On the capability of optical remote sensing imagery with various spatial resolutions on impervious surface estimation

Abstract

Urbanization and population growth led to an increasing demand of land resources worldwide. This global process of land transformation is not only persistent in highly dynamic urban areas in developing countries, but also in cities of the Western world. The exponential increase of impervious surfaces for establishment of settlements and transport infrastructure yields negative consequences in many regards, e.g., increasing surface runoff and flood risk, decreasing groundwater recharge, or intensification of the urban heat island effect. Thus, exact and area-wide estimation of impervious surfaces is of high value and must be repeated regularly in order to provide up-to-date information for policy makers. For example in Germany, surveys on imperviousness are often conducted at irregular intervals with patchy spatial coverage and distribution only. In this regard, suitable remote sensing data offers a cost-effective solution for area-wide surveying and monitoring of impervious surfaces. This study investigates the capability of optical satellite remote sensing imagery with various spatial resolutions for impervious surface estimation. The utilized input data sets range from WorldView-2 imagery with very-high resolution (VHR) of 0.5 m to high resolution (HR) imagery acquired by Sentinel-2 and Landsat-8 with only 30 m spatial resolution. These HR sensors possess the advantage of cost-free imagery and high temporal resolution, which enables effective capabilities for monitoring imperviousness. VHR data is analyzed by means of object-based image analysis (OBIA), while Support Vector Regression (SVR) is employed for estimation of impervious surfaces in HR imagery since individual urban objects cannot be resolved at lower spatial scales. In addition, different input features are evaluated systematically in order to assess the potential of data compression by means of Principal Component Analysis (PCA) as well as spectral indices like the Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation index (SAVI), or Normalized Difference Built-up Index (NDBI). Our results illustrate that the OBIA classification of the WorldView-2 imagery with 0.5 m spatial resolution for the city of Munich achieved estimations of imperviousness with a Root Mean Square Error (RMSE) of about 10.0 % compared to official data of imperviousness provided by the city municipality. Naturally, the classification based on HR data obtained lower accuracies. However, RMSE values in the order of 20.0 % - 25.0 % dependent on the utilized features and spatial resolutions reveal that these remote sensing data sets still allow a general assessment with fair to high accuracy at no cost. In general, the accuracy of impervious surface mapping is highly dependent on the spatial resolution of the input data, i.e. the accuracy increases with spatial detail of the imagery. In addition, the combination of spectral indices or data compression techniques like PCA yielded competitive results compared to the original input values of the multispectral bands.