A multiscale, object-based image analysis approach in assessing biodiversity for Nepal and New Zealand sites

The relationship between remote sensing and biodiversity is well recognised due to the spatial component inherent in the landscape. The landscape phenomena exist and interact in multiple scales. The interaction in multiple scales occurs within the scale and across the scales. To address the issue of this interaction, we develop a framework in multi-scale environment from remotely sensed data of diverse geographical territories (Nepal and New Zealand) by extracting the meaningful image objects, analysing such image objects and relating these image objects to landscape objects. In relating the image objects to landscape objects, we apply thematic, topological and geometric indices such as the Normalised Difference Vegetation Index (NDVI), the Grey Level Co-occurrence Matrix (GLCM), shape index, area, density and asymmetry for image objects. These indices and the developed framework are tested for pertinent scale (the most appropriate scale for analysis) issues using statistical measure of association – Relative Interquartile Range (RIQR). The test result shows that the pertinent scale can be achieved and it is dependent on interpreter’s objective, heterogeneity / homogeneity of the landscape. This methodology shows that pertinent scale issue is promising in the study of biodiversity and associated landscape phenomena.