FAO - The Africover Methodology

Land cover is the most important element needed for the description and the study of the environment; as the basic information for environmental databases, it is considered the only geographically explicit feature that can be used as a geographical reference by other disciplines. The methodology, operational procedures and software tools for the land cover analysis in the Africover program have been developed by the Africover Eastern Africa module under the financing of the Italian Cooperation.

From a technical point of view, the preparation of the Africover products relies essentially on remote sensing data and Geographic Information Systems (GIS); the land cover is mainly derived from visual interpretation of recent high resolution satellite images digitally enhanced. The production chain and the overall project methodology is mainly based on the development of the FAO Land Cover Classification System

AFRICOVER PRODUCTION CHAIN

The different steps of the production chain are:

1. Selection of satellite imagery
2. Collection and Organization of Ancillary Data
3. Preliminary interpretation
4. Field verification
5. Final interpretation
6. Quality control
7. Digitization

1. Selection of Satellite Imagery and Georectification

2. Collection and Organization of Ancillary Data

On the basis of the collected ancillary data, the National Photo interpreters presents the different aspects influencing the land cover in their country. This is the basis for the development of a virtual legend of land cover classes to be used to start the interpretation. The legend is based on the FAO Land Cover Classification system

3. Preliminary Interpretation

The photo interpretation of satellite images is a process of extrapolation based on iterative controls. This process consists of:

marking on the false-color image, boundaries of areas or aerial photographs to assign to this a LCCS code
(identification).

extrapolating this delineation and identification to all parts of the image expressing similar characteristics
(color, structure and texture.)

Iterative controls , which involve constant comparison of extrapolation results with other information such as the ancillary documentation proper, aerial photographs and computerized data displays, are made necessary by the very nature of multispectral satellite data.

4. Field Verification Exercise

After the preliminary interpretation, an inventory is made identifying those areas which have not been definitely delineated or identified or where there are still outstanding doubts. These areas are then divided into two groups:

areas where ground truth surveys are the only solution.
areas where it will suffice to study aerial photographs.

Ground truth surveys are the intermediate phase in the preparation of an AFRICOVER land cover database. Field surveys are conducted for two different purposes:

to provide a conclusive answer to any questions of interpretation still unresolved after the additional processing stage.
to provide a general check on the quality of the results of the photo interpretation.

5. Final Interpretation

The Field Data has to be statistically analyzed and the actual GPS points located on the transparencies.

On the basis of the Field Verification Forms and the photographs, the checked classes have to be reviewed and the result extrapolated on to the interpretation sheets; if the results of the checks are particularly unsatisfactory some or all of the interpretation process has to be started again. The last step of the interpretation is the alignment of the interpretation on the edges of the images. This edge matching should assure that coding and polygons are seamless in the full land cover interpretation coverage.

6. Quality Control

When the Final Interpretation phase is completed the following checks have to be made in order to assure the quality of the interpretation:

homogeneity of the interpretation per scene and between the different ones
consistency of the interpretations of the different interpretation teams
consistency of codification of the various units in comparison with the false-colour images
ensuring that only a single code has been assigned to each unit (closed polygon)
ensuring that adjacent polygons have different codes.
the size limit of the smallest units, according to the Variable Minimal Mappable Area, assuring the level of detail of the interpretation.

7. Digitization

The results of the AFRICOVER land cover interpretation has to be provided in digital form for the following reasons:

for incorporation into the AFRICOVER Eastern Africa database for direct use or cross-referencing with other thematic databases
use in the national databases on natural resources within Eritrea
easy data browsing and delivery in Africover Database Gateway
production of thematic maps at varying scales and with different levels of generalization.

There are two possible approaches to digitizing the land cover interpretation transparencies.

(a) The first consists of digitizing the boundaries of each land cover unit using a series of vector coordinates, with the aid of a digitizing tablet or other suitable peripheral system. During or after digitization, the identifier corresponding to each polygon must be recorded.

(b) The second approach consists of using a scanner to generate a high-definition raster image of the land cover map, and converting this into a vector image using a program for converting raster data into vector data (semi-automated). The different land cover units can be identified either by editing the vector image manually or by creating a transparency for each land cover category class and scanning each of these in turn.