Application of ALOS PALSAR data to classify vegetation in an anthropogenically affected wetland area in Central Spain (Las Tablas de Daimiel)
Thomas Schmid(1), Magaly Koch(2), Jesus Solana(3) and Jose Gumuzzio(3)
(1) CIEMAT, Avda. Complutense 22, 28040 Madrid, Spain
(2) Boston University, 725 Commonwealth Ave., Boston, MA 02215, United States
(3) Autonomous University of Madrid, Campus Cantoblanco, 28049 Madrid, Spain
Semiarid wetlands are very dynamic ecosystems as the different characteristics (areal extension, water depth and salinity, seasonal flooding, vegetation and fauna) that define them vary greatly in the short and long term. They are especially vulnerable to alterations produced at their land/water interface and monitoring these changes represents often a challenge in remote sensing. The main objective of this work is to investigate the capability of ALOS PALSAR in characterizing and monitoring these dynamic ecosystems by utilizing a wetland area in Central Spain for which ample information (obtained from spaceborne, airborne and field data) already exist.
The study site represents a fresh water wetland located in the National Park of Las Tablas de Daimiel in La Mancha Alta (Central Spain), which has been affected by human induced changes. The wetland’s climate is typical Mediterranean and its landscape is karstic with sediments rich in evaporates. Human activities that have substantially changed the wetland in the last three decades include intensive agricultural irrigation, over-exploitation of groundwater, drying and artificial drainage of wetlands, urban development as well as altering and channeling of rivers.
This work presents a methodology based on multipolarized microwave (PALSAR) data to improve wetland vegetation classifications. The main objective is to produce accurate wetland maps that can be used for 1) monitoring changes over time, including wetland disturbances associated with non-native invasive plant species, and 2) identifying and mapping adjacent land-use and land-cover types that may have a direct impact on the wetland health. Inland wetlands in Mediterranean environments such as the study site are usually small and highly fragmented, thus, more vulnerable and likely to disappear more rapidly. From an ecological point of view, these wetlands provide an ideal habitat for nesting and migratory waterfowl. Growing loss of biodiversity stemming from human activities is a major concern in recent years. Biodiversity is defined by both the total number of species and the evenness of species. Therefore, accurate wetland vegetation maps at regular time intervals are needed in order to assess the status and health of wetlands.
A quad-pol PALSAR image (acquired on 5 May 2007) was processed as follows: 1) contrast stretching and speckle removal using adaptive filters, 2) application of directional and textural filters to improve edge enhancements and surface roughness for the interpretation of linear features as well as discrimination of smooth versus rough surfaces (open water and flooded/nonflooded vegetated areas), 3) synthesizing the different polarizations into a color combination for selecting training sites, and 4) performing a partial polar supervised classification of the wetland area.
The PALSAR classification result was validated against a field-based vegetation map of the same season and year, and the result shows that an accurate differentiation of wetland surface texture and spatial distribution of variations in vegetation density and fragmentation was obtained. Furthermore, comparison with classification results obtained from optical sensors such as AVNIR-2 (acquired on 5 November 2007) and Probe1/CHRIS (acquired on 20 July 2007) confirmed that a better discrimination of wetland soils and vegetation in relation to their status (wetland inundation/desiccation, invading vegetation species, soil salinization etc.) could be achieved. This outcome suggests that multipolarized microwave data may enable better separation of vegetation structure and fragmentation than in optical data, which can be used to assess wetland degradation status, so that appropriate measures can be designed for a sustainable management of the wetland.
PALSAR offers several technical enhancements with respect to previous radar sensors. These are the introduction of multiple polarization modes, incidence angles, look directions as well as improved spatial and temporal resolution. These improved capabilities will allow better discrimination and characterization of compositional variations within the wetland, i.e. vegetation types, soil moisture content, as well as better delineation of inundated areas especially in less densely vegetated wetland regions.