Hyperspectral Mapping of Surface Mineralogy in the Lake Magadi Area in Kenya
Hyperspectral mapping of surface mineralogy in the Lake Magadi area in Kenya.
Gayantha Roshana Loku Kodikara March, 2009
Hyperspectral mapping of surface mineralogy in the Lake Magadi area in Kenya.
by
Gayantha Roshana Loku Kodikara
Thesis submitted to the International Institute for Geo‐information Science and Earth Observation in partial fulfilment of the requirements for the degree of Master of Science in Geo‐information Science and Earth Observation, Specialisation: Geo‐hazards
Thesis Assessment Board
Prof.Dr. F.D. van der Meer (Chair) Dr. P.M. van Dijk (External Examiner) Dr. T. Woldai (First Supervisor) Dr. F.J.A. van Ruitenbeek (Second Supervisor)
Observer: Drs. T.M. Loran (Programme Director)
INTERNATIONAL INSTITUTE FOR GEO‐INFORMATION SCIENCE AND EARTH OBSERVATION ENSCHEDE, THE NETHERLANDS
Disclaimer
This document describes work undertaken as part of a programme of study at the International Institute for Geo‐information Science and Earth Observation. All views and opinions expressed therein remain the sole responsibility of the author, and do not necessarily represent those of the institute.
To my loving mother H.B. Kusumawathe.
Abstract
Hyperion and Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER) data were used to identify, map and model the spatial distribution of the mineral precipitates at the Lake Magadi area in Kenya. Mapping was coupled with laboratory analysis, including reflectance and emittance spectroscopic measurements and X‐ray diffraction for selected rock and soil samples. In addition to that, land surface temperature mapping, stratigraphic information and drainage network extraction of the area were carried out using remote sensing techniques and later integrated and modelled.
The spectral responses of 92 rock and soil samples were studied and identified. The spectral signatures of Magadiite (NaSi7O13 (OH) 3 ∙3H2O) and Kenyaite (NaSi11O20.5 (OH) 4 ∙3H2O), which are not known from the work of others, were established in this research. The Magadiite shows diagnostic absorption feature at 1.464 µm. The characteristic absorption features of the Kenyaite are at 1.153 µm and 1.464 µm wavelength regions. In addition to that, spectral signatures of trona, chert, diatomite, basalt/trachyte, erionite, Green bed and High Magadi bed were studied and identified. Chert samples show the broad Si‐OH absorption feature at 2.2 µm while, Green bed, High Magadi bed and diatomite exhibit carbonate absorption feature at 2.35 µm with broad Si‐OH absorption feature at 2.2 µm. Trona exhibits six common absorption features at 1.50, 1.74, 1.94, 2.03, 2.22 and 2.39 µm. Thermal Infrared spectra of trona also exhibit three characteristics features at 6.66, 9.35 and 11.71 µm wavelength regions. The spectroscopic studies of undisturbed soil samples revealed that the evaporites have been restricted to the uppermost part of the surface and the change in mineral phase is possible due to the temperature changes.
Mapping of different stages of evaporites and other surface minerals using combination of ASTER and Hyperion images facilitated by different types of mapping techniques including spectral mapping methods (SAM & MTMF) and band rationing method substantially improved the existing knowledge of the geology of the area. Stratigraphic information extracted from remote sensing methods showed that the mineral precipitates are restricted to the low lying areas associated with water. Drainage network extracted from the ASTER DEM showed the influence of runoff for mineral reactions and formations that were described in existing hydro‐geochemical models of the area. The contribution of spatial and temporal land surface temperature variations for the evaporitic mineral formations in the area was identified after mapping surface temperatures from ASTER TIR (Thermal Infrared) bands. Finally, this study concludes that the usage of remote sensing techniques with existing geochemical knowledge of the area significantly enhanced the capability to derive substantial information related to the distribution and formation of precipitates and evaporites in the area.
Keywords: Hyperion; ASTER; Lake Magadi; Reflectance and emittance spectroscopy; X‐ray Diffraction; Surface mineral mapping; Land surface temperature mapping; Drainage network extraction; Magadiite; Kenyaite; Trona; Chert; Diatomite; Remote sensing
i Acknowledgements
First and fore most, I would like to express my sincere appreciation and thanks to my supervisors Dr. T. Woldai and Dr. F.J.A. van Ruitenbeek for their unlimited and continuous support, critical comments and guidance from the inception to the success of the research. I am very much grateful for their encouragement, advice, friendship and moral support throughout this research and in all hardships met in this study.
I would also like to extend very many thanks to the AES course director Drs. T.M. Loran and the Chairman of the Department of Earth Systems Analysis, Prof. Dr. F.D. van der Meer for their uncountable efforts and assistance during my studies and hardships.
Also many thanks to Drs. J.B. de Smeth for all his help, critical comments, friendship and idea sharing that had a positive effect towards the success of this research. I am also grateful for Mr. Chris Hecker for his support and advice in this research especially during the spectral acquisition of the rock and soil samples using VERTEX 70 FTIR spectrometer in ITC, the Netherlands.
I express my gratitude to all the staff members of AES department especially those of the Geo‐hazard stream for the skills acquired through their effort. I also extend my appreciation to the Earth Resource Exploration staff for the success of my studies and this research in particular. Also many thanks to my PhD advisor, Mr. Zack Kuria for his incredible supports during the field work in Kenya.
I would also like to express my sincere appreciation to Dr. Keith Shepherd and Mr. Elvis Weullow for their guidance and kind support by giving me an opportunity to use MPA FTIR instrument at International Center for Research in Agro Forestry (ICRAF) in Nairobi, Kenya.
My Sincere appreciation also goes to Ing. G.J. van Hummel for his guidance and support by giving me a great opportunity to use XRD instrument at Institute for Nanotechnology (MESA+), University of Twente, The Netherlands.
I would like to thank very much to Magadi Soda company and masai community of the area for their reception and incredible support rendered during my field work. My special thanks go to Mr. S.N. Juguna who took us every where we wanted to go in the field.
I would also like to take this opportunity to thank my fellow students, AES 2007 batch and all ITC friends with Ms. Darani for their good co‐operation, friendship and encouragements. I am also very much grateful for Ghebretinsae Woldu who worked with me during the field work in a same area giving me incredible support during that hard time.
My sincere appreciation also goes to Dr. Jagath Gunathilake (Department of Geology, University of Peradeniya, Sri Lanka), who created my enthusiasm in field of geology and gave me the opportunity to study in ITC.
Of course, all my academic success counts back on the love, help, support and encouragement that I got from my parents, relatives and friends. My mother H.B. Kusuma, most valuable person of my world, how can I dedicate to this work to you? Because you passed away during the last part of my thesis leaving me alone in this world. I would like to dedicate this work to my father L.K. Sumanadasa in particularly in this very hard sorrowful time encouraging me to finish this work successfully. I am very much grateful to my sister Inosha Kodikara and all my relatives, for their sacrifice to stay close to my parents when I am staying away from them. Especially after that sorrowful day every time they encouraged me to start my work again giving me a life and staying in touch with my father when I came back to ITC to complete the Thesis. This success is a result of their encouragement to further my studies.
ii Table of contents
Abstract...... i Acknowledgement...... ii Table of contents...... iii List of figures...... v List of tables...... x List of Abbreviations...... xi 1. Introduction...... 1 1.1. Background ...... 1 1.2. Problem Statement...... 2 1.3. Objectives and Research questions ...... 3 1.4. Research Hypothesis...... 3 1.5. Structure of the Thesis...... 4 2. Literature Review ...... 5 2.1. Lacustrine Evaporitic Sedimentary Process ...... 5 2.2. Principles of Reflectance and Emittance Spectroscopy ...... 11 2.2.1. Electronic Process...... 12 2.2.2. Vibrational Process...... 13 2.3. Principle of Fourier Transform Spectrometer...... 14 2.4. Principle of X‐ray diffraction ...... 15 2.5. Surface mineral mapping using remote sensing...... 17 2.6. Land surface temperature mapping using remote sensing ...... 18 3. Geological and tectonic settings ...... 21 3.1. East African Rift System ...... 21 3.2. Geological and tectonic settings of the Magadi area ...... 22 4. Methodology ...... 25 4.1. Pre‐field work phase ...... 25 4.1.1. Data required before the field work...... 25 4.1.2. Data/Materials acquired during the field work...... 25 4.1.3. Data required after the field work...... 26 4.2. Field work phase ...... 26 4.3. Post‐Field work phase...... 26 4.4. Instrumentation ...... 30 4.4.1. The MPA FTIR Spectrometer ...... 30 4.4.2. The Vertex FTIR Spectrometer...... 31 4.4.3. X‐Ray Powder Diffractometer (XRDP)...... 33 5. Mineral identification using reflectance spectroscopy and X‐ray diffraction methods...... 35 5.1. Magadi Trona (Evaporite Series)...... 35 5.1.1. Spectral Characteristics of Trona...... 36 5.1.2. Spectral Characteristics of Evaporites from undisturbed soil samples ...... 39 5.2. Siliceous rocks...... 46 5.2.1. Spectral Characteristics of Chert ...... 48
iii 5.2.2. Spectral Characteristics of Green beds, Diatomite and High Magadi beds...... 49 5.2.3. Spectral Characteristics of Magadiite ...... 52 5.2.4. Spectral Characteristics of Kenyaite ...... 53 5.3. “Unknown Mineral” ...... 54 5.4. Erionite...... 57 5.5. Volcanic Rocks...... 59 5.6. Other Sedimentary Rocks...... 63 6. Mineral Mapping using Multi‐spectral and Hyperspectal Imaging methods ...... 65 6.1. Surface Mineral Mapping using ASTER ...... 65 6.1.1. ASTER image processing ...... 65 6.1.2. Spectral Properties ...... 68 6.1.3. Mapping Method...... 71 6.2. Surface mineral mapping using Hyperion ...... 76 6.2.1. Pre‐Processing of Hyperion Image...... 77 6.2.2. Calibration to reflectance ...... 80 6.2.3. End member collection...... 84 6.2.4. Accuracy assessment ...... 93 7. Analysis of the spatial distribution of minerals in the Magadi area...... 95 7.1. Extraction of drainage network ...... 95 7.2. Land Surface Temperature (LST) mapping...... 99 7.3. Stratigraphic information from remote sensing ...... 102 7.4. Generation of 3‐D model ...... 104 8. Conclusions and recommendations ...... 107 8.1. Conclusions ...... 107 8.2. Limitations...... 110 8.3. Recommendations ...... 111 References...... 112 Appendices...... 116 Appendix: 01. Chemical analysis of springs and related waters from Lake Magadi area...... 116 Appendix: 02. Geological map of the Magadi area...... 117 Appendix: 03. Collected samples and field descriptions during the field work...... 118 Appendix: 04. XRD Measurement Conditions...... 122 Appendix: 05. XRD Peak list of trona...... 123 Appendix: 06. XRD Peak list of Erionite...... 125 Appendix: 07. Reflectance spectra of collected soil samples...... 128 Appendix: 08. Existence of ground water reservoir below the high Magadi bed...... 132
iv List of figures
Figure: 1.1. Geographical location map of the lake Magadi study area. ………………...... ……………………01 Figure: 2.1. Phase diagram for the sodium carbonate, sodium bi carbonate, water system at.………06 Figure: 2.2. Magadiite chemistry …………………………...…………………………………………………………………………08 Figure: 2.3. Classification of Erionite according to the dominant extra framework cation and their occurrences...... ……………………………………………………………09 Figure: 2.4. Scaled schematic of the distribution of evaporitic lacustrine sediment in the rift valley based on an east‐west transect of Lake Magadi. ………...... ……………….…………………………………….. 10 Figure: 2.5. Optical schematic diagram of a Michelson interferometer. ……………………………………..14 Figure: 2.6. The Interferogram and Reflectance spectrum of Chert sample P006_01. ……………………15 Figure: 2.7. Schematic representation of XRD by regularly spaced planes of atoms in a crystal. ………16 Figure: 3.1. Main lakes and structures in the East African Rift ….…………………………………………………….. 21 Figure: 3.2. Geological Map of Lake Magadi. ………………………………...... …………………….23 Figure: 3.3. Structural interpretation of Lake Magadi basin. ……………………………………...... ……………….23 Figure: 4.1. Processed satellite images to differentiate different surface materials for field sample collection...... ………………………..27 Figure: 4.2. One of the field description data sheets used in the field. ………...... ……………………….. 28 Figure: 4.3. Sample location map. ………………………………………...... …………………28 Figure: 4.4. Flow chart showing the research approach of the study. ……………..…………………………………29 Figure: 4.5. Bruker Multi Purpose Analyzer (MPA) instrument and Optical path diagram of integrating sphere. ………...... ……………………………………………………..30 Figure: 4.6. a) The Optical path diagram showing the optical path going into the external integrating sphere ………...... ……………32 Figure: 4.6. b) Front view of Vertex 70 with external sphere, MCT detector (red) and InGaAs detector (black). ………...... …………………………………………32 Figure: 4.7. Used X‐ray powder diffractometer with its main components. …..………………………………….33 Figure: 4.8. Schematic representation of the components of an X‐ray diffractometer. ……………..……33 Figure: 5.1. Trona and Evaporites samples with all other sample locations ...... …………………………….35 Figure: 5.2. Surface evaporites along shore of the Little Magadi Lake …………...... …………………………..35 Figure: 5.3. Crystallized trona (A001) and its powder form of grain size<2mm (B002) ………………………35 Figure: 5.4. Surface evaporites and collected sample from southern part of Lake Magadi. ………....……36 Figure: 5.5. (a) Trona Crystal Form and (b) Crystal Structure. ………………………...... …………………..36 Figure: 5.6. XRD pattern of the Trona (A001) sample. ……………………...... ……………………………36 Figure: 5.7(a). Normal and (b). Continuum removed reflectance spectra of Trona samples. …………....37 Figure: 5.8. The effects of grain size on the spectral response of Trona. …………………………….…………….38 Figure: 5.9.a) The map of the places where undisturbed soil samples were collected, b), c), d) surfaces of the ground with the sample surfaces before and after dried. ………………………....……………………………39 Figure: 5.10. (a) Normal and (b) Continuum removed reflectance spectra of the undisturbed surface soil sample R01. ………...... ……………….40 Figure: 5.11 (a). Normal and (b) Continuum removed reflectance spectra of the undisturbed surface soil sample that were collected from alluvial plain in northern part of the Lake Magadi. …………...……41
v Figure: 5.12 (a) Normal and (b) Continuum removed reflectance spectra of the trona evaporites from middle part of the Lake Magadi study area. ……………………………...... ……………………………………42 Figure: 5.13. The graph showing the relation between water content and 1.9‐μm and 1.4‐μm absorption features depth. …………………………………...... ……………………43 Figure: 5.14. TIR (3μm‐15μm) emission spectra of Trona and Surface Evaporites with USGS trona reference spectrum. ……………………………………...... …………………44 Figure: 5.15. TIR (3μm‐15μm) emission spectra of Trona and Surface Evaporites. ………………...………..45 Figure: 5.16. Crystal casts in Chert plates from Southern end of Lake Magadi. …………………….....……..47 Figure: 5.17. Laminate Green Chert from Northern part of the Lake Magadi. ……………………………..…..47 Figure: 5.18. Pillow Chert: Near to the Magadi town. ……………………………………………………………………..47 Figure: 5.19. Chert dykes from North‐eastern part of the Lake Magadi. …………………...... …………..47 Figure: 5.20. Green beds from southern part of Lake Magadi. ………………………………...... ……….47 Figure: 5.21. High Magadi Beds from southern part of Lake Magadi. ………………………….....……………….47 Figure: 5.22 (a) The location map of the Chert samples and (b) other silica rich samples. .……………….48 Figure: 5.23 (a) Normal and (b) Continuum removed reflectance spectra of Magadi‐ Chert Series...... …………49 Figure: 5.24 (a) Normal and (b) continuum removed reflectance spectra of Green beds, High Magadi beds and Diatomite. …………………………………...... ………………………50 Figure: 5.25. Comparison between two absorption features. ………………………...... …………51 Figure: 5.26. Diatomaceous earth (diatomite) from northern part of the study area, one bank of river...... …51 Figure: 5. 27 (a) The XRD patterns and (b) reflectance spectra of the several rock samples showing the relation between proportional contribution of mineralogy (eg: SiO2) and the intensities of the XRD pattern and absorption depth of reflectance spectra. …………………………...... ……………………………51 Figure : 5.28. X‐ray diffraction pattern of Magadiite sample (P032) ………….…………...... ………………52 Figure: 5.29 (a) Normal and (b) continuum removed reflectance spectra of Magadiite rock sample and its powder form. ……………………………………...... …………………………………………..52 Figure: 5.30. Kenyaite …………………………………….………………………………………...……………………………………..53 Figure: 5.31 (a) XRD pattern of the Rock sample P036 and (b) its Peak list with peak list of Kenyaite and Quartz. It shows that, the material is composed of Kenyaite and Quartz. …………..…………………….53 Figure: 5.32 (a) Normal and (b) continuum removed reflectance spectra of Kenyaite rock sample and its powder form. …………...... …………………………………………………53 Figure: 5.33. “Unknown” sample (a) and (b) its original location, river bank. Chert bed also can be seen in this Picture. ………………...... ………………………………………………………….54 Figure: 5.34. Normal and continuum removed reflectance spectra of P012 (“Oldhamite”) rock sample and its powder form. …………………………………...... ……………………………………………………..55 Figure: 5.35. XRD pattern of the (a) sample B024 and (b) erionite reference. ……………………...………….57 Figure: 5.36 (a) Normal and (b) Continuum removed reflectance spectra of rock and soil samples which contains Erionite and (c) their sample locations. …………………………………….……………………………..57 Figure: 5.37. Infrared absorption spectra for erionite. ………………………………………………………………….. 58 Figure: 5.38. Location map of the igneous rock samples ……………….………………………………………………….59 Figure: 5.39. Basalt Rocks from Northern part of Lake Magadi looking west. ………………………………….60 Figure: 5.40. One of the Alkali trachyte Sample from near to Little Magadi (looking west). ……....……60
vi Figure: 5.41. Rhyolite sample from sample no P008_02 area‐northern part of Lake Magadi. …….…..60 Figure: 5.42. Scoriaceous basalt sample P015 area from middle part of the Lake Magadi study area...... ……60 Figure: 5.43. Volcanic tuff and color variation of the profile from northern part of the study area...... ….. 60 Figure: 5.44.Basaltic rock (Sample no: P037) with plagioclase phenocrysts. ……….....………………………..60 Figure: 5.45 (a) Normal and (b) Continuum removed reflectance spectra of basaltic rocks samples...... ……….61 Figure: 5.46 (a) Normal and (b) Continuum removed spectra of other igneous rocks samples. ..……….62 Figure: 5.47(a) Rock sample and area. (B) Part of continuum removed reflectance spectra of sedimentary rock sample P031.wihle solid line showing mixed spectrum, dashed lines show chert (SiOH) absorption feature and CO3 absorption feature of Matrix. (C) Sample location map. ……………………...... ……………………………..63 Figure: 5.48 (a). Normal and (b). Continuum removed reflectance spectra of sedimentary rock samples...... ….64 Figure: 6.1. Crosstalk mechanism originates from (a) band 4 detectors and (b) filter boundaries...... ………66 Figure: 6.2. ASTER reclassified end member spectra (*) with its original form. ………………………………..69 Figure: 6.3. ASTER re‐sampled emissivity end member spectra with its original form. ……………….70 Figure: 6.4. (a). Selected reference image spectra for Spectral angle mapper method. (b).Results of spectral mapping method. ………………………………………………...... ……………………………………………………….72 Figure: 6.5. (a) ASTER TIR band ratio 11/12, (b) Classified Quartz‐chert (Red) map after density slicing overlain on ASTER band ratio 11/12, (c) ASTER SWIR false colour composite 4:7:9 (R:G:B) image with extent as same as (a) and (b) to see the spatial distribution. ……………...……………………………………………73 Figure: 6.6. The typical reflectance curves of dry soil, vegetation and clear water. ………....……………..73 Figure: 6.7. (a) ASTER VNIR band 1:2:3N (R: G: B) false colour composite (b) classified water area map (blue) overlain on ASTER band ratio 2/1...... ……....……………………….74 Figure: 6.8. Schematic diagram of ASTER image mineral mapping. ……………...…………………………………74 Figure: 6.9. (a) Mineral abundance map that was derived from ASTER image using SAM and Band ratio techniques with (b) Published geology map of the area...... …………………….75 Figure: 6.10. The Hyperion instrument. ………………...... …………………………………………………………………….76 Figure: 6.11. Part of the Hyperion image of the study area; a) after applying interpolation using ENVI Hyperion tool, b) after removing bad bands and cells using Pushbroom Plugger and c) after removing column stripes using Pushbroom Destriper module. …………………………………………………………………………79 Figure: 6.12.a. First five MNF images before the pre‐processing showing the effect of column stripes...... ……80 Figure: 6.12.b. First five MNF images after the pre‐processing showing increment of quality ...... …80 Figure: 6.13. Image reflectance spectra from chert area after performing (a) IAR reflectance, (b) Empirical line, (c) Log residual and(d) Flaash atmospheric correction method for same image data...... ………………..82 Figure: 6.14. Image reflectance spectra from trona area after performing (a) IAR reflectance, (b) Empirical line, (c) Log residual and (d) Flaash atmospheric correction method for same image data. …………...... ……………………83
vii Figure: 6.15. Selected end members in n‐Dimensional spectral space. ………………………………...………….84 Figure: 6.16. The MTMF rule images 13 (a), 16, (b), 4 (c) and combination of these rule images as false colour composite image (R:4, G:13, B:16)(d). While green, blue and pink colour of image (d) represent the different types/stages of evaporites, red colour represent the chert bed. (e) Image reflectance spectra...... ………85 Figure: 6.17. Spectral plot comparing Laboratory spectra (a) and Hyperion image spectra (b) of Trona and Chert. ……………………………………………………………………...... ………………………………………………..86 Figure: 6.18. (a) Hyperion MNF bands 1, 3 and 5 after masking out Lake Magadi area, (b) Image reflectance spectra of different surface expressions, (c) Hyperion MNF bands 2, 4 and 6, after masking out Lake Magadi area and (d) Continuum removed surface reflectance spectra of selected image spectra within wavelength region 2000nm to 2355nm. ……………...…………………………………………………..87 Figure: 6.19.(a) Hyperion MNF bands 3, 4 and 5 after masking out lake Magadi area, (b) Image reflectance spectra of different surface expressions, (c) Hyperion MNF bands 4, 5 and 6, after masking out Lake Magadi area and (d) Continuum removed surface reflectance spectra of selected image spectra within wavelength region 2000nm to 2355nm. ……………………………….....…………………………..…. 88 Figure: 6.20. Reflectance spectra of green leaves (a) and dry plant materials. ……………………...…………89 Figure: 6.21. (a) Hyperion SWIR bands 196, 208 and 216. (b) Hyperion derived wavelength position of the absorption within 2000nm to 2350nm. …………………………...... ………………………………………………..90 Figure: 6.22. Work flow chart summarizing the surface mineral mapping process of the Hyperion image. ………………………………………………………...... ……………………………90 Figure: 6.23 (a) Published Geology map, (b) ASTER surface mineral map, (c) Hyperion MTMF product, (d) Hyperion MNF bands 2,4,6, with interpreted geology vector map , and (e) Surface mineral map combining information from map b, c, and d. ……………………………………….....……………………………………..91 Figure: 6.24. Spatial subset of the derived mineral map showing some of the geochemical pattern of the study area …………………………………………………………………………………………………………………………………..92 Figure: 7.1. The Graphs show the accuracy of DEM for elevation extraction. …………...……………………..95 Figure: 7.2 (a) Derived drainage pattern and springs overlain on ASTER view shed. (b) Gully erosion from northern part of the Little Magadi Lake. (Looking south) (c) Debris flow southern part of Lake Magadi (looking west). …………………………………………………....……………………………………………………………..96 Figure: 7.3. (a) Field photograph and (b) Subset of surface mineral map overlain on DEM showing the influence of topography for the spatial distribution of Clay minerals. …………………….……………………….97 Figure: 7.4. (a) Drainage network overlain on classified mineral ASTER DEM. (b) Surface mineral map overlain on DEM showing the spatial distribution of Trona, brine, NaAlSi gel and evaporites. …………………...... ………97 Figure: 7.5. (a) Field photographs from Northern part of little Magadi lake shore showing the interaction of water with trachyte debris. (b) Gel near to the hot springs in southern part of Lake Magadi area. …...... ………..98 Figure: 7.6 Derived Land surface temperature maps for different seasons. ………....……………………… 100 Figure: 7.7(a). Yellow color boxes in Little Magadi Lake (LM) and Magadi Lake (ML) showing the selected sites for surface temperature analysis. (b) Systematic distribution of 100 points covering each pixel in 3000m2 of the Magadi Lake area that was used to extract pixel information (Relative surface temperature). The subset shows the margin of water body and dry area at the time that the image was taken (2008‐02‐18). (c). Box plots showing the distribution of relative surface temperature in selected sites in used images. ……...... ………… 101
viii Figure: 7.8 Importance of 3‐D space than for 2‐D map to understand the spatial association ……… 102 Figure: 7.9 Vertical distributions of GPS readings, minerals and rocks. ……………………………...………… 102 Figure: 7.10. Mineral Map overlain on ASTER DEM showing the distribution of Chert in terrain and spatial associations with other minerals. ………………………………………………………………………………… 103 Figure: 7.11. Scaled schematic of the distribution of evaporitic lacustrine sediment in the study area. ………………...... 105 Figure: A. 01.1. Water sample location map of the study area ………………………………………………………. 116 Figure: A.02.1. Geological map of the Magadi area...... 117 Figure: A: 07.1. (a) Normal and (b) continuum removed reflectance spectra of Halloysite and Silica rich minerals in the collected soil samples…………………………………………………………………………………………… 128 Figure: A: 07.2. (a) Normal and (b) continuum removed reflectance spectra of Montmorillonite clay minerals in the collected soil samples………………………………………………………………………………………….. 129 Figure: A: 07.3. (a) Normal and (b) continuum removed reflectance spectra of Montmorillonite clay minerals in the collected soil samples…………………………………………………………………………………………… 130 Figure: A: 07.4. (a) Normal and (b) continuum removed reflectance spectra of Mg clay minerals and Calcite in the collected soil samples……………………………………………………………………………………………….. 131 Figure: A: 07.5. (a) Normal and (b) continuum removed reflectance spectra of undefined minerals (using spectral characteristics) in the collected soil samples………………………………………………….………. 131 Figure: A: 08.1 (a) Locations of the Resistivity Profiles. (b) Used resistivity instrument. (c) Schematic diagram of Main sequence and rollalong method. (d) Resistivity profiles………………………………..……...132
ix List of tables
Table: 5.1. Wavelength positions of the absorption features and common absorption features for sample A001, B002V, P018, B004M, B006M, and B012M ……………..…………………………….…... 38 Table: 5.2. X‐ray Powder Data for Sample no P012 with Oldhamite reference data……………………...... 55 Table: 6.1. Crosstalk parameters for Level‐1A (Default)………………………………………………………………..…..67 Table: 6.2. Gain values for radiance calibration……………………………………………………………………..………….67 Table: 6.3. Flaash input parameters applied on Hyperion radiance image…………………….…….…………….83 Table: 6.4. Error matrix resulting from surface mineral mapping …………………………………………………… 93 Table: 6.5. Error matrix resulting from surface mineral mapping. …………………………………………………… 94 Table: 7.1. Radiometric calibration values for ASTER TIR bands…………………………………………………………99 Table: 7.2. Derived and corrected median elevation values for each class……………………..….…………… 103 Table: A.01. Chemical analysis of springs and related waters from Lake Magadi area...... 116 Table: A.03. Collected samples and field descriptions during the field work...... 118 Table: A.05. XRD Peak list of trona...... 123 Table: A.06.1. XRD Peak list of Erionite (sample no: P010_02)...... 125 Table: A.06.2. XRD Peak list of Erionite (sample no: B020)...... 126 Table: A.06.3. XRD Peak list of Erionite (Sample no: B024)...... 127
x List of Abbreviations
2‐D Two‐ Dimensional 3‐D Three‐ dimensional AR Anti Reflection ASTER Advanced Spaceborne Thermal Emission and Reflectance Radiometer CEA Cryocooler Electronics Assembly DEM Digital Elevation Model EARS East African Rift System EN Emissivity Normalization ENVI Environment for Visualizing Images FPA Focal Plane Array FTIR Fourier Transform InfraRed FWHM Full Width Half Maximum GPS Global Positioning System HAS Hyperion Sensor Assembly HEA Hyperion Electronics Assembly IARR Internal Average Relative Reflectance ICCD International Center for Diffraction Data ICRAF International Center for Research in Agro Forestry ITC International Institute for Geo‐information Science and Earth Observation LST Land Surface Temperature MF Matched Filtering MNF Maximum Noise Fraction MODIS Moderate Resolution Imaging Spectrometer MPA Multi Purpose Analyzer MTMF Mixture Tuned Matched Filtering PDF Powder Data File PPI Pixel Purity Index SAM Spectral Angle Mapper SFF Spectral Feature Fitting SMACC Sequential Maximum Angle Convex Cone SWIR Short Wave InfraRed TIMS Thermal InfraRed Multi‐spectral Scanner TIR Thermal InfraRed TSG The Spectral Geologist VIS Visible VNIR Visible and Near InfraRed XRD X‐ray Diffraction XRPD X‐ray Powder Diffraction
xi
xii HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
1. Introduction
In this section, background of the research, research problem, research objective, research questions and hypothesis is explained.
1.1. Background
The current study area, Lake Magadi, is located in the southernmost part of the eastern Kenya rift; about 100km south west of Nairobi, Kenya. The area is approximately defined by latitude 1041’31.43”S to 1041’32.09”S and longitude 36011’59.62”E to 36019’46.13”E (Fig: 1.1).
Figure: 1.1. Geographical location map of the Lake Magadi study area. Modified after www.thecommonwealth .org.( left) and after Eugster and Jones, (1968) (right)
Lake Magadi has developed some of the most concentrated brines to be found in the alkaline saline lakes of the African rift valleys (Jones et al., 1977). It also contains a deposit of trona (Na2CO3 ∙NaHCO3
∙2H2O), hydrous sodium silicates; Magadiite (NaSi7O13(OH)3 ∙3H2O ) and Kenyaite (NaSi11O20.5(OH)4
∙3H2O ), fine‐grained silica‐rich microcrystalline, cryptocrystalline or micro fibrous sedimentary rocks called chert (SiO2), and authogenic zeolite group minerals such as erionite ((K2, Ca, Na2)2Al4Si14O36
∙15H2O). Mineral formation in the area and their reactions have been studied and cited since 1960s in geological literature (Eugster, 1967; Eugster, 1969; Jones et al., 1977). Hydrothermal circulation associated with the numerous active alkaline volcanoes and their feeder faults, supplies hot alkaline
1 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
brines to the many hot springs situated along the edge of the lake and this system is responsible for the ongoing geochemical evolution of the area (Warren, 2006). Geochemical precipitation and chemical transformation have been studied in the field by looking at their spatial distribution with age of the formation (Behr and Rohricht, 2000; Eugster, 1967; Eugster, 1969) and in the laboratory by changing physical and chemical factors similar to the present and past conditions of the area (Beneke and Lagaly, 1983; Fletcher and Bibby, 1987; Wang et al., 2006). Eugster (1969) and Jones et al., (1976) developed a model for the hydrology and aqueous geochemistry of the Magadi basin with associated mineral precipitation, re‐solution, re‐precipitation and reactions.
1.2. Problem Statement The Pleistocene and Holocene history of the Magadi basin sediments is fully described in the work of Baker (1958) and Eugster (1969). By studying different field sites over the basin, they have developed a (non spatial) model for the aqueous geochemistry of the Lake Magadi basin. The model was based on evaporative concentration and mineral precipitations throughout the different intermediate mineral phases. While Na‐rich alkali trachyte becomes the main source of that geochemical system, trona, zeolite, and chert are some of the end members through intermediate phases such as hydrous sodium silicate and hydrous sodium aluminosilicate gel. As an example, hydrous sodium aluminosilicate gel (Na‐Al‐Si gel) is formed where there is interaction between alkali trachyte with alkaline brine in the lake especially at the shores of Lake Magadi and Little Magadi (Surdam and Eugster, 1976). It is often easier to understand and model these geochemical process, ones the spatial distribution of minerals and geochemical pattern of the area are mapped. The interpretation and reconstruction of these geochemical patterns however, are difficult due to their large spatial extend and complex topography.
Furthermore, some of the mineral phases and type of reactions are limited to restricted places which depend on the spatial associations of the parent materials and other environmental conditions (eg: formation of Na‐Al‐Si gel). To understand the process of mineral formation therefore, necessitate to see each surface mineral with its neighbourhood. This however, does not allow one to perceive the complete picture and understand the spatial distribution of various surface minerals in three‐ dimensional space; as a result, demands laborious investigations in the field.
Alternatively, the access to relatively inexpensive satellite‐borne multi‐spectral and hyper‐spectral data have created new opportunities for the regional mapping of mineralogy, geological structures and rock types including alteration products (Hewson et al., 2005; Vaughan et al., 2005). These techniques provide identification of different surface expressions and mapping possibilities for minerals in the hydroxyl, silicate, sulphate, carbonate and iron oxide groups covering large extents at times with inaccessible terrains. The motivation of this research as such, will be enhance the understanding of geochemical processes in the Lake Magadi area in Kenya by building a spatial distribution model of the surface minerals using integrated remote sensing techniques.
2 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
1.3. Objectives and Research questions
The overall objective of this research is to study and develop a spatial distribution model of the mineral precipitates in the Lake Magadi study area, Kenya.
The following specific objectives and related research questions are defined to meet the general objective of the study;
Specific objective: 01 To understand the formations of mineral precipitates in the study area using previous concepts and models. o What are the existing facts, concepts and models that explain formation of mineral precipitates in the area?
Specific objective: 02 To determine the surface mineralogy using reflectance spectroscopy of rocks and soils. o Is it possible to identify mineral precipitates using reflectance spectra?
Specific objective: 03 To determine which minerals can be mapped using spectral remote sensing. o Is it possible to map all the ground mineralogy that was identified by reflectance spectra?
Specific objective: 04 To map the spatial distribution of minerals and to understand the factors that control their distribution. o What are the main factors that govern the spatial distribution of minerals in the study area?
Specific objective: 05 To interpret and enhance our understanding concerning the formation of the various mineral precipitates by integrating with other data. o To what extent can the distribution and formation of precipitates be understood using remote sensing data?
1.4. Research Hypothesis
o Reflectance spectroscopy allows one to identifying different precipitates such as, hydrous phases of silica and carbonate. o It is possible to map the distribution of different hydrous phases of silica and carbonate minerals using multi‐spectral as well as hyper‐spectral data. o Spatial distribution of the minerals in the study area is highly governed by regional topography, tectonic settings and hot springs of the area.
3 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
1.5. Structure of the Thesis
The thesis will consist of eight chapters,
o Chapter one emphasizes the research problem, the overall and specific objectives, research questions and hypothesis.
o In Chapter two, a review of the available literature including the theoretical aspect of topics crucial for this study will be assessed. In addition, the existing models available in the geochemical processing of Lake Magadi study area will be reviewed.
o Chapter three will deal with information on geology, geomorphology, tectonic and structure of the study area.
o Chapter four concentrates on the methodology and instruments used in the study.
o Chapter five focuses on the spectral analysis of rocks, soils and minerals that were collected from the study area.
o Chapter six deals with surface mineralogical mapping of the study area using multi‐spectral and hyper‐spectral remote sensing.
o Chapter seven deals with the effect of geomorphology, tectonics and surface temperature for the continuous geochemical process and formulates the final model for the spatial distribution of mineral precipitates in Lake Magadi combining all findings.
o The final chapter gives conclusions on the results of the study, recommendations and limitations of the study.
4 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
2. Literature Review
In this section, available literature including the theoretical aspect of topics crucial for this study is reviewed.
2.1. Lacustrine Evaporitic Sedimentary Process The Lake Magadi area is one of the most interesting places to study lacustrine evaporatic sedimentation processes and their mineral reactions in the world. The area is characterized by a trona precipitating saline lake, which is fed by alkaline hot springs (at the perimeter of the basin), hydrothermal alteration and chertification process (Behr and Rohricht, 2000; Surdam and Eugster, 1976; Warren, 2006). Mineral precipitation, re‐solution, re‐precipitation and reactions which are common in this environment can be understood based on the hydrology and aqueous geochemistry of the area. The model that explains these processes is based on evaporative concentration and mixing of waters from three sources: dilute surface inflow; a relatively deep, hot, and concentrated ground water reservoir; and cold, concentrated surface brines (Jones et al., 1977). There are no perennial rivers flowing into the Magadi basin at present (Surdam and Eugster, 1976). However, a large number of springs and seeps are located around the perimeter of the lake. These sources not only range in temperature from 28 to 850C but also vary greatly in discharge. The existence of a hot ground water reservoir was postulated as the principal source of the spring waters. The reservoir must be compositionally uniform long enough to account for the relative constancy of the major hot spring chemistry (Jones et al., 1977) (Appendix: 01). Atmospheric precipitation, overland flow and percolation of water can react with alkaline volcanic terrains; in a process know as silicate hydrolysis.
This can be illustrated by the reaction of albite feldspar with water and aqueous CO2 as follows (Jones et al., 1977),
+ ‐ NaAlSi3O8 + 2H2O + CO2 = Na + HCO3 + 3SiO2 + Al(OH)3 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.1.1
The hydrolysis of fine grained silicate minerals, pyroclastics or glasses in the rift valley volcanics take place rapidly producing bicarbonate‐rich waters with high silica content and leaving alumina clays and residual gels behind (Bish and Chipera, 1991). But another argument shows that the active volcano Ol Doinyo Lengai at the southern end of the Lake Magadi has erupted natrocarbonatite lava and ashes consisting largely of soluble sodium‐potassium carbonate minerals and this ash is readily reached to give a sodium rich solution (Baker, 1986). However, after acquisition of solutes from rainfall and hydrolysis, the waters are subjected to evaporation either at the surface or by capillarity. Precipitation may take the form of efflorescent crusts on top of the sediment surface, or of inter granular caliche‐type films and cements. Efflorescent crusts are the products of complete desiccation and consist predominantly of the most soluble salts. Therefore, they are subjected to complete or differential solution by rain and runoff (Jones et al., 1977). This waters flow out onto the lake as ‐ springs and seeps with low levels of alkaline earths Mg and Ca, and high levels of Na, Cl, and HCO3 .
5 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
This chemistry is ideal for ongoing trona precipitation at higher salinities (Warren, 2006). Trona is the first Na‐carbonate mineral to precipitate in equilibrium with concentrating lake brines. The composition of the Magadi‐trona is straightforward. It consists of trona crystals (sodium sesquicarbonate ‐ Na2CO3∙NaHCO3∙2H2O), Sodium fluoride (NaF), and rarely small amount of common salts (NaCl) (Baker, 1958), that extends some 47km2 in area and 7‐50m thickness. The Trona is locally known as the “Evaporite Series”. It is made up of cm‐scale stacked, upward pointing and growth‐ aligned trona crystals (Warren, 2006). Trona evaporites accumulate as the surface brine is degassing a substantial portion of its CO2 to the atmosphere.
‐ 2‐ 2HCO3 (aq) → CO3 (aq) + CO2(g) + H2O ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.1.2 Where, (aq) = aqueous and (g) = gas.
As that happens the increasingly concentrated brine follows a path similar to the dashed line moving away from point A in Fig: 2.1. At this stage trona precipitation occurs at ambient equilibrium at around point B.
+ 3‐ 2‐ 3Na (aq) + HCO (aq) + CO3 (aq) 2H2O → Na2CO3∙ NaHCO3∙ 2H2O(s) ‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.1.3 (Trona) Where, (aq) = aqueous and (s) = solid.
Once native trona deposition in Lake Magadi has begun, the CO2 re‐supply from the atmosphere is too slow to allow ongoing equilibrium between atmospheric CO2 levels and levels of CO2 in the brine.
‐ ‐ CO3 + H2O + CO2 = 2HCO3 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq 2.1.4
Therefore, trona‐ forming brine is always an alkaline solution with a pH above 9 (Jones et al., 1977; Warren, 2006).
Figure: 2.1. Phase diagram for the sodium carbonate, sodium bi 0 carbonate, water system at 25 C. Trona saturation occurs at B where the brine
trajectory arrow passes into boundary of trona stability field (after Eugster, 1971).
6 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Alkaline brines can store substantial amounts of silica in solution. If such brines come in contact with fresh water, for instance from seasonal rain and runoff, a stratified lake may result and saturation with respect to Magadiite may occur at the interface (Eugster, 1969). This mineral reaction will end‐ up with cryptocrystalline chert. The quaternary sediments of Lake Magadi involve large volume of chert. Origin of the chert can happen in several ways and the process is called silicification. Silicification is a common diagnostic phenomenon in a wide variety of originally non‐siliceous sediments, and its extent ranges from minor to pervasive. In partial and minor silicification in Lake Magadi area includes:
a) Chertification of carbonates and carbonate bearing sandstone and b) Replacement of evaporates, source silica is predominantly biogenic.
Pervasive chertification occurs on the scale of individual layers, beds or entire formations. It has been described in lacustrine, pedogenic and hydrothermal – volcanic environments. Examples are,
c) Magadi‐type cherts which were collected from the study area, d) Other lacustrine cherts and e) Hydrothermal‐ volcanic cherts.
In example (c) to (e) the source of the silica is predominantly inorganic (Hesse, 1989).
In the Lake Magadi area, hydrothermal fluids below the lake floor percolate through the volcanic bedrock (the common lithology here is trachyte lava, an alkaline, intermediate, extrusive igneous rock). The trachytes have some of their components leached out by the hydrothermal fluids. These typically include sodium and silica. As these fluids, rich in dissolved sodium and silica enter Lake Magadi, precipitation occurs, resulting in sodium‐silicate gels on the lake bed. According to Eugster (1967), Magadiite was precipitated from alkaline waters of Lake Magadi study area during the Late Pleistocene, at times when the lake level was more than 10 m higher than today.
The hydrous sodium silicate, Magadiite, with a composition approximating to NaSi7O13(OH)3∙3H2O, was described first by Eugster (1967) in materials from Lake Magadi and subsequently identified in Oregon and California (Eugster, 1967). Conditions associated with the precipitation of Magadiite from Lake brines in Lake Magadi and most of other soda lakes are; 1) elevated alkalinity (pH > 9), 2) high concentration of dissolved silica (up to 2700 ppm) and, 3) in corporation of sodium ions in to the silica lattice that precipitates at the time of super saturation (Fig: 2.2)(Warren, 2006). The precipitation of Magadiite can be represented by the following reaction (Eugster, 1969).
+ + 7 H4SiO4 (aq) + Na (aq) = NaSi7O13(OH)3∙ 3H2O(s) + 9 H2O(aq) + H (aq) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐eq:2.1.5 (Magadiite)
Which is governed by the equilibrium constant K1,
a9 a + a a7 14.3 a (K1)PT = H2O∙ H / Na + H2SO4 = 10 , ( H2O = 1) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐eq:2.1.6
The free energy of formation for Magadiite at 250C, 1 atm is
Δ G0 Magadiite = ‐ 1762.2 kcal/mole.
7 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 2.2. Magadiite chemistry A) Sodium silicates minerals plotted in the upper part of
the ternary diagram. Na2O ± SiO2 ± H2O showing the position of various hydrous sodium silicates related to varying proportion of the constituent water silica and sodic contents. (Ken, Kenyaite; Mag, Magadiite). B) Concentration of [SiO ] at 2 Total equilibrium with Magadiite as a function of
pH at ionic strengths of 0.5, 0.1, and 0.001 at + ‐1 0 0.001 mol Na l , 25 C. (modified after Eugster, 1969 and Warren, 2006)
The conversion of magadiite to quartz‐chert (SiO2) through dehydration and sodium loss may involve Kenyaite (NaSi11O20.5(OH)4∙3H2O) as an intermediate phase,
+ + 22NaSi7O13(OH)3∙3H2O(s) + 8H → 14NaSi11O20.5(OH)4∙3H2O(s) + 8Na + 33H2O ‐‐‐‐‐‐‐‐ eq: 2.1.7 (magadiite) (kenyaite)
+ + NaSi11O20.5(OH)4∙ 3H2O(s) + H → 11SiO2(s) + Na + 5½H2O ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.1.8 (kenyaite) (chert) or it may directly produce a silica phase (Eugster, 1969).
+ + NaSi7O13(OH)3∙3H2O(s) + H → 7SiO2(s) + Na + 5H2O ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.1.9 (magadiite) (chert)
The conversion occurs in near surface layers that are accessible to rain and runoff waters (Warren, 2006). However, according to the field observation and microbiological studies of Behr and Rohricht (2000), they showed that the real inorganic cherts are rear at the locality of magadi‐type cherts. In addition to that they also showed that most of the cherts are older than the high magadi beds which is composed of detrital silicates, saline minerals, calcite, sodium silicates, quartz and zeolites (Surdam and Eugster, 1976) and developed from flat‐topped calcareous bioherms of Pleurocapsa, Gloecocapsa, and other coccoid cyanobacteria, thinly bedded filamentous microbial mats, stromatolites, bacterial slimes, diatoms, Dascladiacea colonies and other organic matter. It means silicification occurred through biogenic process from a silicasol via opal‐A to opal‐C with final crystallization to a chert of quartzite composition. Present study also explored another type of silica rich material that was formed by biogenic process, material called diatomite. A diatom is a single celled organism that is a member of the phylum of algae called Bacillariophyta. They live as individuals or in groups called colonies and exist in all the waters of the earth, both salt and fresh. They form shells made out of silica which they extract from the water. When diatoms die, their silica
8 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
shells accumulate on the floor of the body of water in which they lived. Thick layer of these diatom shells once fossilized is called diatomaceous earth or diatomite (www.mii.org).
Hydrous sodium aluminosilicate gels, another component of this system, can be seen at or near most of the hot springs, as well as at the shores of Lake Magadi and Little Magadi, where alkaline brines are in‐contact with trachytic debris. The reaction responsible for these gels has been elucidated by Eugster and Jones (1968). Trachytic material reacts with alkaline solutions according to trachyte + Na+
+ H2O → Na‐ Al – Si gel + SiO2 assuming the aluminium is conserved during the reaction. Observation of the gel accumulations north of the Little Magadi during four field work of Sudam & Eugster (1976) suggested that the gelatinous material is washed into the lake yearly, particularly during the rainy seasons.
Erionite, (K2, Ca, Na2)2Al4Si14O36∙15H2O, is also a prevalent alteration product of glass in the Lake Magadi region (Surdam and Eugster, 1976). The fibrous erionite is relatively common in zeolite deposits throughout the world (Bish and Chipera, 1991) and it was described and named by Eakle (1898) for wooly masses occurring in welded rhyolite tuff at the Old Durkee Opal quarry, Swayze creek, Baker country, Oregon, USA (www.iza‐online.org). Compared to the other mineral particles, erionite has been shown to have greater pathogenicity than asbestos (Bish and Chipera, 1991). The chemical composition of the erionite varies both in the Si, Al content of the framework and cation content of the erionite cavities. Erionite can be further classified into three groups based on the dominant extra‐frame work cation, namely, Erionite‐Ca, Erionite‐Na, and Erionite‐K. Erionite occurs in several ways. It replaces the rhyolite tuff deposited in alkaline, saline lakes and rarely in deep sea sediments. The process is called diagenetic alteration. There are many occurrences of erionite in cavities of basaltic lavas. Some occurrences show pervasive rock alteration suggesting hydrothermal alteration, while others have limited alteration, consistent with diagenetic reaction with ground water. In addition to that, Erionite has been found in the upper part of drill core in some active hydrothermal areas in yellow stone national park, Wyoming, USA giving clue of hydrothermal alteration (Fig: 2.3) (www.iza‐online.org).
Figure: 2.3. Classification of Erionite according to the dominant extra framework cation and their occurrences. Squares (solid and open) represent samples from cavities in basaltic rocks, and circles represent samples from diagenetically altered pyroclastic rocks. Solid squares
represent erionite from epitaxial overgrowths on levyne, and
open squares from other associations in basalt cavities.
B. H. Baker (1986) noted the presence of erionite bearing tuffs and clays in the Lake Magadi sedimentary basin. Ronald and Eugster (1976) also recognized erionite as the most common zeolite group mineral in the sedimentary deposits of Lake Magadi area. Erionite can form directly from trachytic glass by the addition of H2O only. It is characteristic of the Magadi basin because of the low
9 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
content of alkaline earths in the volcanic glasses and in the solutions interacting with them (Surdam and Eugster, 1976).
The above mentioned chemical evolution of the area can be simply modelled according to their spatial distribution of evaporitic lacustrine sediment in the rift valley based on an east‐west transect of Lake Magadi as shown below.
Figure: 2.4. Scaled schematic
of the distribution of
evaporitic lacustrine
sediment in the rift valley based on an east‐west transect of Lake Magadi. After warren, 2006.
Volcanism in this part of the rift floor had virtually ceased by 0.8 Ma. Therefore volcanic tuff and silt can be seen at the bottom of the rift floor. The oldest sediments in the lake Magadi area are olive‐ green indurated silts, clays and cherts of the Oloronga beds, which began to be deposited more than 0.78 Ma ago, which lied top of the volcanic tuff. High Magadi bed also can be seen with Oloronga bed presenting Magadiite, Kenyaite and cherts. In Fig: 2.4 simply this both beds are called as chert and evaporites. Top of the chert and evaporites bed, Holocene to recent trona bed can be seen.
10 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
2.2. Principles of Reflectance and Emittance Spectroscopy
Reflectance and emittance spectroscopy in the near‐infra red and thermal infra red offers a rapid, inexpensive, non‐destructive tool for determining the mineralogy of rock and soil samples (Gaffey, 1985; Meer and Jong, 2001; Salisbury and D'Aria, 1992). Basically, the thermal infra‐red (TIR) is used to determine composition while Visible and Short Wave Infra‐Red (VIS & SWIR) is used to determine alteration products (Meer and Jong, 2001).
When light interacts with a mineral, some of the photons are reflected from the grain surface, while some pass through the grains and are absorbed. The reflection of light, R, normally incident onto a plane surface is described by the Fresnel equation (Ben‐Dor et al., 1999; Clark, 1999),
R = (n‐1)2+ K2 / (n+1)2 + K2 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.1
Where n is the real part of refraction index. All materials have a complex index of refraction which governs the amount of reflection and refraction. Complex index of refraction, m, has relation with the real part of the index, n and imaginary part of the index (j = (‐1)1/2 and k) of refraction.
m = n‐jk ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.2
According to the Beer’s low, the amount of photons absorbed by the material is directly proportional to the distance traveled through the medium and absorption co efficient, k (Clark, 1999)
‐kx I = Io e ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq:2.2.3
Where I is the observed intensity and Io is the original intensity of light. The absorption coefficient is also related to the complex index of refraction by the equation of;
k = 4πK/λ ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.4
Where, λ is the wave length of light.
The kinetic energy of random motion of particles of matter which has temperature higher than zero Kelvin is known as heat energy. As particles move they collide with each other, causing changes in vibrational, orbital or rotational motion. This change results in the emission of electromagnetic radiation. The material which has maximum efficiency to transform heat energy to radiant energy is called Black Body. The spectral radiation emittance of a black body (M) was derived by plank law;
5 ‐1 MλT = C1/ λ (Exp(C2/λT) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.5
Where λ is the wavelength and T is the absolute temperature in Kelvin. C1 and C2 are the radiation constants with values, 3.742 * 108 Wm‐2μm4 and 1.439 * 104 μmK. The relation between spectral
11 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
radiance, L (Wm‐3Sr‐1) and Spectral radiation emittance, M is given by the following formula (Meer and Jong, 2001).
Lλ = Mλ/π ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.6
But almost all terrestrial materials are not perfect black bodies. Instead they emit radiation depending on their physical and chemical makeup, absorbing some radiation through molecular interactions. The spectral emissivity of the material (ε) is given by the ratio of the spectral radiance to that of a black body at the same temperature;
Ελ = Lλ (material)/ Lλ (black body) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.7
Spectral emissivity is usually predicted from reflectance of materials using Kirchhoff’s low, typically stated in its simplest form as E = 1‐R, where E and R are emissivity and reflectance respectively (Salisbury and D'Aria, 1992).
In the Visible and Short Wave Infra‐Red wave length region(0.4‐ 2.5 μm), iron‐, hydroxyl‐, sulfate‐, water and carbonate bearing minerals display diagnostics spectral features (Hunt, 1980(Meer and Jong, 2001)). Silicate minerals, in contrast, have their spectral features in the thermal Infra‐red than in Visible and Short wave Infra‐red (Salisbury and D'Aria, 1992). The absorption features like position, shape, depth, width are controlled by the particulate crystal structure in which the absorbing species is contained and by the chemical structure of the material (Meer and Jong, 2001). Causes of absorption can be divided in to two categories namely Electronic process and Vibrational process.
2.2.1. Electronic Process
The most common electronic process revealed in the spectra of minerals is due to unfilled electron shells of transition elements like Ni, Cr, Co and Fe. For all transition elements, d orbitals have identical energies in an isolated ion, but the energy levels split when the atom is located in a crystal field. This splitting of the orbital energy states enables an electron to be moved from the lower level in to a higher level by absorption of photons having an energy matching the energy difference between the states (Clark, 1999). On the other hand crystal field varies with crystal structure from mineral to mineral. Therefore amount of splitting varies and the same ion produces obviously different absorptions giving specific mineral identification opportunity from Spectroscopy.
Another electronic process is Charge transfer absorptions. Absorption bands can be caused by inter‐ element transition where the absorption of a photon causes an electron to move between ions or between ions and ligands. This occurs between same metal in different valence status, such as between Fe2+ and Fe3+. Absorption bands that caused by Charge transfers are diagnostics for mineralogy.
12 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Other electronic processes are Conduction bands and Color centers. Conduction bands show diagnostic spectral features in visible region, minerals such as Cinnabar and Sulfur (Clark, 1999). A color centers is caused by irradiation of an imperfect crystal.
2.2.2. Vibrational Process
In the simplest case, a spectral absorption caused by a vibrational transition is due to isolated, single molecules (Harloff, 2000). There are two basic types of vibrational processes: External vibrational process, in which the molecule moves as a rigid unit, and internal vibrational process, in which atoms in a molecules move relation to the each other. Again, the internal vibrational process can be divided in to two components. (1) If there is a change in bond length, the mode is called a stretching vibration, labeled as “v”. (2) If there is a change in bond angle, the mode is called a bending vibration, labeled as “δ” (Harloff, 2000). The frequency of vibration depends on the strength of each bond in a molecule and their molecular mass (Meer and Jong, 2001). Each vibration, mentioned here, can also occur at roughly multiples of the original fundamental frequencies. These additional vibrations are called overtones (Clark, 1999). Each higher overtone or combinations is typically 30 to 100 times weaker than the last (Clark et al., 1990), hence the feature will be more subtle and difficult to sense (Meer and Jong, 2001).
Absorption in a spectrum has two components: Continuum and Individual features. The Continuum is the “Background absorption” onto which other absorption features are superimposed (Clark, 1999). It can be happened due to the different processes in a specific mineral or possibly absorption from a different mineral in a multi‐mineralic surface (Clark and Roush, 1984) or due to the wing of a larger absorption feature (Clark, 1999). To see individual features that are characteristics to the spectrum, the continuum has to be removed. On the other way around, removal of continuum is a mathematical function used to isolate a particular absorption feature for analysis of spectrum (Clark and Roush, 1984). This typically done by estimating the absorptions from the other process by suitable functions such as Straight‐ line Segments, Polynomial and Gaussian.
The definition of a continuum also provides a more consistent definition of band depth. The band depth DH is defined as,
DH = (RC‐RB)/RC ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.8
Where, RB is the reflectance at the band center and RC is the reflectance of the continuum at the band center (Clark and Roush, 1984).
As discussed before, the absorption bands for most rocks and minerals are found in the infrared region. The intensity of absorption band is proportional to the change of the dipole moment and polarizability during the vibration (Harloff, 2000). Intensity of the IR absorption can be described as,
A = log(1/T) = log (l0 / l) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.2.9
13 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Where, l0 and l are Intensity of incident radiation and Intensity of transmitted radiation respectively. The position of the absorption band is a function of absorption frequency of the material. For diatomic molecules, the absorption frequency can be calculated based on Hook’s law,
½ V = 1/2πc (K (m1 + m2)/ (m1 ∙ m2)) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq:2.2.10
Where, m1 and m2 are masses of vibrating atoms, c is velocity of light, v is wave number and K is force constant (bond strength).The Bond strength is changed with bond type.
2.3. Principle of Fourier Transform Spectrometer
The main component in the fourier transform infrared (FTIR) spectrometer is an interferometer. The Michelson interferometer which consists of two mirrors and a beam splitter positioned at an angle of 45 degrees to the mirrors, is most commonly used. While a KBr beam splitter coated with germanium is typically used for mid IR region (2.5 μm – 50 μm), a CaF2 beam splitter coated with silicon is used for near IR region (0.8 μm – 2.5 μm). The radiation of the source is divided by the beam splitter and directed to the fixed (having distance L) and movable mirrors equally (Griffiths and Haseth, 1986)(Fig: 2.5).
Fixed Mirror
Movable Mirror L
Source Δ X L + Δ X
Δ X = 0 Detector
Figure: 2.5. Optical schematic diagram of a Michelson interferometer.
Normally, the moving mirror is scanned at a constant velocity resulting in changing optical path differences (2Δ X) of the two beams as a function of time. The reflected beams interfere at the beam splitter, from where 50% of the radiation returns to the source, and 50% reaches to the detector. At the detector, the intensity of the radiation is measured as a function of the optical path difference of the beams in both arms of the interferometer. The two beams undergo constructive interference, yielding a maximum detector signal, if the optical path difference is an integral multiple of the wavelength λ, I e. if, 2 ∙ 2X = nλ (n = 0, 1, 2,……) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.3.1
14 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Destructive interference, and a minimum detector signal, occurs if 2ΔX is an odd multiple of λ/2. The complete functional relationship between I(ΔX) and ΔX is given by the cosine function,
I(ΔX) = S(v~) ∙ cos (2π∙V~ ∙ ΔX) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.3.2
Where, v~ is an wave number and S(v~) is the intensity of monochromatic spectral line at wave number v~. Since spectrometers are equipped with a polychromatic light source the interference occurs at each wave length as continuous spectrum. This wavelength dependent interference pattern is called the interferogram. The interferogram contains the basic information on frequencies and intensities characteristics of a spectrum but in a form that is not directly interpretable. Thus, the information is converted to a more familiar spectrum form using fourier transform method (Fig: 2.6). The general equation for the fourier transform is applicable to a continuous signal. However, if the interferogram signal is digitized and consists of N discrete equidistant points, the discrete version of fourier transform must be used as shown below.
S(K ∙ Δ v~) = ∑ I (n∙ ΔX ) exp (I 2π n k/ N) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.3.3
The continuous variables X and v~ have been replaced with n. ΔX and K ∙ Δv~ representing the n discrete interferogram points and the K discrete spectrum points (www.brukeroptics.com).
INTENSITY
SIGNAL AMPLITUDE
INTERFEROGRAM SPECTRUMSPECTRUM
WAVENUMBER (Cm‐1) OPTICAL RETARDATION Figure: 2.6. The Interferogram and Reflectance spectrum of Chert sample P006_01.
2.4. Principle of X-ray diffraction
X‐Ray Powder Diffraction (XRPD) is a non‐destructive widely applied technique that provides detailed information about the atomic structure of crystalline substances. X‐ray diffraction analysis can be conducted on single crystals or powders. Of all of the methods available to the analytical chemist only X‐ray diffraction is capable of providing general purpose qualitative and quantitative information on the presence of phases (e.g. compounds) in an unknown mixture (Jenkins, 2000).
Minerals consist of crystals which are themselves made‐up of ordered arrays of atoms, arranged in a periodic or repetitive way (Klein and Hurlbut, 1999). When an X‐ray beam strikes such a three dimensional arrangement, it causes electrons in its path to vibrate with a frequency of the incident X‐
15 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
radiation. These vibrating electrons absorb some of the X‐ray energy and, acting as a source of new wave fronts, scatter this energy as X‐radiation with the same frequency and wave length (Jenkins, 2000). In general scattered waves interfere with and destroy one another. In certain specific directions, however, the scattered waves are in phase with one another and combine to form new wave‐fronts. This constructive interference is known as diffraction (Wilson, 1987; Zussman, 1967).
X‐ray diffraction can be conveniently visualized as a reflection of the incident beam by parallel, closely spaced phases of atoms within a crystal (Fig: 2.7). The condition for reflection is the well known Bragg equation (eq: 2.4.1).
nλ = 2d Sinθ ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.4.1
Crystal Planes
Figure: 2.7. Schematic representation of XRD by regularly spaced planes of atoms in a crystal.
Theta (θ) is the angle that the beam makes with the atomic planes; 2θ is the angle that the diffracted beam deviates from the primary beam; d is the distance between equivalent atomic planes in the crystal (d‐spacing); and λ is wavelength of the radiation. Note that when DE + EF = nλ, where n is an integer (n = 0, 1, 2, 3,.., n), the diffracted beams from each plane of atoms would be in phase, leading to constructive interference which accounts for XRD. In effect, when that condition is met, an XRD peak is observed. The factor in the Bragg equation of interest to mineralogists is the d‐spacing, which can be determined in XRD analysis by fixing λ and measuring the θ angle where a peak in X‐ray intensity occurs (Zussman, 1967). Such measurements can be made for a single crystal, for a mineral in powder form, or for a mixture of minerals in powder form. Information gained from diffraction angles and relative peak intensities for pure minerals can be used to establish structural details of those minerals. X‐ray diffraction can also be used to identify the minerals present in a mixture, such as the minerals present in rock.
The diffraction pattern of a substance can be described in terms of three sets of parameters: 1) the position of the diffraction maxima, 2) the peak intensities, and 3) the intensity distribution as a function of diffraction angle. These three pieces of information can, in principle, be used to identify and quantify the contents of the sample, as well as to calculate the material’s crystallite size and distribution, crystallinity, stress and strain (Jenkins, 2000). The ideal specimen preparation for a given
16 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
experiment depends largely on the information desired. A sample which is used only for the identification of its constituents may be quite different from a sample used to measure strain, which in turn may be different from a sample used in quantitative analysis. A diffraction pattern is characteristic of the atomic arrangement within a given phase and to this extent it acts as a fingerprint of that particular phase. Thus by use of the ICDD (International Centre for Diffraction Data) PDF (Powder Data File) a series of potential matches can be obtained. A complication in the application of this method for the analysis of multiphase materials is that the patterns are superimposed and consequently there may be uncertainties as to which lines belong to which phases.
2.5. Surface mineral mapping using remote sensing
“Remote sensing is the science of acquiring, processing, and interpreting images and related data, acquired from aircraft and satellites or hand held instrument that record the interaction between matter and electromagnetic energy” (Sabins, 1999). Remote sensing images are widely used for geological and or mineralogical mapping using their spectral signatures. While iron‐, hydroxyl‐, sulphate‐, and carbonate‐ bearing minerals display strong spectral features in the VNIR and SWIR portions of the electromagnetic spectrum, silicate minerals exhibit spectral feature in the TIR (Hellman and Ramsey, 2003; Rowan et al., 2006; Thurmond et al., 2006). Therefore SWIR wavelength region 2.0 μm‐ 2.5 μm is mostly used to map clay minerals and hydrothermal alteration systems (Rowan et al., 2006). The earliest use of satellite geological remote sensing was for mineral exploration in the 1970s detection of clay minerals in SWIR region (Hellman and Ramsey, 2003). Remote sensing for lithologic mapping using TIR spectral signatures was first demonstrated using the airborne Thermal infra‐red multi‐spectral scanner called TIMS (Ninomiya et al., 2005; Sabins, 1999). Instruments that are sensitive to TIR wavelengths can measure known spectral features that are ‐ 2‐ related to the fundamental vibrational frequencies of inter‐atomic bonds such as Si‐O, CO3 , and SO4 , within common rock forming minerals (Vaughan et al., 2005). ASTER instrument is the widely used multi‐spectral system to map rock forming minerals and alteration minerals because of its sufficient spectral, spatial and radiometric resolution for geologic application (Ninomiya et al., 2005; Rowan et al., 2005).
Multi‐spectral remote sensing has been used routinely since the lunch of Landsat in the early 1970s in discipline as widely varying as geology, ecology, hydrology, oceanography, military applications and many of others (Meer and Jong, 2001). Many of the surface materials, although not all (ex: Rock forming minerals such as Quartz and Feldspar), have diagnostic absorption features around 0.4 μm to 2.5 μm wavelength regions with 20 to 40 nm wide at half of the band depth (Goetz et al., 1985). But multi‐spectral scanners such as Landsat have band width between 100 and 200 nm can not identify spectral features. Therefore, until recently, the main limitation of remote sensing was that lack of detail surface material identification due to the broad band width and poor spatial resolution of sensors available (Meer and Jong, 2001).
Imaging spectrometry which is defined as “the simultaneous acquisition of images in many narrow, contiguous spectral bands” (Goetz et al., 1985) allows extraction of a detailed spectrum for each
17 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
picture element (pixel) of the image. High spectral resolution reflectance spectra collected by imaging spectrometers allow direct identification of individual materials based upon their reflectance characteristics such as absorption band depth, band position, asymmetry of the absorption features, etc. Sabins, (1999) listed three functional categories that are used for the image prosesing process, namely, 1) Image restoration compensates for image errors, noise, and geometric distortions introduced during the scanning, recording, and play back operations. Hyperion pushbroom hyperspectral image shows column striping and bad bans and dud cells due which must be compensate during this stage; 2) Image enhancement improve the information content of the image. Density slicing is a one of the most common image enhancement method; and finally, 3) Information extraction is performed to display spectral and other characteristics of the scene that are not apparent on restored and enhanced image. Unlike multi‐spectral, hyperspectral remote sensing used lots of advanced information extraction methods such as spectral angle mapper (SAM), spectral feature fitting (SFF), linear spectral unmixing, and SMACC end member extraction (ENVI, 2008).
2.6. Land surface temperature mapping using remote sensing
Land surface temperature (LST) is important for a variety of climate, hydrologic, ecological, geological and biogeochemical studies. The most popular usages that are used in geology in present are coal fire detection and mapping of geothermal activities (Coolbaugh et al., 2006; Gangopadhyay et al., 2005). A surface having a temperature more than absolute temperatures emits energy, depending on the properties of the surface and is a function of the wavelength as described by plank (using plank’s law), is the theory behind the LST mapping. The total amount of radiation emitted over a hemisphere ‐3 from a perfectly emitting blackbody surface at any given wavelength, LBB (in units of Wm ) is,
5 LBB = C1 / λ ∙ (1/π [exp (C2/λT) – 1]) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.6.1
Where C1 and C2 are the radiation constants with values 3.74151 * 10‐16 Wm‐3 and 0.0143879 mK respectively, λ here is wavelength in m and T is the temperature in Kelvin (Mushkin et al., 2005). Most of the surface materials however, do not emit radiance like a perfect blackbody (Kealy and Hook, 1993; Mushkin et al., 2005), the total amount of emitted radiation at any given wavelength, therefore, Lem, is given by:
5 C2/λT Lem = LBB ∙ ε = C1/λ ∙ (ε / [e ‐1]) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 2.6.2
Where the surface emissivity, ε, is defined as the ratio between the measured surface emitted radiation and the radiation expected from a blackbody at the same temperature.
Temperature of the surface material can yield the inversion of eq: 2.6.2, as follows,
5 T = C2/λ ∙ 1/(ln[(C1ε/λ Lem) + 1]) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq : 2.6.3
18 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The classical challenge in quantitative thermal‐IR remote sensing is however, in separating temperature from emissivity (Temperature – Emissivity Separation: TES), thereby retrieving temperature and emissivity (Mccabe et al., 2008). Because ε is wavelength‐dependent and thus measurements at n wavelengths are always accompanied by at least n+1 degree of freedom (i.e., n∙ε + T) even, if the surface material is compositionally homogeneous, the selected pixel is isothermal and atmospheric effects of the image have been successfully removed (Hook et al., 1992; Kealy and Hook, 1993; Mccabe et al., 2008; Mushkin et al., 2005). This problem has led to the development of a variety of techniques which differ according to the assumptions they make. Emissivity normalization method and reference channel method can be introduced as two of commonly used TES techniques in geological literature. Even though, split‐ window technique is commonly used in previous literature to map surface temperature in the ocean, it is not valid more over land surface due to unknown emissivity differences (Kealy and Hook, 1993). One of the qualitative techniques developed to enhance temperature information is the decorrelation stretch (Hook et al., 1992). However, it can be used only for visualization purpose.
The reference channel method which used to separate temperature from emissivity, assumes a pixel has a constant emissivity value in the wavelength region covered by a certain channel. Then the temperature for that channel can calculates using equation 2.6.3 (Kealy and Hook, 1993; Mushkin et al., 2005).
The approach of emissivity normalization method is also similar to the reference channel method and it calculates temperature from each channel using a constant emissivity value (frequently 0.96 is used, averaging emissivity values of most common geological surfaces). The highest of these is assigned as a temperature of the pixel (Kealy and Hook, 1993). The accuracy of the both techniques is strongly dependent on the signal to noise ratio (S/N) of the data, if the input channel indicates any noise, that noise will introduce error in the recorded temperature. Studies of Kealy and Hook (1993) conclude more accurate land surface temperature can be obtained by using the emissivity normalization method than the reference channel.
19 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
20 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
3. Geological and tectonic settings
This section aims to briefly review some of the key geological features of the East African Rift System with Geological and tectonic settings of the Lake Magadi study area.
3.1. East African Rift System
The East African Rift System (EARS) is Ethiopian perhaps one of the best known Dome continental rifts which exist on the present earth surface. It forms a narrow (50‐ 150 km wide), elongated system of normal faults that stretches some 3,500 km in a sub meridian direction and it is Kenya Dome connected to the world wide system of oceanic rifts via the Afar triangle to the Gulf of Aden and the red sea (Morley et al., 1999). The rift itself traverses two areas of continental uplift, the Ethiopian and Kenya domes, separated by the low‐ lying Turkana depression in northern Kenya (Fig: 3.1). Throughout Ethiopia, East African Dome although the rift defines a single zone of extension and volcanic activity termed the Main Ethiopian Rift, south of Thurkana, the rift is manifested in two branches that encircle the mechanically Tanzanian Craton robust Tanzania craton, referred to Dome herein as the Eastern and Western Kenya Rift (Furman, 2007). The eastern branch runs over a distance of 2200 km, Fig: 3.1. Main lakes and structures in the East African Rift from the Afar triangle in the north, (modified after Frostick (1997)). through the main Ethiopian rift, the Omo‐Thurkana lows, the Gregory (Kenyan) rifts, and ends in basins of the North‐Tanzanian divergence in the south. The western branch runs over a distance of 2100 km from Lake Mobutu in the north, to Lake Malawi in the south (Chorowicz, 2005) (Fig: 3.1).
21 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The evolution of East African Rift System is now moderately understood as a result of mapping and dating of volcanic units in the area. The earliest recorded volcanic activity in the African Rift took place 40‐45 Ma in the northern Turkana depression (Furman, 2007). In the northern Kenya rift, basalts and rhyolites were emplaced at 33‐25 Ma (Oligocene), then nephelinites and phonolites at 26‐ 20 Ma, and highly hyperalkaline basalts at 15 Ma in eastern periphery of the region. At beginning of the Pliocene, trachyte, phonolitic, nephelinitic rocks, and basaltic volcanism were accompanied with some rhyolitic activity. Then, in the late Pliocene and Quaternary, volcanism was trachytic in the rift valley floor, basaltic to the east (Baker, 1986; Chorowicz, 2005). In the central Kenyan rift 1 km thick Samburu flood basalts erupted between 20 and 16 Ma, followed by large volumes of trachyte and phonolite between 5 and 2 Ma, and carbonatite and nepheline ‐ phonolite volcanoes around 1.2 Ma (Chorowicz, 2005). In the Virunga (near to the Lake Edouard, see Fig: 3.1), massif fissure volcanism started between 11 and 9 Ma, and after long period of quiescence (9‐3 Ma), large Pliocene‐ Pleistocene central volcanoes were formed, made of highly under‐saturated, potassic, ultra alkaline lavas. In the south Kivu region (see Fig: 3.1), volcanism commenced at 8 Ma with tholeiites, followed by sodium alkaline lavas (Chorowicz, 2005). Baker (1987) identified three major stages in the tectonic development of the Kenya rift; 1) the pre‐rift stage (30‐12 Ma) with the forming of a deposition and minor faulting, 2) the half‐graben stage (12‐4 Ma) with the forming of the main boundary faults, and 3) the graben stage (< 4Ma) with increase and an inward migration of faulting.
The study of sedimentation patterns in the EARS, on the other hand, leads to a more complete understanding of the tectonic and volcanic evolution, and are controlled by structures, with strong influence of climatic environments and occurrences of great lakes (Baker, 1986; Chorowicz, 2005). In general, sedimentation in rift valley is characterized by a combination of organic and detrital deposits in shallow lakes. The first sediments resulting on the basement are sand‐dominated fluviatile deposits, which are followed by clay and silt deposits with high organic content corresponding to swamp environment. In the late stage of rifting, sedimentation changes and is characterized by thick deposits of deep lacustrine detrital sequences forming turbidities (Chorowicz, 2005).
3.2. Geological and tectonic settings of the Magadi area The Lake Magadi, present study area is located in eastern part of the rift system, just near to the conjunction point of Aswa shear zone from north‐east, Kenya dome from north‐west and Tanzanian craton from west. Geological setting of the study area can be grouped in to three main categories; 1) Precambrian metamorphic rocks, 2) Plio‐ to‐ Pleistocene volcanic rocks, and 3) Holocene to recent lake and fluvial sediments (Atmaoui and Hollnack, 2003). The most volcanic activity in the area occurred between 1.4 and 0.7 Ma with the formation of the Magadi plateau trachyte series (Fig: 3.2) (Baker, 1986). The lacustrine sediments are exposed around the Lake Magadi trough in the central axis of the rift floor; a small narrow basin filled by three successions of quaternary fluvio‐ lacustrine sediments: 1) the Oloronga beds were deposited about 0.8 Ma ago in a weakly alkaline lake. It consists of olive green indurated silts, clays, volcaniclastic sands and irragulerly interbedded cherts which are topped with a thick caliche cap. This bed simply assigned as chert series (in Fig: 3.2 this bed simply assigned as chert), because of the occurrence of cherts at the base of the Oloronga beds (Eugster, 1969). 2) Below the oloronga beds are the extensive plateau trachyte flows that cover the
22 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
rift valley floor and are as old as 1.7 Ma. The latter known as the High Magadi beds, are subdivided into an upper and lower unit separated by a black clay rich layer named Thilapia bed (Eugster, 1969). The High Magadi bed mineralogy is characterized by the following group of minerals: i) detrital silicates, ii) saline minerals, iii) calcite, iv) sodium silicate and quartz, and v) authogenic zeolites (Surdam and Eugster, 1976), 3) The Holocene evaporites series, which continuous to accumulate to the present day, resulted from the increasing desiccation of the lake during the Holocene. It consists of alternating trona sheets containing black mud with a Na2CO3 and Chloride‐rich brine (Baker, 1958) (Appendix: 02).
A Kenya Aswa Shear Zone Dome
Tanzanian Craton
Figure: 3.3. Structural interpretation of Lake Magadi basin (a), with 3D structural sketch (b). Interpretation from SPOT XS image. After Turdu et al., 1999. Figure: 3.2. Geological Map of Lake Magadi. Modified after B.H. Baker ,(1954) and B. H. Baker ,(1986)
Tectonic and structural setting of the area is largely influenced by three main features namely, 1) stable Tanzanian craton, 2) Aswa shear zone, and 3) southern fringes of the Kenya dome (Fig: 3.3) (Turdu et al., 1999). The sedimentary environment of the Lake Magadi basin is highly characterized by the dominant influence of nearly continuous tectonic movements and volcanism, which created a multitude of small short lived basins in the main graben and larger longer lived basins in the broad half graben (Baker, 1986).
23 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
24 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
4. Methodology
In order to meet the objectives of the study, the overall work was divided in to three main phases namely, Pre‐field work phase, Field work phase, and Post‐field work phase.
4.1. Pre-field work phase
At this stage, research questions, hypothesis and suitable research approach were formulated in order to meet the objectives of the research reviewing literatures related to the topic. Required data were classified according to their sources and work phase as: 1) Pre‐field work, 2) during fieldwork and 3) Post fieldwork data acquisition and preparation.
4.1.1. Data required before the field work o ASTER Image o Hyperion Image o Geological map of the area o Sample location maps o Other thematic maps such as roads, contours, and locations of the hot springs Sample location maps were derived from ASTER and Hyperion image using image processing techniques such as band ratios and Maximum Noise Fraction (MNF) after the flat field and Log residual atmospheric correction. Resultant tentative sample location maps were combined with other thematic maps (Roads, contours, and places of the hot springs) to get more information. Interesting sample locations were plotted on the map selecting ASTER Band ratio 14/12: 4/3: 7/5 as RGB colour composite after analysing 20 band ratios (Fig: 4.1 (a)). MNF out put image 2, 4, 6 and 2, 5, 7 from Hyperion image was selected as RGB colour composite (Fig: 4.1 (b) and (c)) to see the detectable spectral variations of the ground material that are required to collect during the field work. Preparation of materials for field data collection was done during the pre‐field work phase.
4.1.2. Data/Materials acquired during the field work o Rock samples o Soil samples o GPS locations with field descriptions o Photographs
25 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
4.1.3. Data required after the field work o Reference X‐ray powder diffraction patterns for selected materials o SWIR and TIR spectral reference libraries for rock and minerals
4.2. Field work phase
The field work phase of the study involved collection of rock and soil samples which is related to the study, land cover analysis, photographs, and GPS data collection. The field description data sheet, which was used in the field, is shown in Fig: 4.2. Rock and soil samples were collected from relatively homogeneous bare land areas which has spatial extend higher than 400 m2. Every location was recorded by the GPS keeping 7‐9 m spatial accuracy. 26 Soil samples, 58 Rock samples and 8 undisturbed surface soil samples were collected from the Study area representing all rock and soil formations (Fig: 4.3) (Appendix: 03). Soil samples consist of set of bare land top soil, alluvial deposit, aeolian deposit and lacustrine sediments. Rock samples consist of set of trona beds, basalt, alkali trachyte, quartz‐ chert series rocks and different types of sedimentary rocks. Undisturbed wet soil samples, up to 6 cm, were collected from sedimentary depositional area, near the hot springs and places where evaporites present. Supplementary data such as location, elevation, geology, geomorphology, slope, land cover, dip and strike (if applicable) were recorded parallel to the sample collection.
Short wave Infra‐red (SWIR) reflectance spectra were acquired by 26 collected soil samples during the field work phase at ICRAF (International Centre for Research in Agro Forestry) in Nairobi, Kenya, using Multi purpose analyzer ‐ furrier transformed spectrometer (MPA ‐ FTIR). Crystallized trona bed (A001) and its’ <2 mm crushed powder (B002V) were analyzed to study the variation due to changes in grain size, shape and structural order. Sample No: B004M, B006M and B012M were collected from different locations where evaporites occur. Here, letter M and V followed by sample numbers are assign for the spectra from spectrometer MPA and VERTEX respectively. Those three samples were crushed and sieved into less than 2 mm grain size after open air dry, and spectral acquisition was done at ICRAF (International Centre for Research in Agro Forestry) in Nairobi, Kenya using MPA Infra‐ Red Spectrometer (Instrumentation and Spectral acquisition are discussed in section 4.4.1). Sample No: P018 was not crushed in this study in order to see the spectral difference between crushed product and its original surface form. Spectral acquisition of samples A001, B002V and P018 were done at ITC (International Institute for Geo‐Information Science and Earth Observation) in the Netherlands with same spectroscopic techniques that are described in section 4.4.2.
4.3. Post-Field work phase The final phase of the study involved the acquisition of spectra from collected rock and soil samples, identification of mineralogy using X‐ray diffactrometry, mineral mapping using multi‐spectral and hyper‐spectral images according to the spectral indices that were identified as a member of ongoing geochemical process of study area. Finally, understand the spatial distribution of mineral precipitates in the study area creating a model showing spatial distribution of precipitated minerals with help of
26 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
derived mineral map, digital elevation model, relative temperature map that was derived from ASTER image, and other accessory data.
0 36 20’E
50’S 0 1 (b)
00’S 0 2
(C) (a)
36020’E Figure: 4.1. Processed satellite images to differentiate different surface materials for field sample collection. (a). ASTER band ratio 14/12:4/3:7/5 (RGB). (b).Classified Hyperion MNF band 2:4:6 (R:G:B) from northern part of the area and (c). Classified Hyperion MNF band 2:5:7 (R:G:B) from southern part of the area
27 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 4.2. One of the field description data
sheets used in the field.
Figure: 4.3. Sample location map
The overall research methodology is illustrated in Fig: 4.4.
28 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Preparation for field Pre‐field work Methods for sample/data collection
Pre Generation of sample location map Material for sample collection ‐ Field
Using ASTER image Using Hyperion image Work Figure: 4.4. Flow chart showing the
research approach of the study. Phase Band ratios Maximum Noise
Roads Combine with other thematic data Identification of sample locations
Contours Locations of hot springs Sample and data collection
Field
ASTER image Hyperion image Sample Data Work
Phase Crosstalk correction Bad band removal Rock and Soil Land cover,etc.
Radiometric correction Fix bad pixels/bands Spectral acquisition
Atmospheric correction Correction of column striping TIR SWIR
Atmospheric correction X‐ray Spectral analysis Diffraction
Band Spectral re‐ Effort polishing ratio sampling Post Mineral identification ‐ Field
Spectral angle SAM, MNF & MTMF
mapping Work End member selection
Phase
Surface mineral map of the study area ASTER images
Temperature‐Emissivity separation
Data integration ASTER DEM Average relative temperature map and analysis
Spatial distribution of mineral Geochemical Process of thre Conclusion and Recommendations precipitatesLake inMagadi Lake areaMagadi area
29 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
4.4. Instrumentation
All soil samples were dried in air dry. After that they were crushed and passed through the 2 mm Sieve. While crushing plant materials and gravels were removed while making sure they are not soil aggregates. Then spectral acquisition was done in ICRAF (International Centre for Research in Agro Forestry) in Nairobi, Kenya using MPA Infra‐Red Spectrometer.
4.4.1. The MPA FTIR Spectrometer
Gold coated Intergrating Sample Sphere
NIR Beam Detector
Figure: 4.5. Bruker Multi Purpose Analyzer (MPA) instrument and Optical path diagram of integrating sphere (Baker, 1986; www.brukeroptics.com).
Multi Purpose Analyzer (MPA) is a Fourier‐ transform Infrared (FTIR) spectrometer which operates in the Near‐Infrared (0.8 μm – 2.7 μm) wave length region. The method that was used for present analysis is the Sphere Macro Sample. It has the sample component unit that operates on the transmission mode and the Integrating sphere window that operates on the diffuse reflectance mode (www.brukeroptics.com). The standards used for this study were Kaolinite and ICRAF “Mua Hills” soil. The samples were loaded into the thoroughly cleaned Petri dishes with thickness of 10‐25 mm.
Spectral measurements were performed over the instrument full range, 12500 cm‐1 – 3600 cm‐1, (0.8 μm – 2.77 μm) at 8 cm‐1 resolution. Each sample had 32 sample scan along 2 horizontal right angle positions. The background measurement is normally taken before the sample measurements and this measurement is valid for 8 hours. 26 SWIR spectra were collected from soil samples in this range.
Reflectance spectra of the rock samples were taken by Vertex‐70 FTIR instrument at ITC (International Institute for Geo‐Information Science and Earth Observation) in the Netherlands.
30 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
4.4.2. The Vertex FTIR Spectrometer
The Vertex‐70 is a Fourier transform Infra‐red (FTIR) Spectrometer that covers the spectral range from 30 cm‐1 in the Far IR, through the Mid and Near IR up to the visible spectral range at 25,000cm‐1. The Interferometer is based on Bruker’s RockSoildTM system, a cube corner reflector based interferometer that is insensitive to alignment issues or vibrations (www.brukeroptics.com). Sources, detectors and mirrors to change the beam path inside the instrument are automatically selected through OPUS software. All the mirrors are Gold coated to optimize the system for the MIR region.
Two type of Beam Splitters are used for the NIR and MIR wave length region respectively, Si on CaF2 Splitter (T401/3) with range of 15,000‐ 1,200 cm‐1 and Ge on KBr Splitter (T303/3) with 7,800 – 370 cm‐1 wave number range (Bruker, 2007). These Beam splitters are needed to be changed manually. In order to allow directional‐ hemispherical reflectance measurements of large rock samples, external detectors that were attached through the customized integrating sphere setup was used for this study (Fig: 4.6). This external sphere (150 mm diameter) setup has several advantages. (1) It has extra space for attaching detectors, (2) Lager sample ports (30 mm diameter) allow for lager sampling beam. This is an advantage for large homogeneous materials, such as rock or soils. (3) Sample port of this setup directs the beam to bottom unlike MPA Spectrometer. Therefore large and heavy rock samples can be brought to the sample port with the help of lab jack (Personal communication with Mr.Chris Hecker). However, this large, external integrating sphere attenuates signal strength than in a small sphere. The MCT detector port is not exactly on the pole of the sphere, but slightly closer to the pole. This allows the folding mirror to block the first reflection off the sample and off the calibration spot in reference position (Fig: 4.6.a). The sample has to be placed at the bottom opening of the sphere. The incident beam is reflected on to the sample opening by a centre mounted mirror. This mirror has to be changed into two positions called sample and reference in order to get sample measurement and reference measurement. Short wave Infra Red (SWIR) spectra were collected for the Rock samples using InGaAs on Sphere external detector, NIR internal Tungsten source (Q482/7) and NIR Si on CaF2 beam splitter. The wave number range and resolution were from 10000 cm‐1 to 3000 cm‐1 and 8 cm‐1 respectively with 256 for each sample and background scan times.
31 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
MCT
Detector
InAsGa Detector
Figure: 4.6. a) The Optical path diagram showing the optical path going into the external integrating sphere (Bruker Optics, modified by author).
Figure: 4.6. b) Front view of Vertex 70 with external sphere, MCT detector (red) and InGaAs detector (black).
For the present study, several Short Wave Infra‐Red (SWIR) spectra were acquired from each rock sample representing their different weathering surfaces and one of the unweathered surfaces. While spectrum from the unweathered surface gives information of the mineral composition of the material, weathered surface gives an idea about mineralogy of weathering product and stage of weathering. On the other hand Image acquired from Space borne/ air borne spectrometry record information about the surface of the material that was probably weathered. From 91 SWIR rock spectra, 58 spectra, which were taken from the unweathered surface of each sample, were selected for the basic classification. Then, similar spectra were classified and grouped based on the shape, position and depth of the absorption features visually analyzing in ENVI and with the help of The
32 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Spectral Geologist (TSG) software. Then, 46 Thermal Infra Red (TIR) spectra were collected representing each rock sample categories using MCT on Sphere external detector, MIR external Tungsten source (Q482/7) and MIR Ge on KBr beam splitter. The wave number range and Resolution were from 7000 cm‐1 to 500 cm‐1 and 8 cm‐1 respectively with 1024 for each sample and background scan times.
The X‐ray diffraction analysis was performed to conform the mineralogy of selected 15 samples representing each category that was described in previous chapter, at Institute for Nanotechnology (MESA), university of Twenty in The Netherlands.
4.4.3. X-Ray Powder Diffractometer (XRDP)
The instrument that has been used to perform X‐ray powder diffraction analysis is called X‐ray Diffractometer. A diffractometer consists essentially of an X‐ray generator, a goniometer for rotating the sample and measuring diffraction angles, and an X‐ray counter tube and counting circuits to detect, amplify and measure the diffracted radiation (Fig: 4.7). Fig: 4.8 shows a schematic drawing of a typical vertical powder diffractometer system and illustrate the geometry of the system. This geometric arrangement is known as the Bragg–Brentano parafocusing system and is typified by a diverging beam from a line source X‐ray tube, falling onto the specimen on sample holder, being diffracted and passing through a receiving slit to the detector. Distances from collimators to sample holder and receiving slit to sample holder are equal. The amount of divergence is determined by the effective focal width of the source and the aperture of the divergence slit. Axial divergence is controlled by two sets of parallel plate collimators (Soller slits) placed between focus and specimen, and specimen and scatter slit respectively (Harris and White, 2007; Jenkins, 2000).
X‐ray tube Detector
Slits Sample charger rf
Monochromator
Goniometer
Shutter light
Sample holders
Figure: 4.7. Used X‐ray powder diffractometer with its Figure: 4.8. Schematic representation of the main components. components of an X‐ray diffractometer (after (Harris and White, 2007)).
33 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The X‐ray generator is designed specifically for high out put power with full rectification and high current stability. The maximum power is usually about 3 kW, and is obtained through a variable transformer controlled from a multi position switch on the control unit (Wilson, 1987). Production of X‐rays is accomplished using an X‐ray tube consisting of a filament electron source and a metal target. The tubes are evacuated to minimize absorption of electrons accelerated from the filament (cathode) to the target (anode). Activation of the tube entails passing a current through the filament to establishing a current (in this case 35 mA) under high voltage (in this case 50 kV) between the filament and target. The goniometer rotates the specimen and the X‐ray counter in a fixed 1:2 ratio making the rotation of X‐ray counter at twice the angular speed of the specimen. The radius of the Goniometer is 173 mm. The detector counts the X‐ray radiation while changing the incident angle theta.
The X‐rays used for this study is of the Cu ka wavelength 1.54056 * 10‐10m, the scan was taken between 2theta of 4.00500 and 83.99500 at increments of 0.010 with a count time of 0.5 seconds for each step. This 2‐theta angle range has been selected to cover diagnostic peaks of excepted minerals. The count time was selected as 0.5 second to give a good signal to noise ratio and to finish the analysis within a reasonable period. The data was acquired by using Xpert‐ APD software (Appendix: 04).
Two different approaches were used to identify phases in an unknown mixture. The first method is an analytical approach in which no basic assumptions were made about the sample being analyzed. The three strongest lines in the pattern were used to locate potential matches in the PDF index. Each time a potential candidate was found, a match was made with the complete pattern. If all lines agree, a phase confirmation was assumed and the lines for the match were subtracted from the original pattern. This process was repeated until all significant lines in the pattern were identified. The second method was based on a series of guesses based in turn on preconceived ideas of what phases might be present. The two basic parameters being used in this search/match process are the ‘‘d’’ values which in turn have been calculated from the measured 2‐θ values in the diffractogram using Bragg equation (eq: 2.4.1), and the relative intensities of the lines in the pattern. The ‘‘d’’ value can be accurately measured – perhaps with an accuracy of better than 0.5% in routine analysis, whereas the intensities are by comparison, hence rather unreliable and can be subjected to errors, sometimes running into tens of percent (Jenkins, 2000). Therefore “d” values were used as a first matching parameter before matching intensities.
34 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5. Mineral identification using reflectance spectroscopy and X-ray diffraction methods
In this section, mineral identification using reflectance spectroscopy and X‐ray diffraction pattern of rocks and soil samples that were collected from the field is discussed.
5.1. Magadi Trona (Evaporite Series)
Evaporites can be defined as a salt rock that was originally precipitated from a saturated surface or near surface brine by hydrologies driven by solar evaporation (Warren, 2006). Six evaporite samples, including trona beds (Sample No: A001 and B002) and surface evaporites (Sample No: P018, B002, B004, and B012), were analyzed (see Figs: 5.1‐5.4) to determine spectral characteristics of evaporites.
Figure: 5.2. Surface evaporites along shore of the Little Magadi Lake
Figure: 5.3. Crystallized trona (A001) and its powder form of grain size<2mm (B002).
Figure: 5.1. Trona and Evaporites samples with all other
sample locations.
35 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
a
b
Figure: 5.4. Surface evaporites and collected Figure: 5.5. (a) Trona Crystal Form and (b) sample from southern part of Lake Magadi. Crystal Structure. (www.webmineral.com)
Sample no: A001, crystallized trona was used for the X‐ray diffraction to conform the mineralogy. Crystallography of trona is shown in Fig: 5.5 and it can be described as monoclinic – prismatic (2/m) crystal system with 5.7577: 1: 2.9504 (X: Y: Z) cell dimension ratios (ICDD, 2008). The X‐ray diffraction pattern shows three most intensity peaks at 33.941, 29.1855, and 18.146‐ 2θ angles due to (411), (‐ 602), and (400) crystal planes respectively (Fig: 5.6). XRD pattern of the selected sample was completely matched with the reference data (Appendix: 05).
A001
Position 2θ [2 Theta]
Figure: 5.6. XRD pattern of the Trona (A001) sample.
5.1.1. Spectral Characteristics of Trona
The spectrum (Fig: 5.7 and Table: 5.1) exhibits six common absorption features at 1.50 µm, 1.74 µm, 1.94 µm, 2.03 µm, 2.22 µm, and 2.39 µm for all samples. Crystallized trona bed and its crushed product shows another strong absorption feature at 1.20 µm and two weak absorption features at 1.34 µm and 2.30 µm. Grove et al. (1992) present the spectrum of trona from Wyoming in USA, for grain size of < 45000 nm, 45000‐12500 nm and 12500‐500000 nm (Fig: 5.8). These three samples show a weak absorption feature at 1.3440 µm and relatively deep and broad absorption feature at 1.2040 µm. In addition to that, they showed absorption features at 1.9400 µm, 1.5120 µm, 1.7400 µm and 1.0080 µm, which is similar to the results of this study. The spectrum of the sample with the
36 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
largest grain size (crystals: A001) shows a small absorption feature near 1.008 µm that is not evident in the smaller grain size sample (B002V), and the features within entire wavelength region are best developed in the spectrum (Drake, 1995). Crystallized trona and undisturbed evaporites don’t show hydroxyl absorption feature at 1.41 µm. However, three crushed surface evaporites samples (B002M, B004M, and B012M) show hydroxyl feature at 1.41 µm, probably, due to the mixing of other surface materials such as clay minerals.
Trona Samples Trona Samples
Reflectance
(a) (b)
Figure: 5.7(a). Normal and (b). Continuum removed reflectance spectra of Trona samples.
In most cases, the absorption features seen in evaporites mineral spectra are caused by water molecules that are essential components in each mineral structure (Crowley, 1991). The relatively weak absorption seen near 1.008 µm probably represents a combination of the first overtone of the O‐H stretching and the first overtone of the H‐O‐H bending fundamentals. The slightly more intense feature near 1.20 µm is assigned to a combination of the H‐O‐H bending fundamental and the first overtone of the O‐H stretch (Crowley, 1991). Absorption feature near 1.50 µm is mainly due to the
37 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
first overtone of the O‐H stretching fundamental. Absorption feature near 1.74 µm is assigned to combinations involving the fundamental H‐O‐H bend, the fundamental O‐H stretch, and low frequency liberation modes of the structural water molecules (Crowley, 1991). The presence of 1.94 µm absorption indicates molecular water in the sample (Clark et al., 1990). Absorption feature near 2.21 µm and 2.39 µm are responsible for the carbonates (Crowley, 1991; Gaffey, 1987).
Table: 5.1. Wavelength positions of the absorption features and common absorption features for sample A001, B002V, P018, B004M, B006M, and B012M in μm. A001 B002 P018 B004M B006M B012M Common 1.202 1.210 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 1.344 1.342 1.364 ‐‐‐‐‐‐ ‐‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 1.410 1.414 1.406 ‐‐‐‐‐‐ 1.501 1.508 1.520 1.510 1.508 1.506 1.50‐ 1.732 1.740 1.745 1.742 1.742 1.728 1.74‐ 1.874 1.872 1.872 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 1.942 1.943 1.945 1.943 1.941 1.943 1.94‐ 2.031 2.036 2.044 2.039 2.037 2.034 2.03‐ 2.222 2.220 2.220 2.214 2.214 2.218 2.22‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 2.291 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 2.308 2.314 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 2.312 ‐‐‐‐‐‐ 2.385 2.395 2.390 2.397 2.397 2.398 2.39‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 2.434 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐ 2.474 ‐‐‐‐‐‐ ‐‐‐‐‐‐ ‐‐‐‐‐‐‐ ‐‐‐‐‐‐
Figure: 5.8. The effects of grain size on the spectral response of Trona. (a) 125000‐ 500000 nm, (b) 125000‐ 45000 nm, (c) less than 45000 nm. After Grove et al. (1992).
c
b
a
38 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.1.2. Spectral Characteristics of Evaporites from undisturbed soil samples
Eight undisturbed surface soil samples were collected using soil rings, representing three different environments. One of the soil sample (R01) was collected from southern part of the Lake Magadi, near to the hot springs (Fig: 5.9.a and b). Three samples (Sample No: R22, R23, and R24) were collected from middle part of the Lake Magadi, places where evaporates frequently occur (Fig: 5.9.a and b) While the rest (Sample No: R02, R03, R04, and R05) were collected from an alluvial plain, which is situated in northern part of the Lake Magadi (Fig: 5.9.a and c).
(b)
(c)
(d)
(a)
Figure: 5.9.a) The map of the places where undisturbed soil samples were collected, b), c), d) surfaces of the ground with the sample surfaces before and after dried.
39 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
As the temperature changes many hydrous evaporite minerals gain or lose water forming other minerals with different hydration states (Drake, 1995). Therefore two different spectra were acquired in each sample within SWIR wavelength region at two stages in order to see the compositional variation in spectra. One SWIR reflectance spectrum from the surface of wet sample (Normally indicate by letter “W” after the sample number) and one SWIR reflectance spectrum from the surface after drying same sample at 1050C for 2 days (Normally indicated by letter “D” after the sample number). Water content of the sample was also measured during the process in order to see the relation between water content and the reflectance assuming moisture content (water content) of the samples were equally distributed. Third spectrum was collected from each dried sample around 1.5 cm below the surface, to check if there is any difference between the surface and below the surface according to the spectral point of view (Normally indicates by letter “L” after the sample number).
Reflectance
(a) (b)
Figure: 5.10. (a) Normal and (b) Continuum removed reflectance spectra of the undisturbed
surface soil sample R01.
All wet sample spectra show (Fig: 5.10, 5.11 & 5.12) more deep and wide molecular water absorption feature at 1.9 µm and hydroxyl absorption feature at 1.4 µm wavelengths. After drying, depth of the absorption feature as well as width of the absorption feature has decreased. Depth of the absorption feature and water content of the sample shows weak positive relation (Fig: 5.13). However, number of data points is not enough to give a strong argument.
40 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Heating to 1050C produces more significant changes in the spectral response that must represent the transition to trona from soil sample, R02, R04, R05, R22 and R23 (Fig: 5.11 and 5.12 ). All of them showed absorption features which are characteristics to the trona. However, this trona formation is only limited to the surface. Because any of the reflectance spectra that were taken from 1.5 cm below the soil surface of each sample doesn’t show any absorption features related to the trona. They show one important absorption feature other than 1.4 µm and 1.9 µm, around 2.2 µm, showing probably, absorption by Al‐OH in clay minerals.
Reflectance
(a) (b)
Figure: 5.11 (a). Normal and (b) Continuum removed reflectance spectra of the
undisturbed surface soil sample that were collected from alluvial plain in northern part
of the Lake Magadi.
R01W, wet soil surface spectra of soil ring 01, near to hot springs, exhibits two small absorption features at 1.7 µm and 2.3 µm(Fig: 5.10). Absorption feature at 1.7 µm has disappeared after drying and 2.3 µm feature has changed to doublet increasing depth and width of the feature. The spectrum
41 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
from 1.5 cm below the surface doesn’t show absorption at 1.7 µm but it shows absorption at 2.3 µm similar to wet sample. It shows another small absorption at 2.2 µm probably due to the clay minerals within the sample. According to the absorption feature variation it can be concluded that, probably the doublet near 2.3 µm has formed as a result of phase changes of the mineral. In addition to that, this phase change is limited to the surface which is exposed to the atmosphere. However, this dried surface does not represent the trona minerals and probably it could be another mineral phase.
Samples that were collected from the alluvium plain, showed absorption feature at 2.2 µm (Fig: 5.11). This feature is the only remaining with depth except features around 1.4 µm and 1.9 µm. Sample No: R03D shows different spectral features from others and it shows absorption at 2.35 µm.
Reflectance
(a) (b)
Figure: 5.12 (a) Normal and (b) Continuum removed reflectance spectra of the trona
evaporites from middle part of the Lake Magadi study area.
42 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Samples that were collected from trona evaporates (Sample No: R22, R23, and R24) showed common absorption feature at 2.3 µm (Fig: 5.12). While dry surface sample shows absorption at 1.50 µm, 1.74 µm, 1.94 µm, 2.03 µm, 2.22 µm, and 2.39 µm, which are characteristics to the trona, wet surface and dry “surface” ( below 1.5 cm) doesn’t show any interesting features except features at 1.41 µm, 1.9 µm, 2.30 µm and 2.45 µm.
0.8
0.7 1.9um abs depth 0.6
0.5 1.4um abs depth 0.4 Linear (1.9um abs 0.3 depth) 0.2 Linear (1.4um abs depth) 0.1
absorption band depth at 1.9um at depth band absorption 0 20.00 25.00 30.00 35.00
water content % Figure: 5.13. The graph showing the relation between water content and 1.9‐μm and 1.4 µm absorption features depth.
43 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The TIR spectrum (2.5 µm – 15.0 µm) for the carbonate minerals trona, exhibits a single narrow feature near 11.7 µm (Crowley and Hook). The USGS spectral library trona spectrum exhibits another two clear features at 6.66 µm and 9.53 µm which are almost similar to the collected trona sample,A001 (Fig: 5.14). R02W and R23W, wet surface soil samples show above characteristic trona feature at 11.71 µm. After drying, this feature is shifted towards high frequencies at 11.50 µm. Not only this feature, but also feature at 6.66 µm becomes wider and deep after the drying at 1050C. In addition to that all dried undisturbed surface soil samples show broad feature between 3 µm to 6 µm with sharp peak at 4.27 µm. Probably feature around 4.27 µm is due to atmospheric CO2 which is not completely removed by reference spectra. Probably these changes can be due to the phase changes of minerals according to the high temperature heating.
Trona and other surface evaporites
Figure: 5.14. TIR (3 μm‐15 µm) emission spectra of Trona and Surface Evaporites with USGS trona reference spectrum
44 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
TIR Spectra acquired from undisturbed wet soil samples (R03W, R04W, and R05W) that were collected from alluvial fan in northern part of the Lake Magadi, shows very distinct sharp features near, 6.14 µm, 9.96 µm, 11.91 µm and two broad features centred at 7.44 µm and 14.31 µm (Fig: 5.15). These features however disappeared after drying. SWIR spectra that were acquired from same dry samples don’t show clear difference especially between R03D and R05D although TIR spectra of this samples show clear differences (Fig: 5.15 and Fig: 5.11). Undisturbed soil sample, R05 shows distinct features clearly than other samples. Even though it is interesting to study the features to recognize mineral phase of the sample, the lack of previous knowledge and available time has compelled the researcher to stop the studies at this level.
Trona and other surface evaporites
Figure: 5.15. TIR (3 μm‐15 µm) emission spectra of Trona and
Surface Evaporites
45 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.2. Siliceous rocks
Siliceous rocks were classified using reflectance spectra, XRD pattern, and description of hand specimens and contain; a) Laminated chert (sample no: P017, Fig: 5.16). The holes on the top of the sample are casts of trona crystals presented in the sodium silica gel which later became chert. b) Laminated green chert (Sample no: P024, Fig: 5.17). The chert originates as sodium silicate gels which crystallize to the ephemeral mineral Magadiite; this then dehydrate to chert (Eugster, 1969). c) Pillow chert (Sample no: S047 & P035, Fig: 5.18). On tectonically activated fault blocks, the horizon with microbial biostromes and a silica rich matrix collapsed by liquefaction and was extruded upward to the surface as a breccia mask. The material passed through the overlying bedded and denser cherts as round to oval bodies. The phenomenology is similar to volcanic pillows. The internal structure consists of brecciated and silicified microbial mats that are thinly rolled and deformed fluidly in the outer zones of the pillows (Behr and Rohricht, 2000). d) Dyke cherts (Sample no: P019_01 & P019_02, Fig: 5.19). According to Behr and Rohricht (2000), it is noted that the dyke represent cracks induced by earthquakes that were intruded as sedimentary veins by the soft silica materials. Like the pillows, the internal fabric of the dikes consists of breciated microbial chert that is cemented by the second chert generation of lower viscosity. e) Green beds (Sample no: B015 and P030, Fig: 5.20). This unit contains >90% of all cherts of the Magadi region. The greenish sedimentary matrix of the Green beds consists of tuffs and pyroclastic silts. The volcanic detritus of the silt and tuffs serves as source of silica. The colour of this rocks results from a fibrous and colloform substrate that coats all grains, crack and pore surface of the detritus. It is yellowish‐ brown to olive‐green and enriched in microbial structures and calcites (Warren, 2006). The presence of calcite was conformed by using 6% HCl acid. f) Quartz chert with the colloform habit (Sample no: S008 and P019_02). Eugster(1969) noted that this reticulated cracks (also called crocodile skin) are presumably caused by the shrinkage associated with the Magadiite‐chert conversion. g) Cherts with small cavities filled with carbonate materials (Sample no: P006_01 and S032). Chertification of carbonate involves the precipitation of pore filling silica cement as well as the replacement of carbonate by silica. Both may occur before and/or after carbonate cementation of the host rock (Hesse, 1989). That carbonate materials react with 6% HCl acid. h) High Magadi beds (Sample no: S037, P009, Fig: 5.21) are a sequence of laminated clays and
silts of greatly variable thickness. High Magadi beds show relatively high content of SiO2
(61.49%), Al2O3 (12.42%), and Na2O (9.70%) (Behr and Rohricht, 2000). i) Magadiite (Sample no: P032) and Kenyaite (Sample no: P036) are intermediate products of Lake Magadi Chertification process, which is classified as a one group for this study; and j) Diatomite (Sample no: P028) which is a silica rich material made up of organic silisification process.
46 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 5.16. Crystal casts in Chert plates from Figure: 5.17. Laminate Green Chert from Northern
Southern end of Lake Magadi. The arrow like part of the Lake Magadi. structures may be casts of trona crystals, or they might have formed during the conversion of Kenyaite.
Figure: 5.18. Pillow Chert: Near to the Magadi Figure: 5.19. Chert dykes from North‐eastern part town. of the Lake Magadi.
Figure: 5.20. Green beds from southern part of Figure: 5.21. High Magadi Beds from southern part
Lake Magadi. of Lake Magadi. Numerous Magadiite beds and layers are inter‐bedded with silts and clays
47 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The location map of siliceous rock samples is shown below.
(a) (b)
Figure: 5.22 (a) The location map of the Chert samples and (b) other silica rich
samples.
5.2.1. Spectral Characteristics of Chert
Sample No: P019_01, P019_02, S008, P006_01, P035, S047, S032, P017 and P024 which were categorized as chert, showed same type of spectra (Fig: 5.23). Four main absorption features can be recognized at 1.42 µm, 1.91 µm, 2.21 µm and 2.46 µm in each spectrum. All the samples contain molecular water as well as hydroxyl bonds due to 1.9 µm molecular water absorption feature and 1.4 µm hydroxyl ion absorption feature (Clark et al., 1990). Aines and Rossman (1984) showed that the absorption at 2.22 µm is assigned to Si‐OH groups in silica which are not interacting with other H2O groups, by analyzing amethyst, chalcedony, opal, and citrine with synthetic quartz. Other than above mentioned four (main) common absorption features, Sample No: P019_01, S008 and P006_01 showed another two absorption features at 1.1607 µm and 1.2532 µm. Causes of these two absorption features were not studied in detail.
48 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Chert Chert
Reflectance
(a) (b)
Figure: 5.23 (a) Normal and (b) Continuum removed reflectance spectra of Magadi‐ Chert Series.
5.2.2. Spectral Characteristics of Green beds, Diatomite and High Magadi beds
Green beds (Sample No: P030 and B015), diatomite (Sample No: P028), and High Magadi beds (Sample No: S037, P009 and S019) shows more or less similar spectra as that of chert with slight variations (Fig: 5.24). 1.4 µm hydroxyl ion absorption feature and 1.9 µm molecular water absorption feature are present in green beds, High Magadi beds and in diatomite (Fig: 5.24). Green beds and diatomite show two distinct small absorption features at 2.224 µm and 2.305 µm superimposed with 2.2 µm broad absorption feature (Fig: 5.25). But sample from High Magadi bed (S037) shows spectral features similar to chert (Fig: 5.24). X‐ray diffraction pattern of the Sample No: S037 revealed that the sample is composed of amorphous material and probably amorphous silica. However, Hunt (1973) indicate that the absorptions in the 2.0 µm to 2.5 µm region is combinations of the OH stretch and either the fundamental Si‐OH stretch or the Al‐OH or Mg‐OH bending modes. On the basis of
49 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
empirical evidence, the location of the most intense feature appears to depend upon whether aluminium is present, in which case the absorption occurs near 2.2 µm or whether magnesium is present, in which case the absorption occur at 2.3 µm (Hunt et al., 1973). XRD patterns of several SiO2 rich materials including chert (P006_01), diatomite (Sample No:P028, Fig: 5.26), Kenyaite (P036) and
Green beds (B015, P030) show the existence of SiO2 in the material quantitatively and the reflectance spectrum of each material shows the presents of 2.2 µm absorption feature which is responsible for the Si‐OH bonds of the material (Fig: 5.27). In addition to that, small absorption feature at 2.36 µm is responsible for the CO3 of the material. That was formulated after analyzing each rock/soil sample with 6% HCl acids and spectral absorption features of each material. Kenyaite sample (P036) however, shows CO3 absorption feature, but it did not react with HCl giving clue of another form of carbonate material except CaCO3. XRD pattern of sample P030 and B015 identified secondary mineral
CaCO3 other than SiO2 (Fig: 5.27).
Reflectance
(a) (b)
Figure: 5.24 (a) Normal and (b) continuum removed reflectance spectra of Green beds (P030 & B15), High Magadi beds (S037 & P009) and diatomite (P028).
50 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Reflectance
Figure: 5.25. Comparison between two Figure: 5.26. Diatomaceous earth (diatomite) absorption features. from northern part of the study area, one bank of river.
Reflectance
(a) (b)
Figure: 5. 27 (a) The XRD patterns and (b) reflectance spectra of the several rock samples showing the relation between proportional contribution of mineralogy (eg: SiO2) and the intensities of the XRD pattern and absorption depth of reflectance spectra.
51 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.2.3. Spectral Characteristics of Magadiite
The mineral Magadiite (NaSi7O13(OH)3∙3H2O) was identified in sample no: P032, using XRD. Fig:
5.28 shows the observed and reference reflectance peaks of Magadiite. While most of the reflectance peaks are matched with each other; some of them are deviated a little bit.
Brindley (1969) studying the X‐ray diffraction pattern of the Magadiite concluded that deviation can happen due to the platy morphology and by the reversible swelling and shrinking behaviour of the crystals. In addition to Figure: 5.28. X‐ray diffraction pattern (Red) that he reported that the X‐ray powder pattern of Magadiite sample (P032). Orange colour of Magadiite in normal air condition is indexed lines shows reflectance peaks of the sample with a monoclinic cell with a= b= 7.25Å, c = that was matched with Magadiite reference
15.65Å, β = 96.8, and d(001) = 15.58Å (Baker, peaks. Blue points on some of the peaks
1986; Clark et al., 1990). represent the unmatched peaks with
Magadiite.
Fig: 5.29 display the Magadiite short wave infrared spectrum from 1.0 µm to 2.5 µm. In order to see the changes of spectra due to changes of form, two spectra were acquired from the same sample rock and its powder form. Except for the hydroxyl (1.4 µm) and water (1.9 µm) absorption features, three absorption features can bee seen. Broad absorption feature near 2.22 µm is due to the Si‐OH bond of the material (Aines and Rossman, 1984). Absorption feature at 1.46 µm is unique for Magadiite and Kenyaite sample. No spectra that were collected from soil and rock samples show this 1.46 µm absorption. Teflon which is composed of C (carbon) and F (Fluoride) shows almost similar absorption like that of Magadiite. However, causes of these absorption features were not studied in detail.
(a) (b) Reflectance Figure: 5.29 (a) Normal and (b) continuum removed reflectance spectra of Magadiite rock sample and its powder form.
52 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.2.4. Spectral Characteristics of Kenyaite
The hydrous alkali silicate, Kenyaite (Na2Si22O45 ∙ 10H2O) (Fig: 5.30) occurs as an alteration product of Magadiite. The transformation can be expressed by Equation 2.1.7 (Eugster, 1969). The observed and reference XRD pattern of the Kenyaite is shown in Figure: 5.31. The changes of the observed and reference XRD pattern is possible due to several reasons. The pattern and chemical composition can Figure: 5.30. Kenyaite be changed by variations in the amount of inter layer water, if the physical conditions change (like relative humidity and temperature) or it can be changed by partial or complete exchange of the interlayer sodium ions by protons if Kenyaite comes into contact with water (Beneke and Lagaly, 1983). The sample (P036), that was selected by XRD analysis as Kenyaite also has considerable amount of Silica (SiO2) (Fig: 5.31). The SWIR spectrum also exhibits the broad Si‐ OH absorption feature around 2.2 µm. In addition to that, it shows another two absorption features at 1.15 µm and 1.46 µm. Absorption feature at 1.46 µm is similar and stronger than to that of Magadiite. However, absorption feature at 1.15 µm is only present in Kenyaite rock sample P036 (Fig: 5.32).
(a) (b) P036
Kenyaite
Quartz
Figure: 5.31 (a) XRD pattern of the Rock sample P036 and (b) its Peak list with peak list of Kenyaite and Quartz. It shows that, the material is composed of Kenyaite and Quartz.
(a) (b) Reflectance
Figure: 5.32 (a) Normal and (b) continuum removed reflectance spectra of Kenyaite rock sample and its powder form. 53 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.3. “Unknown Mineral”
Sample No: P012 that was collected from river bank of northern part of the study area was identified as an Oldhamite from X‐ray diffraction pattern. But, Oldhamite has been found exclusively in enstatite‐rich stony meteorites. It is unstable in terrestrial environments and can react easily even with water, releasing rotten‐egg smell of H2S which is a typical reaction of Calcium sulphide with water.
CaS + H2O → Ca(SH)(OH) + H2S ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐eq: 5.3.1
Ca(SH)(OH) + H2O → Ca(OH)2 + H2S ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 5.3.2
As a result it is impossible to survive along the river even if the area is semi arid (Fig: 5.33). The speculation however, is that it is formed as a result of sulfur present during the thermal processing of a calcium rich host material under the extream reducing conditions of the retort (Masona, 2003).
Unknown mineral
(a) Figure: 5.33. “Unknown” mineral P012 (a) and (b) its original location, river bank. Chert bed also can (b) be seen in this Picture.
Table 5.2 lists the observed d‐spacing of sample no P012 and d‐spacings of reference oldhamite,the peak intensities of the X‐ray diffraction pattern and the persentage deviations which were calculated from the equation below,
|d(obs) – d(ref)|/d(obs) * 100 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 5.3.3
The deviations are mainly less than 0.59 percent and for all reflections average 0.22 percent. This is considerably good consistent to say the mineralogy/crystalography of the sample P012 is most likely similar to the mineralogy of Oldhamite.
54 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Table: 5.2. X‐ray Powder Data for Sample no P012 with Oldhamite reference data.
Sample Dffraction Pattern Oldhamite Diffraction Pattern
Pos. d‐spacing Rel. Int. Pos. d‐spacing Rel. Int. Δd/d(obs)% hkl [°2Th.] [Å] [%] [°2Th.] [Å] [%] 27.7005 3.22049 8.53 27.534 3.2395 2.5 0.59 111 32.0435 2.79323 100.00 31.899 2.8055 100 0.43 200 45.7972 1.97969 61.15 45.736 1.9838 57.6 0.20 220 54.1735 1.69310 1.35 54.218 1.6918 1.1 0.07 311 56.7857 1.62127 17.11 56.842 1.6198 16.7 0.09 222 66.5076 1.40476 8.28 66.676 1.4028 6.6 0.13 400 73.5310 1.28803 0.33 73.581 1.2873 0.2 0.05 331 75.5469 1.25755 15.95 75.823 1.2547 16.2 0.22 420
SWIR spectrum of sample No P012 shows several absorption features as shown below; namely, 1.13 µm, 1.42 µm, 1.52 µm, 1.74 µm, 1.91 µm, 1.95 µm, 2.05 µm, 2.22 µm, 2.31 µm, and 2.39 µm
Absorption features at 1.52 µm, 1.74 µm, 1.94 µm, 2.05 µm, 2.22 µm, and 2.39 µm are similar to the absorption features of trona. This spectra shows doublet at 1.91 µm and 1.95 µm and another absorption feature at 2.05 µm which can be overtone of others. It shows Mg‐OH absorption feature at 2.31 µm (Fig: 5.34). But Oldhamite doesn’t show clear absorption feature in SWIR region. As a sulphide mineral, it tends to be characterized by a very sharp absorption feature in the visible wavelength region. Especially Oldhamite has a deep absorption feature centered at ~ 0.49 µm plus a weaker feature centered at ~0.95 µm. The band depth of this feature is ~ 40%. This feature is due to transitions between a high‐energy conduction band, where electrons can move freely throughout the lattice and a low energy valence band, where electrons are attached to individual atoms (Burbine et al., 2002). This feature is unable to see from collected spectra, because, the used VERTEX 70 FTIR spectrometer measure the reflectance from 1.0 µm wavelength onwards.
(a) (b) Reflectance
Figure: 5.34. Normal and continuum removed reflectance spectra of P012 unknown mineral
(“Oldhamite”) rock sample and its powder form.
55 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The author was dissolved sample (P012) a bit of water and could not detect the rotten‐egg smell of H2S which is a typical reaction of Calcium sulphide with water (eq: 5.3.1). When acid is added there is a typical carbonate reaction but still no smell. The solution was aspirated in the flame of the AAS (Atomic Absorption Spectroscopy) and it gave strong yellow sodium colored flame and no crimson red typical Ca color flame. According to the above spectral and chemical analysis, it should be a carbonate mineral with Na. Probably, it could be a mixture. The question then is why it gives XRD pattern same as Oldhamite. Therefore in this research this unknown mineral called as “Oldhamite”.
56 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.4. Erionite Three samples (Sample No: P010_02, B020, and B024) were identified as erionite bearing rock and soils using X‐ray diffraction method (Fig: 5.35). These samples also contain other minerals such as anorthoclase and silica (Appendix: 06).
B024 (a) Figure: 5.35. XRD pattern of the (a) sample B024 and (b) erionite reference (00‐022‐ 0854). Erionite (b)
Reflectance spectra of identified samples exhibit two common absorption features including one deep absorption feature at 2.30 µm and Si‐OH absorption feature at 2.22 µm. With the exception of these two, B024 and B020 show another three absorption features; 1) OH absorption at 1.78 µm, 2) Carbonate absorption at 2.37 µm, and 3) absorption at 2.14 µm. It is difficult to identify diagnostic features for erionite due to the contamination of other minerals.
Reflectance
(a) (b) (c)
Figure: 5.36 (a) Normal and (b) Continuum removed reflectance spectra of rock and soil samples which contains Erionite and (c) their sample locations.
57 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 5.37. Infrared absorption spectra for erionite.
Emissivity
The Thermal Infra‐red spectra have clear absorption features at 2.8 µm and 6.0 µm (Fig: 5.37). Absorption feature at 2.8 µm in the water stretching region are broad and it may be explained if the part of water in this erionite is structurally bound (Harada et al., 1967).
58 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.5. Volcanic Rocks
Igneous rock samples were collected covering entire study area (Fig: 5.38) and they were categorized in to several groups according to their texture and petrology; (a) Basaltic Rocks (Sample No: P020_01, P020_02, P001_02, S027 and P008_01, Fig: 5.39) containing Pyroxene and Plagioclase phenocrysts, (b) Alkali trachytes (Sample No: P027 and S006, Fig: 5.40) which has a fine grained matrix which can be greenish or brownish in colour, (c) Rhyolite (Sample No: P005 and P008_02, Fig: 5.41) is volcanic rock which is composed of (largely) alkali feldspar and free silica with minor amount of mafic minerals. It is characterized by a light brown colour and aphanitic texture. (d) Scoriaceous basalt (Sample No: P033 and P015, Fig: 5.42) is basaltic rock with empty cavities. It is heavier, darker and more crystalline than pumice, (e) Vesicular Basalt (Sample No: P001_01) is basaltic rock with small ellipsoidal cavities which are formed by bubbles of gas trapped during solidification of the rock, and (f) Tuff (Sample No: P010_03, P004_02, and P004_01, Fig: 5.43) is a pyroclastic extrusive igneous rock which is composed of fragments having a diameter less than 2 mm.
(a) (b)
Figure: 5.38. Location map of the igneous rock samples. (a) sample locations related to Fig: 5: 45, (b). sample locations related to Fig: 5.46.
59 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 5.39. Basalt Rocks from Northern part of Figure: 5.40. One of the Alkali trachyte Sample Lake Magadi looking west from near to Little Magadi (looking west).
Figure: 5.41. Rhyolite sample from sample no Figure: 5.42. Scoriaceous basalt sample P015 area P008_02 area‐northern part of Lake Magadi from middle part of the Lake Magadi study area.
Figure: 5.43. Volcanic tuff and colour variation of Figure: 5.44.Basaltic rock (Sample no: P037) with the profile from northern part of the study area plagioclase phenocrysts.
These volcanic rocks had been formed during the middle Miocene, and again in Plio‐ to Pleistocene times. Eruption of voluminous phonolite and trachytic lavas and ash flow tuffs locally filled the rift and overflowed its flanks (Baker, 1986). A general rule for volcanoes of the rift valley near Lake
60 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Magadi is: the younger the volcanism the higher the alkalinity (Warren, 2006). It means younger volcanic materials that were found from study area are more alkaline.
Reflectance
(a) Mg clay minerals (b)
Figure: 5.45 (a) Normal and (b) Continuum removed reflectance spectra of basaltic
rock samples.
Basalts are generally spectrally featureless because of extreme magnetite quenching (Hunt et al., 1974). Weak water absorption feature at 1.9 µm indicate the presence of water (Fig: 5.45). Alkali trachyte (Sample No: S006) shows a weak and moderate hydroxyl absorption feature at 1.4 µm. These 1.9 µm and 1.4 µm absorption features occur only as a result of alteration or weathering of the rock (Hunt et al., 1974). Two absorption features at 2.30 µm and 2.35 µm are observed as a doublet. The calcite introduces carbonate absorption near 2.35 µm (Pontual et al., 1997) and Mg clay minerals show absorption band at 2.30 µm wavelength. This doublet therefore, can be caused as a result of weathering or alteration of basalt and alkali trachyte rocks. Sample No: S027 shows another absorption feature at 2.206 µm. Petrology of the sample can be described as porphyritic basalt with relatively large phenocrysts that is composed of plagioclase feldspar. The 2.206 µm absorption feature reveals the Al‐OH in plagioclase.
61 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Volcanic tuff, lying below the basaltic rock (Sample No: P010_03, P004_02, and P004_01, Fig: 5.43) exhibit six absorption features including 1.4 µm hydroxyl absorption and deep water molecular absorption feature at 1.9 µm.
Reflectance
(a) (b)
Figure: 5.46 (a) Normal and (b) Continuum removed spectra of other Igneous Rock
samples.
Sample No: P010_03, highly porous tuff, shows deepest 2.30 µm absorption feature probably due to its high weathering condition. Two volcanic tuff samples (Sample No: P004_02 and P004_03) were brought from same location but different stratigraphic elevations (Fig: 5.43). Rock sample (P004_03) derived from lower portion of the stratification shows deeper 2.301 µm absorption feature than the rock sample from upper portion (P004_02) (Fig: 5.46 (b)). This could be due to (a) leaching process; ions like Mg2+ can percolate in the lower part or (b) different deposition stages like different volcanic events with different time, different location, etc (Warren, 2006). Un‐weathered Rhyolite (P008_01) and un‐weathered basalt (P008_02) do not show any strong absorption features. They show one weak absorption feature at 1.91 µm (Fig: 5.46). Basaltic rock (Sample No: P037; Fig: 5.44) shows one absorption feature at 2.21 µm except 1.41 µm and 1.91 µm (Fig: 5.46). The rock also contains weathered plagioclase feldspar phenocrysts. It might be responsible for the 2.21 µm absorption feature which is probably related to the Al‐OH.
62 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
5.6. Other Sedimentary Rocks
The sedimentary environment of the Lake Magadi area is characterized by the dominant influence of nearly continuous tectonics movements and volcanism (Baker, 1986). Therefore Magadi basin contains mainly mixture of natrocarbonatite volcanic tuff, clays containing detrital anorthoclase and amphibole, chert, precipitated Magadiite and calcite concretions (Baker, 1986).
Sedimentary rock (Sample P031; Fig: 5.47 (a), see also Fig: 5.47 (c) for location) which is composed of chert pebbles and natrocarbonatite volcanic ash shows mixed absorption features within 2.1 µm – 2.5 µm wave length region (Fig: 5.47 (b)). Wide absorption feature at around 2.2 µm is due to of Si‐ OH groups in chert and absorption at 2.34 µm is most related to the carbonate in the matrix. Spectrum of other sedimentary rocks that were collected from Lake Magadi area show common absorption feature at 2.3 µm indicating Mg minerals which is more common in igneous rock environment (Fig: 5.48). Spectrum of Rock sample S021 and P003 show another absorption feature at 2.37 µm while Spectrum of Rock sample P038 and P006_02 show absorption feature at 2.34 µm which is most related to the carbonate. Reflectance spectrum of rock sample S030, silty clay sedimentary rock shows only one absorption feature at 2.3 µm which is probably related to the Mg‐ OH bond of the minerals of the rock.
Chert
Volcanic ash
(a)
Reflectance
(c) (b)
Figure: 5.47(a) Rock sample and area. (B) Part of continuum removed reflectance spectra of sedimentary rock sample P031.wihle solid line showing mixed spectrum, dashed lines show chert (Si‐OH) absorption feature and CO3 absorption feature of Matrix. (C) Sample location map.
63 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Reflectance
Mg clay minerals (a) (b)
Figure: 5.48 (a) Normal and (b) Continuum removed reflectance spectra of sedimentary rock samples.
64 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6. Mineral Mapping using Multi-spectral and Hyperspectal Imaging methods
Analysis of ASTER and Hyperion data for mineral mapping which is based on determination of the relationship between spectral reflectance and spectral emittance and the mineral composition of the rock and soil units of the interest is discussed in this chapter.
6.1. Surface Mineral Mapping using ASTER
The Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER) operates in three visible and near‐infrared (VNIR) channels between 0.52 µm and 0.86 µm, with 15‐ m spatial resolution; Six short wave infra‐red (SWIR) channels between 1.60 µm and 2.43 µm, with 30‐ m spatial resolution; and five thermal infra‐red (TIR) channels between 8.13 µm and 11.65 µm, with 90‐ m spatial resolution (King et al., 2003; Rowan et al., 2005; Rowan et al., 2006). ASTER‐TIR is the first satellite‐borne multi‐spectral TIR remote sensing system with sufficient spectral, spatial and radiometric resolutions for geologic applications (Ninomiya et al., 2005). The ASTER instrument acquires data over a 60‐km swath whose centre is pointable (Cross‐track ± 8.550 in the SWIR and TIR, while the VNIR pointable up to ± 240) (King et al., 2003). ASTER also has a back looking VNIR telescope with 15‐ m resolution. Thus, stereoscopic VNIR images can be acquired at 15‐ m resolution (Rowan et al., 2006). ASTER images can be acquired at any point of the globe at least once every 16 days in all 14 bands on average and every 4 days in the three VNIR channels (King et al., 2003).
6.1.1. ASTER image processing
Since launch in 1999, the ASTER instrument has performed exceptionally well. No operational faults have been discovered (King et al., 2003). However, analysis of the data have revealed several characteristics not accounted for prior to launch. One of them is called Cross talk phenomena.
SWIR crosstalk is an offset or additive error in radiance due to the anomalous signal leaking into adjacent bands. This is most pronounced for energy falling on detectors for band 4, and a small fraction reflecting on the detectors for band 5 and 9 (King et al., 2003). Its origin and effects have been documented by Iwasaki and Tonooka, (2005). The most likely mechanism of signal leakage from band 4 to adjacent bands is shown in Fig: 6.1. Since the optical absorption of a Pt‐Si schottky detector of the SWIR radiometer of ASTER is small, an aluminium reflection film is attached at the rear of the detectors to double the sensitivity. A silicon substrate is sandwiched between the antireflection (AR) coating and the SiO2 layer. Since the silicon substrate has refractive index about 2.5 times higher than
65 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
that of the coating, light is confined to the silicon substrate when the incident angle is grater than the critical angle. As a result of this, the incident light to band 4 goes to the other bands by multi‐ reflection (Fig: 6.1 (a)). Another stray light is caused by scattering or reflectance at the filter boundaries although the interference filter of each band is completely glued side by side (Fig: 6.1(b)) (Iwasaki and Tonooka, 2005).
(a)
(b) Figure: 6.1. Crosstalk mechanism originates from
(a) band 4 detectors and (b) filter boundaries. After Iwasaki and Tonooka, 2005.
Based on the above mechanism, Iwasaki and Tonooka (2005) have formulated equation of Cross‐talk correction as shown below,
(k) (k) (k) (4) (k) (k) (k) ƒ corrected(x, y) = {ƒ (x, y) – h (x, y) . ƒ (x‐x(k), y – y(k))}/ 1‐ b a c ‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.1
(k) (4) (k) Where, ƒ (x, y) = ƒ (x – x(k), y –y(k)) . h (x, y) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.2 (k = 5, 6, 7, 8,9) and,
(k) (k) (k) (k) 2 (k)2 2 (k)2 h (x, y) = {a / 2πσx σy } exp{‐1/2 [(x /σx ) + (y /σy )} ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.3
(k) (k) (k) (k) Here, ƒ (x, y) is the cross talk component. The parameters, a , σx , σy , x(k) and y(k) are the amplitude, width of the stray light in the cross‐talk and along track directions, and displacement of the ghost image in the cross‐track and along‐track directions respectively (Iwasaki and Tonooka, 2005).
Further more, b(k) and c(k) are,
b (k) = Radiance(4) / Radiance(k) * {Response4’(k) / Response(4)} ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.4
and,
c(k) = Sensitivity(4) / Sensitivity(k) respectively.‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.5
Here, Radiance(k) and Radiance(4) are the input radiance set independently for each band and specially for band 4 during the pre‐flight test. Response4’(k) is assigned for the response of band 4 during the
66 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
measurement of band k. Furthermore, they showed that the correction using default parameters are sufficient for geologic research (Table: 6.1) (Iwasaki and Tonooka, 2005).
Table: 6.1. Crosstalk parameters for Level‐1A (Default).
(k) (k) (k) (k) (k) Band X(k) Y(k) a σx σy b c No: (pixel) (pixel) (pixel) (pixel) 4 ‐ ‐ ‐ ‐ ‐ 1.000 1.000 5 0 65 0.09 28 20 3.282 0.316 6 0 146 0.03 30 26 3.329 0.287 7 0 ‐227 0.02 34 30 3.635 0.285 8 0 ‐146 0.03 30 26 5.224 0.198 9 0 ‐65 0.09 28 20 6.072 0.146
ERSDAC Crosstalk 3 software was applied to the ASTER L1B image to generate corrected ASTER Hierarchical data files (HDF) using input parameters listed in table: 6. 1.
ASTER Level‐1B data contains pixel values of radiance at sensor, and not reflectance. In order to compare multi‐spectral image spectra with reference reflectance spectra directly, the radiance values in the image must be converted to the reflectance and atmospheric correction should be performed to remove or minimize the haze effect. In order to remove haze effect and convert DN values to relative reflectance value of the image, Empirical line correction and Log residual method were used after the radiometric calibration. The calibration to spectral radiance units of the L1B ASTER were obtained using the equation, Radiance = (DN – 1) * Gain. Used gain values are shown in table: 6.2.
Table: 6.2. Gain values for radiance calibration.
Band No: Gain Band No: Gain 1 0.6760 8 0.0417 2 0.7080 9 0.0318 3 0.8620 10 0.0068 4 0.2174 11 0.0067 5 0.0696 12 0.0065 6 0.0625 13 0.0056 7 0.0597 14 0.0052
67 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.1.1.1. Empirical line correction A linear regression is used for each band to equate DN and reflectance. The equation used in empirical method is shown below,
Reflectance (field spectrum) = gain x radiance (input data) + offset ‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.6
For empirical line correction, 2 spectrally feature less field spectra which were collected from relatively homogeneous and large ground target areas, were selected with that of image spectra. When dark target was selected from exposed basaltic rock mountain (Sample no: P015), bright target was selected from high Magadi bed (Sample no: R01).
The Image corrected from Empirical line method was used for mineral mapping in SWIR region and the Image corrected from Log residual method was used for mineral mapping in TIR bands.
6.1.2. Spectral Properties
6.1.2.1. Spectral Reflectance
Representative SWIR laboratory reflectance spectra of the rock and soil samples that were described in chapter 5 were categorized by absorption features that indicate the presents of chert, erionite, basalt, rhyolite, Magadiite, Kenyaite, trona, and “Oldhamite” (Unknown sample). Fig: 6.2 shows the ASTER re‐sampled spectra with its original form in SWIR wavelength region. Even though most of absorption features of earth materials are situated in SWIR region, it is difficult to identify those absorptions after re‐sampling to ASTER image due to the low spectral resolution and low spectral depths of the characteristic absorption features. Most of the rock and soil samples of the study area show Si‐OH absorption feature in the same position as the Al‐OH absorption feature near 2.2 µm. Chert shows clear absorption at band 6 and 7. Trona and chert have clear difference in band 6, 7, 8. Band ratio 7/6 will give positive value for trona while others give negative values according to the spectra in Fig: 6.2. Mixture of quartz and calcite (High Magadi bed), Magadiite, and Kenyaite also show the absorption at band 6 and 7 due to the present of SI‐OH bonds in silica. Erionite shows absorption feature at band 6. According to the general shape and diagnostic absorption position of the spectra, Magadiite and Kenyaite are difficult to identify and mapped from ASTER image; the reason been, the characteristics absorption features of Magadiite and Kenyaite are located at 1.46 µm and 1.15 µm which are not covered by ASTER bands. The absorption feature at 2.3 µm of erionite, High Magadi bed, and “Oldhamite” cannot be easily identified because of the broader spectral band passes (Hellman and Ramsey, 2003). Normally, Al‐OH and Mg‐OH rotational effects associated with clays and other hydroxylated minerals result in absorption in ASTER band 6 (Hellman and Ramsey, 2003; Hewson et al., 2005; Moghtaderi et al., 2004; Rowan et al., 2005; Rowan et al., 2006). In this case ASTER band 6 also shows clear absorption band due to Si‐OH bonds in siliceous materials such as chert, diatomite, and High Magadi bed which are common in the study area. The weathering and alteration product of volcanic rocks of the area also exhibit Al‐OH absorption feature at 2.2 µm due to the Al‐clay minerals (Rowan et al., 2005). Therefore it is difficult to distinguish Al‐clay minerals from silica rich materials only using SWIR wavelength region, especially in this type of area.
68 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
4 ASTER BAND 5 6 7 8 9 4 ASTER BAND 5 6 7 8 9
*
* *
*
*
*
*
*
*
*
(a) (b)
Figure: 6.2. ASTER reclassified end member spectra (*) with its original form
69 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.1.2.2. Spectral Emissivity
Emissivity spectra of the Lake Magadi samples typically show an intense Si‐O vibrational feature which is located 10 11 12 ASTER BAND 13 14 in the 8.0 µm – 10.5 µm wavelength region (Fig: 6.3). The representative chert sample is characterized by emission minima near 8.40 µm and 9.20 µm, separated by a sharp emission maximum at 8.65 µm which is almost similar to quartz. The result of re‐sampling these high resolution laboratory spectra to the 5 ASTER TIR band passes indicates general loss of definition of specific absorption features, except for the main quartz‐ chert reststrahlen feature located in ASTER band 12. The absence of an ASTER band around 9.8 µm resulting from intense atmospheric absorption is an important limitation (Rowan et al., 2006). Because most of the volcanic rocks including rhyolite, basalt, and volcanic tuff show an intense emissivity minimum at around 9.5 µm (Fig: 6.3). However, the general shape differences of the emissivity spectra are useful to broadly categorize rock types. Siliceous rocks typically show intense Si‐O emissivity minima in band 12 relative to band 13 and 14, so that band 13/band 12 or band14/band 12 ratio images are particularly useful for the mapping of sandstone, quartzite, and silicified rocks such as, in this case, chert, diatomite, Kenyaite, and Magadiite (Moghtaderi et al., 2004; Rowan et al., 2006; Tommaso and Rubinstein, 2006). In addition to that, the quartz index Qi also use to identify silicate rocks due to the emissivity minima in bands 10 and 12 and a small emissivity peak in band 11.
Qi = (b11 *b11)/(b10 * b12) ‐‐‐‐‐‐‐‐‐‐ eq: 6.1.7
Qi is expected to be high for quartz and low for K‐feldspars (Ninomiya et al., 2005; Tommaso and Figure: 6.3. ASTER re‐sampled emissivity Rubinstein, 2006). end member spectra with its original form from rock samples representing each groups.
70 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.1.3. Mapping Method
6.1.3.1. Spectral Mapping method
Quantitative spectral mapping techniques, such as the Spectral Angle Mapper (SAM) method, operate by comparing image spectra with field or laboratory reference spectra. It assumes that the data is correctly calibrated to apparent reflectance with dark current and path radiance removed (Meer and Jong, 2003). The spectral similarity between the image spectrum, x, and the reference or field spectrum, y, is expressed in terms of the average angle, α, between the two spectra as calculated for each channel as,
‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.1.8
The result represent by the gray scale rule image representing the angular distance in radian between each pixel spectrum and the reference (mineral) spectrum. Darker pixels in the rule image show greater similarity to reference spectra. Even it was developed for hyper‐spectral images; it can be used in lower resolution system (Tommaso and Rubinstein, 2006). How ever some constraints must be considered due to ASTER spectral resolution. The overall shapes of the ASTER image spectra are different from the ASTER laboratory spectra because of the effects of several factors that influence the spectra, but are obviated in the laboratory. They include the presence of variable mineral mixtures, grain‐size variations, residual atmospheric absorption features, desert varnish top of the rock and soil surface, and calibration errors of laboratory spectrometer and/or ASTER instrument. In addition to that, although, spectra of pure or high identified mineral content rock and soil sample used as representative spectra, the image spectra represent 30‐m pixels within which the accepted mineral content may be less concentrated (Rowan et al., 2006). According to the spectral differences between ASTER laboratory and ASTER image spectra, Image spectra were selected as reference for spectral Angle Mapper (SAM) method.
Five different types of main end member image spectra were selected for spectral angle mapper method representing five main phases of geo‐chemical evolution of evaporite minerals in the area with help of theoretical knowledge, features of collected reference spectra and information of the visited Field locations (Fig: 6.4 (a)). The mean spectra were acquired from five known locations representing alkali trachyte/basalt, hydrous Na‐Al‐Si gel, brines, evaporites and clay minerals (Al) using ROI (Region of interest) tool in ENVI. The general shape and characteristic absorption features which were studied in reflectance spectroscopy of rocks and soils were also considered to select mean spectra for spectral angle mapping method. Threshold value 0.05 radian was used for the SAM techniques in order to increase the accuracy of classification. Classified image is shown in Fig: 6.4 (b). Dark green colour represents the alkali trachyte/basalt outcrop area. While, pink colour represent the hydrous sodium aluminous silicate gels which are in contact with trachytic debris, light green colour and light blue colour represent the evaporites and brines in the area respectively. Al‐clay minerals shown in brown colour and unclassified area is shown in black.
71 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(a)
(b)
Figure: 6.4. (a). Selected reference image spectra for Spectral angle mapper method. (b).Results of
spectral mapping method: ASTER SAM classification showing Alkali trachyte/basalt (Dark Green), hydrous sodium aluminous silicate (Pink), evaporites (light blue), brine (light green) and Al‐clay minerals (brown).
6.1.3.2. Band Ratio Method
Band ratio is a very simple and powerful technique in remote sensing. Basic idea of this technique is to emphasize or exaggerate the anomaly of the target object (Tommaso and Rubinstein, 2006) according to their reflectance variation of the bands. Band ratio 11/12 was used to map the silica rich material, chert of the study area according to the shape of the emissivity spectrum of chert. Density slicing method was applied to delineate the places where there is a high probability of occurrence of quartz‐chert. Fig: 6.5 Shows the band ratio image and interpreted chert image map of the Lake Magadi area. TIR part of the ASTER image shows line striping and this line striping remains in band ratio map but it doesn’t show in SWIR region.
72 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(a) (b) (c) Figure: 6.5. (a) ASTER TIR band ratio 11/12, (b) Classified Quartz‐chert (Red) map after density slicing overlain on ASTER band ratio 11/12, (c) ASTER SWIR false colour composite 4:7:9 (R:G:B) image with extent as same as (a) and (b) to see the spatial distribution.
Bodies of water have a rather different response to radiation than that of water bound up in the molecules of minerals. In the near infrared region water acts almost like a perfect black body, and absorbs virtually all incident energy (Fig: 6.6) (Drury, 2001). If the surface material is covered by very thin water body, such as in evaporitic brine environment, reflectance of the SWIR region is Figure: 6.6. The typical reflectance curves of dry soil, vegetation attenuated by that water surface and clear water. After Drury, (2001). resulting very low reflectance similar to water. Therefore it is difficult to distinguish and map clear water body and trona in this region using the reflectance properties of ASTER SWIR region. In the visible range, water shows high
73 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
reflectance than SWIR region. Therefore ASTER band ratio: band 2/ band 1 was used to map clear water body and trona area within the lake (Fig: 6.7(b)).
(a) (b) Figure: 6.7. (a) ASTER VNIR band 1:2:3N (R: G: B) false colour composite (b) classified water area map (blue) overlain on ASTER band ratio 2/1.
Combination of all the outputs that were derived from ASTER image, were used to create the surface mineral map of the entire study area including other associated members like water bodies. The summary of ASTER image surface mineral mapping process is illustrated in Fig: 6.8.
Water Body VNIR Band Ratio 2/1 Evaporites Series (Trona)
Trachyte/ Basalt
ASTER Brine and Evaporites Spectral angle SWIR mapping Clay minerals
Na‐Al‐Si gel
TIR Band Ratio 12/11 Quartz‐Chert Series
Figure: 6.8. Schematic diagram of ASTER image mineral mapping.
74 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(a) (b)
Figure: 6.9. (a) Mineral abundance map that was derived from ASTER image using SAM and Band ratio techniques with (b) Published geology map of the area (Baker, 1958; Baker, 1986).
75 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.2. Surface mineral mapping using Hyperion
Hyperion is the first earth‐orbiting imaging spectrometer which was launched onboard the Earth Observing 1 (EO‐1) satellite on November 21, 2000, operating across the full solar‐ reflected spectrum with nominal spectral coverage from 0.4 μm‐2.5 µm and 10 nm sampling and spectral response functions (Green et al., 2003). The radiometric range of Hyperion spans from zero to the maximum Lambertian reflected radiance with 12 bits of digitization (Green et al., 2003). The Hyperion push broom instrument captures the image frame spectra from an area of 30 m along‐track by 7.7 km cross‐track (Pearlman et al., 2003). Hyperion has a one telescope, slit and two grating spectrometers: one spectrometer with 70 channels in the VNIR wavelength region (0.4 µm‐ 1.0 µm) with silicon detector array and one spectrometer with 172 channels in SWIR wavelength region (0.9 µm – 2.5 µm) with an HgCdTe detector array (Biggar et al., 2003). The 242 total channels include 21 channels in a region of overlap between about 0.9 µm and 1.0 µm (Hubbard et al., 2003). The SWIR focal plane is cooled to about 110 K by a cryocooler in order to improve the SNR (Signal to Noise Ratio) and enhance the stability. The Signal‐to‐noise ratio (S/N) of VNIR is about 161/1 while in SWIR is about 40/1 (Hubbard et al., 2003). A single telescope and slit are used with a dichroric filter to spit the signal to the two spectrometers (Biggar et al., 2003). Radiation from the slit is split into two beams for each spectrometer using a beamsplitter. Each path is focused onto a row of focal plane array (FPA) pixels (the spatial dimension of the FPA) and the grating disperses the scene spectra onto the other dimension (the spectral dimension) of the 2‐dimensional FPA. As the spacecraft flies over the scene, an image is formed with the spatial pixels imaging the scene in a pushbroom manner. A hyperspectral “image cube” of the scene is formed by the image data from the various spectral pixels. The Hyperion instrument consists of 3 physical units, (1) the Hyperion Sensor Assembly (HSA), (2) the Hyperion Electronics Assembly (HEA), and (3) the Cryocooler Electronics Assembly (CEA). A layout of the Hyperion instrument is shown in Fig: 6.10 (Pearlman et al., 2003).
Solar calibration Baffle Spectrometers
Telescope/Optics
Cryocooler
Figure: 6.10. The Hyperion instrument.
76 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Hyperion has common in‐flight radiometric calibration systems for the two spectrometers. Two independent strings of two incandescent lamps produce a reference signal which can be compared to both solar and lunar calibrations. The solar calibration is a diffuse white paint applied to the back side of the instrument aperture cover. The cover is set to a 370 angle, and the spacecraft is oriented to illuminate the diffuser surface with the direct solar beam (Biggar et al., 2003). Lunar calibration does not involve diffusers, since the instruments view the moon directly through their earth‐viewing optical path (Ungar et al., 2003). Spectral calibration of Hyperion system is largely determined through the presence of atmospheric spectral features, such as the O2 absorption feature at 762 nm, and ground targets with well defined distinct spectral characteristics (Ungar et al., 2003). The Hyperion data is initially processed by EO‐1 product generation system (EPGS) and distributed at different processing levels. Commonly available levels of correction are Level 1R and Level 1G. Level 1R is radiometrically corrected with no geometric correction applied. The image data of this product are provided in 16‐bit radiance values. Level 1G is terrain corrected and provided in 16‐bit radiance values (Beck, 2003). The Level 1 Radiometric product has a total of 242 bands but only 198 bands are calibrated. Because of an overlap between the VNIR and SWIR focal planes, there are only 196 unique channels. Calibrated channels are 8‐57 for the VNIR, and 77‐224 for the SWIR. The reason for not calibrating all 242 channels is mainly due to the detectors' low responsivity. The bands that are not calibrated are set to zero in those channels (Beck, 2003).
6.2.1. Pre-Processing of Hyperion Image
The Hyperion image used in this study is Level 1R, EO1H1680612008182110KF. “EO1” stands for the Earth Observing 1 satellite and “H” stands for Hyperion. The numbers, 168 and 061, are the WRS path and row respectively. 2008 is the year of image acquisition while 182 is the Julian day of acquisition (1st of July, 2008). The first “1” following “182” indicates that the Hyperion sensor is on. The second “1” indicates that the ALI sensor is on. The following “0”, indicates that the AC sensor is off, while “K” is a code for the pointing mode and “F” is the code for the scene length. There are three files that came with the distributed scene: the meta data (.met), the image (.L1R), and the header file (.hdr). The image file is stored in band interleaved (BIL) order. To read the image file using ENVI software, it has to be converted in to the ENVI standard file format. The “Hyperion tools” module in ENVI designed to facilitate the use of Hyperion data in ENVI was used. Its most basic functionality is to convert Level 1R HDF and Level 1G/1T HDF and GeoTIFF data sets in to ENVI standard format files that contain wavelength, full width half maximum (FWHM), and bad band information. In addition to that two special functions inbuilt in Hyperion tools were also used during the conversion. ENVI mask image was created categorizing good data (as “1”) and bad data (as “0”) for the dataset that suppresses the data acquired by the malfunctioning detectors using Output ENVI mask image. Each detector in Hyperion’s pushbroom array has slightly different band centers and Full Width Half Maximum (FWHM) values for each band. The secondly applied Option called “Interpolate data to common wavelength set” performs a linear interpolation across all detectors on a pixel by pixel, spectrum by spectrum, band by band basis to a common set of wave lengths (ENVI, 2008).
The Hyperion pushbroom scanner has two‐dimensional detector array, and each array contains hundreds of thousands of detector cells. The VNIR spectrometer uses a 70(spectral) * 256 (spatial)
77 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
pixel section of the 128 * 256 silicon focal plane array (FPA) and SWIR spectrometer used a 172 (spectral) * 256 (spatial) pixel section of the 256 * 256 mercury‐cadmium‐ telluride array (Pearlman et al., 2003). Therefore, there is a high possibility of something going wrong with a few of these cells. Even a single dud detector cell can cause confusion for the further processing. It can be caused a single‐channel noise spike in one pixel out of practically every scan‐line in the data set showing some kind of vertical striping down the image. There are degrees of dudness. Some dud cells have no response or are very noisy all the time while others just have a moderately different response to the “normal” cells (Mason, 2002).
To remove column (vertical) stripes and to repair “bad” cell values, the Pushbroom Plugger and Pushbroom Striper in MMTG‐A module that was developed by CSIRO Division of Exploration and Mining were used. First, the Pushbroom Plugger was used to correct seriously dud detector cells. It identifies dud cells by looking for outliers in mean response and/or variance in response, given a suitably large image (Mason, 2002). In a calibrated instrument, a detector cell may be termed “bad” if it responds differently to the other cells in the entire detector. This module can identify “bad” cells according to their mean response and variance. A cell is defined “bad” if its response has a mean/or variance that is significantly different to the overall rows, assuming that each cell in the detector row is receiving the same stimuli (Mason, 2002). After identifying “bad” cells module replaces the value of a “bad” cell by interpolation with its nearest left and right specially row neighbours. If the cell has good neighbours on one side only, the replacement is done by extrapolation.
The Pushbroom Destriper was applied secondly, in order to improve moderately dud cells that were not sorted out by the instrument’s most recent calibration. As mentioned before, Hyperion scanner has hundreds of thousands of cells and it is hard to find an exact and stable radiometric gain for lot of them. The temperature differences also might change the detector’s response unevenly (Pearlman et al., 2003). Even if these cells are calibrated properly in the lab, they might respond differently to other cells at signal levels that are different to those that were used in the calibration. Calibration can be divided in to two groups mainly: Pre‐launch calibration and On‐orbit calibration. Pre‐launch calibration included extensive laboratory observations using both lamp‐based and solid‐state detector based measurement. The lamp based and solid state detector based calibrations have showed a 5% to 15% differences in absolute values but similar spectral response profiles (Pearlman et al., 2003). Solar, lunar, and earth‐surface‐ observing “vicarious” measurements (Biggar et al., 2003) have also been used for the on‐orbit calibration. In addition to the absolute calibration, certain features were observed during the pre‐launch calibration. Those were categorized and then removed as part of the image processing to create radiometrically corrected Level 1 data. These included both a SWIR “echo” and a SWIR “smear” (Pearlman et al., 2003). Like the Plugger, Destriper works on one row of detector cells at a time. Given that the image is long enough, it is expected that each cell in a row will have measured the same aggregate surface coverage as those of all the other cells in a row. Hence it is expected that the mean and standard deviation will be the same for all cells in the row. The mean and standard deviation are calculated for each cell and for each row of cells in the detector (Mason, 2002).
78 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
This module were applied to correct gain‐ and offset of the data assuming response mean, MY and standard deviation, SDY of a cell is same as to the response mean, MX and standard deviation, SDX of the row.
If the true value is X and measured value is Y,
Y = A*X + B ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.1 Where, A and B are gain and offset respectively.
MY = S[Y]/N = S[A* X + B]/N = A* S[X]/N + {N * B/N}
= A * MX + B ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.2
2 SDY = sqrt(S [(Y – MY) ]/N) 2 = sqrt(S [(A* X + B – A* MX‐ B) ] / N) 2 = A* sqrt(S [(X – MX) ] / N)
= A* SDX ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.3
Therefore: A = SDY / SDX
B = MY – A* MX And, finally the full linear correction (gain and offset) is given by:
X = (Y – B)/ A ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.4
Fig: 6.11 shows the how effectively the column striping and Bad cells of the Hyperion Image were removed after applying Pushbroom Pugger and Pushbroom Destriper modules.
(a) (b) (c) Figure: 6.11. Part of the Hyperion image of the study area; a) after applying interpolation using ENVI Hyperion tool, b) after removing bad bands and cells using Pushbroom Plugger and c) after removing column stripes using Pushbroom Destriper module.
79 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Quality of the pre‐processed data was checked by using MNF (Maximum Noise Fraction) algorithm in ENVI. The MNF algorithm is a method for ordering data cubes in to components of image quality using a cascaded principle components transform. That selects new components in order of decreasing signal‐to‐noise ratio (Meer and Jong, 2001). The series of output Images of the MNF showed increment of quality of the image after the pre‐processing (Fig: 6.12).
Figure: 6.12.a. First five MNF images before the pre‐ processing showing the effect of column stripes
Figure: 6.12.b. First five MNF images after the pre‐ processing showing increment of quality.
For the spectral analysis of Hyperion data, Band 8 (wavelength 426.81 nm) to Band 220 (wavelength 2355.20 nm) spectral subset (158 bands) and spatial subset covering the study area were selected. The Radiometrically corrected radiance image is then stored as a 16‐bit signed integer values dividing a scaling factor of 40 for VNIR bands and scaling factor 80 for SWIR bands.
VNIR (band 8‐57) = Digital Number/40 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.5.
SWIR (band 79‐220) = Digital Number/ 80 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.6.
6.2.2. Calibration to reflectance The most critical and difficult step in most imaging spectrometer data analysis methods is to convert radiance at sensor data to reflectance in order to compare directly the laboratory spectra with image spectra for rock and mineral identification. For this study three relative and one absolute atmospheric correction methods were used to select best reflectance data to map the surface mineralogy.
6.2.2.1. Relative reflectance In this type of methods, reflectance is measured relative to a standard target from the scene. Relative atmospheric methods used are internal average relative reflectance correction, log residual correction and empirical line correction.
80 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Internal average relative reflectance correction Internal average relative reflectance (IAR reflectance) is calculated by determining an average spectrum for an entire scene and dividing each spectrum in the dataset by that average spectrum. The resulting spectra represent reflectance relative to the average spectrum. It may cause, however, artefacts due to the degree of contribution of some cover types which has strong absorption features and it may reduce information from image reflectance spectra (Meer and Jong, 2001).
Log residual correction The log residual correction was performed by MMTG‐A module in ENVI masking water body of the area using Outlier mask generation option in same module. The log residual algorithm is based on a simple model that relates a radiance spectrum, Xi(b) to the corresponding reflectance spectrum ri(b).
Xi(b) = Ti ∙ ri(b) ∙ I(b) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.7 Where, Ti is a topographic factor which is constant for all wavelengths while different for each pixel. I(b) is an illumination factor, which is constant for the whole scene and varies across wavelengths. “i “ and “b” represent the pixel (i = 1….(ns*nl)) and spectral channels (b = 1…nb) of the image cube respectively.
This module, first removes the effect of topography and shading by normalizing each spectrum albedo. Then the second step removes the atmospheric absorptions that are common to the whole scene (Mason, 2002).
Empirical line correction The functionality of empirical line correction method is introduced in section 6.1.1. High Magadi bed (R01) and basaltic rock (P015) laboratory spectra were selected with that of image spectra to derive the gain and offset for the correction. This calibration was done for spectral subset of the image covering wavelength 2000 nm‐ 2355 nm range.
6.2.2.2. Absolute reflectance Flaash atmospheric model Flaash absolute reflectance atmospheric correction model can be described by a standard equation for a spectral radiance at a sensor pixel, L , that applies to the solar wavelength range (excluding thermal emission) and flat, lambertian materials or their equivalents (ENVI, 2008).
L = {(Aρ)/(1‐ρeS)} + {(Bρe)/(1‐ρeS)} + La ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 6.2.8
Where, ρ is the pixel surface reflectance. ρe is an average surface reflectance for the pixel and the surrounding region. S is the spherical albedo of the atmosphere. La is the radiance backscattered by the atmosphere. A and B are coefficients that depend on atmospheric and geometric conditions but not on the surface. While the first term of the above equation corresponds to radiance that is reflected from the surface that is scattered by the atmosphere into the sensor. The values of A, B, S and La are determined from MODTRAN4 calculations that use the viewing and solar angles and the
81 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
mean surface elevation of the measurements, and they assume a certain model atmosphere, aerosol type and visible range. The parameters used for Flaash module are summarized in table: 6.3.
Table: 6.3. Flaash input parameters applied on Hyperion radiance image. Parameter Used value/type Parameter Used value/type Sensor type Hyperion Initial visibility 40 km Pixel size 30 m Aerosol model Rural Ground elevation 0.738 km Aerosol retrieval 2‐Band (K‐T) Scene centre latitude ‐1.994429 Water absorption feature 1135 nm Scene centre longitude 36.277371 Spectral polishing Yes Sensor altitude 705 km Width (No of Bands) 9 Acquired date 2008/06/30 Wavelength recalibration Yes Acquired time 07:23:05 Output 10000 Atmospheric model Tropical MODTRAN resolution 15 cms‐1 Atmospheric zone Rural MODTRAN multiscattering Scaled DISORT 8 stream model
CO2 390 ppm Aerosol scale height 2.00 km
The Flaash module was performed with and without spectral polishing. The output image that was derived with spectral polishing was used for further analysis after analyzing reflectance spectra.
Chert IARR Chert Empirical
(a) (b)
Chert Log res Chert Flaash
(d) (c)
Figure: 6.13. Image reflectance spectra from chert area after performing (a) IAR reflectance, (b) Empirical line, (c) Log residual and (d) Flaash atmospheric correction method for same image data. Note: Empirical line correction was done only for the small wavelength range due to absence of field spectra covering entire wavelength range of the image
82 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Trona IARR (b)
(a) Trona Empirical
Trona Log res Trona Flaash
(c) (d)
F igure: 6.14. Image reflectance spectra from trona area after performing (a) IAR reflectance, (b) Empirical line, (c) Log residual and (d) Flaash atmospheric correction method for same image
data. Note: Empirical line correction was done only for the small wavelength range due to
absence of field spectra covering entire wavelength range of the image
Fig: 6.13 and Fig: 6.14 show the image reflectance spectra for chert and trona after performing above mentioned four atmospheric correction methods. The Flaash atmospheric correction method gives reflectance spectra more or less similar to the laboratory spectra of known locations than other methods. Even though, general shape of the spectra of log residual is relatively similar to the reflectance of Flaash, the log residual enhances more noise due to noisy original hyperion image. The Maximum noise fraction (MNF) and pixel purity index (PPI) were used for each atmospheric corrected image to determine the inherent dimensionality and to identify extreme end members of the image data. The image that was corrected by Flaash module showed end members, however, other atmospherically corrected images, didn’t show end members due to its noise. Therefore output image from Flaash atmospheric correction was selected for mineral mapping process. After the atmospheric correction, Effort polishing was performed to reduce the systematic linear errors in the data by finding a reasonably large set of calibration pixels for which the “true” spectra can be estimated. It calculates a correction gain and/or offset using least‐squares regression between the calibration pixels (true) and actual spectra (Mason, 2002). 1000 calibration spectra with 14 polynomial order were used to correct the data.
83 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.2.3. End member collection
The first process in analyzing the Hyperion data was to perform a maximum noise fraction (MNF) transformation to determine the inherent dimensionality of the image data. The MNF transform uses two principle component transformations. While the first transformation decorrelates and rescales the noise in the data, the second is performed for the noisiest data. Based on MNF results, the lower order MNF bands are usually set aside and the higher order bands are selected for further processing. These were used for selection of pure end members. The pixel purity index (PPI) in MMTG‐A module in ENVI was used to identify the most spectrally pure or extreme pixels in the imagery. The PPI works by projecting the n‐dimensional MNF spectra to a random 1‐dimensional line, and tagging the pixels that are within a given threshold of the extremes of this line (Mason, 2002). Auto cluster in n‐dimensional visualizer was then used to find the end member image spectra with help of region of interesting tool (roi tool). The n‐dimensional visualizer is used to locate, identify, and cluster the purest pixels and the most extreme spectral responses in a data set (ENVI, 2008). Fig: 6.15. shows the n‐dimensional spectral space with end members that were selected for the next stage. Mean spectra were then extracted for each ROI to act as end members for spectral mapping.
Figure: 6.15. Selected end members in
n‐Dimensional spectral space
6.2.3.1. Mixture Tuned Matched Filtering
The Matched Filter (MF) technique has long been used by electrical engineers for the detection of known signals in mixed backgrounds, especially in radio and radar applications (Boardman, 1998). Mixture tuned matched filtering is an enhancement of the MF algorithm (Mason, 2002). This process is designed to maximise the response of known end members and suppress the results of the complex unknown background (ENVI, 2008). MTMF combines the best parts of the Linear Spectral Mixing model and Statistical Matched Filter model while avoiding the drawbacks of each parent method. The main advantage of MTMF method is the mapping ability of a single known target without knowing the other background end member signatures, unlike traditional spectra mixture modelling (Kruse, 2003).
MTMF algorithm was applied to MNF data with help of image end member spectra that was derived from n‐dimensional space. MTMF rule images 4:13:16 false colour composite image clearly discriminates the evaporites series and chert bed of the area according to their general spectral shape and characteristics spectral features. Spectra MF03, MF04 and MF06 show spectral shape and features similar to the trona and evaporites. While MF02 shows spectral features, doublet at 2.2 µm which is similar to High Magadi bed, MF05 shows broad 2.2 µm feature giving information of presents of chert. MF01 shows Al‐clay minerals according to its 2.2 µm narrow absorption feature. According
84 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
to the above spectral analysis it can be conclude that the green, blue and pink represent the different type of evaporites of the area, while red colour represents the chert bed and High Magadi bed in Fig: 6.16 (d) according to the spectral matching of image and field spectra of chert and evaporites (Fig: 6.17).
MF01
MF02
MF03
MF04
MF05
MF06
(a) (b) (c) (d) (e) MTMF 13 MTMF 16 MTMF 4 MTMF 4:13:16
Figure: 6.16. The MTMF rule images 13 (a), 16, (b), 4 (c) and combination of these rule images as false colour composite image (R:4, G:13, B:16)(d). While green, blue and pink colour of image (d) represent the different types/stages of evaporites, red colour represent the chert bed. (e) Image reflectance spectra showing spectral shape and characteristics features of different surface materials that were represented by different colours.
85 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(a) (b)
Figure: 6.17. Spectral plots comparing the laboratory spectra (a) and Hyperion image spectra (b) of trona (A001) and chert (S047).
The Lake area shows very high spectral variations and this attenuate the spectral dimensionality of surrounding area. As a result of this, MNF eigen images don’t show more clear features in surrounding area. Therefore the Lake area was masked out from the entire image, and surrounding remaining area was subjected to the MNF transformation to see more discriminative land cover variations of the area based on inherent dimensionality of the data. Fig: 6.18 and 6.19 show the different MNF image colour composites with image spectra for different surface covers which have different spectral behaviours. The mean image spectra were acquired using ROI tool representing more than 10 pixels.
86 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(d) (a) (b) (c) MNF 1:3:5 MNF 2:4:6
Figure: 6.18. (a) Hyperion MNF bands 1, 3 and 5 after masking out Lake Magadi area, (b) Image reflectance spectra of different surface expressions, (c) Hyperion MNF bands 2, 4 and 6, after masking out Lake Magadi area and (d) Continuum removed surface reflectance spectra of selected image spectra within wavelength region 2000 nm to 2355 nm.
87 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(d)
(a) (b) (c) (d) MNF 3:4:5 MNF 4:5:6
Figure: 6.19.(a) Hyperion MNF bands 3, 4 and 5 after masking out Lake Magadi area, (b) Image reflectance spectra of different surface expressions, (c) Hyperion MNF bands 4, 5 and 6, after masking out Lake
Magadi area and (d) Continuum removed surface reflectance spectra of selected image spectra within
wavelength region 2000 nm to 2355 nm.
Image end member spectra IS02, IS04, IS09, IS10, IS13, IS14 and IS16 show the absorption wings in the 0.4 µm to 0.9 µm wavelength region which is highly characteristics to the dry plant materials. Absorption wings in this position are generated by intense blue and ultraviolet absorptions in
88 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
compounds such as lignin, tannins and pectins (Elvidge, 1990). Area represented by IS09 and IS10 show more green leaves than dry plant material and it can be identified due to the general shape of the spectra with specific features (Fig: 6.20).
(a) (b)
Figure: 6.20. Reflectance spectra of green leaves (a) and dry plant materials. (after Elvidge, 1990) The spectral features of lignin and holocellulose dominate the reflectance spectra of dry plant material which exhibit diagnostic spectral features in the NIR (2.0‐2.5 µm) region. Lignin spectral features frequently predominate in decayed plant materials with a broad absorption between 2.05 µm and 2.14 µm, a sharply defined absorption at 2.27 µm, and distinct absorptions at 2.33 µm and 2.38 µm (Elvidge, 1990). IS02, IS09, IS13 and IS14 show those absorption features clearly. Even though dry plant material shows another distinct absorption feature at 1.72 µm, due to strong atmospheric absorption this 1.72 µm absorption is disappeared. According to the spectral characteristics, not only the presence of vegetation, but also vegetation type and its status can also be mapped using this data. In this study, however, it was limited to map spatial distribution of vegetation with geology of the area.
IS05, IS06 and IS01 show spectral absorption features in 2000 nm to 2355 nm wavelength region which are very similar to the absorption features of silica (Si‐OH) rich material simply called in this study, as high Magadi bed. IS03 shows a very clear broad absorption feature around 2.2 µm which is similar to the quartz‐chert series rocks of the area. IS07 and IS08 show broad absorption feature between 2.2 µm and 2.3 µm and more close to the 2.3 µm due to mixture of silica rich materials and volcanic tuffs. The spectral shape and occurrences are described in section: 5.6 in more detail. IS11 and IS15 show a broad absorption feature at 2.3 µm which is more similar to the absorption feature of volcanic tuff of the area.
According to the above analysis, most of the geological material and vegetation have diagnostic absorption feature in 2.00 µm to 2.35 µm wavelength region. Therefore spectral indices tool in MMTG‐A module was used in order to study the wavelength at maximum absorption in this
89 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
wavelength region for all the pixels of the image. The output image indicates the wavelength location of the maximum band depth. This tool works on one spectrum at a time and it returns an attribute of one feature in the spectrum (Mason, 2002). Fig: 6.19 shows the pixels that have minimum absorption feature within 2000 nm to 2355 nm wavelength region. Green coloured pixels of the Fig: 6.21 (a) show the pixel that have 2.2μm absorption band. (Fig: 6.21 (b) uses to show the spatial extend of the image (a)) According to this image chert area can be identified. Not only the chert, but also most of the Al‐clay minerals show the absorption feature around 2.2 µm due to the Al‐OH vibrational bond of the mineral. Therefore this image commonly shows the clay minerals and quartz‐chert series rocks of the area. Line striping remaining in lake area on the image creates wrongly interpreted features. However it can be identified by its characteristic linear distribution. Summary of Hyperion image analysis part is shown in Fig: 6.22. Final Surface mineral map with vegetation was created combining all the information that was derived from ASTER and Hyperion images as shown in Fig: 6.23.
Hyperion image (Reflectance)
Maximum Noise Spectral Fraction reduction
Data Dimensionality Without Determination Lake area
area
With No Lake Derive End members from data?
Input user supplied Yes end members? Spatial Pixel Purity Index (PPI) reduction
Examine PPI Results
n‐dimensional visualization and end member selection
Input user supplied Yes End member end members? collection No Classification MTMF
Mapping Results
Figure: 6.22. Work flow chart summarising the surface mineral mapping process of the Hyperion image
Figure: 6.21. (a) Hyperion derived wavelength (a) (b) position of the absorption within 2000nm to 2000nm 2350nm 2113:2234:2314 2350nm. (b) Hyperion SWIR bands 196, 208 and 216.
90 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
(a) (b) (c)
Figure: 6.23 (a) Published Geology map overlain on ASTER DEM, (b) ASTER surface mineral map, (c) Hyperion MTMF product, (d) Hyperion MNF bands 2,4,6, with interpreted geology vector map , and (e) Surface mineral map combining information from map b, c, and d.
WGS‐84 UTM ZONE‐37S
Location of the subset in Fig: 6.24
(e) (d)
91 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Fig: 6.23(e) is a surface mineral map of the Lake Magadi area that combines all the mineralogical information derived from both ASTER and Hyperion images showing the surface minerals related to the local geology and geochemical precipitation process. Spatial distribution of chert and High Magadi bed in mineral map is superimposed with the published geology map of the area. Although different types of evaporites and solutions have been categorized as one (trona) in published geology map, mineral mapping process in remote sensing clearly identified and mapped the different types/stages of evaporites and solutions such as trona, brines and Na‐Al‐Si gel. Spatial distribution of mineral precipitates and related solutions improve the understanding of mineral formations of the area with help of hydro‐geochemical literature. Subset of the surface mineral map in Fig: 6.24 is used to demonstrate two of the mineral formation reactions that can easily understood from the derived surface mineral map. Reaction of water with alkali trachyte/ basalt introduced clay minerals through the hydrolysis process and those minerals can be accumulated due to the sedimentary process at the shore of the lake area. The surface mineral pattern of this reaction/process shows in part A. According to the derived surface mineral map, Na‐Al‐Si gel is very often found in between the Alkali trachyte and brine area, around the lake shore, as shown in Part B giving clue of the way of Na‐Al‐Si gel formation.
Water ← Clay minerals ← Alkali trachyte/Basalt → Na‐Al‐Si gel → Brine → Trona/Evaporites
193063.57 E Part A Part B 9782405.70S
Trona/Evaporites Water Body Al‐Clay minerals Alkali trachyte NaAlSi gel Brine Unclassified
194338.57 E 9781880.70S Fig: 6.24. Spatial subset of the derived mineral map showing some of the geochemical pattern of the study area.
Volcanic tuff, which is not recorded in published geological map, can be seen in southern part of the study area in derived mineral map, close to the Lenderut volcano. Spatial distribution of exposed surface clay minerals is also another new finding from surface mineral mapping in remote sensing which is difficult (impossible) to map in the field. Highly vegetated areas and water bodies are also shown in Fig: 6.23 (e) with the Alkali trachyte and Basalts. However, Alluvium terraces in northern part of the Lake Magadi study area was not identified and categorized using remote sensing due to their inconsistency of surface minerals (materials). Some of the areas (e.g., contact areas) were leaved as unclassified the reason being to map only the homogeneous surface materials.
92 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
6.2.4. Accuracy assessment Assigned class name for each pixels, which were superimposed with 101 point locations of the ground data, were extracted from the classified raster surface mineral map (Fig: 6.23 (e)). Ground data and related classified pixel of the surface mineral map also listed in Appendix: 03. The corresponding error matrix is shown in table: 6.4. The overall accuracy of the classification is 51.48%. Although the primary purpose of this research was to map spatial distribution of the precipitated minerals such as trona, High Magadi bed and the chert, the producer accuracy of those classes were poor except the High Magadi bed (77.81%). Therefore the accuracy of the High Magadi bed would potentially lead one to the conclusion that, it can be used for the purpose of mapping this class although the overall accuracy of the classification was poor. The problem of this conclusion is the fact that the user’s accuracy of this class is only 53.9%. That is, even though 77.81% of the High Magadi bed has been correctly identified as “High Magadi bed”, only 53.9% of the areas identified as “High Magadi bed” within the classification are truly of that category. A more careful inspection of the error matrix shows that there is a significant confusion between the “High Magadi bed” and “chert” classes. This confusion is acceptable due to two main reasons, o Though chert is not always observed in the field with large spatial extend, it is always found with silica rich High Magadi bed (spatial association). o The reflectance and emittance spectral properties of chert and High Magadi bed are relatively similar due to presence of SiO2 as the main constituent; hence it makes difficulties to differentiate each other by remote sensing methods (spectral resemblance).
Table: 6.4. Error matrix resulting from surface mineral mapping
Ground Truth data Trona water Basalt/ High Clay Na‐Al‐ Chert Total Error of User trachyte Magadi Minerals Si gel Commission accuracy bed (%) (%) Unclassified 3 1 11 0 3 0 3 21 100 0 Trona 5 0 0 0 0 0 0 5 0 100
Water 0 2 0 0 0 0 0 2 0 100
Basalt/Trachyte data 4 0 16 0 3 0 1 24 33.3 66.7
High Magadi 1 0 0 7 1 0 4 13 46.1 53.9 bed Clay Minerals 0 0 6 2 5 0 3 16 68.7 31.3
Na‐Al‐Si gel 0 0 0 0 0 2 0 2 0 100 Classification Chert 0 0 1 0 2 0 15 18 16.6 83.4
Total 13 3 34 9 14 2 26 101 Error of 61.5 33.3 52.9 22.2 64.3 0 42.3 Omission (%) Producer 38.5 66.6 47.1 77.8 35.7 100 57.7 Overall accuracy = accuracy (%) 51.48%
The Na‐Al‐Si gel in the error matrix shows highest (100%) in user’s and producer’s accuracy. The reason for this is the fact that, the reflectance information of Na‐Al‐Si gel was acquired at the same places where the ground data is available, while laboratory reflectance information was used for the
93 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
others for SAM and MTMF spectral mapping methods. It is therefore does not yield acceptable information as others. In fact, the most reliable categories associated with this classification from both a producer’s and user’s perspective are “water”, “High Magadi bed” and “chert”.
According to the value of the overall accuracy (around 50%), one can argue that this accuracy can be resulted by chance. The KHAT statistic therefore, was applied to validate the accuracy of the classification further. This statistic serves as an indicator of the extent to which the percentage correct values of an error matrix are due to “true” agreement versus “chance” agreement. Table: 7.8. shows the modified error matrix for KHAT accuracy assessment.
Table: 6.5. Error matrix resulting from surface mineral mapping
Ground Truth Trona water Basalt/ High Clay Na‐Al‐Si Chert Total trachyte Magadi bed Minerals gel
Trona 5 0 0 0 0 0 0 5
data Water 0 2 0 0 0 0 0 2
Basalt/Trachyte 4 0 16 0 3 0 1 24 High Magadi bed 1 0 0 7 1 0 4 13 Clay Minerals 0 0 6 2 5 0 3 16 Classification Na‐Al‐Si gel 0 0 0 0 0 2 0 2 Chert 0 0 1 0 2 0 15 18 Total 10 2 23 9 11 2 23 80
The k^ (“KHAT”) statistic is a measure of the difference between the actual agreement between ground data and an automated classifier and the chance agreement between the reference data and a random classifier (Lillesand et al., 2004). Conceptually, k^ can be defined as,
K^ = (observed accuracy – chance agreement)/ (1‐ chance agreement) = 80 (5 + 2+ 16 + 7 + 5 + 2 + 15) – ((5 ∙ 10)+(2 ∙ 2)+(24 ∙ 23)+(13 ∙ 9)+(16 ∙ 11)+(2 ∙ 2)+(18 ∙ 23)) (80)2 ‐ ((5 ∙ 10)+(2 ∙ 2)+(24 ∙ 23)+(13 ∙ 9)+(16 ∙ 11)+(2 ∙ 2)+(18 ∙ 23)) = 0.56
The KHAT value, 0.56, obtained here is somewhat higher than the overall accuracy (0.51) computed earlier. This value, 0.56, can be considered as an indication that an observed classification is 56 percent better than one resulting from chance.
94 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
7. Analysis of the spatial distribution of minerals in the Magadi area
In this chapter, building of a 3‐d model using remote sensing and geophysical methods to understand the spatial distribution of mineral precipitates of the study area is discussed. This is mainly based on the mineral reactions with water. The chemical process, described in this model starts from mineral reactions with water during the atmospheric precipitation, overland flow and percolation. This ion rich (e.g., carbonate) water flows into the lake basin and concentrated according to the environmental settings of the area such as surface temperature variations. Springs recharged from subsurface reservoir also highly contribute for the mineral formation of Lake Magadi area. The entire model was divided into several phases for easy reference.
7.1. Extraction of drainage network There are no perennial rivers flowing into the Magadi basin at present. The Little Magadi Lake and Magadi Lake are fed by continuous flowing from springs and seasonal atmospheric precipitation. Drainage network of the area was created using ASTER 15‐m spatial resolution DEM, to study the behaviour of overland flow of rain water in this tectonically rugged terrain. Before using the DEM to create drainage network, the applicability of ASTER DEM for elevation was assured by using 149 GPS locations that were collected during the field work (Appendix: 03). It is assumed that the GPS elevations are accurate enough for this study. The pearson correlation coefficient is used to measure the strength and direction of the linear relationship between the elevation of DEM and the elevation value from GPS. The reasonably high pearson’s correlation coefficient r (0.92) and the coefficient of determination r2 (0.86) permitted the use of ASTER DEM for catchments extraction.
r = 0.924898 r2 = 0.855436
Figure: 7.1. The Graphs show the accuracy of DEM for elevation extraction. (Residual standard error: 12.8 on 149 degrees of freedom.)
95 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The drainage network extraction process was done in ILWIS 3.31 as follows. First of all, the digital elevation model was cleaned using the fill sinks operation. It removes the local depressions that consist of single pixel/multiple pixels which has/have smaller height values than all of its neighbouring pixels. Then the flow direction operation was performed to determine into which neighbouring pixel any water in a central pixel will flow naturally in a sink‐free DEM that was created in the previous step. For this, the lowest height method was used. After that the flow accumulation operation was performed to get cumulative count of the number of pixels that naturally drain into outlet. Flow accumulation map was used to extract a basic drainage network with stream threshold 1000. Depending on the flow accumulation value of the pixel and the threshold value for that pixel, it is decided whether that pixel contributes for drainage or not. If the flow accumulation value of the pixel exceeds the threshold value, the output pixel value will be true (value = 1) indicating drainage pixel; if not, false (value = 0). Finally, drainage network ordering operation was performed to extract drainage network map using flow direction map and drainage network map giving minimum drainage length as 100 m. The segment vector map that was created in this step was used as a drainage network map of the area.
(b)
(c) Figure: 7.2 (a) Derived drainage pattern and springs (Hot and Normal) overlain on ASTER DEM. (b) Gully erosion from northern part of (a) the Little Magadi Lake. (Looking south)
(c) Debris flow southern part of Lake
Magadi (looking west).
Influence of topography for drainage network is shown in Fig: 7.2(a). The ASTER surface mineral map that was created in surface mineral mapping stage, overlaid on the DEM shows the clay minerals in
96 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
valley areas and flat terrains and spectrally featureless less weathered alkali trachyte rocks in steep slopes and in mountain range (Fig: 7.3). Fig: 7.3(a) shows the stepwise mountain range with depositional flat and gentle slope area covered with soils and exposed alkali trachyte in steep slope area.
(a) (b)
Figure: 7.3. (a) Field photograph and (b) Subset of surface mineral map overlain on DEM showing the
influence of topography for the spatial distribution of Al‐clay minerals. Clay minerals (Al) = light brown, alkali trachyte = green, chert = red, High Magadi bed = dark brown, water = blue, trona = white, evaporites = light blue, high vegetation = light green, and unclassified areas in the gray scale of the DEM.
These transported or in‐situ Al‐clay minerals are formed probably due to weathering product of alkaline volcanic rocks. Spectral analysis of the soil sample conformed the presents of Al and Mg rich clay minerals according to their spectral features (Appendix: 07). The drainage network of the area mostly follows the valleys which are covered by Al‐clay minerals (Fig: 7.2 (a)). Due to the presence of large catchment areas, it can be assumed that a large amount of water flows in to the Magadi lake during the rainy season and it is conformed by the gullies and debris flows observed in the field (Fig: 7.2 (b & c). N
N
(b) Figure: 7.4. (a) Drainage network (blue line) overlain on classified mineral ASTER DEM. (b) Surface mineral map overlain on DEM showing the spatial distribution of trona, brine, Na‐Al‐Si gel and evaporites. (a)
97 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Climate in the lowermost part of the rift valley, in Lake Magadi area is arid, with high temperatures and an adiabatically reduced mean annual rainfall that is typically less than 500 mm in non El‐Nino years (Warren, 2006). This facilitates the precipitation of Na‐Al‐Si gel in perennial brine pools about the lake edges and trona in the more central saline pan areas (Fig: 7.4(b)). Hydrous sodium aluminium silicate gels may have formed by either precipitation from the spring waters by cooling or interaction of the bicarbonate waters with trachyte debris (Eugster and Jones, 1968; Surdam and Eugster, 1976).
Trachyte debris + Na2CO3 (in water) → gel + SiO2 (in lake) ‐‐‐‐‐‐‐‐‐‐‐‐‐ eq: 7.01
(a) (b)
Figure: 7.5. (a) Field photographs from Northern part of Little Magadi lake shore showing
the interaction of water with trachyte debris. (b) Gel near to the hot springs in southern
part of Lake Magadi area.
Field observations from northern part of the Little Magadi Lake and southern part of Magadi lake show both formations (Fig: 7.5 (a) & (b)) and ASTER mineral map shows clearly spatial distribution of Na‐Al‐Si gel of the study area. The water originated from the hot springs, seepages and overland flow after atmospheric precipitation, is subjected to evaporation on the flat Magadi basin concentrating more ions forming brines and more concentrated gels as described in section 2.1. Formation of evaporates is limited to the surface probably due to the neediness of atmospheric CO2. Some minerals which are formed during wet conditions are changed to another mineral phases due to increase of surface temperature. This phenomenon was studied using reflectance spectral characteristics of SWIR and TIR wavelength region of undisturbed soil samples as described in detail in Section 5:1.2. Naturally this process is taken place in this area due to changes of land surface temperature during the day and within the seasonal changes.
Air temperatures around the lake can reach up to 400C in the dry season, which results the temperature as high as 66‐680C in dark colored mud‐flat sediment that makeup the edges of the moat facies and the thin brine sheets in the trona areas (Warren, 2006). Therefore it is important to study the land surface temperature variation throughout the year and changes within few years back. Five TIR bands of ASTER image facilitate mapping of surface temperature over the study area with high spatial resolution.
98 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
7.2. Land Surface Temperature (LST) mapping
Eight ASTER images from 2000‐ 2008 with limited cloud cover were used to analyze the surface temperature variation with the time. Thermal atmospheric correction was done after the radiometric calibration using below Gain and Offset values (Table: 7.1).
Table: 7.1. Radiometric calibration values for ASTER TIR bands. Band No Gain & Offset (Offset = ‐ Gain) Band 10 0.006882 Band 11 0.006780 Band 12 0.006590 Band 13 0.005693 Band 14 0.005225
This atmospheric correction algorithm assumes that the atmosphere is uniform over the data scene and that a near black body surface exists within the scene. A single layer approximation of the atmosphere is used. No reflected downwelling radiance is also assumed (ENVI, 2008). The algorithm first determines the wavelength that most often exhibits the maximum brightness temperature. This wavelength is then used as the reference wavelength. At this wavelength, the reference blackbody radiance values are plotted against the measured radiances for each wavelength. A line is fitted to the highest points in these plotted data and the fit is weighted to assign more weight to regions with denser sampling. The atmospheric compensation for this band is then applied using the slope and offset derived from the linear regression of these data with their computed blackbody radiances at the reference wavelength (ENVI, 2008).
The Emissivity Normalization method in ENVI was used to derive land surface temperature of the area. In this method, temperatures are calculated from each channel using a constant emissivity value. The highest of these is assigned as the temperature of the pixel (Kealy and Hook, 1993). The performance of the Emissivity Normalization (EN) method is therefore a function of the accuracy of atmospheric corrections method and the actual difference between assumed emissivity value and true emissivity value (Mushkin et al., 2005). The emissivity value 0.96 was selected for each channel, which representing a reasonable average of likely values for exposed geological surface. The highest NDVI value of the images that were selected for temperature‐emissivity separation is less than 0.3. Vegetation cover of the area creates more difficulties due to changes of vegetation cover with time which is more sensitive to the emissivity. Therefore the selection of 0.96 as a constant emissivity value is reasonable for this type of less vegetated arid environment.
99 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Rainy Season (a) 2007‐01‐23 (b) 2008‐02‐18 (c) 2003‐08‐15 (d) 2004‐08‐17
Low Temp. High Temp.
Dry (e) 2006‐08‐20 (f) 2000‐09‐23 Season (g) 2002‐10‐15 (h) 2005‐12‐10
Figure: 7.6 Derived Land surface temperature maps for different seasons.
Using these series of LST images (Fig: 7.6 (a)‐(h)), the Lake dry periods and the spatial distribution of water in those periods can be studied. As an example, in the January and February (Figure: 7.6 a. and b.), rainy season of the area, the lake is filled by water. But later in the year, it is dried and the water is only remaining in several places near hot springs and Little Magadi Lake. Hot springs around Little Magadi Lake and Magadi Lake are responsible for the water in those places. Series of LST images (e), (f), (g) and (h), show drying process of the Lake with time after rainy season; even though those images are from different years. Two sites covering 3000 m2 were selected representing Little Magadi Lake and Magadi Lake for further analysis. Relative temperature variations were calculated using 100 pixel points for each site in every image and results are shown in Fig: 7.7. However, Images 7.6 (b), (d), (g) and (h) show the influence of cloud cover for the Land surface temperature (LST) mapping.
100 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Figure: 7.7(a). Yellow color boxes in Little Magadi Lake
LM (LM) and Magadi Lake (ML) showing the selected sites for surface temperature analysis. (b) Systematic distribution of 100 points covering each 2 pixel in 3000 m of the Magadi Lake area that was used to extract pixel information (Relative surface
(b) temperature). The subset shows the margin of water body and dry area at the time that the image was taken ML (2008‐02‐18). (c). Box plots showing the distribution of relative surface temperature in selected sites in used images. (a)
20051210ML 20051210LM 20021015ML
20021015LM
year 20000923ML 20000923LM
the 20060823ML of 20060823LM 20040817ML Month 20040817LM 20030815ML 20030815LM 20080218ML 20080218LM 20070123ML 20070123LM (c) Low Relative Surface Temperature ‐‐‐‐‐‐‐> High
In the box plot, Little Magadi Lake and Magadi Lake are denoted in each box plot followed by the date of acquisition followed by letter “LM” and “ML” respectively. As an example, relative temperature variation of Little Magadi Lake site in 10th of December 2005 is represented by Box plot followed by 20051210LM. In January both lakes have relatively low temperatures. The image that was acquired in February shows relatively high temperature values in both lakes probably due to the haze effect. In addition to that, in this image, Magadi Lake site shows a wide temperature range within the selected area (3000m2). The reason for this is shown in Fig: 7.7 (b). The selected area includes both water and dry portions at this time giving clue of drying process. After this transition, relative temperature of Magadi Lake site has increased while the relative temperatures of Little Magadi Lake remain constant. This is a good statistical indication for different behaviours of Magadi Lake and Little Magadi Lake through the time. By analyzing series of Land surface temperature with time, the precipitation process and contribution of the rain water for formation of evaporites can be understood. Due to lack of availability of ASTER images for every month in one year, it is difficult to make a good series of LST maps to interpret Land surface temperature changes. Although MODIS (moderate resolution imaging spectrometer) could have facilitated the time series maps with good temporal resolution, its low spatial resolution limits the importance of mapping in this small study area.
101 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
7.3. Stratigraphic information from remote sensing
Spatial association of different hydro Y Z chemical and mineral members and their availability are important to study the X mineral forming process of the area. Surface mineral mapping from remote Y sensing gives spatial association of rocks and minerals in the area in 2‐D (two X dimensional) space. However, Figure: 7.8. Importance of 3‐D space than for 2‐D map to neighbouring rock and minerals in 2‐D understand the spatial association space is not always real neighbours in reality due to the third component z axis (elevation) (Fig: 7.8). Spatial association of rocks and minerals mapped from satellite images therefore does not directly give indication of direct contact between them especially in this type of rugged terrain. The study of rocks and minerals in 3‐ dimentional space is crucial to understand the geochemical process of the area including mineral reaction, precipitation, re‐solution, and re‐precipitation. ASTER Level3, 15‐m resolution DEM (Digital Elevation Model) was used to extract the height information of pixels of the ASTER mineral map which indicate the exposed lithological layers. Distribution of elevation for each classified mineral category was used to study the influence of geomorphology for the mineral precipitates of the area (Fig: 7.9). During the field, 149 GPS elevation readings were taken between the elevation ranges from 571 m to 712 m with field descriptions. This GPS values were used to calculate the gain (1.0736) and offset (‐46.2) of the ASTER DEM to derive the elevation of all the classified pixels that has information for different lithologies which were not covered during the field work. Except Box plot one, all others were derived from ASTER DEM extrapolating mineral information from ASTER optical and thermal bands.
1. GPS readings 2. Trachytic/Basaltic rocks 3. Clay minerals 4. Chert 5. Magadi Lake water 6. NaAlSi gel
7. Evaporites
(m) 8. Brine
9. Trona
Elevation
Figure: 7.9. Vertical distributions of GPS readings, minerals and rocks.
102 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Table: 7.2. Derived and corrected median elevation values for each class.
Median Elevation from DEM (ED) Elevation, E = 1.0736ED – 46.2 Magadi Lake water 609m 607.62m NaAlSi gel 608m 606.54m Brine 608m 606.54m Evaporites 607m 605.42m Trona 608m 606.54m Chert 604m 602.25m
Median elevation value has a meaning especially for flat sedimentary environments such as depositional and evaporitic, to get elevation value of the horizontal bed. According to the above analysis most of the trona, brines, evaporites, and Na‐Al‐Si gels are laid at same elevation (606m) giving clue of contribution of water for formation of evaporites. The mineral map shows the spatial association of evaporite series with water in 2‐D space (Fig: 6.8). Combination of these two adds an advantage to study the chemical reactions in 3D space. Most of exposed quartz‐chert series rocks are laid at ~602m, relatively low elevation than evaporite series due to high contribution of cherts from southern part of the study area which is relatively lowers elevation than middle and northern part. In addition to that, most of basaltic/ trachytic rocks and their weathering products, clay minerals are mostly presented higher levels than the evaporitic series and chert series and they cover most likely entire elevation range of the study area. Quartz‐chert series rocks are laid in middle part of the valley, depositional area, just above the trona formation, spatially associated with High Magadi beds (Fig: 7.10).
N
Figure: 7.10. Mineral Map overlain on ASTER DEM showing the distribution of Chert and High Magadi
beds in the terrain and their spatial associations. Blue Lines = Extracted Drainage Network
103 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
7.4. Generation of 3-D model To understand the process 3‐d model was created combining all the information that was extracted from remote sensing and field work. Entire process, which was studied in previous chapters, can be demonstrated in this model. Model shows the contribution of precipitation and overland flow of precipitated water through the drainage network, evaporation process due to the land surface temperature and changes of environmental condition throughout the year as well as within the day and precipitation of different types of evaporites and precipitates including chert series and trona evaporites series mostly in depressions and low lying terrains. Geochemical processes of the system are adopted by literatures and previous models. Spatial association of rocks and minerals in 3‐ dimentional space with chemical evolution of same rocks and minerals from literature revealed the formation and distribution of precipitates in the study area as demonstrated in the following simplified model.
104 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Understand the process Study of Land Surface temperature combining all data in 3‐D variation with time
ASTER and Hyperion surface mineral mapping
Dem creation and Drainage
network generation
Rain
Rain
A
Figure: 7.11. Scaled schematic of the B distribution of evaporitic lacustrine sediment in the study area (vertical exaggeration of the DEM: x10)
Rain Rain N
W E Evaporation ‐‐ 750m Meters above + Na + mean sea level K ‐‐ 600m ++ + Ca Mg Legend ‐ Spring inflows HCO3 SiO2 ‐ ‐ Cl SO4 ‐ ‐ HCO3 ↔ CO3 + Ca CaCO3 ++ Mg CaMg(CO3)2 Deep ground water reservoir
A 15.578 km B 188180.797m (E) 203759.529m (E) 105 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
106 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
8. Conclusions and recommendations
8.1. Conclusions
The formation of mineral precipitates and their spatial distribution were studied in Lake Magadi area incorporating remotely sensed data, field data and laboratory analysis which were later integrated and modelled. The important lessons learned and the research findings are summarized below,
o The spectral responses of 92 field samples (rock and soil) including trona, Magadiite, Kenyaite, chert, diatomite, basalts/trachyte, green beds and High Magadi beds were studied and identified in this research. Mineralogy of those categories was conformed using X‐ray diffraction method and the physical properties of the sample. The spectral characteristics of
Magadiite (NaSi7O13(OH)3 ∙3H2O ) and Kenyaite (NaSi11O20.5(OH)4∙ 3H2O) were established in this study. The spectral characteristics of these minerals are not known from the work of others (at least from literatures searched done by the author) not from the study area itself. In addition to that, the study of reflectance and emittance spectroscopy of undisturbed surface soil samples showed that the evaporites are only restricted to the upper most part of the surface and that the mineral phase can be changed due to the changes of temperatures.
o Several spectral mapping techniques have been tested to extract information on the surface mineralogy of the area using Hyperion, hyper‐spectral and ASTER, multi‐spectral data. Using these data it was possible to map the spatial distribution of surface minerals including chert, High Magadi bed, alkali trachyte, trona/ evaporites, brine, Na‐Al‐Si gel, volcanic tuff and Al‐ clay minerals of the area. Highly vegetated areas and water bodies were also mapped in addition to the geology. Mapping of different stages of evaporites using remote sensing data substantially improved the existing knowledge of the geology of the area. An improved geology map of the area is given in Figure: 7.23.
o Stratigraphic information revealed from remote sensing methods shows that the mineral precipitates and evaporites are lying in depressions and low lying terrains while most of the higher areas are made of volcanic rocks and their weathering product of clay minerals. Drainage network extracted from the ASTER DEM shows the contribution of runoff water for mineral reaction and formations that were described in existing Hydro‐geochemical models of the area. The contribution of spatial and temporal land surface temperature variations for the evaporitic mineral formations in the area was identified after mapping surface temperatures from ASTER TIR (Thermal Infrared) images for the last eight years.
107 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
The objectives and specific objectives of this research are fully outlined in Section: 1.3. This research however, started with the following research questions in mind;
i. What are the existing facts, concepts and models that explain formation of mineral precipitates in the area? ii. Is it possible to identify mineral precipitates using reflectance spectra? iii. Is that possible to map all the ground mineralogy that was identified by reflectance spectra? iv. What are the main factors that govern the spatial distribution of minerals in the study area? v. To what extent distribution and formation of precipitates be understood using remote sensing data?
With regard to question i. the following conclusions can be drawn: While chemical equations of the mineral formations revealed the required cations, anions and environmental conditions for certain mineral/ rock formation, existing geochemical model shows the source of ions and how associate the rock and minerals to form new mineral precipitates. The published geology maps of the area gave the general information related to the important cations and anions bearing minerals and rocks with their spatial distributions. Study of the existing models and concepts gave the overall idea regarding the spatial distribution and formation of mineral precipitates of the area as outlined below,
When rain water flows over the surface and through the fault system, it reacts with alkaline volcanic terrains giving high amount of anions and cations to the water. Same water flows into the Magadi and Little Magadi Lakes along the temporal drainage or from seepages and hot spring coming from deep ground water reservoir. Evaporation process of the lake area leads to precipitation of different types of evaporites and precipitates based on spatial associations of required cations and anions. Not only the evaporation process, but also hot springs around the lake introduced new precipitates according to their high ionic concentrations and cooling effect after water comes from the hot springs.
With regards to question ii. the following conclusions can be drawn: Even though most of the precipitates and evaporites show characteristic spectral features reflecting mineralogy of the material, some of them could not be identified because of lack of spectral information related to the mineralogy of some of the evaporites.
While the fine‐grained silica rich sedimentary rock, chert exhibits broad absorption feature at 2.2 µm, other siliceous materials such as Green beds, diatomite and High Magadi beds exhibit combination of broad absorption feature at 2.2 µm and narrow absorption feature at 2.3 µm. Mineral precipitates Magadiite and Kenyaite show very distinct small absorption feature at 1.464 µm wavelength region. Main evaporite mineral in the study area, trona exhibits common six absorption features at 1.50‐, 1.74‐, 1.94‐, 2.03‐, 2.22‐ and 2.39‐ µm. Thermal infrared spectra of trona exhibit three characteristics absorption features at 6.66 µm, 9.35 µm and 11.71 µm wavelength regions. Not only the position and shape of the absorption feature, but also general shape of the reflectance spectra can be used to identify mineral precipitates and evaporites in the study area.
108 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
With regards to question iii. the following conclusions can be drawn: Chert, diatomite, High Magadi beds, Green beds, basalt/trachyte, Kenyaite, Magadiite, trona and Clay minerals such as Montmorillonite were identified using reflectance spectra. Chert was mapped based on its 2.2 µm broad absorption feature in SWIR region and 9.8 µm feature in TIR region using band ratio from ASTER image and selected MNF bands from Hyperion image. Diatomite, Green beds and High Magadi beds show relatively same general spectral shape and absorption features. In addition to that diatomite and Green bed didn’t show large spatial extends and those were identified in the field as small spots within the High Magadi beds. Diatomite was identified only in one location covering approximately 2 m* 8 m area. Therefore diatomite, Green beds and High Magadi beds was mapped as one unit naming High Magadi bed. Basalt/trachyte of the area was mapped based on their featureless general spectral shape using the Spectral angle mapper technique. Trona was mapped according to their general spectral shape and characteristics spectral absorption features. Magadiite and Kenyaite was not able to map due to their small spatial extent and according to their small diagnostic absorption features which are less than noise of the Hyperion image spectra. Kenyaite always occurs as nodule and never forms as continuous bed. Even though, Magadiite precipitates as a bed due to its small diagnostic absorption feature it is difficult to map even from hyper‐spectral image. Clay minerals were identified and mapped using ASTER image and different types of clay minerals were not mapped in this stage.
Therefore all the ground mineralogy that was identified by reflectance spectra could not be mapped according to the spectral resolution of the image, spectral resolution of diagnostic absorption feature of the field reflectance spectra, noise of the image, spatial extend of ground target, spatial resolution of the image and accuracy of image prepossessing stages. Furthermore, Natural geologic surfaces are often partially covered with non‐geologic materials or composed of mixtures of minerals with varying grain sizes and differing degree of compaction and weathering. These factors are influence remote spectral measurements and limit the number of pixels that can be classified and mapped.
The overall shape of the laboratory/library spectra was different from the image spectra probably, due to the several reasons. The reasons for these differences can be drawn as the presence of variable mineral mixtures, grain size variations, residual atmospheric absorption features, desert varnish top of the rock surface, and calibration error of laboratory spectrometer and/or ASTER/Hyperion instrument. These spectral differences point to the advantage of using image spectra rather than library/laboratory as reference in SAM and MTMF processing.
The area which had very high spectral variations attenuated the spectral dimensionality of the surrounding low spectral variation area of the Hyperion image. Image processing in two steps, with and without (masking out) the high spectral variation area, could be successfully overcome this problem allowing to extract information from the entire image.
109 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
With regards to question iv. the following conclusions can be drawn: Stratigraphic information revealed from remote sensing methods shows that the mineral precipitates and evaporites are lying in depressions and low lying terrains while most of the higher areas are made of volcanic rocks and their weathering product of clay minerals. Geomorphology of the area therefore acts as one of the main factor of the mineral precipitation. Spatial and temporal land surface temperature variations of the area also highly contributed to the spatial distribution of minerals. Presents of hot springs on the other hand governs the spatial distribution of mineral precipitates of the area.
With regards to question v. the following conclusions can be drawn: The usage of remote sensing techniques allowed substantial information to be derived related to the distribution and formation of precipitates and evaporites of the area. Most of the remote sensing methods give information on 2‐ dimensional and 3‐dimensional space. Combination of this spatial information (in this case mineral precipitates) and non spatial information (e.g., literatures related to the geochemical processes of formation of mineral precipitates) can be successfully used to interpret and understand the formation and spatial distribution of various mineral precipitates and evaporites of the study area. Because human brain easily understand the process which is shown in 2‐D or 3‐D space than in non spatial information such as equations.
8.2. Limitations
Following limitations were encountered during the research.
o Small part of the study area has to be left out due to inadequate time and difficulties in accessing and only about 80% of the entire study area could be covered during the field work to collect Rock and soil samples with field descriptions.
o XRD analysis was carried out only for few rocks and soil samples (15 samples out of 92). Even though reflectance spectroscopy of most of the soil samples exhibit different mineral assemblages, their mineralogy could not be conformed by XRD analysis due to the time limitation.
o Due to the lack of spectroscopic literature related to the evaporitic minerals, most of the evaporitic and precipitated minerals could not be identified.
o Only two samples (one each from Magadiite and Kenyaite) were found in the study area during the field work and mineralogy of those were conformed by XRD analysis. However this is insufficient to draw a conclusion for the spectral characteristics of those two minerals.
o In land surface temperature mapping process, different surface materials with different physical properties such as thermal inertia, emissivity, moisture content and local topographic slope orientation respond differently to solar radiation, resulting different surface temperature at different time. This is a big limitation to map absolute LST from the remote sensing method.
110 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
o The ASTER image could cover the entire study area whereas the Hyperion image covered only a part of it. As a result, the geological interpretation of some part of the study area was restricted only to ASTER image and, where the coverage was available the interpretation was done using both the Hyperion image and the ASTER image.
8.3. Recommendations
Based on the above results, conclusions and limitations of the research, the following recommendations are given. o A detail study should be performed to investigate the mineralogy of evaporites and precipitates with their characteristic spectral features. o Building of systematic approach is necessary to study the changes of mineralogy of the precipitation process of evaporites during the drying process with help of XRD and reflectance/emittance spectroscopy.
o Different types of chert found in the field did not exhibit clear differences for SWIR reflectance spectroscopy. Therefore it is recommended to study TIR emittance spectroscopy to explore the possibilities to identify different types of chert.
o Spectral characteristics of Magadiite and Kenyaite were introduced only using one sample from each category. It is recommended to collect more Magadiite and Kenyaite samples to understand unique spectral absorption features for those two minerals.
o One of the samples (Sample no: P012) found during the field work was identified as an Oldhamite from X‐ray diffraction even though it is not originally Oldhamite. Therefore geochemical and crystallographic studies of this sample should be carried out to identify the mineralogy.
o Surface mineral interpretation using Hyperion image should be repeated for the entire study area, since this study could cover only part of the area due to the partial coverage of available Hyperion image.
111 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
References
Aines, R.D. and Rossman, G.R., 1984. Water in Minerals ? : A peak in the Infrared. Journal Of Geophysical Research, Vol. 89(No. B6): 4059‐4071. Atmaoui, N. and Hollnack, D., 2003. Neotectonics and extension direction of the Southern Kenya Rift, Lake Magadi area. Tectonophysics, Vol. 364: 71‐83. Baker, B.H., 1958. Geology of the Magadi area:Report No. 42, Geological Survey of Kenya. Baker, B.H., 1986. Tectonics and volcanism of the southern Kenya Rift Valley and its influence on rift sedimentaion Sedimentation in the African Rifts. Geologycal Society Special Publication, No. 25: 45‐57. Beck, R., 2003. EO ‐ 1 User Guide, Version: 2.3. USGS Earth resources Observation systems Data Center (EDC). Behr, H.J. and Rohricht, C., 2000. Record of seismotectonic events in siliceous cyanobacterial sediments (Magadi cherts), Lake Magadi, Kenya International Journal Of Earth Science, Vol. 89: 268‐283. Ben‐Dor, E., Irons, J.R. and Epema, G.F., 1999. Soil Reflectance. Remote Sensing for the Earth Sciences: Manual of Remote Sensing, 3 rd ed., Vol. 3: 111 ‐ 188. Beneke, K. and Lagaly, G., 1983. Kenyaite‐Synthesis and properties. American Mineralogist, Vol. 68: 818‐826. Biggar, S.F., Thome, K.J. and Wisniewski, W., 2003. Vicarious Radiometric Calibration of EO‐1 Sensors by Reference to High‐ Reflectance Ground Targets. IEEE Transactions on Geoscience and Remote Sensing, Vol. 41(No. 6): 1174 ‐ 1179. Bish, D.L. and Chipera, S.J., 1991. Detection of trace amounts of Erionite using x‐ray powder diffraction: Erionite in tuffs of Yucca mountain, Neveda, and Central Turkey. Clays and Clay Minerals, Vol. 39(No. 4): 437‐445. Boardman, J.W., 1998. Leveraging the High Dimensionality of AVIRIS Data for Improved Sub‐Pixel Target Unmixing and Rejection of False Positives: Mixture Tuned Matched Filtering, AVIRIS Proceedings 1998. JPL Publication 87‐38. Bruker, O., 2007. VERTEX 70, User Manual, 1st updated edition. Burbine, T.H. et al., 2002. Spectra of extreamely reduced assemblages: Implications for Mercury. Meteoritics & Planetory Science Vol. 37: 1233 ‐ 1244. Chorowicz, J., 2005. The East African rift system. Journal Of African Earth Sciences, Vol. 43: 379 ‐ 410. Clark, R.N., 1999. Spectroscopy of Rocks and Minerals, and Principles of Spectroscopy. Remote Sensing for the Earth Sciences: Manual of Remote Sensing, 3 rd ed., Vol. 3: 3‐52. Clark, R.N., King, T.V.V., Klejwa, M. and Swayze, G.A., 1990. High Spectral Resolution Reflectance Spectroscopy of Minerals. Journal Of Geophysical Research, Vol. 95(No. B8): 12,653‐12,680. Clark, R.N. and Roush, T.L., 1984. Reflectance Spectroscopy: Quantitative Analysis Techniques for Remote Sensing Applications. Journal Of Geophysical Research, Vol. 89(No. B7): 6329‐6340. Coolbaugh, M.F., Kratt, C., Fallacaro, A., Calvin, W.M. and Taranik, J.V., 2006. Detection of geothermal anomalies using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) thermal infrared images at Bradys Hot Springs, Nevada, USA. Remote Sensing Of Environment, Vol.106: 350‐359. Crowley, J.K., 1991. Visible and Near‐Infrared (0.4‐2.5um) Reflectance Spectra of Playa Evaporite Minerals. Journal Of Geophysical Research, Vol. 96(No. B10): 16,231‐16,240. Crowley, J.K. and Hook, S.J., Mapping playa evaporite minerals and associated sediments in Death valley, California, with multispectral thermal‐infrared images. Drake, N.A., 1995. Reflectance spectra of evaporite minerals (400‐2500 nm): applications for remote sensing. International Journal Of Remote Sensing, Vol. 16(No. 14): 2555‐2571. Drury, S., 2001. Image Interpretation in Geology. Blackwell Science, 290 pp. Elvidge, C.D., 1990. Visible and near infrared reflectance characteristics of dry plant materials. International Journal of Remote Sensing, Vol. 11(No. 10): 1775 ‐ 1795.
112 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
ENVI, 2008. ENVI Version 4.5, ITT industrie.inc. Eugster, H.P., 1967. Hydrous Sodium Silicates from Lake Magadi, Kenya: Precursors of Bedded Chert. Science, Vol. 157: 1177‐1180. Eugster, H.P., 1969. Inorganic Bedded Cherts from the Magadi Area, Kenya. Contrib. Mineral. and Petrol, Vol. 22: 1‐31. Eugster, H.P. and Jones, B.F., 1968. Gels Composed of Sodium‐Aluminium Silicate, Lake Magadi, Kenya. Science, Vol. 161: 160 ‐ 163. Fletcher, R.A. and Bibby, D.M., 1987. Synthesis of Kenyaite and Magadiite in the presence of various anions. Clays and Clay Minerals, Vol. 35(No. 4): 318‐320. Furman, T., 2007. Geochemistry of East African Rift basalts: An overview. Journal Of African Earth Sciences, Vol. 48: 147 ‐ 160. Gaffey, S.J., 1985. Reflectance spectroscopy in the visible and near‐infrared (0.35‐2.55 um): Applications in carbonate petrology. Geology, Vol. 13(270‐273). Gaffey, S.J., 1987. Spectral reflectance of carbonate minerals in the visible and near infrared (0.35‐ 2.55 um): Anhydrous carbonate minerals. Journal Of Geophysical Research, Vol. 92(No. B2): 1429‐1440. Gangopadhyay, P.K., Maathuis, B.H.P. and van Dijk, P.M., 2005. ASTER ‐ derived emissivity and coal ‐ fire related surface temperature anomaly : a case study in Wuda, north China. International Journal Of Remote Sensing, 26(24). Goetz, A.F.H., Vane, G., Solomon, J.E. and Rock, B.N., 1985. Imaging Spectrometry for Earth Remote Sensing. Science, Vol. 228(No. 4704): 1147‐1153. Green, R.O., Pavri, B.E. and Chrien, T.G., 2003. On‐Orbit Radiometric and Spectral Calibration Characteristics of EO‐1 Hyperion Derived With an Underflight of AVIRIS and In Situ Measurements at Salar de Arizaro, Argentina. IEEE Transactions on Geoscience and Remote Sensing, Vol. 41(No. 6): 1194 ‐1202. Griffiths, P.R. and Haseth, J.A.d., 1986. Fourier Transform Infrared Spectrometry. John Wiley & Sons, 656 pp. Harada, K., Iwamoto, S. and Kihara, K., 1967. Erionite, Phillipsite and Gonnardite in the Amygdales of altered Basalt from Maze, Niigata Prefecture, Japan. The American Mineralogist, Vol. 52: 1781‐ 1792. Harloff, J., 2000. Near ‐ infrared reflectance spectroscopy of bulk analog materials for the igneous Martian crust. Forschungsbericht DLR : Deutsche Forschungsanstalt fur Luft‐ und Raumfahrt;2000‐07. Deutsches Zentrum fur Luft‐ und Raumfahrt (DLR), Koln, 218 pp. Harris, W. and White, N., 2007. Methods of Soil Analysis: Mineralogical Methods. Soil Science Society of America, Parts 5. Hellman, M.J. and Ramsey, M.S., 2003. Analysis of hot springs and associated deposits in Yellowstone National Park using ASTER and AVIRIS remote sensing. Journal of Volcanology and Geothermal Research, Vol.135: 195‐219. Hesse, R., 1989. Silica Diagenesis: Origin of Inorganic and Replacement Cherts. Earth‐Science Reviews, Vol. 26: 253‐284. Hewson, R.D., Cudahy, T.J., Mizuhiko, S., Ueda, K. and Mauger, A.J., 2005. Seamless geological map generation using ASTER in the Broken Hill‐Curnamona province of Australia. Remote Sensing Of Environment, Vol 99: 159‐172. Hook, S.J., Gabell, A.R., Green, A.A. and Kealy, P.S., 1992. A Comparison of Techniques for Extracting Emissivity Information from Thermal Infrared Data for Geologic Studies. Remote Sensing Of Environment, Vol. 34: 123 ‐ 135. Hubbard, B.E., Crowley, J.K. and Zimbelman, D.R., 2003. Comparative Alteration Mineral Mapping Using Visible to Shortwave Infrared (0.4 ‐ 2.4 um) Hyperion, ALI, and ASTER Imagery. IEEE Transactions on Geoscience and Remote Sensing, Vol. 41(No. 6): 1401 ‐ 1410. Hunt, G.R., Salisbury, J.W. and Lenhoff, C.J., 1973. Visible and Near‐Infrared Spectra of Minerals and Rocks: VI. Additional Silicates. Modern Geology, vol. 4: 85‐106.
113 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Hunt, G.R., Salisbury, J.W. and Lenhoff, C.J., 1974. Visible and Near‐Infrared Spectra of Minerals and Rocks: IX. Basic and Ultrabasic Igneous Rocks. Modern Geology, Vol. 5: 15‐22. ICDD, 2008. International Centre for Difffraction Data Iwasaki, A. and Tonooka, H., 2005. Validation of a Crosstalk Correction Algorithm for ASTER/SWIR. IEEE Transactions on Geoscience and Remote Sensing, Vol. 43(No. 12): 2747‐2751. Jenkins, R., 2000. X‐ray Techniques: Overview. Encyclopedia of Analytical Chemistry. John Wiley & Sons Ltd. Jones, B.F., Eugster, H.P. and Rettig, S.L., 1977. Hydrochemistry of the Lake Magadi basin, Kenya. Geochimica et Cosmochimica Acta, Vol 41: 53‐72. Kealy, P.S. and Hook, S.J., 1993. Separating Temperature and Emissivity in Thermal Infrared Multispectral Scanner Data: Implications for Recovering Land Surface Temperatures. IEEE Transactions on Geoscience and Remote Sensing, Vol. 31(No. 6): 1155 ‐ 1163. King, M.D. et al., 2003. EOS Data Products Hand book, Vol. 1. NASA/Goddard Space Flight Center. Klein, C. and Hurlbut, C.S., 1999. Manual of Mineralogy. John Wiley & Sons. inc. Kruse, F.A., 2003. Preliminary Results‐ Hyperspectral Mapping of Coral Reef Systems using EO‐1 Hyperion, Buck Island, U.S. Virgin Islands. JPL Publication 04‐6: 157 ‐ 173. Lillesand, T.M., Kiefer, R.W. and Chipman, T.W., 2004. Remote Sensing and Image Interpretation. John and Wiley. Mason, P., 2002. MMTG A‐List Hyperspectral Data Processing Software. Manual version 1.0, CSIRO Division of Exploration and Mining: pp.103. Masona, G.M., 2003. Oldhamite in processed oil shale. Eastern Oil Shale Symposium, 6‐8 November 1990, Lexington, KY, USA. Mccabe, M.F., Balick, L.K., Theiler, J., Gillespie, A.R. and Mushkin, A., 2008. Linear mixing in thermal infrared temperature retrieval. International Journal Of Remote Sensing, Vol. 29(No. 17‐18): 5047 ‐ 5061. Meer, F.D.v.d. and Jong, S.M.d. (Editors), 2001. Imaging Spectrometry : Basic principles and prospective applications. Kluwer Academic Publisher. Meer, F.v.d. and Jong, S.d., 2003. Spectral Mapping Methods: Many Problems, Some Solutions, EARSeL Workshop on Imaging Spectroscopy, Herrsching, pp. 146 ‐ 161. Moghtaderi, A., Moore, F. and Mohammadzadeh, A., 2004. The application of advanced space‐borne thermal emission and reflection (ASTER) radiometer data in the detection of alteration in the Chadormalu paleocrater, Bafq region, Central Iran. Journal Of Asian Earth Sciences, Vol. 30: 238 ‐ 252. Morley, C.K., Ngenoh, D.K. and Ego, J.K., 1999. Introduction to the East African Rift System. Geoscience of Rift Systems‐ Evolution of East Africa: AAPG Studies in Geology(No. 44): 1 ‐18. Mushkin, A., Balick, L.K. and Gillespie, A.R., 2005. Extending surface temperature and emissivity retrieval to the mid‐infrared (3 ‐ 5 um)using the Multispectral Thermal Imager (MTI). Remote Sensing of Environment, Vol. 98: 141 ‐ 151. Ninomiya, Y., Fu, B. and Cudahy, T.J., 2005. Detecting lithology with Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) multispectral thermal infrared "radiance‐at‐ sensor" data. Remote Sensing of Environment, Vol. 99: 127‐139. Pearlman, J.S. et al., 2003. Hyperion, a Space‐Based Imaging Spectrometer. IEEE Transactions on Geoscience and Remote Sensing, Vol. 41(No. 6): 1160 ‐ 1173. Pontual, S., Merry, N. and Gamson, P., 1997. Spectral Interpretation Field Manual: G‐MEX. AusSpec International pty. Ltd. Rowan, L.C., Mars, J.C. and Simpson, C.J., 2005. Lithologic mapping of the Mordor, NT, Australia ultramafic complex by using the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). Remote Sensing of Environment, 99: 105‐126. Rowan, L.C., Schmidt, R.G. and Mars, J.C., 2006. Distribution of hydrothermally altered rocks in the Reko Diq, Pakistan mineralized area based on spectral analysis of ASTER data. Remote Sensing Of Environment, Vol. 104: 74‐87. Sabins, F.F., 1999. Remote sensing for mineral exploration. Ore Geology Reviews, Vol. 14: 157‐183.
114 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Salisbury, J.W. and D'Aria, D.M., 1992. Emissivity of Terrestrial Materials in the 8‐14 um Atmospheric Window. Remote Sensing Of Environment, Vol. 42: 83‐105. Surdam, R.C. and Eugster, H.P., 1976. Mineral reactions in the sedimentary deposits of the Lake Magadi region, Kenya. Gelogical Society of America Bulletin, Vol. 87: 1739‐1752. Thurmond, A.K., Abdelsalam, M.G. and Thurmond, J.B., 2006. Optical‐radar‐DEM remote sensing data integration for geological mapping in the Afar Depression, Ethiopia. Journal Of African Earth Sciences, Vol. 44: 119 ‐ 134. Tommaso, I.D. and Rubinstein, N., 2006. Hydrothermal alteration mapping using ASTER data in the Infiernillo porphyry deposit, Argentina. Ore Geology Reviews, Vol. 32: 275 ‐290. Turdu, C.L. et al., 1999. Influence of Preexisting Oblique Discontinuities on the eometry and Evolution of Extensional Fault Patterns: Evidence from the Kenya Rift Using SPOT Imagery. Geoscience of Rift Systems‐ Evolution of East Africa: AAPG Studies in Geology(No. 44): 173‐191. Ungar, S.G., Pearlman, J.S., Mendenhall, J.A. and Reuter, D., 2003. Overview of the Earth Observing One(EO‐1) Mission. IEEE Transactions on Geoscience and Remote Sensing, Vol. 41(No. 6). Vaughan, R.G., Hook, S.J., Calvin, W.M. and Taranik, J.V., 2005. Surface mineral mapping at Steamboat Springs, Neveda, USA, with multi‐wavelength thermal infrared images. Remote Sensing Of Environment, Vol. 99: 140‐158. Wang, Y.‐R., Wang, S.‐F. and Chang, L.‐C., 2006. Hydrothermal synthesis of magadiite. Applied Clay Science, Vol. 33: 73‐77. Warren, J.K., 2006. Evaporites: Sediments, Resources and Hydrocarbons. Springer, 1034 pp. Wilson, M.J., 1987. A hand book of determinative methods in clay mineraloy. Blackie, 308 pp. www.brukeroptics.com, Bruker Optics. Access date: 2008.11.15. www.iza‐online.org, Erionite. Access date: 2008.12.08. www.mii.org, Diatomites. Web, acsess date: 2008. 11.10. Zussman, J., 1967. Physical Methods in Determinative Mineraloy. Academic press.
115 Appendices
Appendix: 01. Ghemical analysis of springs and related waters from Lake Magadi area.
LocatJon and sanole Tenp. Fleld Lab Dlrsolved HCOI. co3 Por, Dace nunber!./ (oc)- p.!.. tH :11 1oo Egl_tds .Densl ty Hot sDrlnrs (t50oc) LlTl u686 (N of NE lagoon) 68 8.95 99 11,300 169 14,80O 3,O70 185 5,480 t52 37 5.0 l. l 28,600 1.02r L973 MrO23 (M686) 66 9.2O 8.90 80 11.200 r90 12,300 4r170 r59 5,210 l4l 24 8.t t.3 27,?00 1.026 1973 U1017 (81E,Ur4)---- 80 9.-lO 8.90 85 r1,900 2h3 l1,lo0 4,300 195 5,540 161 25 8.9 6.6 28,900 r.027 L97t MrOrS (Br9)- 83 9.44 9.O5 85 10.50o 198 r0,40o 4,450 168 4,890 t45 20 8.3 5.5 25,600 1.024 1973 Hl0lg (hot ri.ver)'- 60 9.48 9.24 93 11,300 193 12,000 4,480 15l 5,550 l5l 22 8.3 5.0 28,000 L,026 t973 lllo2r (N of !ll5)--- 60 9.36 9.13 83 1r,800 220 13,700 4,130 160 5,7E0 166 22 7.8 5.E 29,500 7.A27 1973 MrO22 (821)- 71 9.22 8.90 8l 11,90o 212 l3,O0O 4,630 154 5,550 ls6 24 8,8 5.8 29.LOO 1.026 !e_r'*"g!scJ 5019J 7170 M5(7 (Er7)-- 4l .8 9.60 9.45 46 12, too L27 8,820 7,A1O 206 6,540 t32 20 5.1 17 12.700 r.025 rlto M583------43. S 9. 50 82 r J, 7o0 L46 9 ,280 7,94O 229 6,820 146 30 1.9 16 34,900 r.026 7170 M588 (Br4)-- 44.2 9.60 79 1J,200 190 8,640 7,720 Z3L 6,390 141 33 2.5 30,700 1.025 7/70 v27 (Br3)--- 9.35 59 l?,70a t4l 6.220 8,040 232 6, 190 L25 5.3 12 3r,900 t.0?7 LLl72 ltSo (Blrd Rock)---- :::_ ::: 9.80 77 l5,5oo 175 8,480 9,850 246 7 ,77O r55 37,700 1,03r 5173 r{1004 (Br2)- J7 9.90 9.65 i4 9,880 rl5 ?,100 5,42O 249 4, E20 100 17 6.7 l0 24,700 r.022 5173 Mr005 (8r3,M27)---- 4l 9 .88 9.5? 74 12,3O0 133 9 ,040 6,680 225 6,110 t26 28 6.7 ll 30,400 l.O28 5l7t x1006 (M80)- J8.5 9.8J 9.55 49 12,200 121 9,42a 6,450 2O2 5,090 r22 20 6.6 r0 30,400 L.o27 5173 u]007 (8r7,!{547)--- 42.5 9.84 9.31 4t t2,500 121 8, 740 I t2'lo 196 6,240 L26 22 6.9 lr 3l r 20o r.028 5113 Mr008 (B8)-- 37 9,66 9, 30 55 3.870 49 4 ,080 1,590 122 1,550 50 6.8 3.0 2.8 9,600 1.008 5171 Xr009 (86,Mlu)---- 12 8.98 8.93 r04 8,060 104 11,200 r,810 227 3,930 69 15 6,3 3-r 20,000 r.018 5/7J M10tl (t{84)_ 36 I .Ltt L25 6l 6,320 9E 7 ,23A 2,r20 196 3,0o0 69 L7 4.3 3.0 15,500 r.0[3 5/7t H10r3 (82)-- 36.5 9.60 9. J0 59 4,500 96 5,200 1,680 9t 2,L2O 62 9.9 h.0 3.1 ll,200 r.010 5/73 H1025 (Frsh Spring) 3J 10.18 9.78 )1 7,480 85 4,090 4,860 73 3,560 69 l4 2.8 5.o r8,400 1.016 Spring outflow 7l70 r.{s50------29.8 9.62 9.60 1.9 tJ,200 12? 9,4ro 7,2rO 200 6,630 Llz 24 6.9 L2 34,100 7-o25 ?,24O L46 r8 1.029 7 /70 M55l------26.3 9.60 9.45 5l 14,200 t36 9-750 7,820 2t7 24 7.0 36,400 ? l7O u555------30. 7 9.60 9.55 66 1s,000 169 10,700 8,950 253 7,940 r69 26 5.3 rl 40,600 t.o3l 7170 H55t------t2 9.78 9,90 168 1E.300 222 8,130 r1,900 321 9, 150 202 62 5.5 L7 45,400 1.038 517f U1024 (NE lagoon)-- 39.6 9.60 9.22 lrl r6,500 212 16.000 7,440 2L4 8,120 202 30 ll l.t1 4l,200 1.038 Pi! seepa8e ground $ater 7/70 u5028 (1.8 m)------3l 9. 65 82 16,?00 r98 r0,100 9,650 237 8roEo 206 z0 ,.9 18 41,50O 1,034 7l70 H596 (2.9 m)------28 3_3', 9. 50 198 23.100 234 7 ,2t10 13,200 562 16, 100 33E 21 7.5 6E 58,900 1.047 LlTI x621 (1.7 m)------32 10. 20 448 25,900 402 3,O10 22t2OO 3O2 13,800 343 s8 32 64,800 1.058
36.10' 36'20'
Locations are given in Fig: A: 01-.1-. Number of
GEI lrono nearby or equivalent spring previously EX Surlccr briar El tudllotr W) Allwi$n sampled by Baker (B) or Eugster (M) is given in /Norrl forr 0 Str. ')"o*? parentheses. lon contents are given by mg/kg.
l'50'
Figure: A. 01.1
1,16 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 02. Geological map of the Magadi area.
Figure: A.02.1. Geological map of the Magadi area (after Baker, 1952).
117 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 03. Collected samples and field descriptions during the field work.
Date Coordinates Coordinates GPS Field discription Sample Rock M. Type N E Elevation No type/Mineralogy U. (m) 20080918 9807910 197705 588 Laggons B012 Trona e 1 20080918 9808141 197756 590 Basalt gravels in gully B013 Montmorillonite* e 1 20080918 9780089 196130 590 Silty clay deposite B032 Opal* a 1 20080918 9792876 194103 615 Basalt gravels topsoil B034 Halloysite* e 1 20080918 9807699 197779 586 Hot spring ‐‐ c 0 20080918 9807699 197779 662 Hot spring ‐‐ c 0 20080918 9785404 198631 600 Chert bed ‐‐ i 0 20080918 9785388 198630 588 Clay bed ‐‐ i 0 20080918 9794950 190309 632 Slican slide place P001 Basalt e 2 20080918 9778557 194227 580 surface cracks (no water) P018 Trona f 2 20080918 9779992 195887 582 Chert bed P019 Chert e 2 20080918 9779945 195869 580 Basalt P020 Basalt e 2 20080919 9777660 193998 648 Basalt ‐‐ e 0 20080919 9777730 192194 589 Chert gravel topsoil B001 Montmorillonite* i 1 20080919 9791869 196224 590 Magadi trona beds B002 Trona b 1 20080919 9778394 192017 585 Lagoonal top soil B004 Trona b 1 20080919 9778211 191870 584 Cracks with evaporites (s part) B006 Trona e 1 20080919 9777756 192126 589 Chert gravel topsoil B015 Green beds i 1 20080919 9779014 192688 605 Dry lagoon B018 Halloysite* i 1 20080919 9777638 192223 584 Area covered with chert B024 Erionite i 1 20080919 9777782 191950 584 Chert and basalt topsoil B030 Mg clay minerals* i 1 20080919 9777750 192862 603 Dry lagoon B033 Calcite* f 1 20080919 9777522 192241 586 Bare land top soil B035 Montmorillonite* i 1 20080919 9779292 194352 571 Chert bed ‐‐ f 0 20080919 9778159 192365 587 Chert pebbles on surface ‐‐ f 0 20080919 9778393 191997 590 Hot spring ‐‐ a 0 20080919 9778247 191564 580 Gravels top of the basalt ‐‐ a 0 20080919 9778392 192035 580 Red algee,white stuff, hot ‐‐ h 0 springs 20080919 9780492 195788 609 Hot spring ‐‐ h 0 20080919 9777784 193864 617 Erosional deposite of the slop ‐‐ e 0 20080919 9778426 191980 609 Hot spring ‐‐ e 0 20080919 9778643 194024 588 Chert intrusion ‐‐ f 0 20080919 9779706 194462 581 Clay bed ‐‐ f 2 20080919 9792928 194182 609 Basalt P015 Basalt e 2 20080919 9778557 192055 585 Hot spring ‐‐ a 0 20080919 9778206 191928 588 Near hotsprings Ring 01 HighMagadiBed g 3 20080919 9785448 198603 608 Recent clay deposite S030 Sedimentatary i 2 rock 20080919 9780624 195721 604 Alluvial fan S085 Chert, Basalt a 2 pebbles 20080920 9804272 200395 629 White color intrusion B011 Montmorillonite* a 2 20080920 9805682 201002 657 Yellow color rock B020 Erionite e 2 20080920 9804388 200419 629 White color intrusion B023 Mg clay minerals* e 2 20080920 9777622 193462 600 Ash color surface B028 Halloysite* f 1 20080920 9803000 199778 613 Basalt rock bolders B029 Montmorillonite* i 1 20080920 9806014 201122 653 Basalt with clay B037 ………………… e 1
118 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
20080920 9805410 200618 633 Basalt rock with different ‐‐ a 0 oreantation 20080920 9804604 200419 629 White color intrusion ‐‐ a 0 20080920 9804652 200454 625 Basalt intrusion ‐‐ e 0 20080920 9804848 200413 633 Chert bed ‐‐ g 0 20080920 9804542 200358 624 Dry river bed with bolders ‐‐ a 0 20080920 9805566 200781 625 River bed with basalt bolders ‐‐ e 0 20080920 9801714 200021 641 Young basalt intrusion ‐‐ g 0 20080920 9801666 199996 634 Chert bed ‐‐ g 0 20080920 9780500 195607 601 Alluvial fan ‐‐ f 0 20080920 9780544 195768 606 Chert and basalt sediments ‐‐ e 0 20080920 9806014 201122 653 Basalt with clay P005 Basalt e 2 20080920 9801086 200529 625 Highly weathered chert or P006 Chert g 2 basalt 20080920 9805438 200664 633 Basalt S006 Basalt e 2 20080920 9801086 200529 625 Quartz with Na feldspar S008 Chert g 2 20080923 9805942 201101 660 Ersional deposite B008 Montmorillonite* e 1 20080923 9806374 200786 634 Red and yellow color clay B010 Montmorillonite* a 2 stone 20080923 9805966 201023 649 Red color silty clay deposite B017 Montmorillonite* a 2 20080923 9798434 201409 711 Basalt/trachyte rock ‐‐ e 0 mountain 20080923 9806054 201014 648 Yellow color clay stones ‐‐ a 0 20080923 9778200 191915 588 Hotspring ‐‐ a 0 20080923 9789922 198304 631 Petrol shed ‐‐ a 0 20080923 9798374 201427 703 contact between basalt and P004 Volcanic tuff e 2 yellowish color rocks 20080923 9798374 201427 703 contact basalt and yellowish P008 Basalt e 2 color rock 20080923 9806374 200786 634 Red and yellow color clay P009 Opal* a 2 stone 20080923 9799860 200700 618 Green clay chert series S021 Sedimentatary i 2 rocks 20080923 9809836 201130 674 Basalt/trachyte rock S027 Basalt g 2 mountain 20080924 9806044 201454 665 Red color silty clay deposite B003 Montmorillonite* e 2 20080924 9808654 201274 684 Ash color surface B005 Halloysite* e 1 20080924 9808544 201279 687 Yellowish brown color rock B007 Montmorillonite* e 2 20080924 9807770 200395 678 Ash color surface B009 Montmorillonite* a 1 20080924 9806766 200141 646 Ash color rocks B014 Montmorillonite* a 2 20080924 9809904 201079 680 Bare land top soil B016 Opal* g 1 20080924 9810312 201268 672 Ash color surface B019 Montmorillonite* a 1 20080924 9809122 201322 682 Bare land top soil B022 Montmorillonite* e 1 20080924 9807812 201023 700 Ash color top soil with basalt B025 Montmorillonite* a 1 20080924 9806654 201543 654 Sand deposit near to the rive B026 Montmorillonite* e 1 20080924 9806766 200141 646 Yellow color rocks B027 Opal* a 1 20080924 9810134 201316 671 White color intrusion B031 Opal* g 1 20080924 9808486 201274 694 Earthquake crack ‐‐ e 0 20080924 9808854 201209 689 Volcanic rock bolders ‐‐ a 0 20080924 9810294 201359 674 Grass with small trees ‐‐ a 0 20080924 9801228 201096 673 Ash gray color surface ‐‐ g 0 20080924 9807568 201100 712 Trachyte/ basalt mountain ‐‐ g 0 20080924 9807372 199968 671 Basalt rock mountain ‐‐ a 0
119 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
20080924 9808218 200825 704 Basalt rock mountain ‐‐ a 0 20080924 9808296 201125 712 Trachyte /basalt mountain ‐‐ g 0 20080924 9777643 191804 605 Clay bed ‐‐ g 0 20080924 9777645 191737 600 Near to the dry river ‐‐ e 0 20080924 9777646 192088 598 On weathered chert ‐‐ i 0 20080924 9777785 191948 587 Subsurface ‐‐ a 0 20080924 9780091 195939 605 Boundary ‐‐ a 0 20080924 9789909 198293 630 Petrol shed ‐‐ a 0 20080924 9808358 201270 690 Earthquake crack S‐end ‐‐ a 0 20080924 9808403 201262 690 Earthquake crack middle ‐‐ e 0 20080924 9808426 201265 690 Earhquake crack N‐end ‐‐ e 0 20080924 9807134 200217 648 Surface exposure P003 Sedimentary rock a 2 20080924 9809182 201222 685 Gravels on topsoil P007 Basalt a 2 20080924 9806896 200214 648 Gully erosion area P010 Erionite g 2 20080924 9807370 201040 712 Basalt rock mountain P038 Mg clay minerals* a 2 20080924 9810414 201100 672 Silty clay deposite S019 Montmorillonite* g 2 20080924 9806874 200501 682 Chert bed S032 Chert a 2 20080924 9806672 201585 656 Silty clay deposite S037 Amorphose Silica a 2 20080926 9789730 198704 607 Bare top soil surface B021 Montmorillonite* i 1 20080926 9789748 198917 612 Chert bed ‐‐ i 0 20080926 9789614 199109 613 Chert bed ‐‐ i 0 20080926 9789810 200382 622 White color clay bed ‐‐ a 0 20080926 9790378 198427 633 Crushed Trona ‐‐ e 0 20080926 9790604 198614 624 Basalt rock over the surface ‐‐ a 0 20080926 9788788 199347 600 White color evaporites ‐‐ f 0 20080926 9788800 199402 604 White color evaporites ‐‐ f 0 20080926 9788994 199415 599 White color evaporites ‐‐ f 0 20080926 9777244 192006 616 Green color rock P030 Green beds i 2 20080926 9789250 200123 618 Flat huge area ash color P031 Sedimentary rock g 2 20080926 9789706 198941 611 White color surface P032 Magadiite i 2 20080926 9789890 199921 615 Trachyte basalt mountain P033 Basalt a 2 20080926 9789842 199007 607 Green color leach from white P034 Mg clay minerals* i 2 color rock 20080926 9789758 198800 608 Chert bed P035 Chert i 2 20080926 9789758 198556 609 White color hard rock P036 Kenyaite i 2 20080926 9789904 198722 601 Lake evaporites Ring 22 Trona b 3 20080926 9789904 198722 601 Lake evaporites Ring 23 Trona b 3 20080926 9789904 198722 601 lake evaporites Ring 24 Trona b 3 20080926 9788866 199010 602 Chert bed pieces top of the S047 Chert f 2 surface 20080929 9808080 201190 741 Basalt mountain ‐‐ e 0 20080929 9808323 201271 703 Resis_01_01 RES_101 e 4 20080929 9808406 201266 694 Resis_01_02 RES_102 e 4 20080929 9808496 201257 691 Resis_01_03 RES_103 e 4 20080929 9808584 201254 690 Resis_01_04 RES_104 e 4 20080929 9808678 201244 690 Resis_01_05 RES_105 e 4 20080929 9808763 201238 690 Resis_01_06 RES_106 a 4 20080929 9808264 201181 691 Resis_02_01 RES_201 g 4 20080929 9808348 201195 690 Resis_02_02 RES_202 g 4 20080929 9808429 201216 691 Resis_02_03 RES_203 g 4 20080929 9808523 201239 691 Resis_02_04 RES_204 e 4 20080929 9808608 201261 691 Resis_02_05 RES_205 e 4
120 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
20080929 9808694 201280 691 Resis_02_06 RES_206 e 4 20080930 9778258 191883 608 Bare land top soil B036 ………………… a 1 20080930 9800978 193345 663 Basalt rock ‐‐ e 0 20080930 9802170 199371 612 Green chert bed ‐‐ i 0 20080930 9802168 199450 615 Green chert bed P024 Chert i 2 20080930 9802172 199421 612 Green chert bed with white P012 "Oldhamite" i 2 rock 20080930 9803680 199584 607 Trachyte rock P027 Basalt a 2 20080930 9802172 199421 610 Green chert bed with white P028 Diatomite i 2 rock 20080930 9803824 199190 608 Lagoons Ring 02 Trona a 3 20080930 9804070 199150 611 Lagoons Ring 03 Trona a 3 20080930 9804056 199306 613 Lagoons Ring 04 Trona e 3 20080930 9803926 199359 602 Lagoons Ring 05 Trona a 3
Coordinate System: UTM ‐ WGS 84 – Zone 37S. * = Recognized the mineralogy using The Spectral Geologist Software. But not conformed. Type: 1= Soil Samples, 2 = Rock Samples, 3 = Undisturbed Surface Soil Samples, 4 = Resistivity line locations, 0 = No samples “Oldhamite” = Mineralogically not Olhamite but X‐ray diffraction gives diffraction pattern similar to Oldhamite Column, M.U. = Map units of the classified surface mineral image. a = Unclassified b = Trona c = water d = brine e = Basalt/ Alkali trachyte f = High Magadi beds g = Clay minerals h = Na‐Al‐Si gel i = Chert
121 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 04. XRD Measurement Conditions:
Dataset Name ITC File name F:\XRD\XRD_XRDML\A001_1.xrdml Comment Configuration=XPert‐APD, Owner=System, Creation date=23‐3‐2006 9:10:36 Goniometer=PW3020 (Theta/2Theta); Minimum step size 2Theta:0,005; Minimum step size Omega:0,005 Sample stage=PW1774 Spinner Diffractometer system=XPERT Measurement program=A001, Owner=System, Creation date=9‐12‐2008 15:45:20 trona Measurement Date / Time 9‐12‐2008 15:46:09 Operator xrdadmin Raw Data Origin XRD measurement (*.XRDML) Scan Axis Gonio Start Position [°2Th.] 4.0100 End Position [°2Th.] 83.9900 Step Size [°2Th.] 0.0100 Scan Step Time [s] 0.5000 Scan Type Continuous Offset [°2Th.] 0.0000 Divergence Slit Type Automatic Irradiated Length [mm] 12.00 Specimen Length [mm] 10.00 Receiving Slit Size [mm] 0.1000 Measurement Temperature [°C] 25.00 Anode Material Cu K‐Alpha1 [Å] 1.54060 K‐Alpha2 [Å] 1.54443 K‐Beta [Å] 1.39225 K‐A2 / K‐A1 Ratio 0.50000 Generator Settings 10 mA, 10 kV Diffractometer Type 0000000021250696 Diffractometer Number 0 Goniometer Radius [mm] 173.00 Dist. Focus‐Diverg. Slit [mm] 91.00 Incident Beam Monochromator No Spinning No
122 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 05. XRD Peak list of trona Trona samples from different areas in the world show different XRD patterns even though, the chemical composition is same. Therefore crystal structure of the trona can be change according to the crystallographic defects or due to the presents and degree of impurities.
Pos. Height FWHM d‐spacing Rel. Int. Tip width Matched by [°2Th.] [cts] [°2Th.] [Å] [%] [°2Th.] 9.0624 264.82 0.6298 9.75846 27.08 0.7557 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 18.1460 408.91 0.9446 4.88886 41.81 1.1336 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 22.1077 15.69 1.2595 4.02091 1.60 1.5114 00‐029‐1447; 00‐002‐0601; 00‐011‐0643 29.1855 354.15 0.9446 3.05992 36.21 1.1336 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 32.1353 50.86 0.6298 2.78546 5.20 0.7557 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 33.9417 373.85 0.9446 2.64124 38.22 1.1336 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 36.8667 250.61 0.9446 2.43812 25.62 1.1336 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 44.6168 978.03 0.6298 2.03096 100.00 0.7557 00‐029‐1447; 00‐002‐0601; 00‐011‐0643; 00‐001‐1077 46.2295 137.23 0.6298 1.96380 14.03 0.7557 00‐029‐1447; 00‐002‐0601; 00‐011‐0643 48.0908 43.82 0.9446 1.89205 4.48 1.1336 00‐029‐1447; 00‐002‐0601; 00‐011‐0643 52.0568 79.82 1.2595 1.75686 8.16 1.5114 00‐029‐1447; 00‐002‐0601; 00‐001‐1077 55.9455 176.71 0.9446 1.64361 18.07 1.1336 00‐029‐1447; 00‐002‐0601; 00‐001‐1077 66.7063 79.29 0.6298 1.40222 8.11 0.7557 00‐029‐1447; 00‐002‐0601 70.4225 62.21 0.6298 1.33706 6.36 0.7557 00‐002‐0601; 00‐001‐1077 75.6466 41.88 0.9446 1.25718 4.28 1.1336 00‐002‐0601 82.0977 21.73 0.7680 1.17298 2.22 0.9216 00‐002‐0601
123 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Mineral name ‐ Trona PDF index name ‐ Sodium Hydrogen Carbonate Hydrate
Empirical formula ‐ C2H5Na3O8
Chemical formula ‐ Na3H(CO3)2. 2H2O
Reference code 00‐002‐0601 00‐029‐1447 00‐011‐0643 00‐001‐1077 Crystallographic parameters Crystal system Monoclinic Monoclinic Monoclinic Monoclinic Space group C2/c I2/a I2/a C2/c Space group number 15 15 15 15 a (A0) 20.4100 20.1060 20.1100 20.4100 b (A0) 3.4900 3.4920 3.4900 3.4900 c (A0) 10.3100 10.3330 10.3100 10.3100 Alpha (0) 90.0000 90.0000 90.0000 90.0000 Beta (0) 106.3300 103.0500 103.1000 106.3300 Gamma (0) 90.0000 90.0000 90.0000 90.0000 Calculated density 2.13 2.12 2.13 2.13 (g/Cm3) Volume of cell (106 704.76 706.74 704.77 704.76 Pm3) Z: 4.00 4.00 4.00 4.00 RIR ‐ 0.96 ‐ ‐ Status, Subfiles and Quality Status Marked as ‐ Marked as deleted Marked as deleted deleted by ICDD by ICDD by ICDD Subfiles Inorganic Mineral Inorganic Mineral Inorganic Mineral Inorganic Mineral Quality Blank (B) Star (S) Blank (B) Indexed (I) Comments Deleted by Berry parcel June ‐ Deleted by 29‐1447 Berry parcel June 26, 1959 26, 1959. Color Colorless Colorless Colorless, dirty white Colorless Sample source Lake Magadi, Sweetwater Searles Lake, ‐ (Specimen from) Kenya Colony. County, Wyoming, California, USA USA. Melting point 195 ‐ 195 Unit cell data source Brown, Peiser, Powder Dtffraction ‐ Brown, Peiser, Tumer‐Jones., Tumer‐Jones., Acta Acta Crystallogr., Crystallogr., 2, 167, 2, 167, (1949) (1949) Primary reference British Museum Natl. Bur. Stand. Pabst., Am. Mineral., ‐ (Natural History) (U. S.) Monogr. 25, 44, 274, (1959) 15, 71, (1978)
124 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 06. XRD Peak list of Erionite: 01. Sample No: P010_02
Pos. Height FWHM d‐spacing Rel. Int. Tip width Matched by [°2Th.] [cts] [°2Th.] [Å] [%] [°2Th.] 7.8701 21.35 0.2362 11.23394 11.44 0.2834 00‐042‐0373 13.5068 25.73 0.3936 6.55578 13.78 0.4723 00‐010‐0361; 01‐083‐1618; 00‐042‐0373 20.0822 4.49 0.9446 4.42165 2.41 1.1336 20.6484 50.80 0.1181 4.30167 27.22 0.1417 00‐042‐0373 21.7812 38.37 0.2362 4.08045 20.56 0.2834 00‐010‐0361; 01‐083‐1618 23.8098 80.68 0.1968 3.73719 43.23 0.2362 00‐010‐0361; 01‐083‐1618; 00‐042‐0373 27.6374 186.63 0.1181 3.22770 100.00 0.1417 00‐010‐0361; 01‐083‐1618 27.9687 153.57 0.0590 3.19021 82.29 0.0708 00‐010‐0361; 01‐083‐1618 31.3102 44.60 0.3149 2.85695 23.90 0.3779 01‐083‐1618; 00‐042‐0373 33.5798 10.75 0.2362 2.66887 5.76 0.2834 00‐042‐0373 35.3700 29.67 0.2362 2.53778 15.90 0.2834 00‐010‐0361; 01‐083‐1618 37.3065 10.09 0.4723 2.41038 5.41 0.5668 01‐083‐1618 40.9678 10.57 0.2362 2.20303 5.67 0.2834 01‐083‐1618; 00‐042‐0373 41.8571 35.31 0.1968 2.15826 18.92 0.2362 01‐083‐1618 48.5161 9.42 0.6298 1.87646 5.05 0.7557 01‐083‐1618; 00‐042‐0373 49.7183 10.62 0.2362 1.83387 5.69 0.2834 01‐083‐1618 51.2735 6.51 0.9446 1.78184 3.49 1.1336 01‐083‐1618 53.4289 0.29 0.9446 1.71493 0.16 1.1336 01‐083‐1618 55.6041 31.83 0.3149 1.65289 17.06 0.3779 01‐083‐1618 58.2718 10.34 0.4723 1.58341 5.54 0.5668 01‐083‐1618 61.3873 12.71 0.6298 1.51032 6.81 0.7557 01‐083‐1618 72.5928 2.75 0.9446 1.30234 1.48 1.1336 01‐083‐1618 74.4643 13.16 0.4800 1.27312 7.05 0.5760 01‐083‐1618
Ref. Code Compound Name Displacement [°2Th.] Scale Factor Chemical Formula
00‐010‐0361 Anorthoclase, syn 0.000 0.596 Na0.71 K0.29 Al Si3 O8
01‐083‐1618 sodium alumotrisilicate 0.000 0.460 Na Al Si3 O8
00‐042‐0373 Erionite‐K, syn 0.000 0.234 K0.34 Na0.066 Al2 Si7.79 O18.78 ∙x H2 O
125 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
02. Sample No: B020
Pos. Height FWHM d‐spacing Rel. Int. Tip width Matched by [°2Th.] [cts] [°2Th.] [Å] [%] [°2Th.] 8.0348 75.74 0.0984 11.00404 52.20 0.1181 00‐042‐0369 9.8117 17.63 0.3149 9.01486 12.15 0.3779 00‐042‐0369 13.6776 108.60 0.0984 6.47433 74.84 0.1181 00‐042‐0369 18.0240 22.75 0.1968 4.92167 15.68 0.2362 00‐003‐0187 19.7828 28.48 0.4723 4.48790 19.63 0.5668 00‐042‐0369; 00‐003‐0187 20.7923 145.10 0.1181 4.27223 100.00 0.1417 00‐042‐0369 21.6404 23.32 0.2362 4.10669 16.07 0.2834 00‐042‐0369 23.5456 60.87 0.1181 3.77852 41.95 0.1417 00‐042‐0369; 00‐031‐0633 23.9647 120.45 0.1378 3.71339 83.01 0.1653 00‐042‐0369 25.1483 67.39 0.1574 3.54123 46.45 0.1889 00‐042‐0369; 00‐003‐0187; 00‐031‐0633 26.2442 21.72 0.2362 3.39580 14.97 0.2834 27.2135 80.58 0.1574 3.27700 55.54 0.1889 00‐042‐0369 28.4990 22.52 0.3936 3.13205 15.52 0.4723 00‐042‐0369 30.9239 116.21 0.1574 2.89176 80.09 0.1889 00‐042‐0369 31.4592 137.74 0.1378 2.84376 94.93 0.1653 00‐042‐0369 31.7089 103.50 0.0590 2.82193 71.33 0.0708 00‐042‐0369; 00‐031‐0633 32.0382 88.79 0.1181 2.79368 61.19 0.1417 00‐042‐0369; 00‐031‐0633 33.6843 31.75 0.1968 2.66083 21.88 0.2362 00‐042‐0369 34.8403 50.63 0.3542 2.57515 34.90 0.4251 00‐003‐0187 36.1554 43.88 0.1574 2.48443 30.24 0.1889 00‐042‐0369; 00‐031‐0633 39.7492 6.23 0.6298 2.26771 4.29 0.7557 41.1698 21.20 0.2362 2.19269 14.61 0.2834 00‐042‐0369; 00‐031‐0633 42.9627 14.82 0.2362 2.10525 10.21 0.2834 43.7390 17.51 0.3149 2.06966 12.07 0.3779 00‐031‐0633 48.4633 18.71 0.2362 1.87838 12.89 0.2834 00‐042‐0369 49.8546 17.89 0.2755 1.82917 12.33 0.3306 00‐031‐0633 50.9846 17.77 0.3149 1.79125 12.25 0.3779 51.7535 36.49 0.1968 1.76643 25.15 0.2362 00‐031‐0633 53.5694 17.23 0.2362 1.71076 11.88 0.2834 00‐003‐0187 55.7344 66.03 0.2362 1.64933 45.51 0.2834 00‐031‐0633 58.4111 21.83 0.3936 1.57997 15.04 0.4723 61.3944 44.81 0.4723 1.51016 30.88 0.5668 00‐031‐0633 63.6420 16.13 0.3149 1.46214 11.11 0.3779 66.7358 3.86 0.4723 1.40167 2.66 0.5668 00‐031‐0633 68.4274 7.28 0.9446 1.37109 5.02 1.1336 00‐031‐0633 70.5616 12.99 0.4723 1.33476 8.95 0.5668 00‐031‐0633 72.5707 15.17 0.7872 1.30268 10.45 0.9446 00‐003‐0187 74.6367 18.57 0.3149 1.27166 12.80 0.3779 00‐031‐0633 76.6968 6.98 0.6298 1.24256 4.81 0.7557 00‐003‐0187 78.4255 14.01 0.5760 1.21844 9.65 0.6912
Ref. Code Compound Name Displacement [°2Th.] Scale Factor Chemical Formula
00‐042‐0369 Erionite‐K, syn 0.000 0.400 K0.28 Na0.192 Al2 Si10.43 O25.86 ∙x H2 O
00‐003‐0187 Halloysite‐7A 0.000 0.154 Al2 Si2 O5 ( O H )4
00‐031‐0633 Fayalite, magnesian 0.000 0.232 ( Fe , Mg )2 Si O4
126 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
03. Sample No: B024
Pos. Height FWHM d‐spacing Rel. Int. Tip width Matched by [°2Th.] [cts] [°2Th.] [Å] [%] [°2Th.] 7.8044 127.03 0.0689 11.32841 27.57 0.0827 00‐022‐0854 9.7285 16.68 0.1574 9.09179 3.62 0.1889 00‐022‐0854 11.8110 27.58 0.1574 7.49296 5.99 0.1889 00‐022‐0854 13.4748 109.74 0.1771 6.57130 23.82 0.2125 00‐022‐0854 15.4807 15.97 0.4723 5.72404 3.47 0.5668 00‐022‐0854
16.5952 18.88 0.2362 5.34205 4.10 0.2834 00‐022‐0854
19.5128 48.45 0.3149 4.54939 10.52 0.3779 00‐022‐0854 20.5749 167.15 0.0984 4.31688 36.28 0.1181 00‐022‐0854 21.4403 44.86 0.1378 4.14455 9.74 0.1653 00‐022‐0854 23.3539 72.63 0.1181 3.80910 15.77 0.1417 00‐022‐0854 23.7483 234.04 0.0984 3.74672 50.81 0.1181 00‐022‐0854 24.9319 113.71 0.1378 3.57148 24.68 0.1653 00‐022‐0854 26.9958 95.35 0.1378 3.30293 20.70 0.1653 00‐022‐0854 28.4126 65.92 0.0984 3.14138 14.31 0.1181 29.0977 46.66 0.1574 3.06895 10.13 0.1889 01‐076‐0739 30.5791 22.13 0.1574 2.92357 4.80 0.1889 00‐022‐0854
31.2804 141.83 0.1378 2.85960 30.79 0.1653 00‐022‐0854
31.5043 187.30 0.0984 2.83979 40.66 0.1181 00‐022‐0854; 01‐076‐0739 31.8011 460.67 0.0689 2.81396 100.00 0.0827 00‐022‐0854 33.5038 44.29 0.1574 2.67475 9.62 0.1889 00‐022‐0854 33.9090 71.44 0.1181 2.64370 15.51 0.1417 01‐076‐0739 34.7580 41.49 0.3936 2.58105 9.01 0.4723 01‐076‐0739 35.9427 46.97 0.1181 2.49865 10.20 0.1417 00‐022‐0854; 01‐076‐0739 36.2399 49.96 0.1181 2.47884 10.85 0.1417 00‐022‐0854; 01‐076‐0739 36.8035 27.94 0.1574 2.44216 6.06 0.1889 01‐076‐0739 40.0447 30.33 0.1574 2.25166 6.58 0.1889 01‐076‐0739 40.9390 25.89 0.1574 2.20451 5.62 0.1889 00‐022‐0854 42.7489 25.37 0.1968 2.11528 5.51 0.2362 00‐022‐0854; 01‐076‐0739
43.5095 19.78 0.2362 2.08005 4.29 0.2834 00‐022‐0854
44.7129 17.36 0.3936 2.02682 3.77 0.4723 01‐076‐0739 45.5104 231.64 0.1181 1.99314 50.28 0.1417 01‐076‐0739 46.3486 10.60 0.2362 1.95903 2.30 0.2834 01‐076‐0739 48.3335 38.32 0.1968 1.88312 8.32 0.2362 00‐022‐0854; 01‐076‐0739 49.6874 24.70 0.1181 1.83493 5.36 0.1417 00‐022‐0854
Ref. Code Compound Name Displacement Scale Chemical Formula [°2Th.] Factor
00‐022‐0854 Erionite 0.000 0.486 ( Na , K )8 ( Si , Al )36 O72 ∙23 H2 O
01‐076‐0739 trisodium 0.000 0.152 Na2 C O3 Na H C O3 ( H2 O )2 carbonate hydrogencarbonate dihydrate
127 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 07. Reflectance spectra of collected soil samples.
samples.
soil
(b) collected
the
in
minerals
rich
Silica
and
(a) Halloysite
of
spectra
reflectance
removed
(b) continuum
(b)
and
Normal (a)
07.1.
A:
(a) Figure:
128 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
samples.
soil
(b)
collected
the
minerals in
clay
(a) Montmorillonite
of
spectra
reflectance
removed
(b) continuum
(b)
and
Normal
(a)
07.2.
A:
(a) Figure:
129 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
samples.
soil
(b) collected
the
in
minerals
clay
(a) Montmorillonite
of
spectra
reflectance
removed
(b) continuum
(b)
and
Normal
(a)
07.3.
A:
(a) Figure:
130 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
soil
spectra
visually
“The by collected
using the reflectance
done
in
and
removed
basically
(b) is features
characteristics)
continuum
(b)
software. spectral
(absorption) and identification
(TSG) (using
Normal
spectral mineral
(a)
minerals
their Geologist”
07.5.
section,
A:
this (a) undefined
In Spectral analyzing samples. of Figure:
samples.
soil reflectance
removed collected
(b) (b) the
in continuum
Calcite
(b)
and
and
Normal minerals
(a)
clay
Mg 07.4.
of A:
(a) (a) spectra Figure:
131 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
Appendix: 08. Existence of ground water reservoir below the high Magadi bed
Two resistivity profiles were acquired in southern part of the Lake Magadi area (Fig: A: 08.1(a)) during the fieldwork using SYSCAL R1 PLUS resistivity tomography instrument (Fig: A: 08.1(b)). This instrument automatically acquired 759 readings from 72 electrodes which were installed within 5m spacing. The Wenner‐schlumberger array was used with one rollalong for each line to cover more cross sectional area with 304 extra readings (Fig: A: 08.1(c)). The data was imported in to RES2Dinv software through PROSYS11 (software) and apparent resistivity profile was derived from least square inversion method in RES2Dinv. The apparent resistivity profiles are shown in Fig: A: 08.1(d).
ERI 2 (b)
ERI 1
(a) (c)
W-E ERl 1
W-E
ERl 2 (d)
Figure: A: 08.1 (a) Locations of the Resistivity Profiles. (b) Used resistivity instrument. (c) Schematic diagram of Main sequence and rollalong method. (d) Resistivity profiles.
132 HYPERSPECTRAL MAPPING OF SURFACE MINERALOGY IN THE LAKE MAGADI AREA IN KENYA.
According to above two resistivity profiles, presence of ground water reservoir in the Lake Magadi area can be conformed. Because, very low apparent resistivity value is mostly responsible for presence of very high conductive medium in subsurface, it is most frequently due to the ground waters. The size of the subsurface water body has increased with the closeness to the Lake as well as the closeness to the hot springs. Even though, surface of the area very dry, below few meters (~2‐m) low resistive ground water reservoir can be find giving indication of shallow ground water reservoir for this model.
133