Development of an indicator to monitor mediterranean wetlands

Abstract

Wetlands are sensitive ecosystems that are increasingly subjected to threats from anthropogenic factors. In the last decades, coastal Mediterranean wetlands have been suffering considerable pressures from land use change, intensification of urban growth, increasing tourism infrastructure and intensification of agricultural practices. Remote sensing (RS) and Geographic Information Systems (GIS) techniques are efficient tools that can support monitoring Mediterranean coastal wetlands on large scales and over long periods of time. The study aims at developing a wetland indicator to support monitoring Mediterranean coastal wetlands using these techniques. The indicator makes use of multi-temporal Landsat images, land use reference layers, a 50m numerical model of the territory (NMT) and Corine Land Cover (CLC) for the identification and mapping of wetlands. The approach combines supervised image classification techniques making use of vegetation indices and decision tree analysis to identify the surface covered by wetlands at a given date. A validation process is put in place to compare outcomes with existing local wetland inventories to check the results reliability. The indicator´s results demonstrate an improvement in the level of precision of change detection methods achieved by traditional tools providing reliability up to 95% in main wetland areas. The results confirm that the use of RS techniques improves the precision of wetland detection compared to the use of CLC for wetland monitoring and stress the strong relation between the level of wetland detection and the nature of the wetland areas and the monitoring scale considered.

Fig 1. Study area. PACA region: location and departments.

Fig 2. Topographic map of the PACA region and samples of satellite images (false color) of Dep. 13 (littoral) and Dep. 05 (mountainous).

Fig 3. Study area: Departments of the PACA region, their relative wetland surfaces, and % covered by wetland.

Fig 4. Diagram of the classification process developed for the wetland indicator.

Fig 5. Decision tree generated for a Landasat 7 ETM+ image of the PNRC of July 2001.

Red surface correspond to water bodies.

Fig 6. Calculation process of flux accumulation lines for Department 4.

Fig 7. Image change process for rice fields detection in the PNRC between July and October, 2001.

White areas correspond to a high change on greenness values representing rice fields.

Fig 8. Example of the layer union process for Dep. 13.

Water bodies, wetland vegetation and rice fields are combined to obtain the wetland indicator layer.

Fig 9. Illustrative example of error A and error B calculated from the layers of the indicator and the inventories of wetlands.

Fig 10. Indicator results in hectares. Years 1984 and 2001 are compared with the wetland inventories.

Different scales are used in order to show in more detail the differences in surface detected by RS in each department.

Fig 11. Results of Error A and B (%) obtained in the analysis of the wetland indicator in years 1984 and 2001.

Fig 12. Map of wetlands of the PACA region.

To the left, surface detected by the wetland indicator; to the right, inventory of wetlands.

Fig 13. Cases of wetland classifications between 1984 and 2001.