Using the Landsat Archive for the Monitoring of Mediterranean Coastal Wetlands: Examples from the GlobWetland-II Project
Christian Hüttich1, Julia Reschke2, Manfred Keil3, Stefan Dech3, Kathrin Weise4, Coralie Beltrame5, Eleni Fitoka6, Marc Paganini7
(1) Department of Remote Sensing, University of Würzburg, Würzburg, Germany; now at Institute of Geography, Department of Earth Observation, Friedrich-Schiller-University Jena, Germany
(2) Department of Remote Sensing, University of Würzburg, Würzburg, Germany; now at Institute of Photogrammetry and Remote Sensing, Vienna University of Technology, Vienna, Austria
(3) German Remote Sensing Data Center, German Aerospace Center, Oberpfaffenhofen, Germany
(4) Jena-Optronik GmbH, Jena, Germany
(5) La Tour du Valat, Research centre for the conservation of Mediterranean wetlands, Arles, France
(6) The Goulandris Natural History Museum Greek Biotope Wetland Centre (EKBY), Thessaloniki, Greece
(7) European Space Agency (ESA), Frascati, Italy
Abstract – The monitoring and assessment of the status and trends of wetlands is of major concern for long-term biodiversity conservation initiatives. In particular, the coastal wetlands in the Mediterranean have undergone considerable land use and land cover changes in recent decades, by way of urban growth and increasing tourism infrastructure. A rise in sea level would result in a major ecological pressure for natural and human-made wetlands in deltas and coastal lagoons. Satellite observation techniques allow for an objective monitoring of the Earth on large scales. Aiming at the demonstration of the current capabilities of satellite Earth observation applications to support inventorying, monitoring, and assessment of wetland ecosystems, the GlobWetland-II project makes use of the 35 years of Landsat archives for a basic identification and delineation of wetlands during1975, 1990, and 2005. We present results of wetland identification and delineation mapping from 1975 to 2002, for the test sites of the Menderes Delta and Güllük Bay. Methodically, a decision tree approach is used integrating multi-temporal spectral (Landsat MSS, TM, ETM+) and topographic (SRTM) satellite-based indicators for the occurrence of wetlands. While comparing the spatial distribution of wetlands over 27 years, a remarkable decrease of natural wetlands became obvious, particularly at the expense of increasing agricultural land use, artificial areas, or dam constructions. The detected land change processes show the capabilities of satellite-based Earth observation time series. In particular, the potential of the freely available Landsat archives for the quantification of the status and trends of Mediterranean wetlands in the framework of the RAMSAR Convention on Wetlands is demonstrated.
1 Spatial information needs for wetland monitoring
Wetlands serve as important resources for water supply, water quality, recharge of groundwater aquifers, and flood and shoreline protection, and provide a number of important ecosystem services. Wetlands are biodiversity hotspots, hosting large numbers of threatened species  and play an important role in regional economics through things like reed production, fishing, and tourism  . Beside those provisioning ecosystem services, coastal and inland aquatic ecosystems provide a number of important regulating services, such as climate regulation , water runoff and erosion regulation, water purification, and pollination .
Despite the economical and ecological importance of wetland systems, they are being affected by distinct change processes triggered by global climate change (e.g. global sea level rise, droughts, and flooding) and human-induced land use pressure, mainly through the intensification of agriculture and urban growth. Wetlands have a high degree of bio-complexity, making them vulnerable to human induced climate and land use change. Most conversions of coastal and inland wetlands,with increasing trends, were due to habitat changes which had the highest impact on biodiversity in the last century .
The standardized and continuous monitoring of wetlands is essential for understanding the status and trends of the direct drivers of change, such as changes in local land use and land cover, species migration, external inputs due to agriculture, and harvesting. Satellite-based Earth observation technologies have been well recognized as a method for mapping and monitoring wetland ecosystems. Recent developments of satellite data policies of the national space agencies increased the availability of multi-scale Earth observation data and lead to the situation that the remote sensing community lives in a data-rich world. Because historical data archives, such as the Landsat data archive , are freely available for scientific purposes, it significantly increases the role of satellite-based Earth observation data for ecosystem monitoring and nature conservation purposes . As recently stated by MacKay , Earth observation techniques are essential for standardized wetland monitoring mechanisms, and cover the information needs for managing wetlands in the context of the Ramsar Convention.
2 The GlobWetland-II Project
Spatial information needs are addressed during several reporting mechanisms as a base for management planning and implementing policy responses concerning sustainable management and nature protection. Spatial information is necessary for baseline and status assessments and for deriving trends in several kinds of inventories and monitoring tasks.
Fitoka & Keramitsoglou  pointed at the capabilities of satellite-based wetland observation initiatives to derive spatially explicit information on the status and dynamics of wetlands.
A number of projects and programs cover wetland monitoring tasks using Earth observation techniques in the Mediterranean, such as the Pan-Mediterranean Wetland Inventory (PMWI), the Land and Ecosystem Accounting (LEAC) of the European Environmental Agency (EEA), the Mediterranean Wetland Observatory (MWO) and the GlobWetland initiative. Within GlobWetland, the European Space Agency (ESA) in 2003 launched a joint wetland monitoring project in collaboration with the Ramsar Secretariat aiming at the demonstration of Earth observation capabilities for supporting inventorying, monitoring, and assessment of 50 selected wetland test sites on the globe. ESA continues supporting this effort in close collaboration with the Ramsar Secretariat, its Scientific and Technical Review Panel, and the Mediterranean Wetlands initiative. The first and major continuation of GlobWetland-I is the GlobWetland-II project,a regional pilot project of the Ramsar Convention on Wetlands. GlobWetland-II aims at developing a Global Wetland Observation System (G-WOS), with about 200 test sites throughout Africa and the Middle East. All test sites will be located less than 100 kilometers from the Mediterranean coast.
This pilot information system, also called the GlobWetland-II information system, includes maps and system software. The GlobWetland-II system software consists of three components: A remote sensing component for tasks like satellite image pre-processing, land use and land cover classification and change detection; a Geographic Information System (GIS) component for the wetland indicator computation,describing the status and trends of wetlands; and an online GIS component providing permanent access to the maps and information data produced during the project or provided by users and partners . The GlobWetland-II geo-information maps will be produced to highlight 1975-1976, 1990-1991, and 2005-2006, on the 200 wetland sites, at a geographical scale of 1 to 50,000 to 1 to100,000 in the Mediterranean, taking full advantage of the time series of Landsat data (MSS, TM and ETM). Besides these main components, a demonstration task in GlobWetland-II is to cover the inventory and delineation of wetland areas in catchment and sub-catchment areas in the Eastern Mediterranean, especially in Turkey and Egypt.
3 Wetland Inventory and Delineation Mapping: Examples from GlobWetland-II
This article aims at the presentation of the methodological framework of the GlobWetland-II wetland inventory and delineation mapping approach. First results are demonstrated on the Turkish test sites of the Menderes Delta and Güllük Bay region, located at the Turkish west coast of the Aegean Sea.
One of the major aims of developing the wetland identification and delineation map product is to support the Ramsar implementation mechanisms with the support of traceable Earth observation techniques. The wetland delineation maps shall aim to detect the occurrence of wetlands as a decision support to update Ramsar databases on wetlands of international importance and provide spatial information of the temporal spatial dynamics of those wetlands. Knowledge about the historic spatial distribution of threatened wetlands, either through urban growth or agricultural expansion, can help local administrators to define the spatial extent for wetland re-naturation projects.
3.1 Data and Methods
Multi-temporal Landsat acquisitions for 1975, 1990, and 2002 were achieved from the ESA and USGS data archives covering the hydrological cycle of coastal Mediterranean wetland systems. Different water cycle regimes become apparent by incorporating multi-temporal acquisition dates. Time series of the normalized difference vegetation index (NDVI) capture the temporal characteristics of the Ramsar wetland types, as shown in Figure 1 of the Ramsar classes of coastal lagoons, marshland, and seasonal flooded marshland. Flooding events within the marshlands are indicated by decreasing NDVI values caused by increasing absorption rates in the near infrared band. The temporal dimension, and the availability of multi-temporal imagery are mandatory for the spectral discrimination of natural and human-made wetlands from the surrounding land cover.
A schematic presentation of the wetland delineation process is shown in Figure 2. The pre-processing of multi-temporal Landsat imagery includes an atmospheric correction and ortho-rectification done by the CATENA image processing chain hosted at the German Aerospace Center DLR . In most cases, the geocoded products delivered by ESA also fulfilled the necessary quality for multi-temporal investigations. A multi-data feature space was generated prior to the classification process consisting of a near infrared mean layer, which is computed from multi-temporal Landsat imagery, a normalized difference seasonal NDVI index based on the wet and the dry NDVI image, typically a summer and a winter acquisition, and a slope and an elevation layer derived from the Shuttle Radar Topography Mission digital elevation model with a 90-meter spatial resolution (SRTM DEM).
The classification process itself consists of a simple rule- and knowledge-based decision tree classification. The composition of a set of multi-temporal spectral and topographic features combines key parameters indicating a wetland, the main one of which is a distinct hydrological cycle caused by alternating wet winter and spring conditions, and hot and dry summers in the coastal Mediterranean region. A very low NIR mean threshold indicates the occurrence of open water bodies or wet soils such as mud flats where high variations indicate seasonally flooded marshlands or Ramsar-type natural wetlands, and irrigated agricultural lands , or Ramsar-type human-made wetlands. Low elevation and slope thresholds are applied to delimit wetlands to be detected to the flat areas of river deltas, coastal lagoons or other depressions.
Validation was performed on the presented test areas of the Menderes delta and around Güllük Bay. For that task, more than 700 reference points were collected during an intensive field campaign in February 2009, and integrated in an accuracy assessment of the most recent wetland identification and delineation mapping product of 2002.
3.2 Results and Discussion
Exemplary results of the spatiotemporal dynamics of coastal wetlands in western Turkey between Izmir and Bodrum are shown in Figure 3. The mapping results show the basic wetland delineation 1975, 1990, and 2002, in shades of green over a false-color composite of the original Landsat image, showing green vegetation in red, with forest and shrub lands in dark red, and agricultural lands in brown-beige and red.
While comparing the spatial distribution of wetlands over a period of 27 years, a remarkable decrease of natural wetlands became obvious, particularly at the expense of increasing agricultural land use, artificial areas, and dam constructions. The first example shows different land use pressures on natural wetlands. The situation in the 1970s shows a more natural and intact wetland system that decreases in its areal extent through agricultural intensification in the hinterland and coastal lagoon of the river delta. The wetland map of 2002 shows that the entire geo-ecosystem has changed by the dam construction. The Tahtali dam is now an important water reservoir for the nearby city of Izmir. However, the inland wetlands and coastal lagoons have significantly decreased. The second example shows a similar land change process of an airport construction (the airport of Bodrum) in a former wetland area of Güllük Bay south of Izmir. The wetland area decreased by 729.38 km2 between 1975 and 2002 (7.3 percent of the study area). The two examples demonstrate that besides agricultural intensification, urban sprawl and growing demands of artificial areas for tourism, second housing and infrastructure as shown in Figure 4 are the major land-use modification processes affecting the decrease and ongoing pollution of coastal wetland systems.
Results of the accuracy assessment –performed on the 2002 wetland delineation map product– show an overall classification accuracy of 81.49 percent. Wetland areas achieve producer’s and user’s accuracies of 67.26 percent and 72.9 percent. The non-wetland class resulted in 88.17 percent and 85.05 percent.
The detected decreasing trends in natural and semi-natural wetland areas were demonstrated at exemplary GlobWetland-II test sites in western Turkey. Retrieving reliable wetland delineation results with a simple decision tree classifier shows the capabilities of high resolution Landsat time series for longer-term monitoring purposes. In particular, the potential of the freely available Landsat archives for the quantification of the status and trends of Mediterranean wetlands in the framework of the Ramsar Convention of Wetlands is demonstrated, showing high potential for the implementation of the Ramsar Convention. Area-wide wetland identification maps of past time imagery bares high potential for the support of local scale wetland protection planning, keeping in mind that the procurement of historic land cover and land use data and the analyses of former ecosystem conditions is challenging.
This study has been conducted by DLR and the University of Würzburg in preparation of the GlobWetland-II project. The application of a threshold-based mapping framework covers both the detection of wetland geo-ecosystems with acceptable mapping accuracies, and a comprehensible method which can be used by local wetland managers and local administrative staff. The method is already being applied in the project.
The consistent temporal tracking of coastal Mediterranean wetlands is of major importance for all water-related ecosystem services affecting both land and sea ecosystems. A standardized wall-to-wall monitoring of the coastal geo-ecosystems of the Mediterranean Sea is still outstanding. Coastal wetlands affect an important buffer zone function in terms of water de-pollution, storage, and biodiversity balancing. The cross-border identification of land cover impact, including land cover and land use types and processes, on semi-natural wetlands, is of major concern and a basic parameter for the quantification of the Ramsar indicators like the status and trends of wetland extent. The combination of deriving large-area Earth observation products (1:90.000) of Wetland Identification and Delineation Maps and regional scale (1:50.000) land cover and Ramsar wetland typology maps at three points in time will be the basic spatial information products to derive the Ramsar indicators of effectiveness as defined by the Scientific and Technical Review Panel (STRP) and the Mediterranean Wetlands Observatory (MWO). The most important Earth observation-derived indicators are change in wetland area, inundation in the ecosystem, and change in wetland area due to urbanization. Within Globwetland-II, the capabilities and technical limitations of Earth observation for the Ramsar reporting mechanisms are being evaluated. Recent results highlight the needs to make a bigger picture of the ecological state of Mediterranean coastal wetlands.
 BirdLife International, Important Bird Areas and potential Ramsar Sites. 2001.
 E. B. Barbier, J. C. Burgess, and C. Folke, Paradise lost?: the ecological economics of biodiversity. Earthscan, 1995, p. XVI, 267 p.
 D. Dudgeon et al., “Freshwater biodiversity: importance, threats, status and conservation challenges.,” Biological Reviews of the Cambridge Philosophical Society, vol. 81, no. 2, pp. 163-182, 2006.
 E. Fitoka and I. Keramitsoglou, Inventory, assessment and monitoring of Mediterranean Wetlands: Mapping wetlands using Earth Observation techniques. 2008, p. 140.
 Millenium-Ecosystem-Assessment, Ecosystems and human wellbeing:Wetlands and Water Synthesis. Washington: , 2005, p. 80.
 J. Ju and D. P. Roy, “The availability of cloud-free Landsat ETM+ data over the conterminous United States and globally,” Remote Sensing of Environment, vol. 112, no. 3, pp. 1196-1211, Mar. 2008.
 P. Leimgruber, C. A. Christen, and A. Laborderie, “The Impact of Landsat Satellite Monitoring on Conservation Biology,” Environmental Monitoring and Assessment, vol. 106, no. 1-3, pp. 81-101, Jul. 2005.
 H. MacKay, C. M. Finlayson, N. Davidson, D. Pritchard, and L.-M. Rebelo, “The role of Earth Observation ( EO ) technologies in supporting implementation of the Ramsar Convention on Wetlands,” Journal of Environmental Management, vol. 90, no. 7, pp. 2234-2242, 2009.
 K. Jones, Y. Lanthier, P. van der Voet, E. van Valkengoed, D. Taylor, and D. Fernández-Prieto, “Monitoring and assessment of wetlands using Earth Observation: the GlobWetland project.,” Journal of environmental management, vol. 90, no. 7, pp. 2154-69, 2009.
 K. Weise, E. Fitoka, H. Hansen, and M. Paganini, “Executive Summary of the GlobWetland-II Project,” 2010. [Online]. Available: www.globwetland.org.
 P. Reinartz, “The CATENA Processing Chain – Multi-Sensor Pre-processing: Orthorectification, Atmospheric Correction, Future Aspects,” in Geoland Forum 6, 24 -25 March 2010, Toulouse, France, 2010.
 G. Sarigül, “Personal communication,” 2009.