Identification of eroded areas using remote sensing in a badlands landscape on marls in the central Spanish Pyrenees
Introduction
Maps of active erosion areas and areas at risk of erosion are of great potential use to environmental (governmental and private) agencies, as they allow erosion prevention efforts to be concentrated in those places where the benefit will be highest. There is no single straightforward method for assessing erosion, and erosion evaluation is highly dependent on the spatial scale and the purpose of the assessment (Warren, 2002). For limited spatial scales (less than 100 ha), field surveys can provide an accurate means of analyzing erosion damage (Herweg, 1996). However, for focal area selection over larger areas other approaches that integrate available spatial data need to be applied. Studies on erosion undertaken at spatial scales covering local to regional areas (Vrieling et al., 2006) have provided both quantitative information (e.g., erosion rates) and qualitative information (e.g., erosion risk areas).
Methods for evaluating erosion risk on catchment and regional scales (10 to 10,000 km2) include the application of erosion models or qualitative approximations using remote sensing and geographic information (GIS) technologies. Merrit et al. (2003) have exhaustively described current erosion models. However, in most cases erosion models have been created for use at small scales, so their extrapolation to larger scales (catchment or regional) is very complex and sometimes leads to errors (Kirkby et al., 1996, Schoorl et al., 2000, Yair and Raz-Yassif, 2004). The use of remote sensing and GIS techniques has been shown to have potential for erosion assessment on regional scales, including identification of eroded surfaces, estimation of factors that control erosion, investigation of soil and vegetation characteristics, and monitoring the advance of erosion over time (Muchoney and Haack, 1994, Lambin, 1996).
In most cases remote sensing techniques have been applied simply to identify the characteristics (or the absence) of the vegetation cover, largely because of limited visibility of the soil surface in humid and sub-humid environments (Vrieling, 2006). Other studies have demonstrated the usefulness of remote sensing techniques in determining temporal and spatial erosion patterns (Pilesjo, 1992, Rode and Frede, 1997, Metternicht and Fermont, 1998, Szabo et al., 1998, Millward and Mersey, 1999, Reusing et al., 2000, Haboudane et al., 2002, Metternicht and Gonzalez, 2005). Calculation of the percentage of bare ground has also been used to estimate erosion risk (e.g., De Jong, 1994, Paringit and Nadaoka, 2003). Other methodologies applied to inventories and monitoring of erosion processes include band ratios (Pickup and Nelson, 1984, Frazier and Cheng, 1989), vegetation indices (Pickup and Chewings, 1988; Tripathy et al., 1996), combinations of reflective and microwave data (Koopmans and Forero, 1993, Singhroy, 1995, García-Meléndez et al., 1998, Metternicht and Zinck, 1998, Singhroy et al., 1998), and combinations of remote sensing data and other ancillary data (Floras and Sgouras, 1999, Mati et al., 2000, Giannetti et al., 2001, Shrimali et al., 2001, Zinck et al., 2001, Haboudane et al., 2002, Ma et al., 2003, Symeonakis and Drake, 2004, Beguería, 2006a). No studies have explored the application of the receiver operating characteristic (ROC) curve analysis, which allows for selecting the optimum classification map based on the omission and commission errors, to the generation of an erosion map.
Several studies have estimated erosion in the Spanish Pyrenees using remote sensing at the regional (Beguería, 2005) and the catchment scales (Fargas et al., 1997). These studies have shown that the badland systems developed on Eocene marls constitute the main sediment sources in the Pyrenees, with important consequences for the silting of reservoirs (Valero-Garces et al., 1999). The term badlands is used to describe areas of unconsolidated sediment or poorly consolidated bedrock, with little or no vegetation. They are typically associated with accelerated erosion and consequent unstable landscapes, so that their fixation requires considerable effort. Badlands develop in a wide range of climatic zones, but particularly in semiarid areas. In sub-humid and humid regions the development of badlands is linked to lithological and topographical factors (Regüés et al., 1995, Morgan, 1997, Oostwoud-Wijdenes et al., 2000), and to climatic conditions such as freeze–thaw cycles in winter, and wetting–drying in spring–summer. In the Spanish Pyrenees, a combination of favorable relief and climatic conditions is coupled with highly erodible marls outcrops, explaining the presence of badland systems with intense soil erosion processes (Regüés et al., 1995, Gallart et al., 2002, Nadal-Romero et al., 2007, Nadal-Romero et al., 2008).
The objective of this study was to test a method for identification of areas of severe erosion (badlands) and areas of erosion risk by means of remote sensing classification techniques. The method involved several steps: i) application of a supervised classification algorithm to areas with active erosion features to obtain a map of the spectral distance; ii) selection of a classification threshold based on the ROC curve and error statistics; and iii) assessment of the uncertainty associated with the predictions. The study area corresponded to the corridor of Eocene marls in the middle section of the Ésera River basin, in the central Spanish Pyrenees.
Section snippets
Study area
The study area is located approximately 23 km north of the Barasona reservoir, in the Spanish Pyrenees, and is an integrated badlands landscape developed on Eocene marls orientated north−southeast (Fig. 1) at 620 m to 2149 m (Fig. 2) altitude. The badlands system is conformed by a group of typical hillside badlands developed on sandy marls with clay soil, and is strongly eroded over convex hillsides with a moderately inclined slope. Runoff from this area enters the Viu and Rialvo rivers in the
Data selection and preparation
In this study Landsat data (spatial resolution 30 m) from August 1st 2006 were used because of the lower frequency of cloud cover in this month (Fig. 1A). The image was geometrically corrected using control points and the algorithm developed by Pala and Pons (1996) implemented in the Miramon software, which accounts for topographic distortion by incorporation of a digital terrain model (DTM).
The atmospheric effect on the electromagnetic signal was corrected using the radiative transfer code 6S (
Selection of classes and training areas
The definition of thematic classes and selection of training areas were based on visual analysis of aerial orthophotos (SIGPAC, 2003). Land cover variability of the study area comprised five classes: erosion active areas (badlands), scrubland, grassland, conifers, and deciduous forest. The training sample was used to obtain spectral signatures for each thematic class (Fig. 4). The badlands–the class of interest to this study–were characterized by high brightness values for all bands, especially
Conclusions
This study has demonstrated the utility of remote sensing data in basic and applied geomorphological research, both at watershed and regional scales (study areas between 10 and 10,000 km2). The use of a supervised classification method based on the maximum likelihood algorithm plus the ROC curve analysis for choosing the most appropriate classification threshold enabled reliable mapping of areas with active erosion. Selection of training areas using aerial orthophotos (SIGPAC, 2003) enabled
Acknowledgments
This research was financially supported by the project “Processes and sediment balances at different spatial scales in Mediterranean environments: Effects of climate fluctuations and land use changes” (CGL2006-11619/HID), funded by CICYT, Spanish Ministry of Education and Science. The contribution of the first author has been possible thanks to a scholarship granted by The National Council for Science and Technology of Mexico (CONACYT). This work would not have been possible without the
References (60)
- et al.
Identification of critical sediment source areas at regional level
Physics and Chemistry of the Earth
(1997) An introduction to ROC analysis
Pattern Recognition Letters
(2006)- et al.
Use of geoinformation techniques in identifying and mapping areas of erosion in a hilly landscape of central Greece
International Journal of Applied Earth Observation and Geinformation
(1999) - et al.
Remote sensing of soils in the eastern Palouse region with Landsat Thematic Mapper
Remote Sensing of Environment
(1989) - et al.
Multisensor data integration and GIS analysis for natural hazard mapping in a semiarid area (Southeast Spain)
Advanced Space Research
(1998) - et al.
Integrated use of satellite images, DEMs, soil and substrate data in studying mountainous lands
International Journal of Applied Earth Observation and Geoinformation
(2001) - et al.
Airborne SAR and Landsat MSS as complementary information source for geological hazard mapping
ISPRS Journal of Photogrammetry and Remote Sensing
(1993) - et al.
Assessment of erosion hazard with the USLE and GIS: a case study of the upper Ewaso Ng'iro North basin of Kenya
International Journal of Applied Earth Observation and Geoinformation
(2000) - et al.
Estimating erosion surface features by linear mixture modelling
Remote Sensing of Environment
(1998) - et al.
FUERO: foundations of a fuzzy exploratory model for soil erosion hazard prediction
Environmental Modelling and Software
(2005)