Land-use and land-cover (LULC) studies can provide some of key indicators of desertification (i.e. land degradation, change of rainfall pattern, disturbance of hydrological cycle, etc.). Remote sensing, with the ability to cover wide areas, is valuable for assessing and monitoring the status of the natural resources, detecting the changes in spatial and temporal scale and predict them for the future. To this aim, multiple data collections are necessary to monitor environmental processes, and the choice of a reliable change detection technique on radiometrically homogeneous data can help identify land cover transformations. This paper discusses the development and implementation of a method that can be used with multi-decadal LANDSAT data for computing LULC maps on the study area located in northern Apulia region (Italy), in order to improve knowledge about land degradation processes and to support further studies in the field of hydrological modelling. More precisely, the territory includes the Gargano area, the Candelaro, Cervaro, Carapelle, Ofanto river basins and the reclamation ground of Margherita di Savoia. The area has faced large scale human induced LULC changes, but quantitative data is to date lacking. In this study case the acquisition dates of 1984, 1987, 2001, 2003, 2009 and 2011 of imagery were selected to cover a time trend of 27 years. The historical archive of LANDSAT imagery dating back to the launch of ERTS in 1972 provides a comprehensive and permanent data source for tracking change on the planet’s land surface due to image availability and accessibility. After processing auxiliary synthetic bands to improve training pixel separability and to maximize information variance of data, the and Multi-Layer Perceptron (MLP) feed-forward neural network was chosen as classification method, aimed at the characterization of land use over the time according to the level of surface imperviousness. Among many land use classification methods, the Multi-Layer Perceptron (MLP) neural network approach was found the best reliable and efficient method in the absence of historical ground reference data. The Artificial Neural Networks (ANN) approaches have a distinct advantage over statistical classification methods in that they are non-parametric and require little or no a priori knowledge on the distribution model of input data. Additional superior advantages of ANNs include parallel computation, the ability to estimate the non-linear relationship between input data and desired outputs, and fast generalization capability. Many previous studies on the classification of multispectral images have confirmed that ANNs perform better than traditional classification methods in terms of classification accuracy.
Attenzione! Scheda prodotto non ancora validata
I metadati della pubblicazione sono in fase di verifica da parte dell'Ateneo
|Titolo:||Characterizing land cover changes using multitemporal LANDSAT imagery in Northern Apulia (Italy)|
|Data di pubblicazione:||2014|
|Nome del convegno:||First joint Workshop of the EARSeL Special Interest Group on Land Use & Land Cover and the NASA LCLUC Program on "Frontiers in Earth Observation for Land System Science"|
|Appare nelle tipologie:||4.2 Abstract in Atti di convegno|