Spectrally unmixed percent of impervious surface, soil, and vegetation cover in central Arizona-Phoenix, year 2000
Publication date: 2008
Author(s):
- Soe Myint, Global Institute of Sustainability
Abstract:
Urban land covers (e.g., cement parking lots, asphalt roads, shingle rooftops, grass, tress, exposed soil) can only be recorded as either present or absent in each pixel when using traditional per-pixel classifiers. Sub-pixel analysis approaches that can provide the relative fraction of surface covers within a pixel may be a potential solution to effectively identifying urban impervious areas. Spectral mixture analysis approach is probably the most commonly used approach that models image spectra as spatial average of spectral signatures from two or more surface features. However, spectral mixture analysis does not account for the absence of one of the surface features or spectral variation within pure materials since it utilizes an invariable set of surface features. Multiple endmember spectral mixture analysis (MESMA) approach addresses these issues by allowing endmembers to vary on a per pixel basis. The MESMA technique was employed in this study to model Landsat ETM+ reflectance in the Phoenix metropolitan area. Field spectra of vegetation, soil, and impervious surface areas collected with the use of a fine resolution Quickbird image and pixel purity index tool in ENVI software were modeled as reference endmembers in addition to photometric shade that was incorporated in every model. This study employs thirty endmembers and six hundred and sixty spectral models to identify soil, impervious, vegetation, and shade in the Phoenix metropolitan area. The mean RMS error for the selected land use land cover classes range from 0.003 to 0.018. The Pearson correlation between the fraction outputs from MESMA and reference data from Quickbird 60 cm resolution data for soil, impervious, and vegetation were 0.7052, 0.7249, and 0.8184 respectively.
Keywords:
Land-Use, Land-Cover and Land Architecture,
disturbance patterns sonoran desert,
phoenix,
urban,
metropolitan area,
land architecture,
land use changes,
gis,
database remote sensing gis applications,
remote sensing,
human environment feedback,
landsat,
Climate, Ecosystems and People,
impervious,
vegetation cover,
spectral unmixing,
spectral mixture modeling,
multiple endmember spectral mixture analysis,
Land-Use, Land-Cover and Land Architecture urban,
land use,
land cover,
vegetation,
abundance,
methods,
remote sensing,
modeling,
soil,
reflectance caplter,
central arizona phoenix longterm ecological research,
arizona,
caplter created,
az,
cap,
arid land
Temporal Coverage:
2000-04-19
Geographic Coverage:
Geographic Description: Central Arizona Phoenix
Bounding Coordinates:
Longitude:-112.786534 to -111.571667
Latitude:33.844955 to 33.191274
Contact:
Information Manager,
Global Institute of Sustainability,Arizona State University,POB 875402,TEMPE
caplter.data@asu.edu
Methods used in producing this dataset:
Show
Background information on subpixel analysis
The traditional hard classifiers (e.g., minimum distance, Mahalanobis distance, maximum likelihood) can label each pixel only with one class. Information on the fractional amount of spatially mixed spectral signatures from different ground-cover features is not possible with the per-pixel classifiers (hard classifiers). Hence, the traditional classification of mixed pixels may lead to information loss, degradation of classification accuracy, and degradation of modeling quality in successive applications. Sub-pixel analysis that can provide the relative abundance of surface materials within a pixel is a potential solution to per-pixel classifiers especially when dealing with medium to coarse spatial resolution satellite sensor images.
Linear Spectral Mixture Analysis (SMA)
Linear spectral mixture analysis (SMA), which provides sub-pixel endmember abundance information, is probably the most commonly used approach of all subpixel analysis techniques. The approach is based on the assumption that the spectrum at each pixel is a linear combination of the spectra of all ground components within the pixel, and that the linear mixture coefficients are equal to the fractional area of each ground component in a pixel. The mathematical model of linear spectral mixture analysis can be defined as Xi =SUM(fkXik +ei) wher Xi = Total spectral reflectance of band i of a pixel k = number of endmembers fk = fraction of an endmember k within a pixel Xik = known spectral reflectance of endmember k within the pixel in band i ei = error term for band i The root mean square (RMS) error is given by: RMS = ((SUM ((ei)^2)))^0.5) / m where ei are the error terms for each of the m spectral bands considered. The above constrained least-squares estimate assumes the followings. SUM(fk) = 1 and 0≤ fk ≤1 Limitations of Linear SMA (1) linear spectral mixture classifier does not permit number of representative materials (endmembers) greater than the number of spectral bands. (2) An invariable set of endmembers to model the spectra in all pixels. This assumption could potentially fail to account for the fact that the number and type of land cover components within each pixel are highly variable. The endmembers used in SMA are the same for each pixel, regardless of whether the materials represented by the endmembers are present in the pixel.
Multiple Endmember Spectral Mixture Analysis
Multiple Endmember Spectral Mixture Analysis (MESMA) method, enables to establish for each pixel the best mixture model, corrects errors often inevitable in SMA. The MESMA differs from SMA because allows to vary, pixel by pixel, the number and the abundance of endmembers in the image. While SMA method considers only one mixture model, MESMA algorithm generates from endmember all possible subset models. The best-fit model is defined by the linear combination of the endmembers with least error when compared with the spectrum image. For each pixel model is calculated: (a) endmember fraction and (b) the root mean squared (RMS) error. This procedure defines the elements that constitute each pixel plus its relative abundance. The method adjustments a mixture model for each pixel with lowest error and generate abundance images of the endmembers. However, a difficulty to use MESMA method is the computational time that demands supercomputers
Data Files (1) :
Raster:
Surface spectral analysis of vegetation cover
Description: Spectrally unmixed percent impervious surface, soil, and vegetation cover in CAPLTER
Horizontal Coordinate System:WGS_1984_UTM_Zone_12N
Rows:2499
Columns:3931
Column |
Description |
Type |
Units |
Cell Value |
Value for Red Green and Blue bands
|
Integer |
|