Reconstructing Historical Land Cover Type and Complexity by Synergistic Use of Landsat Multispectral Scanner and CORONA

Shahtahmassebi, Amir Reza and Lin, Yue and Lin, Lin and Atkinson, Peter Michael and Moore, Nathan and Wang, Ke and He, Shan and Huang, Lingyan and Wu, Jiexia and Shen, Zhangquan and Gan, Muye and Zheng, Xinyu and Su, Yue and Teng, Hongfen and Li, Xiaoyan and Deng, Jinsong and Sun, Yuanyuan and Zhao, Mengzhu (2017) Reconstructing Historical Land Cover Type and Complexity by Synergistic Use of Landsat Multispectral Scanner and CORONA. Remote Sensing, 9 (7). ISSN 2072-4292

Full text not available from this repository.


Survey data describing land cover information such as type and diversity over several decades are scarce. Therefore, our capacity to reconstruct historical land cover using field data and archived remotely sensed data over large areas and long periods of time is somewhat limited. This study explores the relationship between CORONA texture—a surrogate for actual land cover type and complexity—with spectral vegetation indices and texture variables derived from Landsat MSS under the Spectral Variation Hypothesis (SVH) such as to reconstruct historical continuous land cover type and complexity. Image texture of CORONA was calculated using a mean occurrence measure while image textures of Landsat MSS were calculated by occurrence and co-occurrence measures. The relationship between these variables was evaluated using correlation and regression techniques. The reconstruction procedure was undertaken through regression kriging. The results showed that, as expected, texture based on the visible bands and corresponding indices indicated larger correlation with CORONA texture, a surrogate of land cover (correlation >0.65). In terms of prediction, the combination of the first-order mean of band green, second-order measure of tasseled cap brightness, second-order mean of Normalized Visible Index (NVI) and second-order entropy of NIR yielded the best model with respect to Akaike’s Information Criterion (AIC), r-square, and variance inflation factors (VIF). The regression model was then used in regression kriging to map historical continuous land cover. The resultant maps indicated the type and degree of complexity in land cover. Moreover, the proposed methodology minimized the impacts of topographic shadow in the region. The performance of this approach was compared with two conventional classification methods: hard classifiers and continuous classifiers. In contrast to conventional techniques, the technique could clearly quantify land cover complexity and type. Future applications of CORONA datasets such as this one could include: improved quality of CORONA imagery, studies of the CORONA texture measures for extracting ecological parameters (e.g., species distributions), change detection and super resolution mapping using CORONA and Landsat MSS.

Item Type:
Journal Article
Journal or Publication Title:
Remote Sensing
Uncontrolled Keywords:
ID Code:
Deposited By:
Deposited On:
19 Jun 2018 13:24
Last Modified:
23 Sep 2020 04:23