Post-fire landscapes are complex systems to analyze through remote sensing data. An important challenge of this paradigm is to understand how textural features change with scale in a real variety of landscapes. The complexity of these systems can best be explored through their partition into objects or patches. The challenge and flexibility of the multi-scale segmentation/object algorithm lies in defining the parameters that relate to different organization levels. Quantitative changes in measurements across space scales differ depending on how the scale is defined.
The purpose of this study is to evaluate how different textural features contribute most to discriminating vegetation types and status after fire. The textural indices analyzed were: homogeneity, dissimilarity, entropy and contrast.
The methodology involves:
Multiscale dataset generation using an object-based image analysis and an upscaling technique.
Semi-automatic hierarchical classification.
Multi-scale analysis of textural features of the objects.
A scale effect evaluation using discriminant analysis.
Results shown in Fig.1 the responses of different textural indices with changing scales. Semi-automatic hierarchical classification showed the capacity of discriminating plant communities at different landscape levels. The interaction of texture features in a region-based multi-scale image segmentation derived from high resolution imaging gave very promising results in plant community identification of degraded ecosystems.
As Fig.2 shows, bject-based analysis of textural features such us entropy or homogeneity is closely related with the organization level of shrubs communities after fire. Those results suggest the capacity of textural features analysis in post-fire vegetation evolution assessment.
Figure 1. Object-based jeralquical segmentation.
Figure 2. Multiresolution segmentation. Composition of homogeneity criterion: Shape 0.1 and Compacness 0.5
References
Cerrillo, R. M., Rosenfeld, M., Pérez-Aranda, J., Padrón, E., Guzmán, J. R., Clemente, R. H. et al. (2008). Evaluación de la mortalidad de bosques de ñirre (nothofagus antárctica) en la patagonia chilena mediante imágenes Landsat TM y ETM+_1.
Padrón, E. & Cerrillo, R. M. (2007). Aboveground biomass in Prosopis pallida (Humb. and Bonpl. ex Willd.) H. B. K. ecosystems using Landsat 7 ETM+ images. Revista chilena de historia natural, 80, 43-53.