Comparison of landscape metrics for three different level land cover/use maps generated from SPOT6/7 images and object based classification approach

Abstract : High resolution (HR) satellite images are very important geospatial data sources to create up-to-data. Land cover/land use (LCLU) maps to be used for city and regional planning processes and determining spatio-temporal changes of cities. Land cover/use maps with different scales could be produced using satellite images obtained from variety of sensors having different spatial and spectral characteristics and classification methods. This research aims to create three different land cover/land use (maps) based on three different levels of Coordination of Information on the Environment (CORINE) nomenclature. CORINE includes five main classes at first level which are artificial surfaces, agricultural areas, forest and semi-natural areas, wetlands and water bodies. These five classes become thematically more detailed at second level with 15 classes and the richest thematic classes are at level 3 with a total number of 44 land cover/use classes. Class definitions become more complicated toward to third level and classification of third level classes is becoming a challenging task. This study aims to develop decision trees for different classes at three different hierarchical level to be used in object based classification. Izmir metropolitan city of Turkey was selected as study area considering its complex and various landscape characteristics. High resolution SPOT 6/7 images having 1.5 m spatial resolution was used as main earth observation data for this research.In order to identify second and third level LCLU classes, additional geospatial data from open sources will be included in the study. Several landscape metrics such as Patch Density (PD), Edge Density (ED), Largest Patch Index (LPI), Euclidean Nearest Neighbor Distance (ENN), Area-Weighted Mean Fractal Dimension Index (FRAC_AM) and Contagion (CONTAG) metrics etc. were calculated for three different level LCLU maps and results were compared.