A comparative study for unmixing based Landsat ETM+ and ASTER image fusion

https://doi.org/10.1016/j.jag.2009.11.007Get rights and content

Abstract

The mineral environment of the Bouaouane–Jebel (Hill) Hallouf mine, in the north of Tunisia, is monitored and analyzed by making use of the laboratory analysis and remote sensing images, ETM+ as well as ASTER VNIR and SWIR data, acquired in the same period. The main contribution of this paper consists of a methodology using multispectral multi-sensor fusion for the refinement of the mine tailing cartography around the studied mine. The developed methodology is based on the linear spectral unmixing approach which is applied to a multispectral hybrid image. This image was generated from the fusion of Landsat ETM+ and ASTER SWIR data. A comparative study is made between the hybrid and ASTER (VNIR and SWIR) images classification with respect to laboratory analysis. The given results show that the fusion of Landsat ETM+ and ASTER SWIR multispectral image yields the best mineral detection.

Introduction

The mine of Bouaouane–Jebel (Hill) Hallouf which is exploited for the lead and zinc ores and forsaken since 1986 is among several types of mines in the Mejerda river watershed. The Mejerda river is the most important river exploited for the agriculture irrigation and drinking water supply of the north of Tunisia. Mine tailing cartography becomes fundamental in order to follow the environmental changes and pollution.

The mining and smelting of these ores since 1986 has left over an important quantity of mine tailings. Mapping and monitoring these mine tailings is necessary to understanding and minimizing their impact in the environment. Several studies have used remote sensing for mineral and mine tailing mapping. They used multispectral and hyperspectral data (Peters, 1988, Windeler, 1993, Kruse et al., 2003, Wang et al., 2003). The multispectral data were used particularly for tailing coarse map generation. Detailed mineral fraction maps are not feasible in the case of low spectral resolution image data.

In Mezned et al. (2009) we developed a coarse map of mine tailing around the test site, which was based on Landsat (Enhanced Thematic Mapper Plus) ETM+ multispectral data. A constrained linear spectral unmixing method was applied using image derived endmember spectra. This method gives the distribution and abundance images of surface cover endmembers (mine tailing, vegetation and soil) constituting the area of a pixel. The resulting map represents a coarse mine tailing distribution due to the relatively low spectral and spatial resolution of the data. Thus, we could not assess quantitatively the mineralogical composition of the mine tailings.

In this work, we adapted the Multi-sensor Multiresolution algorithm (MMT) to the multispectral multi-sensor image fusion for the refinement of the mine tailing cartography. Considering previous works, dealing about mine tailing, the MMT algorithm is for the first time applied to ETM+ and ASTER data. This allows the amelioration of the spectral resolution of the processed data and so the refinement of the tailing map. Thus the results give detailed fraction maps of the principal minerals existing in the test site.

The MMT algorithm uses the detailed information of the high resolution multispectral image to unmix the lower resolution image. We propose to use the multispectral (MS) Landsat ETM+ data combined with the Landsat ETM+ panchromatic (Pan) as the high resolution image and the multispectral ASTER Level 2B Short Waves Infra Red (SWIR) as the lower resolution image. The classification of the resulting hybrid multispectral image was performed with the spectral unmixing method (Haertel and Shimabukuro, 2005) using the endmember library spectra for mine tailing mapping. The results were compared with those of ASTER classification, and the laboratory analysis used as field truth.

The paper is structured as follows. Section 2 presents the test site and the used data. Section 3 describes the tailing cartography approach. In Section 4 an overview of the first tailing coarse cartography was presented. Sections 5 Multispectral classification refinement, 6 Results analysis and validation describe and evaluate the method used to refine the tailing map. Section 7 ends the paper by presenting some concluding remarks.

Section snippets

Test site and data preprocessing

The abandoned mine of the Bouaouane–Jebel Hallouf (Fig. 1) was exploited for lead and zinc ores (Mansouri, 1980). It is located near the Kassab Ouad (36°42N, 9°05E) which discharge directly into Mejerda river. In the “napes and Mejerda medium” zone, there are three groups of mineralization associated with Trias, Cretaceous superior and Miocene. The mineralization of Jebel Hallouf–Bouaouane is concentrated in the Miocene (Mansouri, 1980, Slim, 1981) (Fig. 2). The mining of these ores has left

Tailing cartography approach

The overall flowchart illustrating the different steps of the proposed methodology is depicted in (Fig. 3). It is composed mainly of four steps. An initial correction step which consist of geometrical and atmospherical correction, so as to have the same geodetic reference and to estimate surface spectral reflectance (Fig. 3). The radiometric and atmospheric correction were performed using the “Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes” FLAASH model to estimate the spectral

Multispectral coarse classification

A first constrained spectral unmixing classification of Landsat ETM+ MS/Pan fused subset image (196×196 pixels), with six bands and 15 m spatial resolution, was performed using image-derived endmember spectra (vegetation, soil and mine tailings). An overview of this methodology is presented in Mezned et al. (2006). The resulting mine tailing fraction map was used to highlight the mine tailing area and masking the surrounding zone in the fused (hybrid and ASTER) images. Only mine tailing pixels

Multispectral classification refinement

The tailing classification refinement includes two main parts: the first one deal with the multispectral Landsat ETM+ and ASTER images fusion based on unmixing method. The resulting Hybrid and ASTER fused images present an ameliorated spectral resolution. The second part, focus on the fused images classification and comparison. We used besides the (MS) and (Pan) ETM+ data, the Aster Level 2B VNIR and SWIR surface reflectance product acquired on June 26, 2000. The results of masked Hybrid and

Results analysis and validation

In this section we compare the endmember abundances estimated from both image classification to those determined with laboratory analysis. Indeed, we present the comparison of field-based determined mineralogy with satellite-based interpretations at spatially coinciding point. This comparison allowed validating the results and determining which results yields best results.

The considered statistical regression analysis consists of calculation of the coefficient of determination (R2). The

Conclusion

In this paper, we have proposed a methodology using multispectral multi-sensor fusion for the refinement of the mine tailing cartography around the Bouaouane–Jebel Hallouf mine. The developed methodology was based on the fusion of Landsat ETM+ and ASTER SWIR data using the constrained linear unmixing algorithm. It aims to generate detailed mineral fraction maps in the case of low spectral resolution image data. The resulting masked Hybrid image was classified using the unmixing algorithm.

Acknowledgments

The authors would like to thank the NASA Land Processes Distributed Active Archive Center and the User Services USGS Earth Resources Observation and Science (EROS) for providing numerous remotely sensed data.

References (17)

  • P.S. Chavez et al.

    Comparaison of three different methods to merge multiresolution and multispectral data: Landsat TM and SPOT panchromatic

    Photogramm. Eng. Remote Sens.

    (1991)
  • T.G. Cooley et al.

    FLAASH, a MODTRAN4-based atmospheric correction algorithm, its application and validation

  • V.F. Haertel et al.

    Spectral linear mixing model in low spatial resolution image data

    IEEE Trans. Geosci. Remote Sens.

    (2005)
  • F.A. Kruse et al.

    Comparison of airborne hyperspectral data and EO-1 Hyperion for mineral mapping

    IEEE Trans. Geosci. Remote Sens.

    (2003)
  • A. Mansouri, 1980. Gisements de Pb-Zn et Karstification en milieu continental: le district du Jbel Hallouf-Sidi Bou...
  • MATLAB, 2004. MATLAB Users Guide, MATLAB 7.0 Release...
  • N. Mezned et al.

    Mine tailings mapping using Landsat multispectral imagery of the versant basin amount of Medjerda river in the north of Tunisia

  • Mezned, N., Abdeljaoued, S., Boussema, M.R., 2009. Unmixing based Landsat ETM+ and ASTER image fusion for hybrid...
There are more references available in the full text version of this article.

Cited by (21)

  • Unmixing-based Sentinel-2 downscaling for urban land cover mapping

    2021, ISPRS Journal of Photogrammetry and Remote Sensing
    Citation Excerpt :

    Besides, due to the lack of enforcing proper constraints to the unmixing solution, unrealistic reflectance values are often generated by this unmixing-based image fusion (Amorós-López et al., 2013). Commonly applied constraints are to force the reflectance values to be positive (Mezned et al., 2010), lower than the corresponding coarse band reflectance value (Zurita-Milla et al., 2008) and/or sum to unity (Delalieux et al., 2014; Wei et al., 2016). Amorós-López et al. (2013) further limited the range of possible spectral shapes based on a group of pre-selected training spectra.

  • Hyperspectral field spectroscopy and SENTINEL-2 Multispectral data for minerals with high pollution potential content estimation and mapping

    2020, Science of the Total Environment
    Citation Excerpt :

    The traditional mine tailings and soils studies require field and laboratory analyses of a great number of collected samples from various point location and at various depths (Ferrier, 1999; Kemper and Sommer, 2003). This approach, providing the present pollution status, is time consuming and requires large resource investments (Mezned et al., 2012). Mapping the minerals with high pollution potential and predicting their contents within soils constitutes an automated solution for pollution characterization and assessment.

  • Mapping of ferric <sup>(Fe3+)</sup> and ferrous <sup>(Fe2+)</sup> iron oxides distribution using band ratio techniques with ASTER data and geochemistry of Kanjamalai and Godumalai, Tamil Nadu, south India

    2020, Remote Sensing Applications: Society and Environment
    Citation Excerpt :

    These works indicate the credibility of using remote sensing datasets for hydrothermally altered mineral mapping. Magendran and Sanjeevi (2014) and Mezned et al. (2010) employed EO–1 Hyperion, Landsat ETM+ and ASTER images and obtained the most reliable mineral detection including iron ore grade estimation. Many successful studies on iron ore grade estimation using spectral sensing and geochemistry (Morais Maria Carolina de et al. (2010), Magendran et al. (2011), Magendran and Sanjeevi (2014), Gopinathan et al. (2015a,b), Sadeghi et al. (2013) encouraged us to attempt this study for Kanjamalai and Godhumalai regions.

  • Spatio-spectral fusion of satellite images based on dictionary-pair learning

    2014, Information Fusion
    Citation Excerpt :

    Through integrating the spectral information of LSHS data and the spatial information of HSLS data, we expect to extend the applications of available satellite images, thereby meeting various demands of data users. To address this spatial and spectral fusion problem, one category of classic method is spatial-unmixing-based algorithms [5–9]. The processing steps of these methods are: (1) geometrical co-registration between the HSLS and LSHS images; (2) the multispectral classification on the high spatial resolution image (HSLS) to unmix the low spatial resolution image (LSHS); and (3) determination of the class spectra via regularized linear unmixing.

View all citing articles on Scopus
View full text