Institut für Ökologie (IE)
Filtern
Institut
Sustainable landscape development is the main goal of decision makers worldwide. Achieving this goal in the long term leads to achieving social, economic and environmental sustainability. Remote sensing has been playing an essential role in monitoring remote areas. This study has employed part of the role of remote sensing in supporting the direction of decision makers towards sustainable landscape development. The study has focused on some of the main elements affecting sustainable environment: land uses, specifically agricultural land uses, water quality, forests, and water hazards such as floods. Three research programs were undertaken to investigate the role of Terrasar-x imagery, as a source of remote sensing data, in monitoring the environment and achieving the previous stated elements. The investigation was intended to investigate the effectiveness of TSX imagery in identifying the cropping pattern of selected study areas by employing a pixel-based supervised maximum likelihood classifier, as published in Paper I, assessment of the efficiency of using TSX imagery in determining land use and the flood risk maps by applying an object-based decision tree classifier as published in Paper II, and determination of the potential of inferential statistics tests such as the two samples Z-test and multivariate analysis, for example Factor Analysis, for identifying the kind of forest canopy, based on the backscattering coefficient of TSX imagery of forest plots, as presented in Paper III. Papers I and II covered two pilot areas in the Lower Saxonian Elbe Valley Biosphere Reserve “das Biosphärenreservat "Niedersächsische Elbtalaue" around Walmsburger Werder and Wehninger Werder. Paper III focused on the Fuhrberger Feld water protection area near Hanover in Germany. The inputs for this research were mainly SAR Imagery and the ground truth data collected from field surveys, in addition to databases, geo-databases and maps. The study presented in Paper I used two filters to decrease speckle noise namely De-Grandi as multi-temporal speckle filter, and Lee as an adaptive filter. A multi-temporal classification method was used to identify the different crops using a pixel-based maximum likelihood classifier. The classification accuracy was assessed based on the external user accuracy for each crop, the external producer accuracy for each crop, the Kappa index and the external total accuracy for the entire classification. Three cropping pattern maps were produced namely the cropping pattern map of Wehninger Werder in 2011 and the cropping pattern maps of Walmsburger Werder in 2010 and in 2011. The study showed that image filtering was essential for enhancing the accuracy of crop classification. The multi-temporal filter De-Grandi enhanced the producer accuracy by about 10% compared to the Lee filter. Furthermore, gathering and utilizing large ground truth data greatly enhanced the accuracy of the classification. The research verified that using sequence images covering the growing season usually improved the classification results. The results exposed the effect of the polarization and demonstrate that the majority of the classifications produced according to the crop calendar had higher total producer accuracy than using all acquisitions. The study demonstrated undertaken in Paper II applied the decision tree object-based classifier in determining the major land uses and the inundation extent areas in 2011 and 2013 using the Lee-filtered imagery. Based on the maps produced for the land uses and inundation areas, the hazard areas due to the floods in 2011 and 2013 were identified. The study illustrated that 95% of the inundated area was classified correctly, that 90% of vegetated lands were accurately determined, and around 80% of the forest and the residential areas were correctly recognized. The research undertaken in Paper III statistically analyzed the backscattering coefficient of the Lee-filtered TSX in some forest plots by the Factor Analysis and two sample Z-test. The study showed that Factor analysis tools succeeded in differentiating between the coniferous forest and the deciduous forest and mixed forest, but failed to discriminate between the deciduous and the mixed forest. On one hand, only one factor was extracted for each sample plot of the coniferous forest with approximately equal loadings during the whole acquisition period from March 2008 to January 2009. On the other hand, two factors were extracted for each deciduous or mixed forest sample plot, where one factor had high loadings during the leaf-on period from May to October, and the other one had high loadings during the leaf-off period from November to April. Furthermore, the research revealed that the two sample Z-test enabled not only differentiation between the deciduous and the mixed forest against the coniferous forest, but also discrimination between deciduous forest and the mixed forest. Statistically significant differences were observed between the mean backscatter values of the HH-polarized acquisitions for the deciduous forest and the mixed forest during the leaf-off period, but no statistically significant difference was found during the leaf-on period. Moreover, plot samples for the deciduous forest had slightly higher mean backscattering coefficients than those for the mixed forest during the leaf-off period.