EVALUATING THE CARTOGRAPHIC CAPABILITY OF ASTER AND RADARSAT
IMAGERY FOR MONITORING SOIL EROSION INDICATORS IN THE ARGENTINEAN
Blanco1, G. Metternicht2, H.
1 - National Patagonic Center (CENPAT-CONICET), Puerto
2 - Department of Spatial Sciences, Curtin
Soil erosion is a key factor contributing to land degradation processes worldwide. Recent reports of the United Nations Environmental Programme mention that around 2 billion ha of soil, equal to 15 per cent of the Earth’s land cover or an area bigger than the United States and Mexico combined, is now classed as degraded as a result of human activities.
About one-sixth of world’s degraded land is classed as strongly or extremely degraded, making efforts for their restoration largely ineffective. The adoption of measures to remediate or decrease land degradation require access to up-to-date spatial information on the intensity, rate and spatial distribution of soil erosion processes, to curb degradation trends and secure sustainable land use and management. Multi-sensor optical and microwave remote sensing can contribute significantly to map patterns of surface features related to wind erosion at regional level.
Areas like the Peninsula Valdes (southern Argentina), declared a UNESCO World Heritage site in 1999, are subject soil degradation by wind and overgrazing. Mapping and monitoring the presence of landscape erosion indicators such as stabilized and active dunes is crucial to improve prediction, monitoring and planning of areas threatened by sand encroachment. To this end, this paper investigates and compares the contribution of optical sensors like the Terra-ASTER and the microwave Radarsat ASAR to the discrimination of these land degradation features. Furthermore, an evaluation is undertaken to compare the classification accuracy achieved by specific regions of the spectrum (e.g. optical or microwave), and their synergistic use in an object oriented fashion.
The research approach encompasses:
a) Identification of landscape features related to the presence of active and stabilised dunes. Two vegetation types are considered as dune stabilisers: scrub and grass;
b) Calibration and georeferencing of the Terra ASTER imagery, including the computation of spectral indices and principal component analysis for the removal of redundant spectral information;
c) Despeckle and georeferencing of a precision mode Radarsat imagery;
d) Creation of a geo-spatial soil database to store field observations and spectral characteristics of wind-erosion related features in the optical and microwave regions of the spectrum;
e) Selection of textural measures based on the grey level co-occurrence matrix that best characterise patterns of stabilised and active dunes in the Radarsat ASAR imagery. Extraction of Radarsat derived textural measures;
f) Segmentation and object-oriented classification of the selected imagery using eCognition. This step requires the definition of proper fuzzy membership functions for characterising stabilised and active dunes, as well as main vegetation classes of the study area.
g) Accuracy evaluation of the spectral ASTER data, the ‘textural’ Radarsat ASAR data and combined spectral and textural data provided by ASTER and Radarsat data sets using an error matrix and Kappa statistics.
Initial results suggest an improvement in the classification of active dunes and stabilised dunes (vegetated by either scrub or grass) is achieved by using an object oriented classification that integrates textural information derived from microwave imagery and optical/IR data from Terra-ASTER. It appears that changes in surface roughness caused by different vegetation types stabilising the dunes is a major influence in radar backscattering. For instance, Radarsat imagery enables a clear separation of long and narrow dunes stabilised by scrub against those stabilised by grass, the former showing a higher classification accuracy. On the other hand, ASTER optical and infrared wavebands show superior performance in the cartography of grass-stabilised dunes. It appears that the synergistic use of microwave, optical and infrared data increases substantially the accuracy in the discrimination and mapping of soil degradation features related to wind erosion.