Main question

Basin scale water resources analysis requires land cover classification as one of the basic parameters. The use of public domain data sets and the application of remotely sensed techniques were used to derive a land cover map at basin scale.

Methodology

For the Gediz Basin in western Turkey NOAA-AVHRR monthly NDVI images were used to produce a 1 km land cover map. An innovative approach to classification was used where temporal changes were expressed by a Principle Component Analysis, maximum vegetation growth was included using the NDVI for August and a Digital Elevation Model was employed to represent non-vegetative characteristics.

These three datasets were combined in a False Color Composite and an unsupervised classification technique was used to derive the six major land covers.

Figure 1. Methodology and datasets used to derive the land cover map.

Result and conclusions

The datasets and the methodology are presented in Figure 1. It is clear that the three datasets used represent distinguished information, making the classification unambiguous. Some field visits were made to verify the derived map, resulting in a qualitative indication of the accuracy (Figure 2)

Figure 2. Final land cover map after field verification.

Reports

    Droogers, P., G.W. Kite, and W. Bastiaanssen. 1999. Land cover classification using public domain datasets: example for Gediz Basin, Turkey. In Proceedings of International Symposium on Arid Region Soils. Menemen, Turkey, 21-25 September 1998. p 34-40

Please feel free to contact the Project Leader of this project for more information.