Abstract
Integration of agricultural and environmental policy in the EU has claimed multivariate and multiscale examination of the agricultural environment throughout the EU territories; landscape conservation, environmental impact assessment, and good farming practices must be seen in a common framework. In parallel, EO data of very high spatial resolution has opened up a new insight in the field of landscape pattern analysis and techniques appropriate for their processing, such as object-oriented analysis, are receiving increased attention. Site-specific management in Agriculture is the management of heterogeneity of the agricultural environment, with the aim of improving the economics of the agricultural system and the environmental protection and can be applied on different scales, such as basin, neighbourhood and parcel scale. The scale, on which heterogeneity management is applied, is very important, because spatial data carry different information when observed on different scales. The f ...
Integration of agricultural and environmental policy in the EU has claimed multivariate and multiscale examination of the agricultural environment throughout the EU territories; landscape conservation, environmental impact assessment, and good farming practices must be seen in a common framework. In parallel, EO data of very high spatial resolution has opened up a new insight in the field of landscape pattern analysis and techniques appropriate for their processing, such as object-oriented analysis, are receiving increased attention. Site-specific management in Agriculture is the management of heterogeneity of the agricultural environment, with the aim of improving the economics of the agricultural system and the environmental protection and can be applied on different scales, such as basin, neighbourhood and parcel scale. The scale, on which heterogeneity management is applied, is very important, because spatial data carry different information when observed on different scales. The framework of site-specific management and the control mechanisms available today, more specifically ‘precision agriculture’, ‘best agricultural practices’, ‘environmental impact assessment’, and ‘landscape management’, shows that any isolated practice in sustainable thresholds and scale change in the Mediterranean agricultural system is problematic. Heterogeneity of the system is due to a series of factors, such as the fire, erosion, grazing, construction or abandonment of terraces and intensification of cultivation. This thesis approaches the need for an integrated site-specific management in the Mediterranean agricultural system, using the hierarchy theory for the analysis of the structural character of the landscape. Thus, the main aim of the thesis was the multi-scaled analysis of the landscape, as a background for site-specific management on any scale. In order to carry out this target, very high spatial resolution data was used. This data, such as satellite images, can supply all the basic parameters of the environment and the landscape, such as vegetation, cultivations, land uses, drainage network, and erosion. Though, concentration of big volume of data, does not always reduce uncertainty; therefore, as an alternative, segmentation of data, i.e. hierarchical clustering of pixels into objects with natural meaning, was employed. The hierarchical model that was developed contained 21 classes totally, structured in 3 levels (6 in the 1st, 5 in the 2nd, and 10 in the 3rd level), that were described by 19 fuzzy membership functions in some decades of combinations. The 1st level describes the form of vegetation, the 2nd level describes linear features of the landscape, and the 3rd level describes land uses. The model is scale-independent and automated, because it was constructed by non-parametric terms and more specifically by using fuzzy sets and not training sites. The model proved to be quite stable, while the overall spatial accuracy was moderate, i.e. 74.0% for the 1st level, 50.0% for the 2nd level, and 64.0% for the 3rd level. Though, some specific classes in all levels, such as tree canopy, smooth vegetation, bare land, linear vegetation, non-linear features, buildings, tree plantations, and natural vegetation, were better classified. For the quantitative analysis of the biophysical main structure, vegetation, and land uses of the landscape of a Mediterranean agricultural system (Archanes, Crete), some indices used in Landscape Ecology were adopted, while some new metrics were invented. Thematic information from three hierarchical level, together with topographic and hydrological metrics were employed in order to formulate preliminary management zones on several scales. The methodology was open in parameter selection and flexible with regard to the resulting zones. The method achieved to define and locate areas of interest –on all scales-, that met different criteria each time. The results were optically confirmed. Future research will improve the model for the classes of low accuracy and will validate it for different study areas. Also, it will evaluate deeply the used indices, with regard to their stability in different sampling schemes, and their input in prediction models. Last, it will check stability of different sampling schemes in formulating management zones.
show more
