Using Cartographic Modelling and Regression Trees for Assessing Spatial Distribution of Cadastral Parcels and Agricultural Holdings on Rural Areas
ISBN 978-85-88783-11-9
Authors
1GÄ…siorowski, J.
1INSTITUTE OF GEODESY AND CARTOGRAPHY Email: jedrzej.gasiorowski@igik.edu.pl
Abstract
The paper presents a method of assessing the quality of spatial distribution of cadastral parcels and agricultural holdings within a village depending on environmental conditions. The method is based on plenty of indicators representing in quantitative way both spatial distribution and environmental determinants (such like terrain relief, land use structure, hydrographic conditions, road network etc.). The indicators values have been determined for the study area of four villages in Poland localized in different country regions and representing various terrain relief conditions, land use structure and farming culture. Moreover all the villages have been a subject of land consolidation in recent years, so the cadastral parcels and agricultural holdings distribution can be treated as optimal on the study area. The indicators values have been calculated with the use of cartographic modelling and more or less complex spatial analyses, including inter alia the analysis of geographic objects spatial distribution focusing on the measuring of density, clustering and dispersion of point, linear and surface objects and availability analysis. To develop a model of relationships between environmental conditions and optimal parcels and holdings spatial distribution within the village, the method of regression trees was used, where agricultural holdings of the study area were the observation set, indicators representing environmental conditions were explanatory variables and indicators representing spatial distribution were response variables. The whole set of agricultireal holdings was randomly divided into two subsets: training set (80%) and test set (20%). The model was developed on the basis of training subset and validated using the test subset. The model has a form of three decision trees (one tree for each indicator representing a specific aspect of agricultural holdings spatial distribution within a village), which nodes represent decisions based on specific environmental conditions indicators and leafs represent the expected value of holding spatial distribution indicator. The method includes also a proposal of cartographic visualization of cadastral parcels and agricultural holdings spatial distribution assessment, that can be a graphical support for identifying objects which distribution is not optimal and requires to improved.