Detection of Rapid Erosion in SE Spain: A GIS Approach Based on ERS SAR Coherence Imagery

This paper presents an integrated remote sensing—GIS approach for identifying areas vulnerable to rapid erosion in Almería Province, Southeast Spain. Earth Resources Satellite (ERS) Synthetic Aperture Radar (SAR) multi-temporal interferometric coherence imagery has been used to detect rapid erosion that causes random changes in the micro-topography of a land surface. These small scale changes result in reduced coherence of the radar signals between the initial and eroded states. However, the detection solely based on coherence imagery is not exclusive because other factors may cause losing coherence as well. A Geographic Information System (GIS) is then used to derive criteria for a set of environmental conditions favourable to rapid erosion from multi-datasets including Landsat Enhanced Thematic Mapper (ETM ) imagery and geological maps. The areas predicted to be most vulnerable to rapid erosion across the imagery are then identified where hard evidence of low coherence coincides with favourable conditions. Introduction Soil erosion is a widespread and long-term problem in many Mediterranean countries. Irregular and often intense rainfall in semi-arid regions can trigger rapid erosion in areas where the slopes are relatively steep, the lithologies are soft or unconsolidated, and vegetation is sparse. Tremendous damage to ecosystems can also be triggered by inappropriate land use (e.g., excessive cultivation) and engineering work (e.g., road cutting). To improve environmental management, it is important to detect and monitor the areas subject to such erosion. Multi-temporal interferometric SAR (INSAR) coherence imagery is an effective and relatively novel technique for the automatic detection of random changes on land surface achieved by measurement of temporal decorrelation. Since erosion processes randomly change the micro-topography of a slope’s surface, the technique can be used for monitoring erosion. This geometric change alters the radar scatterers, randomly changing the phase and amplitude of the signal in each pixel. The signal coherence, a measure of local correlation of phase and magnitude, between the initial and eroded states, estimated over a convenient number of neighbouring pixels, will be gradually reduced. Eventually coherence will be lost completely when the average random phase change, among pixels in a neighbourhood, exceeds half of the radar wavelength in the slant direction (Liu, et al., 2001). It has been shown that, as a type of random change, erosion can be detected effectively using ERS INSAR coherence Detection of Rapid Erosion in SE Spain: A GIS Approach Based on ERS SAR Coherence Imagery Jian Guo Liu, Philippa Mason, Fiona Hilton, and Hoonyol Lee imagery, based on the above principles, in our previous study in Sorbas basin in Almería Province, SE Spain (Liu, et al., 1999a, b; Pearson, 1999). The detection has not been exclusive however, and the loss of coherence is actually the result of several factors. Additionally, coherence imagery has limitations in terms of spatial decorrelation (Lee & Liu, 2001), poor resolution and geometric distortion. A reliable result cannot therefore be achieved without the aid of other data sources. This paper reports a further development to improve reliability and expand the study to a regional scale, based on a combined remote sensing—GIS approach using integrated geospatial data and coherence imagery. The study area covers most of Almería Province, SE Spain (Figure 1) which is one of the most arid areas in Europe (annual precipitation less than 300 mm). In this semi-arid region, badlands are a typical landform which develops where poorly-cemented Tertiary molasse debris and calcareous mudstones or marls become barren and heavily dissected by periodic torrential storms which can deposit more than 100 mm of rain in one event. Such events can also cause substantial damage through landslides and mudflows (Harvey, 1984; 1987). Rapid erosion will occur especially when the natural land surfaces are disturbed and vegetation is stripped. Multi-Temporal Coherence Imagery for Change Detection ERS SAR Data Three ERS-2 SAR raw data scenes, from descending orbits with 70, 140 and 210 days separation, taken during 1996–1997, were used in this study (Table 1). The scenes were chosen based on the temporal separation, perpendicular baseline and weather conditions. Dry weather conditions at the time of each data acquisition is desirable in order to eliminate the possible decorrelation caused by moisture change which is a key factor affecting the dielectric properties of the surface scatterers. The nearest weather station from which historical meteorological records are available is at Almería. No precipitation was recorded there on any of the dates when images were acquired. The ERS SAR raw data were processed to produce Single Look Complex (SLC) images and then three coherence images were generated by ensemble averaging, over a 4 16 pixel window, in the range and azimuth directions, after phase ramp correction. These coherence images are named Coh21, Coh23, and Coh31, where Coh21, for example, is derived using ALM2 as the master scene and ALM1 as the slave scene P H OTO G R A M M E T R I C E N G I N E E R I N G & R E M OT E S E N S I N G October 2004 1 1 7 9 Department of Earth Science and Engineering, Imperial College of Science, Technology and Medicine, Prince Consort Road, London SW7 2AZ, UK ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 70, No. 10, October 2004, pp. 1179–1185. 0099-1112/04/7010–1179/$3.00/0 © 2004 American Society for Photogrammetry and Remote Sensing ISR-05.qxd 9/10/04 17:06 Page 1179