INFERENTIAL METHODS FOR RURAL LANDSCAPE CHANGE ASSESSMENT

The analysis of geographical data describing rural landscape features referred to manifold time steps is necessary for the formulation of a proper knowledge framework for land-use planning tools. When detailed measurements of the actual variations of main landscape features are required, GIS analyses are time consuming; thus they can be carried out only on a sample of the whole territory. The study aims at comparing alternative sampling methods for rural landscape at the wide area scale, and assessing which one is the most appropriate for a detailed measurement of the actual variations of main landscape features. We have chosen a target study area (787 km2) in the eastern part of the Bologna province, Italy. The parameter we have estimated is the change in building cover density, that is the difference between the area covered by buildings at the end and the beginning of the study period 1975-2005, divided by the land area. We have set up an area frame that subdivides the territory into area units. We have decided to take advantage of the enumeration areas used for the most recent population and housing census. The need to maximize the precision of the estimate of the landscape change suggested the adoption of a stratification which identifies zones characterised by different rates of change in the main landscape features. The stratification we have created is based on a combination of appropriate land-use/land-cover classes - referred to the end of the studied period - and of land suitability for agricultural and forestry use. The stratification has been refined reclassifying the enumeration areas according to the attribution of the majority of their surface to areas already urbanized at the beginning of the study period or to areas urbanized during that period. A further subdivision has been made depending on whether the final destination is predominantly residential, productive, or any other type of urban use. First, we have compared the number of sample units needed for reaching a chosen precision of the estimate with two methods; then we have taken into consideration operational disadvantages through a cost function and we have compared the precision of the estimate given the same total budget. According to our results, we should suggest to use the two-steps selection procedure (TSPRN) for estimating the change in the main landscape features on wide areas. However, the selection and survey process can be optimized in an operational project, reducing some of the costs specified in the cost function.