Estimating Wetland Vegetation Abundance from Landsat-8 Operational Land Imager Imagery: A Comparison between Linear Spectral Mixture Analysis and Multinomial Logit Modeling Methods

Document Type

Article

Publication Date

1-2016

Keywords

Vegetation, Earth observing sensors, Landsat, Reflectivity, Imaging Systems, Visualization, Remote Sensing, Associative arrays, Data modeling, Soil Science

Digital Object Identifier (DOI)

https://doi.org/10.1117/1.JRS.10.015005

Abstract

Mapping vegetation abundance by using remote sensing data is an efficient means for detecting changes of an eco-environment. With Landsat-8 operational land imager (OLI) imagery acquired on July 31, 2013, both linear spectral mixture analysis (LSMA) and multinomial logit model (MNLM) methods were applied to estimate and assess the vegetation abundance in the Wild Duck Lake Wetland in Beijing, China. To improve mapping vegetation abundance and increase the number of endmembers in spectral mixture analysis, normalized difference vegetation index was extracted from OLI imagery along with the seven reflective bands of OLI data for estimating the vegetation abundance. Five endmembers were selected, which include terrestrial plants, aquatic plants, bare soil, high albedo, and low albedo. The vegetation abundance mapping results from Landsat OLI data were finally evaluated by utilizing a WorldView-2 multispectral imagery. Similar spatial patterns of vegetation abundance produced by both fully constrained LSMA algorithm and MNLM methods were observed: higher vegetation abundance levels were distributed in agricultural and riparian areas while lower levels in urban/built-up areas. The experimental results also indicate that the MNLM model outperformed the LSMA algorithm with smaller root mean square error (0.0152 versus 0.0252) and higher coefficient of determination (0.7856 versus 0.7214) as the MNLM model could handle the nonlinear reflection phenomenon better than the LSMA with mixed pixels.

Was this content written or created while at USF?

Yes

Citation / Publisher Attribution

Journal of Applied Remote Sensing, v. 10, issue 1, art. 015005

Share

COinS