Southern Taiga forest. Gentle gouging that occurred during the most recent glacial period essentially shaped the watershed's relief and drainage pattern. The watershed also contains rather large areas cleared and drained for a variety of agricultural, lumbering, peat extraction, and industrial purposes.
The purpose of this research, utilizing state-of-the-art GIS tools and image processing techniques, is to clearly identify the land use/land cover and the cultural resources (archaeological sites, historic structures, etc.) of the watershed. This analysis, when completed, will ultimately enable those individuals in the Decision Support System (DSS)federal, regional, and local government officials, land use managers, park directors, and park/reserve scientiststo better develop and implement management and use policies for the preservation and protection of the delicate watershed resources.
Advances in remote sensing and GIS techniques enabled the entry point in this research project. The project is initially focused on the natural resources of the watershed. With funding from the University of Central Arkansas Research Council, Green obtained two Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images of the eastern portion of the lowlandthe area that contains the Pra River Watershed. The Landsat 7 ETM+ images contained bands 1, 2, 3, 4, 5, 6h, 6l, 7, and 8 (the panchromatic band). A Hewlett-Packard CPU with a Pentium 4 processor on a Windows XP platform was used to apply the appropriate image processing and GIS analytical techniques.
Because of the large size of the two merged satellite images (12,741 rows by 9,761 columns), Green and his colleagues used the Analysis Mask option in ArcGIS 8.3 (ArcView, ArcEditor) to delimit and extract the watershed. That process enabled subsequent operations to be completed more quickly and efficiently.
Once the watershed had been masked and extracted from the larger image set, they completed a land use/land cover classification utilizing the ArcGIS 8.3 Classified option in the Layer Properties dialog box. Because members had good knowledge of the watershed, they completed a supervised manual classification with known training sites. (A supervised classification is one where the person classifying the image is familiar with the area and can choose sitescalled training sitesthat are known. An unsupervised classification is one where the person classifying the image is not familiar with the area and simply permits the program to identify similar pixels of the image.) That manual classification process enabled the researchers to carefully isolate and highlight specific ranges of data to select specific land use/land cover types.
Although the resulting land use/land cover classification map appeared to be very good, it was determined, for comparative purposes, that an unsupervised classification would be conducted. In this case, 10 classes were selected.
The unsupervised classification analysis utilized the ArcGIS 8.3 Classified option in the Layout Properties dialog box. The results of this analysis were found to be superior to the manual classification technique. The research team concluded that the 10 classes more accurately portrayed the land use/land cover found in the Pra River Watershed, although they were unsure of specific differences in forest types and agricultural classifications.
The first stage, that is the classification stage, is now successfully completed.
The Next Stages
The next stage will be ground truthing (conducting fieldwork at the site) the classification. To guide the research team in its collection of field data on the various land use/land cover classifications in the watershed, it will use the unsupervised classification map created in ArcGIS 8.3. Following ground truthing, the classification will be corrected where needed so that actual land use/land cover in the classified image matches what is really on the ground (the actual land use/land cover). At this juncture, cultural features will be added for analysis and policy management and development. The final stage will be to test the impact of changes in land use and land cover.
DSS will be given the updated maps and data sets so its members can begin to make sound adjustments in management policies and more informed and improved decisions regarding activities that are now occurring within the watershed or will occur in the future.
For more information, contact Dr. Brooks Green, associate professor of geography, University of Central Arkansas (tel.: 501-450-5636; fax: 501-450-5185, e-mail: email@example.com).