Urban land use detection and change using an enhanced monitoring system with remote sensing tools and GIS in the EUREGIO Meuse-Rhine area

  • Urbaner Landnutzungswandel und -detektion unter Verwendung eines weiterentwickelten Monitoringsystems mit Fernerkundung und GIS in der Euregio Maas-Rhein

Mönnig, Carsten; Lehmkuhl, Frank (Thesis advisor)

Aachen : Publikationsserver der RWTH Aachen University (2007)
Dissertation / PhD Thesis

Aachen, Techn. Hochsch., Diss., 2007

Abstract

The urban areas of Central Europe changed considerably during the past three decades. The spatiotemporal changes, induced by expansions and conversions, are monitored and registered in cadastral maps and statistical data. These national data-sets are indexed with specific criteria in every single country and end at the national borders. A cross-national monitoring of urbanised regions in a market area like the Euregio Meuse-Rhine (EMR) is not possible without considerable effort. This research paper will accomplish for the first time the detection and differentiation of urban areas utilising remote sensing tools within the EMR. The changes are documented in a time frame of the late eighties to the beginning of the new millennium. The usage of remote sensing tools allows for the possibility of trans-national urban land use/land cover analysis, the detection of its changes and the interaction between other types of land use. Spatial data like statistics, digital topographic maps and information systems was used as an additional source of information in this research project. Different remote sensing sensors were tested which represent the earth surface in diverse resolutions. The utilised satellite platforms were QuickBird with 4 metres, ASTER with 15 metres and Landsat with 15 respectively 30 metres resolution. Additional aerial photographs with a pixel size of 4 metres were employed as well. Different classification methods and extra synthetic channels were evaluated in order to increase the classification accuracy. An urban land use detection of different EMR communities was completed between the years 1988 and 2001 using a resolution of 30 metres. Higher resolution data was not obtainable from the past years. The usability of meso- to high- resolution remotely sensed data and the corresponding classification results was assessed in regions of different population density. The classification results of the meso-scale satellite imagery were evaluated with supplementary information like national statistics, topographic information systems and digital maps for their usability and accuracy as a land use detection instrument. The results were implemented and visualised in a Geographic Information System (GIS). The main output of this study is the good applicability of meso-scale satellite images to detect urban land use alterations. A classification accuracy of 70 to 80% was achieved, even though it was not possible to accomplish such a fine result in every part of the EMR. In particular more satellite images in between the years 1988 and 2001 were desirable in order to improve the land use change detection. Bonus synthetic channels like diverse texture analyses and a vegetation index enhanced the accuracy to a great extent. After evaluating all utilised resolutions, the 4 metre and the 30 metre imagery provided the best electronic reproduction of urban land use. The high resolution sensors offered the most precise product with a wealth of detail without accumulated mixed pixels. The different urban classes “Continuous and Discontinuous Urban Area” and “Urban Fringe” were differentiated with an accuracy of more than 80%. The comparison of meso-scaled satellite data with official statistics was not possible in a straight approach, because the conditions of inventory of the three EMR countries are different. However, a consistent trend of increasing urbanised districts was observed in selected communities. The classification accuracy of meso- to high resolution data did not show a distinctive discrepancy in densely populated and less densely populated areas, but the high-resolution imagery provides superior results. Only un-altered Landsat 30 metre classifications prove a slightly increasing error quotient in densely populated areas. The assessment of meso-scaled data with existing topographic information systems prove a high correlation, the 15 metre did not show this precision. Concluding, it could be said that the high-resolution imagery, together with synthetic channels, currently provides the best option for the urban land use detection. They will offer a precise image of urban trans- formations in the near future. Due to the currently high costs and limited availability of this data, imagery with inferior resolution will be utilised in the next years, at least in order to provide comparability with remotely sensed data of the past three decades.

Identifier