A GIS FOR RURAL LANDSCAPE SAMPLING

The strategic goal of the research is to provide a statistically sound method for efficiently estimating changes in rural landscape over wide areas. The specific objective of the study is the definition of a sampling design based on a two-steps selection procedure for the assessment of changes in the built environment of rural areas, and its application over a significant study territory. The study area is in the Emilia-Romagna region (Italy) and about 800 km2 wide. We assumed the 1975-2005 time span as the reference period. The estimated variable is the variation of building cover density. The survey procedure adopted consists in GIS backward editing of the cadastre buildings database of the final time reference, through overlay on the raster technical map of the initial time reference. The discretization of the study area for the definition of the areal units was carried out by adopting a sampling frame consisting in the enumeration areas. Strata identification was carried out as a preliminary analysis based on variables already known over the whole territory: classes of land-use/land cover, and land suitability for agricultural and forestry use. Permanent random numbers (PRN) techniques and optimal (Neyman’s) allocation were adopted for sample selection. The final sample allowed us to achieve a Coefficient of Variation (CV) of 8%, thus a precision of estimates greater than the expected one (CV=10%). The results have highlighted that the method allows to obtain a reliable estimate of the changes in the built environment and the identification of the different dynamics in the development of the built areas.