METHODOLOGY

1. Satellite imagery

Satellite imagery has the following characteristics:

2. Definition of vegetation and land use classification

This classification was defined using two information sources. We first examined the TREES project classification and selected all categories found in the study area. Then, we reviewed the INEGI (National Institute of Statistics, Geography and Informatics) land use and vegetation map of this zone in order to have a background.

Classes used were:
Seasonal agriculture
Irrigated agriculture
Grassland
Oak forest
Pine-Oak forest
Low tropical drought deciduous forest
Aquatic vegetation
Shrubland
Mater bodies
Urban area
Bare soil

We identified on images other categories not included in the vegetation and land use classification which are not necessarily vegetation types but were included because they appear quite a lot on imagery:

Shadow
Clouds
Fires
Smoke
Lava flows
Mosaic (zone with a combination of elements, mainly agriculture and forest)
Degraded forest (on the 1998 image)
Burnt zones

3. Imagery rectification

The 1998 image was rectified with 1:250 000 topographical cartography edited by INEGI; the 1993 image was co-registered with the 1998 image. The RMS error was 0.42 for 1998 and 0.45 for 1993.

4. Classification

Methods employed to classify imagery were: hierarchical classification, vegetation indices, thresholds and visual interpretation.

• Vegetation and land use were classified with a supervised classification; training fields were taken from INEGI's cartography and field work.
• Bodies of water were selected using masks made with vegetation indices.
• Urban zones were selected with a population centers map, considering cities, towns and villages with at least 2 500 inhabitants; information was provided by the INEGI 1995 population count.
• Fires were delimited selecting pixels with values in band 6 of image greater than or equal to 170 and values in band 7 greater than or equal to 200. (Values 170 and 200 are digital counts)
• The burnt zones class was delimited by a visual interpretation.

5. Field work

In September 1999, a 5-day field work was completed. Various areas of interest for vegetation and land use identification were examined, getting 94 verification points (see Figures 2 and 3; Figure 3 is interactive). With the ground truth, training fields were adapted and changes and adaptations were made to the predefined images classification.

Figure 2

FIELD WORK

Figure 3

EXAMPLE OF VERIFICATION POINTS OVER THE 1998 SATELLITE IMAGE

This is an interactive figure; for further information, select a point.

6. Change detection

Results of this classification were grouped in eight classes according to categories required in the TREES project.

All resulting images were transformed into a vector format to carry out the change analysis in the geographic information system Arc/Info. Zones with a surface of less than 50 ha - the minimum cartographiable area for the scale used in the TREES project- were eliminated by reason of the project's needs.

Home page | Michoacán Case | Guerrero Case | Valle de Bravo Case | Naturale desasters casoe Hurricane Pauline

For doubts, questions or comments
Geographical Information Systems Sub-director