51 Lithological mapping using remote sensing depends, in part, on the identification of rock types by their spectral charac- teristics. Chemical and physical properties of minerals and rocks determine their diagnostic spectral features through- out the electromagnetic spectrum. Shifts in the position and changes in the shape and depth of these features can be ex- plained by variations in chemical composition of minerals. Detection of such variations is vital for discriminating min- erals with similar chemical composition. Compared with multispectral image data, airborne or spaceborne hyperspec- tral imagery offers higher spectral resolution, which makes it possible to estimate the mineral composition of the rocks under study without direct contact. Arctic environments provide challenging ground for geo- logical mapping and mineral exploration. Inaccessibility commonly complicates ground surveys, and the presence of ice, vegetation and rock-encrusting lichens hinders remote sensing surveys. This study  addresses the following objec- tives: 1. Modelling the impact of lichen on the spectra of the rock substrate; 2. Identification of a robust lichen index for the deconvolu- tion of lichen and rock mixtures and 3. Multiscale hyperspectral analysis of lithologies in areas with abundant lichens. Modelling the impact of lichen cover Spectral mixing of lichens and bare rock can shift the wave- length positions of characteristic absorption features and complicate the spectral mapping of minerals and lithologies. Salehi et al. (2017) investigated how surficial lichen cover affects the characteristics of shortwave infrared mineral ab- sorption features and the efficiency of automated extraction of absorption features. For this purpose, mixed spectra were synthetically generated from laboratory spectra of common rock-forming minerals and lichens. Wavelength displace- ments of characteristic absorption features for each mixed spectrum were then analysed as a function of lichen cover percentage (see an example in Fig. 1). By quantifying lichen Hyperspectral analysis of lithologies in the Arctic in areas with abundant lichen cover Sara Salehi 2000 2100 2200 2300 2400 Wavelength (nm) 0.15 0.02 0.25 0.30 0.35 0.40 Re fle ct an ce Kimberlite Lichen Hull A B quotient 2240 2260 2280 2300 2320 2340 Wavelength (nm) 0.85 0.90 0.95 1.00 H ul l q uo tie nt 0 Kimberlite / 100 Lichen 50 Kimberlite / 50 Lichen 100 Kimberlite / 0 Lichen Fig. 1. A: Averaged spectra of pure rock and lichen for a kimberlite sub- strate in shortwave infrared range. B: The corresponding hull quotient (Clark & Roush 1984) and band centres of mixed spectra associated with the antigorite absorption feature. The 10% spectral intervals used to in- vestigate the main absorption features are highlighted. x: wavelength posi- tions of local minima (Salehi et al. 2017). © 2018 GEUS. Geological Survey of Denmark and Greenland Bulletin 41, 51–55. Open access: www.geus.dk/bulletin http://www.geus.dk/bulletin 5252 cover effects on mineral absorption features, this study high- lights the importance cautious interpretation in areas char- acterised by abundant, lichen-covered outcrops. This can be of significant importance for mineral and deposit identifica- tion, because slightly shifted features for a given spectrum caused by lichen cover can be erroneously identified as a path to a deposit. Salehi et al. (2017) showed that spectral shifts caused by lichens are not constant, i.e. each mineral spectral feature may be affected differently depending on the shape of the lichen spectrum. For example, the absorption feature re- lated to the chlorite mineral group around 2254 nm is shifted towards longer wavelength, while the one around 2320 nm is shifted towards shorter wavelength and the 2380 nm band maintains its spectral characteristics. Spectral shifts are not only related to rock/lichen proportions but also to the modal abundance of minerals in certain rock types. Background minerals and associated overlapping features will have an ef- fect on the related absorption depth and play a critical role in the scale of wavelength displacement. Identification of a robust lichen index The ability to distinguish a lichen cover from its rock/miner- al substrate is important, and decomposition of a mixed pixel into a collection of pure ref lectance spectra can improve the use of hyperspectral methods for mineral exploration. In or- der to identify spectral indices that can directly ref lect the ra- tio of the rock and lichen in hyperspectral data, a number of index structures were assigned to an optimisation algorithm, which was tasked to find the best values for the location of the bands along the ref lectance spectra measured in the labo- ratory (Salehi et al. 2016). In order to further investigate the functionality of the indices for the airborne platform, the spectra were resampled to HyMAP resolution. The indices proposed by Salehi et al. (2016) proved robust to the type of the substrate rock and permitted an estimate of the lichen cover with acceptable, albeit varying, levels of error. The results revealed that the ratio between R 894-1246 and R 1110 explains most of the variability in the hyperspectral data at the original laboratory resolution (R 2=0.769). However, the normalised index incorporating R 1106-1121 and R 904-1251 yields the best results for the HyMAP resolution (R 2=0.765). Re fle ct an ce (o ffs et fo r cl ar ity ) Wavelength (nm) Wavelength (nm) olivine (USGS) talc (USGS) anthophyllite (USGS) hornblende (USGS) hornblende (HyMAP) antigorite-anthophyllite (HyMAP) talc-olivine/pyroxene (HyMAP) actinolite (HyMAP) serpentine-olivine (HyMAP) serpentine (USGS) actinolite (USGS) actinolite-hornblende (HyMAP) A B 500 1000 1500 2000 2200 2300 2400 Fig. 2. Spectra of extracted end members compared with selected minerals from the USGS spectral library in ENVI software. A: full spectral range. B: shortwave infrared range. 53 The proposed methodology has the advantage of not re- quiring a priori knowledge about the exact effects of lichens – or any other substance – on the ref lectance of the mixtures. Instead, this information is obtained by an automated trial and error process. Therefore, this technique can also be ben- eficial for identification of sensitive bands and indices for de- convolution of any other mixed spectra, whether synthetic as in this case, or obtained directly from the samples. Multiscale hyperspectral analysis of lithologies with abundant lichen cover Two sets of hyperspectral data acquired by airborne HyMap (350–2500 nm) and light-weight Rikola (500–900 nm) sen- sors were chosen to investigate the potential of visible near infrared and shortwave infrared spectral range for detailed lithological mapping in the Nagssugtoqidian orogen of West Greenland, where an ultramafic rock unit with abundant li- chen cover is exposed. The extent to which geological infor- mation derived from airborne data is retained in the Rikola Serpentine + pyroxene 0 End-member abundance (%) Serpentine + olivine/pyroxene Amphibole (actinolite/hornblende) Lichen Vegetation 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 66 °4 4' 38 ''N 66 °4 4' 10 ''N 52°31'26'' W 52°30'22'' W 500 m Fig. 3. The result of unmixing analysis and the abundance of mafic-ultramafic minerals using HyMAP data. Masked pixels are indicated by black colour. Ricola VNIR B A B C D Fig. 4. A: Supracrustal rocks in the Innarsuaq region comprising a kilometre-sized body of mafic-ultramafic, looking north. B: bright green and black amphibole and biotite at the corner (alteration zone). C: Host ultramafic rock; green amphibole to the left and white talc vein in the middle. D: 50 cm long asbestos fibres. 5454 hyperspectral data, is examined as an insight to future drone- based hyperspectral mapping capabilities and the possibility of extracting valuable mineralogical and lithological infor- mation using such platforms. The airborne hyperspectral dataset is corrected for ab- normal pixels and removal of bad bands (such as water va- pour absorption features and noisy bands) prior to atmos- pheric correction. Dark pixels, snow, clouds and water were filtered out. Next, the spatial–spectral end-member extrac- tion method (Rogge et al. 2007) is used to derive an image end-member set. This makes an assessment of subtle litho- logical variability across a given study area possible. These end members are then sorted based on expert knowledge of known spectral features (water, snow, vegetation, lichen and geological materials) followed by a more detailed sorting into individual classes within each category. Subtle shortwave infrared features related to key minerals in the geological materials are particularly important. The resulting sorted end-member classes are subsequently averaged to produce a final end-member set. A final set of six geological end mem- bers (Fig. 2), and two end members related to vegetation and lichens are deducted from expert-based analysis. Figure 2 Mafic Lichen and vegetation 500 m 50 m Ultramafic A B C W E 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 Actinolite Serpentine/olivine Talc Anthophyllite Lichen and Vegetation Legend (C). End-member abundance (%) Fig. 5. A: True-colour hyperspectral image mosaic generated using the Rikola camera. B: Minimum Noise Fraction false-colour image: Red: Band 6. Green: Band 2. Blue: Band 1. C: result of spectral unmixing analysis and the abundance of mafic-ultramafic minerals. 55 shows a plot of the extracted end members using the spatial– spectral end-member extraction method and the correspond- ing signatures from the United States Geological Survey (USGS) spectral library (Kokaly et al. 2017). The shortwave infrared spectral characteristics of the ultramafic rocks stud- ied here were controlled by amphibole minerals as exemplified by hornblende, actinolite and anthophyllite (Fig. 3). The ab- sorption features in the shortwave infrared region are located at 2320 and 2380 nm and are of the same order of magnitude. The shortwave infrared spectrum of olivine-rich rocks clearly ref lects a mixture of antigorite serpentine with a characteristic stronger absorption feature at 2320 nm. A less distinct absorp- tion feature at 2310 nm is present for rocks enriched in talc. Fractional abundances of the end members within the scene are determined using an iterative implementation of spectral mixture analysis method (Rogge et al. 2007). The interpretation of HyMap data revealed a number of mafic and ultramafic complexes in the border area be- tween the parautochthonous and allochthonous zones of the Nagssugtoqidian orogen. One such complex occurs to the east of the head of the fjord Kangerluarsuk, here referred to as Innarsuaq (see fig. 1 of Salehi & Thaarup 2018, this vol- ume). As can be seen from Fig. 3, the predictive map from the Innarsuaq area displays a complex distribution of exposed bedrock, a feature confirmed during a brief field visit. The results were validated using expert knowledge of spectral characteristics of lichens and mineralogy, as well as spectral measurements of field samples and associated XRD results. The Rikola camera was operated in ground-based mode and panned stepwise to acquire a set of five overlapping im- ages. The images were corrected for geometric, radiometric and topographic effects and stitched to a continuous mo- saic (Figs 4, 5). The distribution of lithological  units  were then mapped using the Minimum Noise Fraction method (Kruse et al. 1993).The information regarding mineral abundances were retrieved using the Spectra Unmixing procedure (Fig. 5). Conclusions 1. Lichen effects on the spectra of their rock substrate have important implications for the geological analysis of air- borne/spaceborne hyperspectral data where rock-encrust- ing lichens partially obscure exposed bedrock. 2. Analysis of airborne hyperspectral data can result in high- quality regional mapping products capable of discriminat- ing geological materials of interest based on subtle spectral differences. The map product generated from the Rikola scenes in this study captures the broad geological patterns and many of the lithologies generated from the airborne data, although some spectral and lithological discrimina- tion is lost due to its more limited wavelength range. 3. The performance of hyperspectral data acquired from different platforms and at various scales is investigated for qualitative mapping of arctic mineral resources in the presence of abundant lichens. The application of such technologies to extract detailed geological information from complex inaccessible regions of Greenland certainly has a very low cost/benefit ratio in comparison to tradi- tional geological fieldwork. Future space-borne hyperspec- tral sensors will offer new possibilities to expand the scale of mapping in Greenland. Integration with other remote sensing datasets such as magnetic data will simplify min- eral exploration and geological mapping in the Arctic. Acknowledgments The Helmholtz Institute Freiberg is thanked for the use of Rikola Hyper- spectral Imager. References Clark, R.N. & Roush, T.L. 1984: Ref lectance spectroscopy: Quantitative analysis techniques for remote sensing applications. Journal of Geo- physical Research: Solid Earth 89(B7), 6329–6340. Kokaly, R.F., Clark, R.N., Swayze, G.A., Livo, K.E., Hoefen, T.M., Pear- son, N.C., Wise, R.A., Benzel, W.M., Lowers, H.A., Driscoll, R.L.& Klein, A.J. 2017: USGS Spectral Library Version 7. U.S. Geological Sur- vey Data Series 1035, 61 pp., http://dx.doi.org/10.3133/ds1035 Kruse, F.A., Lef koff, A., Boardman, J., Heidebrecht, K., Shapiro, A., Bar- loon, P. & Goetz, A. 1993: The spectral image processing system (SIPS) – interactive visualization and analysis of imaging spectrometer data. Remote Sensing of Environment 44, 145–163 Rogge, D.M., Rivard, B., Zhang, J., Sanchez, A., Harris, J., & Feng, J. 2007: Integration of spatial–spectral information for the improved extraction of endmembers. Remote Sensing of Environment 110, 287–303. Salehi, S. & Thaarup, S. 2018: Mineral mapping by hyperspectral remote sensing in West Greenland using airborne, ship-based and terrestrial platforms. Geological Survey of Denmark and Greenland Bulletin 41, 47–50 (this volume). Salehi, S., Karami, M. & Fensholt, R. 2016: Identification of a robust li- chen index for the deconvolution of lichen and rock mixtures using pat- tern search algorithm (case study: Greenland). International Archives of the Photogrammetry, Remote Sensing & Spatial Information Sci- ences XLI-B7, 973–979. Salehi, S., Rogge, D., Rivard, B., Heincke, B.H. & Fensholt, R. 2017: Modeling and assessment of wavelength displacements of characteris- tic absorption features of common rock forming minerals encrusted by lichens. Remote Sensing of Environment 199, 78–92. Author’s address S.S., Geological Survey of Denmark and Greenland, Øster Voldgade 10, DK-1350 Copenhagen K, Denmark. E-mail: ssal@geus.dk. http://dx.doi.org/10.3133/ds1035 mailto:ssal@geus.dk