Zone II Versions EN1 Vol 4 (2) 2019
Download
A dataset of urban impervious surface of Hainan Island, based on Sentinel-1 SAR and Sentinel-2A optical images.
 >>
: 2018 - 11 - 29
: 2019 - 03 - 26
: 2019 - 01 - 07
: 2019 - 05 - 13
148 0 0
Abstract & Keywords
Abstract: Urban impervious surface is an important indicator of urbanization. It is also an important indicator to measure the health of urban ecosystems. In the context of global urbanization, impervious surface extraction has become a hotspot of scholarly attention. This dataset is derived from a fast large-area impervious surface extraction method based on a combined use of optical and SAR data. This dataset has a spatial coverage of 18°10′N–20°10′N, 108°37′E–111°03′E, and a time span of the year 2015. 10,000 randomly selected validation points were utilized to assess the accuracy of this dataset. The overall accuracy (OA) and Kappa coefficients were 90.71% and 0.81, respectively. Additionally, our dataset was compared with existing mainstream land cover products (i.e. GlobalLand30 (30 m), Liu’s products and GHS).The reseach results show that our products have higher extraction accuracy and are more refined in exhibiting internal structures of human settlements.
Keywords: Urban impervious layer; Sentinel-1 SAR; Sentinel-2A; Google Earth Engine
Dataset Profile
TitleA dataset of urban impervious surface of Hainan Island, based on Sentinel-1 SAR and Sentinel-2A optical images
Data corresponding authorZhao Xiangwei (tlzxw1696@163.com)
Data authorsZhu Xiulin, Zhao Xiangwei, Du Wenjie, Sun Zhongchang
Time range2015, 2018
Geographical scope18°10′N—20°10′N, 108°37′E—111°03′E
Spatial resolution10 m
Data volume8.71 MB (26.4 MB after being decompressed)
Data formatTIFF
Data service system<http://www.sciencedb.cn/dataSet/handle/695>
Sources of fundingMajor Science and Technology Program of Hainan Province (ZDKJ2016021)
Dataset compositionThis dataset includes Hainan Island’s 2015–2018 Phase 2 impervious surface distribution data, all data is stored in a compressed file. The compressed file contains 2 TIFF files, which correspond to the distribution of impervious surface of Hainan Island in 2015 and 2018. And the dataset is updated once every three years.
1.   Introduction
Urban impervious surface refers to the land-covered surface that prevents water from seeping into the soil, including roads, parking lots, driveways, humanities, building roofs, and other non-permeable surfaces in urban objects. With the continuous development, the urban natural landscape is destroyed, and the percentage of land covered by impervious surfaces is gradually increasing.[1]Urban impervious surfaces are closely related to many environmental problems, such as water quality, runoff, and urban heat island effects[2,3] so impervious surface has become a key indicator for assessing urban ecological environment.[4,5]
Optical remote sensing images rely mainly on image classification when extracting urban impervious surfaces. This method has subjective single-view data analysis, and it has complicated calculations and takes time. The optical image with low spatial resolution has mixed pixels, foreign matter and spectrum, and it is troublesome for classification of objects. Affected by clouds, fog and water vapor, it is difficult to obtain good quality optical images in low latitudes, which brings great difficulties to urban dynamic remote sensing monitoring,[6]and the optical image is greatly affected by the time when the image is acquired in the impervious surface of the city.1 Compared with optical images, Synthetic Aperture Radar (SAR) has the characteristics of full-time, all-weather, high precision and high efficiency. SAR data has great application potential in urban remote sensing.[7,8] So far, there has been little research on large-scale, high spatial resolution impervious surface mapping based on SAR data and optical image fusion.
2.   Data collection and processing
2.1   Introduction of the study area
Hainan Island is located on the southernmost tip of China. Its location is between 108°37'–111°03′east longitude and 18°10′–20°10′north latitude. The outline of the island is elliptical. The central part of Hainan Island is high and low in the middle. The mountains, hills, terraces and plains form a ring-shaped layered landform, and the cascade structure is obvious. Most of the larger rivers originate from the central mountainous area and form a radial water system. The terrain on Hainan Island is complex, with a large number of towns, villages and different types of features. This dataset is mainly produced for Hainan Island. The Sentinel-1 SAR image of the study area is shown in Figure 1.


Figure 1   Image map of study area
2.2   Data preprocessing
The Sentinel-1 satellite is the Earth observation satellite in the European Space Agency's Copernicus Plan (GMES). The research data in this paper mainly uses the Sentinel-1A to obtain the Ground Range Detected data set for a dual-polarized C-band synthetic aperture radar instrument. On the Google Earth Engine (GEE) platform, SAR images are processed using the Sentinel-1 toolbox,[9] which applies track files (using recovered orbits), thermal noise removal, radiation calibration, terrain correction, and stripe processing. The data set is already generating calibration and orthomorphism and the calibrated product is updated weekly.
During the ascending and descending orbits, Sentinel-1 data is collected through several different instrument configurations, resolutions, and band combinations, and the data is transformed into homogenous subsets. In GEE, the Sentinel-1 image is processed into a backscatter coefficient (\({\mathrm{ }\sigma }^{°})\) in decibels (dB). The backscatter coefficient represents the backscattered area of the target per unit of ground area.
The auxiliary data is Sentinel-2 data, which is a high spatial resolution image data of wide-format, multi-spectral imaging. The Sentinel-2 data contains 13 16-byte spectral bands, as shown in Table 1. At the same time, Sentinel-2 data has three QA bands, of which Q60 is a bit mask band with cloud mask information. In this study, the 2015 impervious data shared 517 SAR data and 1376 scene optical data. In 2018, the impervious data shared 522 SAR data and 2313 scene optical data. The capability of having several imaging modes in a wide spectral range gives us very important analytical information before and during the process of study.
Table 1   Sentinel-2 data spectrum information table
Band nameCenter wavelength /μmGround spatial resolution /m
B10.44360
B20.49010
B30.56010
B40.66510
B50.70520
B60.74020
B70.78320
B80.84210
B8A0.86520
B90.94560
B101.37560
B111.61020
B122.19020
2.3   Data Processing Principle and Procedures
This data set is obtained by using SAR data and optical image extraction of multi-temporal and lifting rails. The technical route is shown in Figure 2. The technical method includes two aspects: one is to perform backscattering processing and inverse calculation on the GEE multi-temporal lifting orbit SAR data to obtain the original backscattering coefficient \({\mathrm{ }\sigma }^{°}\)
\({\mathrm{ }\sigma }^{°}={10}^{\frac{x}{10}}\) (1)
Where x is the initial pixel value of the Sentinel-1 image under GEE; Secondly, based on the pixel, the image texture features of the SAR data are obtained by analysis, threshold segmentation is performed, and the impervious surface is automatically extracted by optical image and digital elevation model (DEM) data mask refinement.


Figure 2   Technology Roadmap
In the first aspect of data processing, the multi-temporal lifting rail Sentinel-1A GRD (multi-view processed ground distance product) data is obtained, and the backscattering coefficient of the SAR data is processed and averaged, so that Solve the effect of shadows or overlays on the impervious surface extraction in mountainous and urban areas; merge the multi-temporal lifting orbit SAR images into a single image layer by mean processing and obtain the averaged parameter file.
In the second aspect of the data processing, the corresponding texture features are obtained according to the averaged SAR intensity image and related parameters, and the potential impervious surface distribution map is obtained by analyzing the intensity features; Based on the analysis of SAR images and their texture features, the histogram statistics of the Hainan Island optical index and the mean-filtered SAR data are analyzed by histogram. The "double peak method" is used to estimate the optimal threshold and processed locally. After the repeated trials, the optimal segmentation threshold is obtained, and the impervious surface distribution data is obtained on the GEE platform by using the impervious surface fast extraction algorithm. Figure 3 is a schematic diagram of obtaining an optimal threshold by analyzing a backscatter coefficient of a SAR image.


Figure 3   Histogram calculated by the bimodal method
In order to reduce the impact of shadows and overlays on the impervious surface extraction, this dataset uses SRTM products with a spatial resolution of 30 m, which is solved by the slope property. At the same time, we will use the NDVI _max index generated by the Sentinel-2 optical image to finely extract the low-density impervious surface area. The NDVI_max index is the maximum value of the vegetation coverage index calculated for each scene image in the same pixel. Hainan Island vegetation mask threshold selection is shown in Figure 4.


Figure 4   Vegetation separation threshold selection map
3.   Sample description
The distribution of the impervious surface of Hainan Island is shown in Figure 5. This dataset is the impervious surface data of Hainan Island. The spatial resolution of the dataset image is 10 m, TIFF format, and the coordinate system of all data is WGS1984.

(a)Hainan Island impervious surface map in 2015


(b)Hainan Island impervious surface map in 2018

Figure 5   Distribution of impervious surface of Hainan Island in 2015–2018
4.   Quality control and assessment
Accuracy verification of Hainan Island classification results, establishment of confusion matrix, quantitative assessment of classification accuracy was based on 10 000 randomly selected reference points. The accuracy calculation method is the ratio of the sum of diagonal elements in the confusion matrix to the total. In general, a Kappa coefficient of 0.6 indicates that the quality of the classification is good. By calculation, the overall accuracy of Hainan Island is higher than 90%, and the Kappa coefficient is higher than 0.81, as shown in Table 2 and Table 3. This shows that the extraction results of Hainan Island are good and the precision is high.
Table 2   Accuracy verification results of Hainan Island impervious surface distribution data in 2015
Non-Impervious surfaceImpervious surfaceTotalUA/%
Non-Impervious surface4637363500092.47
Impervious surface5664434500088.68
Total5203479710 000
PA/%89.121 6692.432 77
Overall Accuracy / %90.71
Kappa0.814 2
Table 3   Accuracy verification results of Hainan Island impervious surface distribution data in 2018
Non-Impervious surfaceImpervious surfaceTotalUA(%)
Non-Impervious surface4716284500094.32
Impervious surface5834417500088.34
Total5299470110 000
PA(%)88.992 9293.95873
Overall Accuracy / %91.33
Kappa0.8266
Each period of data randomly selects 70 block areas, using Google Earth's historical image to calculate the impervious surface area of each block area in 2010, 2015 and 2018, and a linear regression equation was established with the extraction results of this dataset, GlobalLand30 product, globe urban2015 product and GHSL product, are shown in Figure 6.

(a)Impervious surface distribution results of 2015


(b)Impervious surface distribution results of 2018


(c)GlobalLand30 products


(d)urban2015 products


(e)GHSL products

Figure 6   Figure 6 Different data and Google Earth linear regression equation
Comparing the detailed results of the classification, it is found that the extraction results of this data set can visually and clearly show the internal structure of the city and the distribution pattern of the building. Compared with this dataset, GlobalLand30 products are more aggregated and cannot reflect the details of the internal structure of the impervious surface; urban2015 products fail to extract the impervious surface of the low-density area and have rectangular boundaries around the city; GHSL products have low density. The extraction effect of the impervious surface area is poor, and the final extraction result is leaked. As shown in Figure 7.

(a) Sentinel-2 optical image maps


(b) Sentinel-2 optical image maps


(c) Sentinel-2 optical image maps

Figure 7   Comparison of extraction results with GlobalLand30 products, urban2015 product details
Notes: (a), (b), (c) are Sentinel-2 optical image maps

(d) data set results


(e) data set results


(f) data set results

 
Notes: (d), (e), (f) are data set results

(g) GlobalLand30 product results


(h) GlobalLand30 product results


(i) GlobalLand30 product results

 
Notes: (g), (h), (i) are GlobalLand30 product results

(j) urban2015 product results


(k) urban2015 product results


(l) urban2015 product results

 
Notes: (j), (k), (l) are urban2015 product results

(m) GHSL product results


(n) GHSL product results


(o) GHSL product results

 
Notes: (m), (n), (o) are GHSL product results
We have compared the data of Phase 2, and we can clearly see the changes in the impervious surface of the city. Through the analysis of the scope of change, we can understand the progress of urban development. Take Haikou Meilan International Airport as an example, as shown in Figure 8.

(a)Google Earth imagery on April 19, 2015


(b)Impervious surface extraction results of 2015


(c)Google Earth imagery on October 30, 2018


(d)Impervious surface extraction results of 2018

Figure 8   Comparison of the results of 2015 and 2018 at Haikou Meilan International Airport
5.   Value and significance
Based on the GEE cloud computing platform, this paper uses the multi-temporal and lift rail Sentinel-1 SAR data and Sentinel-2A optical image to obtain a high-accuracy impervious data set by using the impervious surface fast extraction method.
The amount and distribution of impervious surfaces are important input parameters of hydrological models, especially in highly urbanized basins. Urban imperviousness is an indicator of the urbanization process and a necessary indicator for monitoring environmental changes and ecosystems. Urban impervious surface will affect the urban water cycle, water quality, surface runoff, climate and other ecological environmental factors. It is of great significance for the analysis and evaluation of climate, environment and water cycle in urban areas, and is widely used in the urban comprehensive assessment process. Impervious data sets not only contribute to the management of the urban environment, but also have a great guiding role in the future planning of the city.
6.   Usage notes
Hainan Island urban impervious data sets are saved in TIFF format. ArcGIS, ENVI, ERDAS and other commonly used GIS and remote sensing software can support the reading and operation of this data.
1.
Sun Z C, Wang C Z, Guo H D, et al. A modification of normalized difference impervious surface index (MNDISI) for automatic impervious surface area extraction[J/OL]. Remote Sensing, 2017, 9(9): 942. DOI: 10.3390/rs9090942.
2.
Weng Q H. Remote sensing of impervious surfaces in the urban areas: Requirements, methods, and trends[J]. Remote Sensing of Environment, 2012, 117(2): 34-49.
3.
Sun Z C , Li X W, Fu W X, et al. Long-term effects of land use/land cover change on surface runoff in urban areas of Beijing, China[J]. Journal of Applied Remote Sensing, 2013, 8(1): 1354-1365.
4.
Fan F L, Fan W, Weng Q H. Improving urban impervious surface mapping by linear spectral mixture analysis and using spectral indices[J/OL]. Canadian Journal of Remote Sensing, 2015, 41(6): 577-586. DOI: 10.1080/07038992.2015.1112730.
5.
Li W L, Wu C S, Choi W. Predicting future urban impervious surface distribution using cellular automata and regression analysis[J/OL]. Earth Science Informatics, 2018,11(1): 19-29. DOI: 10.1007/s12145-017-0312-8.
6.
Yu S S, Sun Z C, Guo H D, et al. Monitoring and analyzing the spatial dynamics and patterns of megacities along the Maritime Silk Road [J]. Journal of Remote Sensing, 2017, 21(2): 169-181.
7.
Lan H N, NGHIEM S V, HENEBRY G M. Expansion of major urban areas in the US Great Plains from 2000 to 2009 using satellite scatterometer data[J/OL]. Remote Sensing of Environment, 2018, 204: 524-533. DOI: 10.1016/j.rse.2017.10.004.
8.
Zhang H S, Lin H, Wang Y P. A new scheme for urban impervious surface classification from SAR images[J/OL]. ISPRS Journal of Photogrammetry & Remote Sensing, 2018, 139: 103-118. DOI: 10.1016/j.isprsjprs.2018.03.007.
9.
VECI L, PRATSIRAOLA P, SCHEIBER R, et al. The Sentinel-1 Toolbox[C]. IEEE International Geoscience & Remote Sensing Symposium DLR, 2014.
10.
CONGALTON R G. A review of assessing the accuracy of classification of remotely sensed data[J]. Remote Sensing of Environment, 1991, 37(1): 35-46.
Data citation
Zhu X, Zhao X, Du W & Sun Zh. A dataset of urban impervious surface of Hainan Island, based on Sentinel-1 SAR and Sentinel-2A optical images. Science Data Bank, DOI: 10.11922/sciencedb.695 (2019).
Article and author information
How to cite this article
Zhu X, Zhao X, Du W & Sun Zh. A dataset of urban impervious surface of Hainan Island, based on Sentinel-1 SAR and Sentinel-2A optical images. China Scientific Data 4(2018).DOI: 10.11922/csdata.2018.0084.zh
Zhu Xiulin
Mainly responsible for work: data analysis and paper writing.
(1994- ), Female, People in Heze City, Shandong Province, Master student, Research direction for spatial data mining and application.
Zhao Xiangwei
Mainly responsible for work: data quality control.
tlzxw1696@163.com
(1974- ), Male, People in Jining City, Shandong Province, Ph.D., Associate professor, Aesearch direction for spatial data processing and analysis.
Du Wenjie
Mainly responsible for the work: design and quality control of the data processing process.
(1994- ), Male, People in Jiangyin City, Jiangsu Province, Master student, Research direction for geodesy and measurement engineering.
Sun Zhongchang
Mainly responsible for the work: pre-processing of data and quality control.
(1983- ), Male, People in Jining City, Shandong Province, Ph.D., Associate researcher, Research direction is urban remote sensing, microwave remote sensing.
Major Science and Technology Program of Hainan Province (ZDKJ2016021)
Publication records
Published: May 13, 2019 ( VersionsEN1
Released: Jan. 7, 2019 ( VersionsZH1
Published: May 13, 2019 ( VersionsZH2
References
中国科学数据
csdata