Bayesian Spectral Mixture Analysisfor Urban Vegetation

2006 ◽  
pp. 91-108 ◽  
Author(s):  
Conghe Song
Keyword(s):  
Urban Vegetation ◽  
Spectral Mixture

scholarly journals Improved Automated Detection of Subpixel-Scale Inundation—Revised Dynamic Surface Water Extent (DSWE) Partial Surface Water Tests

2019 ◽  
Vol 11 (4) ◽  
pp. 374 ◽  
Author(s):  
John Jones
Keyword(s):  
Surface Water ◽  
Mixture Models ◽  
Model Development ◽  
Depth Estimation ◽  
High Temporal Resolution ◽  
Commission Error ◽  
Dynamic Surface ◽  
In Situ Data ◽  
The Us ◽  
Spectral Mixture

In order to produce useful hydrologic and aquatic habitat data from the Landsat system, the U.S. Geological Survey has developed the “Dynamic Surface Water Extent” (DSWE) Landsat Science Product. DSWE will provide long-term, high-temporal resolution data on variations in inundation extent. The model used to generate DSWE is composed of five decision-rule based tests that do not require scene-based training. To allow its general application, required inputs are limited to the Landsat at-surface reflectance product and a digital elevation model. Unlike other Landsat-based water products, DSWE includes pixels that are only partially covered by water to increase inundation dynamics information content. Previously published DSWE model development included one wetland-focused test developed through visual inspection of field-collected Everglades spectra. A comparison of that test’s output against Everglades Depth Estimation Network (EDEN) in situ data confirmed the expectation that omission errors were a prime source of inaccuracy in vegetated environments. Further evaluation exposed a tendency toward commission error in coniferous forests. Improvements to the subpixel level “partial surface water” (PSW) component of DSWE was the focus of this research. Spectral mixture models were created from a variety of laboratory and image-derived endmembers. Based on the mixture modeling, a more “aggressive” PSW rule improved accuracy in herbaceous wetlands and reduced errors of commission elsewhere, while a second “conservative” test provides an alternative when commission errors must be minimized. Replication of the EDEN-based experiments using the revised PSW tests yielded a statistically significant increase in mean overall agreement (4%, p = 0.01, n = 50) and a statistically significant decrease (11%, p = 0.009, n = 50) in mean errors of omission. Because the developed spectral mixture models included image-derived vegetation endmembers and laboratory spectra for soil groups found across the US, simulations suggest where the revised DSWE PSW tests perform as they do in the Everglades and where they may prove problematic. Visual comparison of DSWE outputs with an unusual variety of coincidently collected images for locations spread throughout the US support conclusions drawn from Everglades quantitative analyses and highlight DSWE PSW component strengths and weaknesses.

Sign in / Sign up

Export Citation Format

Share Document