Mapping impervious surface distribution in China using multi-source remotely sensed data

Figures & data

Table 1. Datasets used in research.

Dataset	Description
VIIRS DNB^a	Time range: June–September 2015 Spatial resolution: 500 m Reference system: Albert Equal Area projection
MODIS NDVI time-series data^b	Data product: MOD13Q1 Spatial resolution: 250 m Time range: January–December 2015
Landsat OLI data^c	Over 500 scenes of Landsat 8 OLI images without clouds or less than 1% of clouds were used. These images were mainly acquired in growing season (May–October) in 2015
Validation samples	Collection: from high spatial resolution images of Google Earth Time: 2015

^aDownloaded from National Centers for Environmental Information (https://ngdc.noaa.gov/eog/viirs/download_dnb_composites.html)

^bDownloaded from NASA Goddard Space Flight Center (https://ladsweb.modaps.eosdis.nasa.gov/archive/allData/)

^cDownloaded from United States Geological Survey (http://earthexplorer.usgs.gov/)

Note: VIIRS DNB, Visible Infrared Imaging Radiometer Suite Day/Night Band; MODIS NDVI, Moderate Resolution Imaging Spectroradiometer/normalized difference vegetation index; OLI, Operational Land Imager.

Figure 1. Framework of mapping impervious surface at national scale using the combination of multiple-source remotely sensed data (Note: ISA, impervious surface area; VIIRS DNB, Visible Infrared Imaging Radiometer Suite Day/Night Band; MODIS NDVI, Moderate Resolution Imaging Spectroradiometer/normalized difference vegetation index; NDVI_max, maximum NDVI value of MODIS NDVI time-series data; MNDWI, Modified Normalized Difference Water Index; OLI, Operational Land Imager).

Figure 2. Comparison of different remotely sensed data and products in Urumqi (a) and Shanghai (b). (1) VIIRS DNB with 500-m cell size; (2) (1–NDVI_max) image with 250-m cell size; (3) ISAI image with 250-m cell size; (4) the initial ISA image using the thresholding approach based on ISAI with 250-m cell size; (5) the ISA result from Landsat with 30-m cell size (Note: the pixel values in a1–a3 and b1–b3 are continuous values between 0 and 1; the pixel values in a4–a5 and b4–b5 are categorical values of 1 (ISA in red) and 0 (non-ISA in white)).

Figure 3. Spatial distribution of fractional impervious surface areas in China in 2015 at cell size of 1 km by 1 km, highlighting cities of (a) Urumqi, (b) Lanzhou, (c) Beijing, (d) Chengdu, (e) Guangzhou, and (f) Shanghai (note: the ISA value here is a proportion of ISA in a pixel, its values are within 0–1).

Table 2. Accuracy assessment results of mapping impervious surface area (ISA) in China.

Region	Type	Reference data		Producer’s accuracy	User’s accuracy	Overall accuracy
		ISA	Non-ISA
Southeast	ISA	1611	89	86.2	94.8	93.2
	Non-ISA	257	3143	97.2	92.4
Northwest	ISA	1516	184	80.0	89.2	88.9
	Non-ISA	381	3019	94.2	88.8
Total	ISA	3127	273	83.1	91.9	91.1
	Non-ISA	638	6162	95.7	90.6

Note: ISA, impervious surface area; Non-ISA, non-imperviousness area.