Research paper
Mapping urban impervious surface with dual-polarimetric SAR data: An improved method

https://doi.org/10.1016/j.landurbplan.2016.03.009Get rights and content

Highlights

  • Dual-polarimetric SAR data were exploited for impervious surfaces mapping.

  • Usage of polarimetric SAR data was superior over single polarization data.

  • Some individual polarimetric features were found to provide negative effect.

  • Approximately 3.5% improvement was achieved using all polarimetric features.

Abstract

Synthetic aperture radar (SAR) data can provide complementary information to improve the mapping of urban impervious surfaces. However, most studies have focused on using only single polarization SAR data. This paper presents a comparative study on the combined use of multispectral optical data and dual polarization SAR data to identify urban impervious surfaces. The experimental results using SPOT-5, TerraSAR-X and ALOS PALSAR data were consistent compared with our previous results using single polarization SAR data. The two-fold result showed that polarimetric SAR images were generally superior to single polarization SAR data for extracting impervious surface areas, although not every individual polarimetric feature could provide a positive result for impervious surfaces mapping. Compared with using only optical and SAR data, the separate HH and HV polarization data improved the accuracy of the results. The incorporation of both Entropy and Alpha features also improved the accuracy. However, the HH/HV ratio and the separate use of coherence did not provide positive results. Noticeably, a combination of all of the dual-polarimetric SAR features was capable of obtaining the best accuracy, with an improvement of approximately 3.5% compared with that of only using SPOT-5 images. This result indicates the superiority of dual-polarimetric SAR data over single polarimetric SAR data for the mapping of urban impervious surfaces.

Introduction

The mapping of impervious surface areas and spatial distributions is important for the environmental study of urbanized areas (Ma et al., 2014; Wu & Thompson, 2013). Among all approaches to impervious surfaces mapping, remote sensing is becoming the major technique due to its convenience and low cost from local to global scales. Many methods have been proposed to extract impervious surfaces using optical remote sensing data (Deng & Wu, 2013; Hu and Weng, 2009, Hu and Weng, 2011a, Hu and Weng, 2011b; Van de Voorde, Jacquet, & Canters, 2011; Weng, 2012; Weng & Hu, 2008; Wu & Murray, 2003; Yang & Li, 2013; Yang, Huang, Homer, Wylie, & Coan, 2003; Zhang, Zhang, & Lin, 2014). More recently, other data sources were evaluated and applied to estimate impervious surfaces, such as time-series planimetric data and nighttime light data (Ma et al., 2014, Wu and Thompson, 2013). However, accurate impervious surfaces mapping is still challenging because of the diverse urban land cover classes and the relative complexity of climatology and phenology effects (Deng and Wu, 2013, Weng, 2012, Zhang et al., 2014). For example, separating impervious surfaces from non-impervious surfaces is difficult when their spectral signatures are similar. Moreover, shaded areas from tall buildings and trees in urban areas are often confused with dark impervious surfaces.

While various data sources have been evaluated to improve impervious surfaces mapping, the potential of Synthetic Aperture Radar (SAR) data was much less investigated. Because of the sensitivity to surface geometric properties (e.g., roughness), SAR data can compensate optical data to accurately identify urban areas (Calabresi, 1996; Henderson & Xia, 1997; Stasolla & Gamba, 2008; Zhang, Zhang, & Lin, 2012; Zhang et al., 2014). Many approaches have been proposed to extract urban areas from SAR data. A Markovian classification algorithm was developed to classify urban cover from high-resolution SAR images (Tison, Nicolas, Tupin, & Maitre, 2004). Texture has been widely identified as a beneficial factor for urban classification using SAR data (Dekker, 2003; Gamba & Aldrighi, 2012; Majd, Simonetto, & Polidori, 2012; Voisin, Krylov, Moser, Serpico, & Zerubia, 2013). Multi-temporal SAR data were also employed to improve the extraction of urban areas (Hu and Ban, 2012, Niu and Ban, 2013). Nevertheless, most of these studies considered only SAR data with single polarization. Multi-polarimetric SAR data offer a much better capacity for distinguishing different scattering mechanisms of ground targets; hence, they were applied to urban area extraction in recent years. Touzi decomposition was applied to polarimetric SAR data for extracting urban areas (Bhattacharya & Touzi, 2011). The RR-LL (R-right; L-left) circular-polarization correlation coefficients were developed to detect man-made targets from SAR data in urban areas (Ainsworth, Schuler, & Lee, 2008). Fully polarimetric (FP) features were also derived from Radarsat-2 FP SAR data for urban studies (Guo, Yang, Sun, Li, & Wang, 2014). All of these above-mentioned studies highlight the advantages and necessities of using polarimetric SAR to improve the extraction of urban areas. The FP SAR system transmits both horizontal and vertical polarized signals alternatively and receives them coherently to measure the polarimetric characteristics of the ground target. Based on this information, the scattering matrix of any combination of transmitted and received polarizations can be derived. However, this approach suffers from the disadvantages of system complexity and reduced swath width caused by doubled pulse repetition frequency (Chen & Quegan, 2011). Therefore, the dual polarimetric (DP) SAR mode is often used as a way to balance swath width and polarization capability (Souyris & Mingot, 2002; Souyris, Imbo, Fjortoft, Mingot, & Lee, 2005). Additionally, SAR data were used jointly with optical data to classify urban areas and obtained promising results (Cao and Jin, 2007, Gamba and Houshmand, 2001, Zhang et al., 2014). The results of these studies indicate that a joint use of optical and SAR data in urban land cover classification yields higher quality than those when they are used separately.

However, the potential of SAR data should be further explored for impervious surfaces mapping. Although dark impervious surfaces (e.g., old rooftops, asphalt roads or parking lots) and bright impervious surfaces (e.g., new concrete roads, new rooftops or concrete parking lots) have similar backscattering characteristics in SAR images, they can have different behaviors in optical images because of their differences in spectral reflectance. Therefore, by incorporating SAR data with optical data, dark impervious surfaces and bright impervious surfaces can be better separated. While most of the existing studies that combine optical and SAR data employ only single polarization SAR data (Cao and Jin, 2007, Gamba and Houshmand, 2001, Zhang et al., 2014), the potential of combining polarimetric SAR data with optical data for impervious surface estimations is still under-explored. This study aimed at evaluating different features of polarimetric SAR data to improve the accuracy of urban impervious surface estimations. In this study, Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) data, with HH (Horizontal transmitting, Horizontal receiving) and HV (Horizontal transmitting, Vertical receiving) polarizations, from the area along the boundary of Shenzhen and Hong Kong were used. Moreover, to evaluate the complementary information carried by dual-polarimetric SAR data, TerraSAR-X data with single polarization and Système Pour l’Observation de la Terre 5 (SPOT-5) data over the same area were also employed for comparison.

Section snippets

Study area and data sets

The Pearl River Delta (PRD) in Southeast China was selected as the study region (Fig. 1). During the past three decades, the PRD has been dramatically urbanized, consequently leading to severe environmental pollutions. There is an urgent demand in continuously monitoring the dynamic urbanization process in PRD, while satellite remote sensing provides a cost-effective approach. Nevertheless, the PRD lies in a rainy and cloudy area with large amounts of cloud contamination, making it challenging

Feature extraction

Various features were extracted from both the dual-polarimetric SAR data and the optical data. The extracted features consist of two categories: polarimetric features from dual-polarimetric SAR data and textural features from single polarization SAR and optical data. First, the polarimetric features were extracted, including the HH/HV ratio, the Alpha and the Entropy decomposition parameters, and the coherence between HH and HV channels. These polarimetric features have been widely reported to

Land cover classification using different optical and SAR data

With the different combinations of optical and SAR image features in Table 1, six land covers could be classified in this study area. The confusion matrix-based accuracy was calculated to evaluate and compare the classification result using the test samples described in the methods section. Table 3 demonstrates the confusion matrix of the validation results, where some interesting findings can be observed. First, compared with using only SPOT data, using only ALOS or TSX data resulted in a

Discussion

The overall objective of this study is to explore the theoretical contribution of polarimetric SAR data to urban impervious surface mapping. To attain this goal, we applied a statistical experiment, such as a trial-and-error test, at the beginning stage to acquire more comprehensive polarimetric SAR data, such as fully polarized and compact polarized data. Our project team collected most of the required polarimetric SAR data in the urban areas of the PRD, with different polarimetric modes,

Conclusions

Urban impervious surfaces have played an important role in various urban studies, such as urban environment, urban ecosystem and urban planning. Therefore, the accurate estimation of impervious surfaces is significant for these studies from local and regional to global scales. Due to the urban land cover diversity and the spectral confusion of these land covers, fusing optical and polarimetric SAR data has become a promising approach to improve urban impervious surfaces mapping. However, most

Acknowledgements

This study is jointly supported by the Research Grants Council, General Research Fund (CUHK 14601515), National Natural Science Foundation of China (41401370), National Basic Research Program of China (2015CB954100) and CUHK Direct Grants (4052093). The authors would like to thank two anonymous reviewers and the editor for providing critical comments and suggestion that have significantly improved the original manuscript.

References (42)

  • A. Bhattacharya et al.

    Polarimetric SAR urban classification using the Touzi target scattering decomposition

    Canadian Journal of Remote Sensing

    (2011)
  • G. Calabresi

    The use of ERS data for flood monitoring: an overall assessment

    (1996)
  • G.Z. Cao et al.

    A hybrid algorithm of the BP-aNN/GA for classification of urban terrain surfaces with fused data of Landsat ETM+ and ERS-2 SAR

    International Journal of Remote Sensing

    (2007)
  • J. Chen et al.

    Calibration of spaceborne CTLR compact polarimetric low-frequency SAR using mixed radar calibrators

    IEEE Transactions on Geoscience and Remote Sensing

    (2011)
  • S.R. Cloude et al.

    A review of target decomposition theorems in radar polarimetry

    IEEE Transactions on Geoscience and Remote Sensing

    (1996)
  • S.R. Cloude et al.

    An entropy based classification scheme for land applications of polarimetric SAR

    IEEE Transactions on Geoscience and Remote Sensing

    (1997)
  • R.J. Dekker

    Texture analysis and classification of ERS SAR images for map updating of urban areas in the Netherlands

    IEEE Transactions on Geoscience and Remote Sensing

    (2003)
  • P. Gamba et al.

    SAR data classification of urban areas by means of segmentation techniques and ancillary optical data

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

    (2012)
  • P. Gamba et al.

    An efficient neural classification chain of SAR and optical urban images

    International Journal of Remote Sensing

    (2001)
  • H.D. Guo et al.

    Synergistic use of optical and PolSAR imagery for urban impervious surface estimation

    Photogrammetric Engineering and Remote Sensing

    (2014)
  • R.M. Haralick et al.

    Textural features for image classification

    IEEE Transactions on Systems Man and Cybernetics Smc3

    (1973)
  • Cited by (54)

    • Automatic impervious surface mapping in subtropical China via a terrain-guided gated fusion network

      2024, International Journal of Applied Earth Observation and Geoinformation
    • Investigation the seasonality effect on impervious surface detection from Sentinel-1 and Sentinel-2 images using Google Earth engine

      2021, Advances in Space Research
      Citation Excerpt :

      For that, three commonly used classification methods have been compared; maximum likelihood (ML), neural network (NN), and support vector machine (SVM) (Fig. 2). Based on the previous studies, pixel-based classification methods have achieved acceptable accuracies in urban feature extraction (Myint et al., 2011; Pandey et al., 2013; Zhang et al., 2016). The Maximum likelihood classification is a parametric supervised method that assumes normal distribution for the data.

    View all citing articles on Scopus
    View full text