Integration Of Reflectance To Study Glacier Surface Using Landsat 7 ETM+: A Case Study Of The Petermann Glacier In Greenland Félix O. Rivera Santiago Department Of Geology, University Of Puerto Rico, Mayaguez Campus, P.O. Box 9017 Mayaguez Puerto Rico,00681 Abstract- Glaciers are masses of ice and granular snow formed by compactation and recrystallization of snow. The Petermann Glacier located in the northwestern part of Greenland has suffered 4 major scalping events in the past 50 years : 1) 1959-1961 (153km2), 2)1991 (168km2), 3)2001 (71km2), 4)2010 (270km2).Snow and Ice have high reflectance in the visible range and low reflectance in infrared region. Our objective was to identify zones of accumulation and abrasion, identify areas of snow and ice and underline any geomorphologic features found in the glacier using two images: one from June 2009 and another one from June 2011 using the Landsat 7 ETM+ sensor. Images were analyzed using ENVI 4.8. Two math equations were applied NDSI (Normalized Snow Different Index) and SIR (Snow and Ice Ratio). Since the acquired images from the Landsat 7 ETM+ had the Scan Line Corrector (SLC) problem; a subset of a specific area was used. With the NDSI we were able to determine areas with snow and areas without snow indicating a drastic change in 2009 with more exposed bedrock (ablation) and 2011 with more snow accumulation, SIR was able to underline ice in the glacier and boundaries. NDSI determined that higher temperatures could have affected the glacier in 2009 due to more exposed bedrock and more snow accumulation due to lower temperatures in 2011. Geomorphologic features were useful to determine the north moving direction of the glacier and compression areas. Landsat 7 ETM+ images with the SLC problem are still useful for scientific purposes, more research need to be done to determine temperature changes from this glacier using other sensors. Keywords : Petermann, Glacier, NDSI, K-Means, SIR, Greenland 1
INTRODUCTION Glaciers in one way or another is responsible for altering the landscape especially in high latitudes. These are created by the compaction and crystallization of snow and granular snow, this change from snow to ice is thick enough to create motion in land. Dynamic classification on glaciers is based on their mass balance that determines if a glacier is advancing, neutral or retrieving. Mass balance can be determined by two processes: accumulation and ablation. Accumulation is the process in which snow and ice are added to the glacier; and ablation refers to the process in which a snow and ice is lost from a glacier. Calving or mechanical loss of ice in glaciers is a major contributor in mass loss from Greenland Ice Sheet and other glaciers (Benn et al., 2007). Rapid retreat followed by a climate change increases the potential of glacier to contribution to sea level rise. (Benn et al., 2007). Fast ice discharge and melt have been more common from marine ending boundary of glaciers in western and southeastern Greenland (Johannessen et al., 2011). Increasing changes in temperature have drawn attention in the Arctic due to climate models that predict Arctic warming and change in mass balance of the glaciers. One of the effective ways to monitor the cryosphere is through satellite remote sensing and reflectance curves. This study focuses in analyzing Landsat 7 ETM + images from 2009 and 2011 from Greenland s Petermann glacier that has been studied throughout the past years because of four major scalping events in the past 50 years: 1) 1959-1961 (153km2), 2)1991 (168km2), 3)2001 (71km2), 4)2010 (270km2) (Johannessen et al., 2011).This evets may indicate higher temperatures of affecting this glacier. The objective of this study was to apply reflectance to: 1) Identify areas of accumulation and ablation in the glacier; 2) Identify areas of snow and ice in the glacier and; use a true color image to: 3) underline and interpret any geomorphologic features found in the glacier. I intend to answer a series of question: 1) How Remote Sensing can help monitor this glacier? 2) Is it possible to speculate different temperature changes with just two images from two different years? Study Site METHODOLOGY This study is focused on satellite images from one glacier located in the northern part of Greenland know as the Petermann Glacier (Latitude: N 80 30' 0" Longitude: W 59 30' 0"). In Greenland the Ice sheet cover about 1.7 million km 2 that are drained by glaciers that cease in land, sea and by runoff by summer melting (Johannessen et al., 2011). This glacier connects the Greenland Ice sheet with the Arctic Ocean. The ~70 km long and ~15 km wide glacier is a floating ice shelf that drains through some of the largest and widest fjords (Kelly and Lowell, 2009). On August 5 2010 a major calving event removed around 28 km of the 70 km ice, producing an ice island that measured ~270 km 2 and it 2
was considered one of the biggest calving events observed in the past 1-2 decades (Johannessen et al., 2011). Method Methodology (Figure 2) can be divided in five main different sections: 1) Image acquisition, 2) ENVI program 3) spatial subset 4) Processing 5) Results/Interpretation. I will describe in detail every step. First two images from the Landsat 7 ETM+ were ordered from USGS website, one from June 2009 and another from June 2011. Images from 2010 were not useful because they had clouds. Once the images were acquired, I used the program Environment for Visualizing Images 4.8 (ENVI 4.8) to analyze in true color the images the images. Since the Landsat 7 has the scan line corrector (SLC) problem shown in Figure 1 a spatial subset was done at the boundaries where the SLC problem was not interfering. Using ENVI I made a spatial subset of one image based on a Region of Interest (ROI). This ROI was then applied to the 2011 image in order to focus on the same region in both images. Landsat Images ENVI Sub set via ROI Processing NDSI Result/Interpretations FIG. 1. Methodology used to analyze and interpret images from the Petermann Glacier. Processing SIR K Means classification True color Image Data processing can be sub-divided in four sections in which I will first mention and then explain how each process was applied: 1) Normalized Snow Differentiation Index (NDSI), 2) Snow and Ice Ratio (SIR) 3) K-means classification, 4)True color Image. 3
map snow, that uses the green band (B2) and the mid-infrared (B5) defined as: NDSI= B2-B5/B2+B5 FIG. 3. Reflectance vs. Wavelength curves for different types of snow and Ice. Notice how the snow and ice have similar reflectance in the visible spectra (400-800 µm) but differ in the infrared region (800-2000 µm) (Xiao et al., 2001) FIG. 2. Landsat 7 ETM + image from June 2009 of the Petermann Glacier (top) and a close up view of the SLC problem (bottom). In order to understand how data processing works we first have to understand different spectral curves from snow and ice (Fig. 3). Snow and Ice have very high reflectance in the visible spectra and but differ in having low reflectance values in the mid-infrared band. Four different types of snow and ice with different sizes, with differences in reflectance are not clearly seen in the visible but are more notable in the mid-infrared (Fig.3). So I will apply both equations using ENVI band math equation. The NDSI to I will also apply the SIR to identify areas of snow and ice that uses the red band (B3) and the near infrared band (B4) defined as: SIR=B3/B4 Once the equations were applied snow had positive pixel values and bedrock had negative values. A mask based on pixel value was applied to differentiate between snow and bedrock A K-Means unsupervised classification with two classes and one iteration was used. Last but not least a true color image was used to identify any structure on the glacier. 4
DISCUSSION Snow Area Cover is difficult to address in the field many equations to identify snow. Here we apply two equations one to identify snow and a second equation to identify snow and ice areas Once the subset was applied to the image we were able to identify Snow Cover Areas (SCA) in both images using pixel values (Fig. 4 and Fig.5). Glacier. Bright and high pixel values from the NDSI indicate high content of snow and low pixel refers to low content of snow. The total Snow Cover Area (SCA) can be identified with NDSI. For example we can see the pixel value ranges from one to cero. Higher pixel values indicate more snow content and low pixel values indicate the opposite. There is a big difference in snow content in June 2009 and June 2011 indicating less SCA in June 2009 and more SCA in 2011. I also noticed that the SCA could be also obtained from a K-means classification with two classes and one iteration (Fig. 6 and Fig. 7). FIG. 4. NDSI for June 2009 of the Petermann Glacier. Bright and high pixel values from the NDSI indicate high content of snow and low pixel refers to low content of snow. FIG. 6. 2009 Petermann Glacier K-means classification with two classes and one iteration. FIG. 5. NDSI for June 2011 of the Petermann Difference between ice and snow in is an important tool to measure any change that might affect the glacier. With a SIR equation I was able to delineate the boundaries of the glacier (Fig. 8 and Fig. 9). In these images brighter or higher pixel values indicate where the ice is located and lower pixel indicate snow or bedrock. There 5
is also a difference in the 2009 image where less ice is concentrated in the glacier. differential movement of the ice. Once the crevasses were outlined (Figure 10) the direction of movement was to the north FIG. 7. 2011 Petermann Glacier K-means classification with two classes and one iteration. FIG. 9. SIR for 2011 Petermann Glacier, notice the bright pixels where the glacier is located indicating the presence of ice. FIG. 8. SIR for 2009 Petermann Glacier, notice the bright pixels where the glacier is located indicating the presence of ice. Geomorphology studies the landforms and landscapes that shape present landscapes. Three main features were outlined (Fig. 10) : moraines, crevasses and movement direction. Moraines are the accumulation of glacier debris transported by the glacier movement; crevasses are produces by tensional fractures produced by FIG. 10. 2009 true color image of the Petermann glacier, moraines, crevasses and direction of movement of the glacier (red thick arrow). CONCLUSION Remote sensing has been widely used to interpret, understand and underline important features in a specific area of interest. Landsat 7 ETM+ proved that even with the SLC problem I was able to determine important information related to the glacier: 6
NDSI and K-means classification highlighted the areas covered with snow and bedrock. SIR focused on the ice cover areas and was able to point out the boundaries of the glacier. True color image outlined sediments transported, direction of movement and compression areas. In conclusion more snow was less snow was accumulated in 2009 and more snow was found in 2011. High temperatures could explain this phenomenon back in 2009 since there was more exposed bedrock and accumulation in this date. Still more research need to be done with other sensors to understand temperature changes. Acknowledgments. - I would like to thank Fernando Gilbes for helping me with the interpretation. Massive calving in 2010 and the Past Half Century. The Cryoshpere, 169-181. Kelly, M. A., & Lowell, T. V. (2009). Fluctuations of local glaciers in Greenland during latest Pleistocene and Holocene time. Quaternary Science Reviews, 2088-2106. Qunzhu, Z., Meisheng, C., Xuezhi, F., Fengxian, L., Xianzhang, C., & Wenkun, S. (1983). A study of Spectral Reflection Characteristics for Snow,Ice and Water in the Norh of China. 145 (pp. 451-462). Hamburg: Hydrological Applications of Remote Sensing and Remote Data Transmission. Roshani, N., Valan, Z. M., Razaei, Y., & Nikfar, M. (2008). Snow Mapping of Alamchal Glacier Using Remote Sensing Data. The International Archives of the Photogrammetry, 805-808. Xiao, X., Shen, Z., & Qin, X. (2001). Assessin the potential of VEGETEATION sensor data for mapping snow and ice cover: a Normalized Differenc Snow and Ice Index. International Journal of Remoite Sensing, 2479-2487. LITERATURE CITED Bamber, J. L., & Rivera, A. (2007). A Review of Remote Sensing Methods fot Glacier Mass Balance determination. Global and Planetary Change, 138-148. Benn, D. I., Warren, R. C., & Mottram, R. H. (2007). Calving processes and the dynamics of calving glaciers. Earth-Science Reviews, 143-179. Gao, J., & Liu, Y. (2001). Applications of Remote Sensing,GIS and GPS in Glaciology: a review. Progress in Physical Geography, 520-540. Hall, D. K., Box, J. E., Casey, K. A., Hook, S. J., Shuman, C. A., & Steffen, K. (2008). Comparison of satellite-derived and in-situ Observation of Ice and Snow Surface Temperatures over Greenland. Remote Sensing of the Environment, 3379 3749. Johannessen, O. M., Babiker, M., & Miles, M. W. (2011). Petermann Glacier, North Greenland: 7