Michishita, R; Jiang, ZB; Gong, P; Xu, B (2012). Bi-scale analysis of multitemporal land cover fractions for wetland vegetation mapping. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 72, 1-15.

Abstract
Land cover fractions (LCFs) derived through spectral mixture analysis are useful in understanding subpixel information. However, few studies have been conducted on the analysis of time-series LCFs. Although multi-scale comparisons of spectral index, hard classification, and land surface temperature images have received attention, rarely have these approaches been applied to LCFs. This study compared the LCFs derived through Multiple Endmember Spectral Mixture Analysis (MESMA) using the time-series Landsat Thematic Mapper (TM) and Terra Moderate Resolution Imaging Spectroradiometer (MODIS) data acquired in the Poyang Lake area, China between 2004 and 2005. Specifically, we aimed to: (1) propose an approach for optimal endmember (EM) selection in time-series MESMA; (2) understand the trends in time-series LCFs derived from the TM and MODIS data; and (3) examine the trends in the correlation between the bi-scale LCFs derived from the time-series TM and MODIS data. Our results indicated: (1) the EM spectra chosen according to the proposed hierarchical three-step approach (overall, seasonal, and individual) accurately modeled the both the TM and MODIS images; (2) green vegetation (GV) and NPV/soil/impervious surface (N/S/I) classes followed sine curve trends in the overall area, while the two water classes displayed the water level change pattern in the areas primarily covered with wetland vegetation; and (3) GV, N/S/I, and bright water classes indicated a moderately high agreement between the TM and MODIS LCFs in the whole area (adjusted R-2 >= 0.6). However, low levels of correlations were found in the areas primarily dominated by wetland vegetation for all land cover classes. (C) 2012 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS) Published by Elsevier B.V. All rights reserved.

DOI:
0924-2716

ISSN:
10.1016/j.isprsjprs.2012.04.006