| dc.description.abstract | Ecotope maps, representing vegetation communities, are a versatile tool in the Netherlands for calculations on floodplain roughness and monitoring biodiversity. Current ecotope maps show low accuracies regarding grass and brushwood ecotope types and are often produced on a low resolution (smallest units 20 x 20 m). Considering fauna ecology and biodiversity, the current resolution of classification is too low to monitor habitats of small floodplain fauna. Important patch characteristics, such as area and shape; and vegetation patterns such as patch connectivity and heterogeneity have yet not been monitored on meter - decimetre scales. Therefore, in this research floodplain ecotope maps are improved, using high resolution UAV images and object based image analysis (OBIA). These approaches are used to divide the landscape in segments which are homogeneous according to a threshold which is determined with trial and error. Results showed that standard deviations, vegetation height and NDVI are valuable object feature variables for extracting patches on a meter scale. The used method leads to (1) overall classification accuracies of almost 90% regarding grasses and brushwoods, (2) extraction of the most ecologically relevant patch characteristics and (3) (temporal) patterns for small floodplain fauna. OBIA with high resolution images is therefore recommended for future ecotope classification cycles. | |
