Impervious surfaces extraction using alos palsar data in Ha Noi city

177 HNUE JOURNAL OF SCIENCE DOI: 10.18173/2354-1059.2018-0043 Natural Sciences 2018, Volume 63, Issue 6, pp. 177-184 This paper is available online at IMPERVIOUS SURFACES EXTRACTION USING ALOS PALSAR DATA IN HANOI CITY Dang Vu Khac 1 , Christiane Weber 2 1 Faculty of Geography, Hanoi National University of Education 2 Joint Research Unit Territories, Environment, Remote Sensing, Spatial Information, The National Center for Scientific Research, France Abstract. For identifying urban expansion and impacts induced, it is necessary to have accurate and timely information about the distribution of impervious surfaces (or other meaning is build-up surfaces - Definition of NGS at .org/encyclopedia/) at different periods [1]. In order to identify impervious surfaces coping with the limitation of spectral imagery in the case of cloudy weather, the purpose of the article is solving this issue by using synthetic aperture radar (SAR) images in order to support urban evolution study. Extracting impervious surfaces by coherent image have been promoted and validated through inter-comparison with another imagery product from Landsat TM. Both sensors data have been captured at the same time period. The results demonstrate the capacity of this multi-sensors approach for urban evolution study. This type of product might help authorities in decision making when it is necessary to develop a sustainable strategy for improving urban life, and reducing the impact of urbanization on agricultural resources and natural environment. There are 5 sections in this paper: introduction, study area, method and data used, results and conclusion. Keywords: Interferometry radar, impervious surfaces, coherence, urbanization. 1. Introduction Since the middle of 20th century, world population increases rapidly and this one will reach 6.25 billions people in the year of 2050 if the rate of urban growth will remain constant [2]. Beside the population growth, considerable urban sprawl with a change of land use occurs in many suburban areas, where human features due to socio-economic activities replace rapidly natural features. Monitoring the urban evolution through land use observation and modelling is an efficient approach to assess the impacts of such processes. A large range of observation techniques from in-situ measurement and photographic interpretation to imagery processing techniques might be selected in this case. Since the 1920s, the photographic interpretation became basic method to map spatial development of cities [3]. However, such approach is expensive and difficult to implement. Wherefore the creation of multispectral images processing methods in the 1970s has allowed to map automatically land use and to provide coherent and objective information about many features on Earth's surface [4]. Received June 21, 2018. Revised August 7. Accepted August 14, 2018. Contact Dang Vu Khac, e-mail address: dangvukhac@gmail.com Dang Vu Khac and Christiane Weber 178 In urbanization studies, a lot of works has been executed, based on the association between multispectral images and geographic information systems. We can list here some representative examples: Weber used multispectral images of SPOT during the period 1986 to 1996 to identify urban growth of Tunis City (Tunisia) [5]. Then results were used to project the location of build- up areas in next 10 years (from 1996 to 2006) regarding urban development probabilities. Similarly, Al-Awadhi developed an expansion model of Muscat city (Oman) in using together geographic information layers and remote sensing products [6]. Liu established a model for supporting the urban planning and decision making processes in order to realize the forecast of Beijing's urban expansion [7]. Yagoub has carried some similar researches out Middle East and Iran. This author investigated the development of Al Ain city (United Arab Emirates) during the period from 1976 to 2000, where changes occurred rapidly in Arabian Peninsula [8]. Rahman analysed the evolution of land use during the period from 1990 to 2013 at Al-Khobar city (Saudi Arabia) in exploiting Landsat TM, ETM and OLI, which were captured respectively in 1990, 2001, and 2013. Then the author identified levels of change for urban area with the calculation of Shannon entropy values in these 3 years [9]. Thus, the utilisation of active radar images is still unclosed in the urban evolution studies. In Vietnam, during the last ten years several researchers also develop some joint utilization of remote sensing data and geographic information systems capacities for studying the urbanization processes and assess urban pressures on natural and agricultural areas. Tran Thi Van detected urban build-up areas from Landsat multispectral images based on extraction of impervious surfaces - one specific type of land use in urban environment [10]. Otherwise, Pham Van Cu realized a study at Hanoi city in the context of rapid urbanization, which causes a huge loss of agricultural land at suburban areas [11]. The author used Landsat multispectral images captured in 1993, 2000 and 2007 for investigating the transformation of agricultural land to human settlements. Impacts are not only remarkable through land use changes but also through food resources capacities and farmers activities in rural districts. Although multispectral image processing has some advantages relating to image capture such as revisiting orbits, wide coverage, or numerical values enabling efficient image processing techniques and quantitative approaches. However, remote sensing imagery has also limitations for instance: (i) relating to cloud cover in tropical region such as in the case of Vietnam, (ii) or to access, it is likely that images are not available at any time for investigating. In order to overcome such limitations, we propose to use radar interferometry analysis to identify the distribution of impervious surfaces through a couple of coherent images. 2. Content 2.1. Study area Hanoi is the capital of the country and one of two "special cities" of Vietnam; it locates in an economic priority region of Northern Vietnam. So this city turns into one of typical metropolization case with an explosion of urbanized area and industrialization development. Before July 2008, Hanoi city covered about 921.8 km 2 including nine urban districts with 84.3 km 2 , and 5 rural districts with 836.67 km 2 . Population total was 3145.3×10 3 people [12]. Population density had mean value of 5000 people/km 2 and especially some central districts can reach 30000 people/km 2 , for instance in Dong Da district [13]. At initial period, the provincial boundary remained stable; there was only a modification between urban territory and rural territory with an integration of some rural communes into urban districts. But in August 2008, Hanoi city has integrated the entire Ha Tay province, one part of Hoa Binh and Vinh Phuc Impervious surfaces extraction using ALOS PALSAR data in Hanoi city 179 provinces at last modification. After the adjustment of the administrative boundary, the city covers an area of 3344,7 km 2 with 6451×10 3 people in 2009 (Figure 1). Figure 1. Central urban region of Ha Noi city (Data: [14]) Together with the expansion of administrative boundary, areas of construction works increase abruptly. Previous evolution studies highlighted some locations in urban districts where the land transformation process was occurring. The observation and quantitative measures of these changes have been calculated through land cover layers obtained from multispectral images classifications [11, 15, 16]. Classification methods proved its efficiency for monitoring urbanization process in time and space. In fact, some researches identify the distribution of impervious surfaces in the central urban region that was defined in the Hanoi capital Construction Master Plan to 2030 and vision to 2050. Since now this central region has concentrated a lot of construction works after these adjustment of administrative boundaries in 2008 (Figure 1). In general term, impervious surface is a surface composed of any material that impedes or prevents natural infiltration of water into the soil - https://definitions.uslegal.com. Urban extension is represented via impervious surfaces, which could be identified by land cover classification from multispectral images. However, multispectral remote sensing has limitations in tropical region due to cloud cover during a raining season. It is interesting to evaluate the capacity to apply radar images for identifying impervious surfaces through interferometry radar technique. The obtained results could be integrated into urban evolution study in order to create information about the change of build-up areas and to project urban sprawl in the future. 2.2. Method and data used Synthetic Aperture Radar (SAR) images have advantages due to active radiation, which allows to observe Earth's surface at any time, inclusive of the night time or raining time. Wherefore, radar image of previous satellites such as ERS1/2, RADARSAT1/2, ENVISAT, ALOS and later satellites such as SENTINEL, ALOS-2, TerrasSAR-X, etc. have been used widely in different fields [17]. Aside some applications relating to geologic structure [18], Dang Vu Khac and Christiane Weber 180 natural hazards [19], land use [20], hydrology [21], urban studies which use multi-temporal radar images for analysing human activities are still limited in Vietnam. In urban studies using interferometry radar technique, intensity values are selected for co- registering one pair of radar images, and phase values are used for creating a coherent image. Lopez-Martinez supposes that interferometric coherence is one quantity, which represents the homologous level between master image and slave image [22]. This quantity is calculated an absolute value of correlation coefficient between two images SLC (Single Look Complex) (Equation A.1) and is used as a characteristic quantity for stability level of interferometric phase [23]. Strozzi considers that the phase stability of artificial structures on two images SAR is one measurement index for presenting the existence of a construction work [24]. It shouldn't have many changes between two consecutive paths of satellite in order to conserve a maximum coherence. This is an appropriate index for mapping the distribution of impervious surfaces and supporting urban change detection. However, coherent values depend a lot on two elements: spatial baseline (distance between two sensors) and temporal baseline (time between two shooting date) [25-27]. (A.1) In this formula, ß is coherent value varying from 0 to 1; U1, U2 are SLC images (Single Look Complex) corresponding with the complex conjugate of signals [28]. Grey and Luckman applied this point of view to a pair of ERS images at wavelength = 5.6 cm (band C) for mapping the urban evolution at South Wales (England) [29]. The author recognized that a pair of ERS images conserves high coherence if temporal baseline must be lower than two months and spatial baseline must be shorter than 300 m in temperate zone. However, Noel A. recognized that results of interferometric analysis was strongly affected due to decorrelation when she realized a land subsidence at Hanoi city with Envisat and ERS radar images [30]. After that, in order to reduce interferometric decorrelation while land subsidence rate is calculated, Dang Vu Khac used multi-temporal approach -MTInSAR for extracting permanent scatterers from 21 ALOS radar images (captured during the period from 2007 to 2011) [31]. Figure 2. Relation of spatial and temporal baseline between SAR image pairs Impervious surfaces extraction using ALOS PALSAR data in Hanoi city 181 In this study, we use raw images, which are captured in band L (wavelength = 23.6 cm) by sensor PALSAR in Japanese satellite ALOS. These images are acquired in ascending orbit of satellite at path 475 and frame 400. In order to choose appropriate pair of radar images for creating interferometric coherent image, a graph, which presents the relation between spatial baseline and temporal baseline of all images, is presented in Figure 2. Moreover, it is necessary to consider the availability of Landsat TM, which is acquired during the same period with SAR images in order to have data for comparing the obtained results by interferometric analysis. Wherefore, we choose one SAR image taken on 25 June 2009 and one SAR image taken on 25 September 2009 with a spatial baseline of 92 m and temporal baseline of 90 days. A global digital elevation model GDEM from ASTER 2 satellite (resolution: 1 arcsec) is also used for removing topographic contribution in interferogram. The precise orbital data (<1m), provided in data header file by Japan Aerospace Exploration Agency -JAXA, is exploited to remove orbital contribution in interferogram. 2.3. Results Obtained interferometric coherent image is presented in Figure 3a. In central urban region of Hanoi city, we can perceive that construction works have higher coherent values corresponding with cyan. Several flat features such as water body or runway of airport correspond to black due to total reflectance of radar signals. Agricultural land corresponds to violet. From this coherent image, we apply a thresholding method to extract pixels, which have coherent values varying from 0.45 to 1.0 and we extracted a group of pixel corresponding to the impervious surfaces (Figure 3b). (a) (b) Figure 3. (a) Coherent image is obtained from two ALOS images captured in 25/6/2009 and 25/9/2009. (b) Impervious surfaces extracted from coherent image These surfaces widely localize on the South bank of Red river, however their distribution has a distinct differentiation. In central urban district such as Ba Dinh, Hoan Kiem, Dong Da, Hai Ba Trung, the impervious surfaces nearly completely cover except some park sites and water bodies. At urban districts such as Thanh Xuan, Ha Dong, Hoang Mai, Cau Giay, Tay Ho, Long Bien, Dang Vu Khac and Christiane Weber 182 impervious surfaces occupy only about 50% area with many construction works which have been implementing. Meanwhile at rural districts belonging to central urban region such as: Me Linh, Hoai Duc, Thanh Tri, Thuong Tin, Gia Lam, Dong Anh; impervious surfaces disperse into small and discrete settlements presenting construction works in suburban villages. In order to evaluate obtained results, impervious surfaces from interferometric coherent image are compared with a respective class from supervised classification (Figure 4a), which is handled by maximum likelihood method of Landsat TM image (path - 127; row - 045, acquired on 08 May 2007). At some positions corresponding to three representative areas: central urban district with high density of construction (Figure 4b), urbanizing district (Figure 4c), and suburban district (Figure 4c); we recognize that impervious surfaces match (75,6%) with specific construction works on this Landsat's classification. (b) (c) (a) (d) Figure 4. Classification image of Landsat (a), impervious surfaces extracted from radar coherence (black dashed lines) are drafted on it: (b) central urban district, (c) urbanizing district (d) suburban district 3. Conclusion When we create interferometric coherent image, the choice of ALOS images based on appropriate spatial and temporal baseline contributes an important role. It could conserve high coherence at build-up areas and low coherence at natural areas. Because wavelength is 23,6 cm, minimum critical baseline between two ALOS images can attain to 3600 m with look angle of 23° [32], so coherent values essentially depend on temporal baseline. For that reason, two ALOS images with 90 days of difference were selected. It is enough to retain high coherence for impervious surfaces and to release considerably the coherence of natural features due to the growth of vegetation, the motion of scatterers, or the change of humidity. These are advantageous conditions to separate impervious surfaces from other features on SAR images in mapping impervious surfaces of Hanoi city at the end of 2009. Impervious surfaces extraction using ALOS PALSAR data in Hanoi city 183 Although it still remains several issues relating to confusion between impervious surfaces and various land cover types due to the characteristics of radar sensor, for instance: mirror reflectance on flat constructed surface, high building's shadow, etc. But the results of this study show the capacity of radar remote sensing for mapping impervious surfaces in tropical region. Then it contributes to calculate urban expansion rate and to evaluate the changes of construction works during different periods at the central urban region of Hanoi city. The obtained impervious surfaces map will provide productive information to scientific community, inhabitants, authorities and responsible institutions for urban planning. This information will contribute to clarify the relation between urban expansions with other factors and to adjust human activities in study area such as planning of resource exploitation (land, water, etc.), to adjust urban planning, etc. REFERENCES [1] Weber, C. 2001. Urban agglomeration delimitation using remote sensing data. “In Remote Sensing and Urban Analysis”, eds. J.-P. Donnay, M. J. Barnsley & P. A. Longley, pp. 145-160. London: Taylor and Francis. [2] ONU. 2010. World Urbanization Prospects: The 2009 Revision. New York: United Nations. [3] Hayden, D. 2004. A Field Guide to Sprawl. New York: Norton and Compagny. [4] Klemas, V., 2001. Remote sensing of landscape level coastal environmental indicators. Environmental Management, 27, pp. 47-57. [5] Weber, C. & A. Puissant, 2003. Urbanization Pressure and Modeling of Urban Growth: Example of the Tunis Metropolitan Area. Remote Sensing of Environment, 86, pp. 341-352. [6] Al-Awadhi, T. 2007. Monitoring and Modeling Urban Expansion Using GIS & RS: Case Study from Muscat, Oman. In Urban Remote Sensing Joint Event, pp. 1-5. Paris, France: IEEE. [7] Liu, H. & Q. Zhou, 2005. Developing Urban Growth Predictions from Spatial Indicators Based on Multi-Temporal Images. Computers, Environment and Urban Systems , 29, pp. 580-594. [8] Yagoub, M., 2004. Monitoring of Urban Growth of a Desert City Through Remote Sensing: Al-Ain, UAE, between 1976 and 2000. International Journal of Remote Sensing, 25, pp. 1063-1076. [9] Rahman, M. T., 2016. Detection of Land Use/Land Cover Changes and Urban Sprawl in Al-Khobar, Saudi Arabia: An Analysis of Multi-Temporal Remote Sensing Data. International Journal of Geo-Information, 5. [10] Tran, T. V., 2011. Using remote sensing and GIS for monitoring the urbanization through impervious surfaces at Ho Chi Minh city. Science and Technology Development Journal (in Vietnamese). [11] Pham, V. C., T.-T.-H. Pham, T. H. A. Tong, T. T. H. Nguyen & N. H. Pham, 2014. The conversion of agricultural land in the peri-urban areas of Hanoi (Vietnam): patterns in space and time. Journal of Land Use Science, 10, pp. 224-242. [12] Lam, Q. D., K. L. Pham, M. T. Nguyen & D. L. Dang. 2000. Geography of Hanoi. Hanoi: Hanoi National University Publishing House (in French). [13] Ngo, D. T. & T. T. L. Do. 2011. Four regularizations of administrative boundary at Hanoi city center during the period of 1954 - 2008: experience and signification. In With Thang - Hanoi, eds. V. H. Vo & D. T. Hoang, pp. 599-615. Hanoi: World Publishing House (in French). [14] PPJ, VIAP & HUPI. 2011. Hanoi Master Plan to 2030 and vision to 2050. 196. Hanoi: Ha Noi's Department of Planning and Architecture. Dang Vu Khac and Christiane Weber 184 [15] Dang, V. K. 1998. Studying urban evolution of Hanoi city - Vietnam using remote sensing and GIS. In GDTA, 63. Paris: Paris 6 University (in French). [16] Rossi, G., V. C. Pham, F. Quertamp & O. Chabert. 2002. Atlas informatique de la province de Ha Noi. Ha Noi: Maison d'édition de Cartographie. [17] Jawak, S. D., T. G. Bidawe & A. J. Luis, 2015. A Review on Applications of Imaging Synthetic Aperture Radar with a Special Focus on Cryospheric Studies. Advances in Remote Sensing, 4, pp. 163-175. [18] Dang, V. K., V. C. Pham & A. T. Vu. 1996. Use of aerial radar images on geological study of That Khe - Lang Son area. In Geology Resource, pp. 354-360. Hanoi: Science and Techniques Publishing House. [19] Raucoules, D. & C. Carnec. 1999. DEM derivation and subsidence detection on Hanoi from ERS SAR. In FRINGE99- Advancing ERS SAR Interferometry from applications towards operations, on CDROM. Liege, Belgium: ESA Publications Division. [20] Dương, V. K. 2011. Applying remote sensing technology and GIS for monitoring the growth situation, development and productivity prediction at Red river delta. 120. Hanoi: Institute of Hydrology, Meteorology and Environmental Studies (in Vietnamese). [21] Pham, Q. S., 2008 Applications remote sensing data and GIS for research and management natural resources and environment of coastal zone and islands in Vietnam. Journal of Water Resources and Environmental Engineering. 23, pp. 321-328. [22] López-Martínez, C. & E. Pottier, 2007. Coherence estimation in synthetic aperture radar data based on speckle noise modeling. Applied Optics, 46, pp. 544-558. [23] Hanssen, R. F. 2001. Radar Interferometry: Data interpretation and Error analysis. New York: Kluwer Academic Publishers. [24] Strozzi, T. & U. Wegmuller, 1998. Delimitation of urban areas with SAR interferometry. In Proceedings of IGARSS’98, pp. 1632-1634. Seattle, Washington: IEEE. [25] Li, F. K. & R. M. Goldstein, 1990. Studies of multibaseline spaceborne interferometric synthetic aperture radars. IEEE Geoscience and Remote Sensing Letters, 28, pp. 88-97. [26] Zebker, H. A., S. N. Madsen, J. Martin, K. B. Wheeler, T. Miller, Y. Lou, G. Alberti, S. Vetrella & A. Cucci, 1992. The TOPSAR interferometric radar topographic mapping instrument. IEEE Transactions Geoscience and Remote Sensing, 30, pp. 933-940. [27] Zebker, H. A., P. A. Rosen, R. M. Goldstein, A. Gabriel & C. L. Werner, 1994. On the derivation of coseismic displacement fields using differential radar interferometry - the Landers Earthquake. Journal of Geophysical Research, 99, pp. 19617-19634. [28] Ferretti, A., A. Monti-Guarnieri, C. Prati, F. Rocca & D. Massonnet. 2007. InSAR Principles: Guidelines for SAR Interferometry Processing and Interpretation. The Netherlands ESA Publications. [29] Grey, W. & A. Luckman, 2003. Mapping Urban Extent Using Satellite Radar Interferometry. Photogrammetric Engineering & Remote Sensing, 69, pp. 957-961. [30] Noel, A. 2008. Contribution of radar interferometry in management of natural risks: Hanoi, Vietnam case. In Faculty of Sciences, 106. Liege: University of Liege (in French). [31] Dang, V. K., C. Doubre, C. Weber, F. Masson & N. Gourmelen, 2014. Recent Land Subsidence Caused by the Rapid Urban Development in the Hanoi Urban Region (Vietnam) using ALOS InSAR Data. Natural Hazard and Earth System Sciences, 14, pp. 657-674. [32] Sandwell, D. & Y. Fialko. 2007. ALOS Cal/Val Support at University of California San Diego: Radar Corner Reflectors for Interferometric Phase Assessment. In The first joint PI Synposium of ALOS Data nodes for ALOS science program in Kyoto. Kyoto: JAXA, pp. 1-4.