Previous Page Table of Contents Next Page


2.1. Satellite data selection
2.2. Satellite data pre-processing
2.3. Satellite data classification
2.4. Satellite data interpretation and vectorization of the resulting units
2.5. LCCS classification
2.6. Field checking
2.7. Composition of final land cover maps
2.8. Generation of the GIS database
2.9. Metadata

For the preparation from recent satellite data of land cover maps specially devised for agricultural applications, an operative methodology was tested and finalized. In consideration of the intended use of the information so generated, the mapped land cover units were associated, into a Geographic Information System (GIS), with available information on soil units and erosion classes, thus creating a comprehensive GIS database for the areas studied.

As indicated, three quite large areas were mapped at 1:50 000 scale using Landsat TM and/or ETM data, one area was mapped at 1:5 000 scale using very high resolution satellite data (IKONOS). This last exercise was undertaken to show to the country’s users the flexibility of satellite mapping and to provide decision-makers with relevant information on costs and time requirements.

The land cover mapping was carried out using the FAO Land Cover Classification System (LCCS), a new methodology especially created for land cover mapping and now used worldwide.

The following list describes the main phases applied in the present study, for the creation of the land cover maps (Figure 2):

1. Satellite data selection;
2. Satellite data preprocessing;
3. Satellite data classification;
4. Satellite data interpretation and vectorization of the resulting units,
5. LCCS classification;
6. Field checking
7. Composition of final land cover maps;
8. Generation of the GIS database;
9. Metadata

2.1. Satellite data selection

As previously indicated, all areas under study were visited in June 1999 for a rapid assessment of local physiography and of the main land cover classes occurring there.

Satellite data were selected on the basis of the crop calendar for the main crops. Although the purpose of the study was to prepare land cover maps and not crop inventories, it was considered of some importance to be able to separate the crops in the field at the time of satellite data acquisition (spring/summer crops) from those already harvested (winter crops). Consequently images acquired in August were preferred.

Some data acquired in different months were also selected, in order to improve the data interpretation (multi-temporal approach). All data were purchased in digital format on CD-ROM.

Fig. 2. Methodological approach

2.2. Satellite data pre-processing

This phase included the standard operations of geometric correction and registration of the satellite image into the Bulgarian National Coordinate System-1970. Ground control points (GCP’s) on the satellite image and on the topographic maps were identified and the linear geometric correction function available in the ERDAS Imagine 8.3.1 package was applied.

Satellite data merging procedures were applied for merging of the new panchromatic Landsat 7 channel 8 (15 m resolution) with the multispectral Landsat 7 channels (30 m resolution) to increase the classification accuracy. Modified Price’s merging method was applied to the data. In the literature for this method, the authors declare an increased classification accuracy of approximately 6 percent.

False colour composites (FCCs) were prepared, and finally, sub-scenes for each topographic map sheet (six in the Montana region, four in the Sofia Intermountain Valley, four in the Plovdiv area and one in the Sandanski area) were extracted.

2.3. Satellite data classification

For unsupervised classification the ISODATA method was applied, and for supervised classification, the maximum likelihood classification (MLC) was preferred. To identify the sample areas for supervised classification, specific procedures and information from topographic maps were used together with thematic maps and expert knowledge of the terrain after field checking.

2.4. Satellite data interpretation and vectorization of the resulting units

Landsat TM enhanced false colour composites RGB (red, green, blue) 4,5,3; 5,3,2; 4,5,7 and 4,3,2 were used for the interpretation and delimitation of the land cover classes.

Interpretation and vectorization on the screen, available in ArcView format was the preferred methodology because polygons created have vector format and can be directly transformed to a land cover map. This exercise was carried out in parallel with the development of a legend using the LCCS approach.

Examples of procedures and of interpretation keys are provided in Figures 3, 4 and 5.

2.5. LCCS classification

The Land Cover Classification System (LCCS) is a comprehensive, standardized a priori classification system, created for mapping exercises and independent of the scale or method used to map. The classification uses a set of independent diagnostic criteria that allow correlation with existing classifications and legends, consequently the system could serve as an internationally agreed reference base for land cover. The methodology is applicable at any scale and is comprehensive in the sense that any land cover identified anywhere in the world can be readily accommodated. The rearrangement of the land cover classes, based on regrouping of the used classifiers, facilitates the extensive use of the outputs by a wide variety of end-users. The Land Cover Classification System (LCCS) has been designed with two main phases: an initial dichotomous phase, in which eight major land cover types are defined, followed by a subsequent modular-hierarchical phase, in which land cover classes are created by the combination of sets of predefined classifiers tailored to each major land cover type in order to use the most appropriate classifiers and to reduce the likelihood of impractical combinations of classifiers. A software program has been developed to assist in land cover interpretation, thus standardizing this process and contributing to its consistency. Despite the huge number of classes that can be generated, the user deals with only one classifier at a time and a land cover class is built up by a stepwise selection in which a number of classifiers are aggregated to derive the class (Figure 6).

2.6. Field checking

Field visits were undertaken in all areas under study to collect tÂrrain information and interpretation keys useful for image interpretation.

Later, field checking was carried out to test accuracy of image interpretation at selected sites and to clarify interpretation assumptions.

GPS were used to precisely locate the ground sites investigated.

2.7. Composition of final land cover maps

Vector shapefiles were created manually, in ArcView using both the original image and the results of the supervised classification in the background to provide a basis for visual interpretation. The main phases for Land Cover Map preparation and the interpretation keys are shown on Figures 3, 4 and 5.

Fig. 3. Main phases of land cover map preparation

Topographic map

Satellite image

Supervised classification

Land cover map

Fig. 4. Colour photographs of different locations and their appearance on Landsat TM (interpretation keys)

Artificial lakes




Fig. 5. Colour photographs of different locations and their appearance on Landsat TM (interpretation keys).

Low density residential areas/Gardens


Very large herbaceous fields (active crop)

Very large herbaceous fields (not active)

Fig. 6. Overview of the Land Cover Classification System, its two phases and the classifiers (from Di Gregorio; Jansen; 2000)

The interpretation was made on the basis of the previously defined legend, created using the LCCS software. The appropriate ‘User Label’ code from this legend file was added to each polygon as an attribute.

Link between ArcView and LCCS

The legend file generated in the LCCS was imported into an ArcView table. The ‘User Label’ field in this table was used to link the information from the LCCS to the ArcView attribute table, which already contained the same ‘User Label’ field. In this way the GIS Code, Land Cover class names and other information from the LCCS were added to the GIS database.

Various statistical parameters were then calculated, such as the area covered by each land cover class, the percentage of the total area covered by each class and the ratio between the areas covered by different classes.

2.8. Generation of the GIS database

Each polygon of the resulting land cover maps was then coded according to the FAO Soil Classification System and assigned an attribute to indicate the type of erosion (water, wind, none etc).

For creation of the soil types and erosion database, information from the 1:10 000 and 1:25 000 scale soil maps was used. These maps were scaled to 1:50 000 topographic map sheets and printed on transparent paper. By overlaying of the land cover, soil and topographic maps, dominant soil units and types of erosion for each polygon of the land cover maps were defined and filled as values in the attributive table.

The soil units were defined and correlated according to the “Revised Legend of the Soil Map of the World. FAO, 1997”.

For erosion, according to the LCCS manual, the first distinction was made between classes: “no visible erosion” and “visible evidence of erosion”. Inside the class “visible evidence of erosion” distinction was made by factors causing erosion: water erosion, wind erosion, mass erosion. The two classes water and wind erosion occur in the test areas. Class “wind erosion” has no division by types and intensity. Class “water erosion” is divided into: sheet erosion, rill erosion, gully erosion. For determination of types and intensity of erosion, information concerning soil types and soil texture (from soil maps) and evaluation of slopes (from topographic maps) were used. In the majority of the polygons within which erosion occurs, there is sheet erosion.

Data on administrative boundaries was acquired in digital format from the Ministry of Agriculture and Forestry. These were then combined with the land cover maps using the ArcView Intersect command. This provided the opportunity to calculate statistical parameters on the basis of some administrative units rather than by map sheet - the preferred method for many potential users of the data. The parameters calculated were similar to those calculated for each map sheet (e.g. the area of vineyards or orchards within a particular administrative unit).

2.9. Metadata

Data describing the contents of the database for each map sheet was prepared in MS Excel format, to preserve information for future users of the database. The following parameters were included:

Data format
Map sheet name
Map sheet code number
Image used for land cover map
Date of data acquisition
Published by /Source
Location of files
Date of last editing

Previous Page Top of Page Next Page