Multitemporal soil erosion mapping from Remote sensing and geomorphological data: the case-study of Saccione River basin area

The aims of this study are the assessment of the multitemporal spatial distribution and the basinal mapping of the soil erosion geomorphological evidences desumed by integration of Landsat ETM 7+ data, interpretation resultances of panchromatic color orthophotos and geomorphological surveyed data. These latter ones are consisting of recognition of landsurfaces interested by areal (sheet and solifluxion processes) and linear (rill and gully erosion) erosional landforms (ELs). The study has been performed in the Saccione river basin (North Apulia), wide 228.6 km2, located in southern Italy. The first step of the study has been the digital thematic mapping of lithological, pedological and land-use features coming from literature data and characterizing the studied area. Then, the whole remotely sensed dataset regarding the studied area was co-registrated and processed using the RSI ENVI 4.4 software. For the registration of the first Landsat image (master image), ground control points (GCPs) were collected by GPS measurements. The other images (slave images) were co-registrated by GCPs image-to-image registration. The dataset was processing by application of contrast stretching, Principal Component Analysis (PCA), decorrelation stretching and RGB false colour compositing tecniques. A field survey was carried out to characterize the geomorphological features detected on the imagery. Particular attention was given to the ELs, which were located using a GPS. In the second step have been delimited the Regions of Interest (ROI) on the Landsat ETM 7+ imagery, i.e. polygons representing the “ground-truth” related to the several land cover class occurring in the imagery. A simple statistical analysis was then conducted on the digital number (DN) values of the pixels enclosed in the ROI of the Els class to determine their spectral response pattern. The whole image dataset has been then classified using a maximum likelihood classification algorithm. The Els distribution maps of 1992, 1999, 2000, 2001 and 2004 years were generated respectively based on the classification results. The results of the classification process have been checked in the field and by photointerpretation on the base of random sampling. Finally, a spatial analysis was performed to temporal monitoring of the ELs shape and location. The field survey and the image processing have allowed to highlight that some classified Els-like “objects”, even if showing the same spectral response, could not be Els: this confirm the strategic importance of the field checking on the semi-automatically produced data. These mistaked “objects” have been deleted during the production of the multitemporal mapping of the Els, as a final result of the study. The multitemporal spatial analysis have show the changes in the shape and position of the Els and, furthermore, how the highest frequency of the ELs are occurring on southern slopes. These slopes, displaying slope angle values ranging from 12° to 20°, are cut on clayey-marly deposits covered by fine-textured and carbonate-rich Inceptisols. Furthermore, the comparison between seasonal different Landsat data of the same area has clearly evidenced the presence of ELs during the end of the summer period. Such results, verified in the field, are connect with bad and non conservative agricultural practices, suggesting that these practices could play an important role as triggering of erosional processes and their evolution.