Analysis of the soil sealing in the urban area of Rome through automatic processing of satellite data with neural networks

Satellite data represent a very powerful tool for monitoring soil sealing in metropolitan areas with a significant level of detail and with a possibility of systematic updating. However, the full exploitation of such capabilities in the application domains requires that two main conditions are satisfied: availability of the data and use of automatic and reliable procedures for data processing. Moreover, to make the methodology largely used, it is important to set up low-cost processing chains that can be afforded by different user communities. This paper presents the first results of a study aiming at the design of new and low-cost procedures for soil sealing monitoring in the city of Rome, Italy. The procedure relies on the use of the software BNeumapper,^ based on neural network algorithms, and BLandsat^ multispectral satellite data. Both the software and the data are freely distributed, so theircombined exploitation represents an interesting method for systematically obtaining thematic maps on an issue of great environmental importance. The obtained results show a classification accuracy of 78.46 %. Overall, we found that 29.36 % of the study area, of about 939 km2, was occupied by a sealed surface. The advantages of reproducibility, and consequently of exportability, of the method open important perspectives both on the free sharing of the geographic data as a powerful factor of democracy and on the development of geospatial data market.