Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Impervious Surface Mapping with Multi-Spectral Remote Sensing Dr. Qihao Weng Associate Professor of Geography; Director, Ctr. Urban & Environmental Change.

Similar presentations


Presentation on theme: "1 Impervious Surface Mapping with Multi-Spectral Remote Sensing Dr. Qihao Weng Associate Professor of Geography; Director, Ctr. Urban & Environmental Change."— Presentation transcript:

1 1 Impervious Surface Mapping with Multi-Spectral Remote Sensing Dr. Qihao Weng Associate Professor of Geography; Director, Ctr. Urban & Environmental Change Indiana State University qweng@indstate.edu

2 2 Acknowledgement This research is sponsored by NSF (BCS- 0521734), and by the NASA ISGC program (NGTS-40114-4), and the USGS IndianaView program for a project entitled “Indiana Impervious Surface Mapping Initiative”.

3 3 Impervious Surfaces and Watershed Quality Impervious surfaces: Anthropogenic features through which water cannot infiltrate into the soil, such as roads, driveways, sidewalks, parking lots, rooftops, and so on. A major indicator of the degree of urbanization, and environmental quality (Arnold and Gibbons, 1996).

4 4 Impervious Surfaces and Watershed Quality Watersheds: Natural integrator of hydrological, biological, and geological processes. Human- watershed interactions (planning and policy). Scale dependency. Require an integrated approach to data analysis, in which GIS, remote sensing, and GPS are ideal tools. Impervious surfaces: A unifying theme for all participants – planners, engineers, landscape architects, scientists, social scientists, local officials, and others at all watershed scales (Schueler, 1994).

5 5 Impervious Surfaces and Watershed Quality Impervious surfaces relate to watersheds in: hydrology (Brun and Band, 2000; Weng, 2001), water quality (Brabec et al. 2002; Hurd and Civco, 2004 ), habitat structure (Booth, 1991; Shaver et al. 1994), biodiversity of aquatic systems (Black and Veatch, 1994; Gillies et al. 2003), land surface temperature (Weng et al. 2006; Lu and Weng, 2006), and water temperature (Galli, 1991).

6 6 Impervious Surfaces and Watershed Quality Transport-related vs. roof-related impervious surfaces: land use zoning emphasizes the latter, but the former has a greater hydrological impact. The magnitude, location, geometry and spatial pattern of impervious surfaces, and pervious/impervious ratio (landscape structure) in a watershed. Threshold: stream quality declines at 10%-15% of imperviousness (Schueler, 1994).

7 7 Impervious Surfaces and Watershed Quality Ranking of Steam Health (Arnold and Gibsons, 1996) Less than 10% imperviousness – protected; 10%-30% - impacted; Over 30% - degraded. Figure created by Prisloe et al. 2001.

8 8 Estimating and Mapping Impervious Surfaces Field survey with GPS - expensive, time- consuming, but accurate. Manual digitizing from hard-copy maps or remote sensing imagery (especially aerial photographs) - become increasingly automated (e.g., scanning and feature extraction). Remote sensing methods using spectral data.

9 9 Estimating and Mapping Impervious Surfaces Most traditional studies correlate impervious surface percentage with land use and land cover (LULC) type/class. Limitations of this approach – Intra-variation of imperviousness within the same class; Vary with use density (Brabec et al. 2002); Inconsistent and not replicable. When LULC data derived from per-pixel image classification, imperviousness data would be limited by the pixel resolution (Clapham, 2003).

10 10 Remote Sensing Methods (1) Multiple regression - relates percent impervious surface to remote sensing and/or GIS variables (Chabaeva et al. 2004; Bauer et al. 2004). (2) Sub-pixel algorithms - decompose an image pixel into fractional components (Ridd, 1995; Ji and Jensen, 1999; Wu and Murray, 2003; Lu and Weng, 2004). (3) Artificial neural network - applied advanced machine learning algorithms to derive impervious surface coverage (Flanagan and Civco, 2001. Output: per-pixel impervious predictions; Training data: Landsat TM spectral reflectance values) (4) Classification and regression tree (CART) algorithm - produced rule-based models for prediction based on training data, and yielded estimates of subpixel percent imperviousness (Yang et al. 2003).

11 11

12 12 Landscape as a Continuum A continuum model better suited -Continuously varying landscapes, e.g. agricultural land in Midwest USA; residential areas; semi-arid areas, urban and suburban areas, etc. A continuum model – pixel measurement regarded as a sum of spectral interactions among the elements weighted by their concentrations.

13 13 Landscape as a Continuum A continuum model suited for L-resolution image scenes (Strahler et al. 1986). Scene elements not detectable. Medium resolution (10-100 m) imagery for heterogeneous landscapes. Description/quantification vs. classification: e.g., Composition of soil - sand, silt, and clay. A continuum model can provide description/quantification of landscapes, but can also be used for classification (Adams et al. 1995; Roberts et al. 1998; Lu and Weng, 2004).

14 14 Linear Spectral Mixture Analysis LSMA: Remote sensing implementation of the continuum model. LSMA assumes: The spectrum measured by a sensor is a linear combination of the spectra of all components (fractions, endmembers) within a pixel. Fraction images used for: biophysical description, landscape characterization, classification, change detection, etc.

15 15 LSMA Model Where: i is the number of bands used; k = 1, …, n is number of endmembers; is the spectral reflectance of band i of a pixel; is proportion of endmember k within the pixel; is the spectral reflectance of endmember k within the pixel on band i, and is the error for band i.

16 16 Research Objective To develop an approach for estimating and mapping impervious surfaces from multi-spectral Landsat imagery

17 17 Study Area – Marion County, Indiana, USA

18 18 Remote Sensing Data Used Landsat ETM+ image of June 22, 2000 (11:14 AM). High resolution air photos: 2002 digital orthophotography.

19 19 Methods Spectral mixture analysis of optical bands (endmembers calculated: green vegetation, soil, low albedo, and high albedo). Impervious surface estimation. LST calculation from Landsat thermal infrared data.

20 20 Fraction images computed from six ETM+ reflective bands using LSMA (A: high albedo; B: low albedo, C: soil; and D: green vegetation)

21 21 Feature spaces between the minimum noise fraction components, illustrating potential endmembers

22 22 ETM+ spectral features of the selected four endmembers

23 23 A: based on combination of high-albedo and low-albedo fraction images. B: Improved impervious surface image by combined use of land surface temperature and fraction images. The values of impervious surface range from 0 to 1, with the lowest values in black and highest values in white.

24 24 Procedure for refining impervious surfaces based on data integration of land surface temperature and fraction images.

25 25 Sample selection for accuracy assessment of impervious surfaces. Digitized impervious surface polygons within the selected sample plot on the digital orthophoto.

26 26 Overall RMSE = 9.22%, system error = 5.68%. 76 sample plots (300m*300m) For plots with less than 30% impervious surface, RMSE = 9.98%, system error = 8.59%. For plots with greater than or equal to 30% impervious surface, RMSE = 8.36%, system error = 2.77%.

27 27 Conclusions The continuum model suitable for estimating and mapping impervious surfaces from multispectral satellite imagery; Impervious surfaces derived from satellite imagery applicable to various applications.

28 28 Reference Books Remote Sensing of Impervious Surfaces, by Qihao Weng, CRC Press/Taylor & Francis, ISBN: 1420043749, $129.95, to be published in Sept. 2007. Urban Remote Sensing, 2006, By Qihao Weng and Dale Quattrochi, CRC Press/Taylor & Francis, ISBN: 0849391997, $99.95. Order at www.crcpress.com


Download ppt "1 Impervious Surface Mapping with Multi-Spectral Remote Sensing Dr. Qihao Weng Associate Professor of Geography; Director, Ctr. Urban & Environmental Change."

Similar presentations


Ads by Google